You may notice some images loading slow across the Open Collections website. Thank you for your patience as we rebuild the cache to make images load faster.

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Design and scheduling of agricultural biomass supply chain for a cellulosic ethanol plant Ebadian, Mahmood 2013

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2013_fall_ebadian_mahmood.pdf [ 2.02MB ]
Metadata
JSON: 24-1.0073710.json
JSON-LD: 24-1.0073710-ld.json
RDF/XML (Pretty): 24-1.0073710-rdf.xml
RDF/JSON: 24-1.0073710-rdf.json
Turtle: 24-1.0073710-turtle.txt
N-Triples: 24-1.0073710-rdf-ntriples.txt
Original Record: 24-1.0073710-source.json
Full Text
24-1.0073710-fulltext.txt
Citation
24-1.0073710.ris

Full Text

DESIGN AND SCHEDULING OF AGRICULTURAL BIOMASS SUPPLY CHAIN FOR A CELLULOSIC ETHANOL PLANT by Mahmood Ebadian B.Sc., Amirkabir University of Technology, 2003 M.Sc., University of Tehran, 2006  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in  The Faculty of Graduate Studies (Forestry)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) May 2013  © Mahmood Ebadian, 2013  Abstract The overall objective of this dissertation is to design and schedule a highly constrained agricultural biomass supply chain to meet the daily biomass demand of a commercial-sized cellulosic ethanol plant at the minimum delivery cost possible. To this end, an integrated simulation/optimization model is developed. The developed simulation model plans and schedules a flow of multi-biomass in the supply chain to meet the daily demand subject to the dynamics and stochasticity of the supply chain. The developed optimization model is used to meet the annual demand at the minimum delivery cost by prescribing the design of the supply chain. The design includes the selection of farms, the location of storage sites, and the assignment of the farms to the storage sites. It also determines the flow of biomass between farms, storage sites and the plant. The integration of the models is made via an iterative procedure. In this procedure, the design is used in the simulation model to manage the flow of biomass in the supply chain. On the other hand, the outputs of the simulation model are used as the inputs of the optimization model to adjust the design. The iterative procedure continues until no improvement can be made in the design. The integrated model is applied to a proposed ethanol plant in Prince Albert, Saskatchewan. The numbers of selected farms and the established storage sites in the integrated model are reduced by 6% and 10%, respectively, compared to the optimization model. Compared to the simulation model, the integrated model leads to the reduction in number of farms (15%), number of storage sites (57%), amount of purchased biomass from farmers (7%), harvested area (13%), supply radius (13%), number of maximum trucks (2 trucks), supply costs (6-12%), energy input (19%), and emitted CO2 (12%). The results of the sensitivity analysis reveal that the most influential parameter on the design is biomass yield. In addition, bale bulk density and in-field and road transportation operations have the highest impacts on the total supply cost compared to other input parameters.  ii  Preface This dissertation provides a clear explanation of the research problem, objectives to tackle this problem, critical review of literature, a case study, data gathering and evaluation, development and application of decision support models to the case study, and the analysis of the obtained results. The academic and industry experts in Canada and the US were consulted during the course of this study. All of the stages of this study were conducted by the author, Mahmood Ebadian under the supervision of his academic advisers Dr. Taraneh Sowlati and Dr. Shahab Sokhansanj. They advised him in the process of defining the research topic, data gathering, model development and validation and the manuscript publication. They are co-authors on all the published manuscripts. This dissertation includes two peer-reviewed manuscripts:   A version of Chapter 4 was published. Ebadian, M., Sowlati, T., Sokhansanj, S., Stumborg, M., Townley-Smith, L., 2011. A new simulation model for multi-agricultural biomass logistics system in bioenergy production. Biosystems Engineering, 110(3), 280-290.    A version of Chapter 5 was published. Ebadian, M., Sowlati, T., Sokhansanj, S., TownleySmith, L., Stumborg, M., 2012. Modeling and analyzing storage systems in agricultural biomass supply chain for cellulosic ethanol production. Applied Energy, 102, 840-849.    A version of Chapter 5 will be submitted. Development of an integrated tactical and operational planning model for supply of feedstock to a commercial-scale bioethanol plant    A version of Chapter 2 will be submitted. Literature review on modeling and analyzing the agricultural biomass supply chain.  iii  Table of Contents Abstract ........................................................................................................................................... ii Preface............................................................................................................................................ iii List of Tables ................................................................................................................................ vii List of Figures .............................................................................................................................. viii Acknowledgements ......................................................................................................................... x Dedication ...................................................................................................................................... xi Chapter 1. Introduction ................................................................................................................... 1 1.1  Motivation ........................................................................................................................ 1  1.2  Problem description.......................................................................................................... 4  1.3  Research objectives and contributions ........................................................................... 12  1.3.1 Research objectives ...................................................................................................... 12 1.3.2 Research contributions ................................................................................................. 12 1.4  Case study ...................................................................................................................... 13  1.5  Organization of the dissertation ..................................................................................... 15  Chapter 2. Literature review ......................................................................................................... 16 2.1  Synopsis ......................................................................................................................... 16  2.2  Static modeling in the biomass supply chain ................................................................. 16  2.3  Dynamic modeling in the biomass supply chain ............................................................ 20  2.4  Optimization modeling in the biomass supply chain ..................................................... 25  2.5  Combined modeling methods in the biomass supply chain ........................................... 31  2.6  Discussion and conclusions ............................................................................................ 33  Chapter 3. Case study ................................................................................................................... 36 3.1  Synopsis ......................................................................................................................... 36  3.2  Ethanol plant data ........................................................................................................... 36  3.3  Supply area data ............................................................................................................. 38  3.4  Farm data ........................................................................................................................ 39  3.5  Crop data ........................................................................................................................ 42  3.6  Weather data ................................................................................................................... 45 iv  3.7  Harvest schedule data ..................................................................................................... 46  3.8  Structure of the supply chain in the case study .............................................................. 48  3.9  Equipment data............................................................................................................... 49  Chapter 4. Development of a new simulation model .................................................................... 55 4.1  Synopsis ......................................................................................................................... 55  4.2  Framework of the developed simulation model ............................................................. 55  4.3  Modules in the developed simulation model.................................................................. 61  4.4  Assumptions ................................................................................................................... 70  4.5  Verification of the simulation model ............................................................................. 71  4.6  Outputs of the simulation model for the case study ....................................................... 72  4.6.1 Flow of biomass in the supply chain ............................................................................ 74 4.6.2 Daily delivery scheduling ............................................................................................. 75 4.6.3 Supply costs .................................................................................................................. 83 4.6.4 Energy input and the associated emitted CO2 .............................................................. 87 4.7  Sensitivity analysis on farmer participation rate ............................................................ 88  4.8  Validation of the simulation model ................................................................................ 91  4.9  Discussion and conclusions ............................................................................................ 96  Chapter 5. Development of an integrated simulation /optimization model ................................ 100 5.1  Synopsis ....................................................................................................................... 100  5.2  Structure of the optimization model ............................................................................. 100  5.3  Integrated simulation/optimization model ................................................................... 115  5.4  Comparison of the integrated model with the simulation and optimization models .... 122  5.4.1 Number of farms and storage sites ............................................................................. 124 5.4.2 Supply costs ................................................................................................................ 126 5.4.3 Energy input and the associated emitted CO2 ............................................................ 129 5.5  Sensitivity analysis ....................................................................................................... 130  5.6  Discussion and conclusions .......................................................................................... 134  Chapter 6. Conclusions, strengths, limitations and future research ............................................ 137 6.1  Conclusions .................................................................................................................. 137  6.2  Strengths and limitations of the study .......................................................................... 142  6.3  Future research ............................................................................................................. 145 v  References ................................................................................................................................... 147 Appendices.................................................................................................................................. 155 Appendix A.......................................................................................................................... 155 Appendix B .......................................................................................................................... 157 Appendix C .......................................................................................................................... 158  vi  List of Tables Table 1-1: Canadian ethanol plants (Canadian Renewable Fuels Association, 2010) ................... 3 Table 1-2: Crop type, number of farms and cultivated area for each crop in Prince Albert......... 14 Table 2-1: Literature review on the static modeling of biomass supply chain ............................. 19 Table 2-2: Literature review on the dynamic modeling of biomass supply chain ........................ 23 Table 3-1: Characteristics of the proposed cellulosic ethanol plant ............................................. 37 Table 3-2: Farm size range in Saskatchewan (Saskatchewan Agriculture and Food, 2006) ........ 42 Table 3-3: Specifications of equipment pieces (Hess et al., 2009, Sokhansanj et al. (2008), Sokhansanj and Turhollow (2002)) .............................................................................................. 51 Table 3-4: Specifications of equipment pieces (Hess et al., 2009, Sokhansanj et al. (2008), Sokhansanj and Turhollow (2002)) .............................................................................................. 52 Table 3-5: Cost data of the machines and equipment (Turhollow and Sokhansanj, 2007) .......... 53 Table 4-1: Construction cost and average dry matter loss of different storage regime (Brummer et al., 2000) ....................................................................................................................................... 66 Table 4-2: Annual recovered biomass and dry matter loss (DML) in the supply chain ............... 74 Table 4-3: Distribution of contracted farms within 160-km radius .............................................. 75 Table 4-4: Number of required trucks for three different weeks .................................................. 80 Table 4-5: Average supply costs ................................................................................................... 86 Table 4-6: Impact of farmer participation rate on the transportation system ............................... 90 Table 4-7: Comparison of logistics costs ($ t-1) in different studies ............................................. 93 Table 4-8: Comparison of energy input (MJt-1) in different studies ............................................. 94 Table 4-9: Comparison of emitted CO2 (kg t-1) in different studies ............................................. 95 Table 5-1: Input parameters of the optimization model estimated by the simulation model ..... 116 Table 5-2: Total delivery cost ($/t) in different iterations of the integrated model .................... 122 Table 5-3: Comparison of the supply area .................................................................................. 125 Table 5-4: Average number of created items and the computational time ................................. 125 Table 5-5: Supply costs ($/t) ....................................................................................................... 127 Table 5-6: Energy input and emitted CO2................................................................................... 130 Table 5-7: Impact of farmer participation rate on the outputs of the integrated model .............. 134  vii  List of Figures Figure 1-1: World ethanol production, 1975-2010 (Brown, 2010) ................................................ 1 Figure 1-2: Sequence of operations in agricultural biomass supply chain ..................................... 5 Figure 1-3: Distribution of produced wheat straw on Canadian Prairies (by permission from Agriculture and Agri-Food Canada) ............................................................................................... 6 Figure 1-4: Sources of complexity and uncertainty in an agricultural biomass supply chain ...... 10 Figure 3-1: Location of the cellulosic ethanol plant, Prince Albert, Saskatchewan (by permission from Iogen Corp.) ......................................................................................................................... 37 Figure 3-2: 160-km supply area considered for the proposed cellulosic ethanol plant ................ 39 Figure 3-3: Crop districts and rural municipalities in the province of Saskatchewan .................. 41 Figure 3-4: Distribution of wheat grain (spring wheat, winter wheat and durum) yield in three rural municipalities inside the 160-km supply radius ................................................................... 44 Figure 3-5: Weekly harvest progress for different wheat crops (Saskatchewan Ministry of Agriculture, 2011) ......................................................................................................................... 47 Figure 3-6: Monthly harvest percent for different wheat crops .................................................... 47 Figure 3-7: The modeled wheat straw supply chain ..................................................................... 48 Figure 4-1: Schematic of the simulated agricultural supply chain ............................................... 56 Figure 4-2: Structure of the simulation model .............................................................................. 59 Figure 4-3: Delay logic flowchart in the simulation model .......................................................... 63 Figure 4-4: Stack configuration in a closed storage site according to the fire codes.................... 67 Figure 4-5: Contribution of logistics operations to the total DML in the supply chain................ 76 Figure 4-6: Daily delivered biomass to the ethanol plant in a week ............................................. 77 Figure 4-7: Daily inventory of the at-plant storage site ................................................................ 77 Figure 4-8: Number of daily truckloads delivered to the ethanol plant in a year ......................... 79 Figure 4-9: Annual delivered truckloads to the ethanol plant from different distance ranges ..... 80 Figure 4-10: Daily biomass inventory level of a roadside storage with 540 t capacity ................ 82 Figure 4-11: Range of moisture content of biomass in the supply chain...................................... 83 Figure 4-12: The components of the total supply cost ................................................................. 84 Figure 4-13: Histogram of the total supply cost ........................................................................... 86 Figure 4-14: Energy input of different operations in the supply chain ......................................... 87 Figure 4-15: Emitted CO2 in different operations in the supply chain ......................................... 88 Figure 4-16: Daily delivered biomass to the ethanol plant during the harvest season (25% participation rate) .......................................................................................................................... 89 Figure 4-17: Number of annual truckloads delivered to the ethanol plant from different distance ranges (50% participation rate) ..................................................................................................... 90 Figure 5-1: Mathematical symbols representing the annual flow of biomass in supply chain ... 105 Figure 5-2: General structure of the integrated simulation/optimization model......................... 118 Figure 5-3: Flowchart of the integrated simulation/optimization model .................................... 121 Figure 5-4: Total delivery cost in different iterations of the integrated model ........................... 124 viii  Figure 5-5: daily biomass delivery ............................................................................................. 126 Figure 5-6: Average and maximum hauling distance ................................................................. 128 Figure 5-7: Total supply cost ($/t) in the integrated model ........................................................ 129 Figure 5-8: Sensitivity of the supply design to the input parameters ......................................... 131 Figure 5-9: Sensitivity of the total supply cost to the input parameters ..................................... 132  ix  Acknowledgements The preparation of this thesis would not have been possible without the support, hard work and endless efforts of a number of individuals and institutions. First and foremost, I offer my enduring gratitude to my supervisors, Dr. Taraneh Sowlati and Dr. Shahab Sokhansanj, for offering direction, technical advice and constructive criticism. Their encouragement, guidance and support, from concept to completion, enabled me to develop an understanding of the subject.  I would also like to express my gratitude to another member of my supervisory committee, Dr. Paul McFarlane for his extremely helpful ideas and suggestions for improving the work presented in this dissertation. I am indeed grateful to Mr. Mark Stumborg and Dr. Lawrence Townley-Smith from Agricultural and Agri-Food Canada (AAFC) for providing relevant information and technical support for this thesis. I also extend my appreciation to Ms. Tamara Rounce from AAFC and Mr. Glenn Payne from the Saskatchewan Ministry of Agriculture for providing the data on the Prince Albert region.  I express my deepest appreciation to all of the institutions that support me financially. This study is funded in part through the University of British Columbia’s Graduate Fellowship, the Natural Sciences and Engineering Research Council of Canada, Agriculture and Agri-Food and the BC Ministry of Forest, Lands and Natural Resource Operations. The Oak Ridge National laboratory is acknowledged for providing data and helping in the validation of the developed models.  Last but not least, I am thankful to all graduate students in the Industrial Engineering Research Group (IERG) and the Biomass and Bioenergy Research Group (BBRG) for their support and encouragement throughout my program.  x  Dedication To my family for their unconditional and unceasing love and support.  xi  Chapter 1. Introduction 1.1 Motivation The production of biofuels has increased because they are an environmentally attractive and technologically feasible replacement for conventional fuels (Judd et al., 2010). The primary reasons for such a rapid growth are: (1) reductions in greenhouse gas (GHG) emissions, (2) production of a new income stream for farmers and economic growth in rural communities, and (3) enhancement of energy security by diversifying energy sources and utilizing local sources (Klein and LeRoy, 2007). Figure 1-1 indicates the fast development of the ethanol production market in the world over the last three decades. 25,000  Million gallons  20,000 15,000 10,000 5,000 0 1975  1980  1985  1990  1995  2000  2005  2010  Year  Figure 1-1: World ethanol production, 1975-2010 (Brown, 2010) Despite the rapid growth of the biofuel market on a global scale, the development of the biofuel industry in Canada has been far slower than in other countries. In 2005, ethanol production in Canada was less than 2% of that in the United States and less than several other countries such as South Africa and Ukraine (Klein and LeRoy, 2007). To boost this emerging industry, federal and provincial governments have put several measures in place including incentive programs, research assistance, and consumption mandates. The federal government in Canada has created the Renewable Fuels Strategy (RFS) to increase their share of biofuels in the 1  national transportation fuel basket. This strategy aligns with the federal commitment to reduce Canada’s total GHG emissions to 83% of 2005 levels by 2020. The RFS mandates a blend of 5% ethanol in gasoline and 2% biodiesel in the distillate pool. In addition to the federal government, several provinces have developed mandates on the consumption of biofuels (Canadian Renewable Fuels Association, 2010). For instance, the provinces of British Columbia and Saskatchewan approved 5% and 7.5% ethanol-blended gasoline at gas stations, respectively. The production capacity of the Canadian ethanol plants in operation is over 1.7 billion liters (Canadian Renewable Fuels Association, 2010). In contrast, the total federal and provincial renewable fuels requirements will result in an ethanol demand of about 2 billion liters per year. Although imported ethanol can be used to meet this demand, most of the demand for ethanol under the federal and provincial requirements can be met with domestically produced ethanol (Canadian Renewable Fuels Association, 2010). This is due to the abundance of cellulosic materials in Canada including agricultural and forest residues which can be used as feedstock for cellulosic ethanol production. Canada has about 36.4 million hectares (Mha) of crop lands available for agricultural production. More than 85% of these lands are located in the Canadian Prairies including Saskatchewan, Alberta and Manitoba and a small portion in northeastern British Columbia (Sokhansanj et al., 2006a). After harvesting the grain as the primary agricultural product, tonnes of crop residues are left on the fields which could be utilized as renewable energy sources (Kumar and Sokhansanj, 2007). Sokhansanj et al., (2006b) estimated that there would be an annual average of 15 million tonnes (Mt) of straw available for industrial uses like ethanol production on the Prairies after consideration of soil conservation and livestock requirements. Despite of the apparent large quantity of crop residues, the list of operational and under construction ethanol plants in Table 1-1 shows that only one plant uses agricultural residues to produce cellulosic ethanol. This plant is a demonstration facility constructed by Iogen Corporation, a Canadian biotechnology firm specializing in cellulosic ethanol. Currently, there is no commercial-sized cellulosic ethanol plant in Canada that exploits agricultural residues as feedstock.  2  Table 1-1: Canadian ethanol plants (Canadian Renewable Fuels Association, 2010) Plant  Feedstock  Annual Capacity Plant  Feedstock  (Million liters)  Annual Capacity (Million liters)  Amaizeingly Green Products L.P.  Corn  58  Husky Energy Inc. Minnedosa  Wheat and Corn  130  Husky Energy Inc. Lloydminster  Wheat  130  IGPC Ethanol Inc.  Corn  162  Enerkem Alberta Biofuels  Municipal Solid Waste  36  Iogen Corporation  Wheat and barley straw  2  Enerkem Inc.– Sherbrooke Pilot Plant  Various types of feedstock  0.475  Kawartha Ethanol Inc.  Corn  80  Enerkem Inc. – Westbury Commercial  Wood waste  5  NorAmera BioEnergy  Wheat  25  Wheat  25  Wheat, wheat starch, corn, barley  42  GreenField Ethanol Inc.  Corporation Corn  195a  Chatham  North West Terminal Ltd.  GreenField Ethanol Inc. Johnstown  Corn  230  GreenField Ethanol Inc.  Corn  27  Pound-Maker Agventures Ltd.  Wheat  12  Corn  155  Suncor St. Clair Ethanol Plant  Corn  400  Wheat  150  International, L.P.  Tiverton GreenField Ethanol Inc. Varennes Terra Grain Fuels Inc. a  Permolex  Volumes include industrial alcohol production  3  There are several obstacles restricting the development of this emerging industry. These obstacles encompass the inefficiencies associated with immature feedstock production practices, marketing and logistics systems, and conversion processes (Fales et al., 2007). These inefficiencies make the cellulosic ethanol industry suffer from the lack of the economies of scale (Klein and LeRoy, 2007). In other words, all aspects of the industry are new and inherently inefficient. Cost reductions in feedstock production, supply systems, and conversion processes will pave the way to establish large-sized cellulosic ethanol plants. Among the above-mentioned inefficiencies, the focus of this research is on the biomass supply chain. The dynamics and stochastic nature of the biomass supply chain make it an unreliable system to deliver large volumes of bulky cellulosic materials to the conversion facility at low costs throughout the year (Cundiff et al., 2009a). The supply system accounts for 35-65% of the total cellulosic ethanol production cost (Fales et al., 2007). In contrast, the feedstock supply costs associated with corn grain-based ethanol made up 8% of the total ethanol production cost in the reference year of 2008 (Hess et al., 2009). Therefore, to accelerate the commercialization of cellulosic ethanol in Canada, the agricultural biomass supply chain must be designed, planned and scheduled in a way that a secure supply of crop residues are delivered to the conversion facility at a minimum cost throughout its business life.  1.2 Problem description A typical agricultural biomass supply chain encompasses all of the operations from biomass producers (farmers in this study) to the gate of the ethanol conversion facility. A general illustration of an agricultural biomass supply chain is depicted in Figure 1-2. The first operations are the harvesting and collection of biomass in which biomass is removed from fields and transported to the nearby storage sites. These operations include cutting, in-field drying, collecting biomass, densifying and transporting it to storage. Biomass can be kept in roadside storage or satellite storage located between the farms and the ethanol plant. Handling and transporting operations include loading biomass onto the transportation vehicles and shipping it to the plant.  4  The next operation is receiving in which the arriving truckloads are unloaded at an at-plant storage site. Preprocessing is the last operation in the supply chain before the conversion facility. It may consist of one or more processes including size reduction, fractionation, sorting, and densification (Sokhansanj and Fenton, 2006). It is noteworthy that preprocessing can also take place earlier in the supply chain, at a location between the farmlands and the ethanol plant. For instance, roadside or satellite storage can be a depot in which preprocessing is performed in addition to storing biomass.  Biomass producers  Harvesting and collecting  Storing  Start Point  Handling and transporting  Receiving  Preprocessing  Conversion process  End Point  Figure 1-2: Sequence of operations in agricultural biomass supply chain Previous research indicates that in order to achieve economy of scale, large biorefineries capable of handling 5,000-10,000 tons of biomass per day are necessary (Carolan et al., 2007). Daily planning and scheduling of operations to deliver such large quantities of biomass from rural areas to commercial-sized biorefineries would be a cumbersome logistics task. In addition, the specific characteristic of agricultural practices in the Canadian Prairies such as crop rotation, and climatic, geographical and biological factors place additional constraints on the supply chain. For example, the large distribution of the produced residues, low and variable yields, short and variable harvest season, frequent crop rotations, and unfavorable and uncertain weather conditions would result in a highly constrained supply chain with a great level of uncertainty and complexity in the Canadian Prairies. Cereal crops, oilseeds and pulse crops dominate the seeded area in the Canadian Prairies (Sokhansanj et al., 2006a). These crops and their residues are widely distributed across the Prairies. Figure 1-3 illustrates the distribution of produced wheat straw as one of the primary cellulosic materials on the Canadian Prairies. In addition to the large distribution, Figure 1-3 shows the low yield of wheat straw ranging from 1 to 3.5 oven dry tonnes per ha. Wheat straw has also relatively lower bulk density compared to other major cellulosic materials such as corn  5  stover, switchgrass and miscanthus (Hess et al., 2009). Similar conditions exist for other crop residues produced in the region.  Figure 1-3: Distribution of produced wheat straw on Canadian Prairies (by permission from Agriculture and Agri-Food Canada)  These characteristics result in the need for a large supply area to meet the annual feedstock demand of a commercial-scale ethanol plant. A large supply area complicates the management of the supply chain since many farms would be required. In addition, low-bulk density of biomass causes the low utilization rate of storage area, handling and transportation equipment which causes the high costs of these operations. Most of the crops grown in the Canadian Prairies are annual crops, and thus there is only one harvest season in a year. Harvest season is a narrow window constrained by the weather conditions (Cundiff et al., 2009a). Local climate influences the start and the length of grain harvest season (Sokhansanj et al., 2006b). The wheat harvest season may begin early in August to early September and it may take from one month to three months in Saskatchewan (Saskatchewan Ministry of Agriculture, 2011). A short and variable harvest season for harvesting and collecting agricultural residues complicates resource management. For a commercial-sized ethanol plant, many harvesting and collecting machines must operate on a tight schedule to guarantee year-round availability of biomass for the conversion process. Since massive volumes of biomass must be processed during the harvest season, breakdown of machines is likely to occur. 6  It also complicates the storage management since large volumes of crop residues must be stored during this short period to secure the year-round delivery to the plant. Thus, crop residues may require being stored for several months. This could result in the significant dry matter losses (DML) of up to 25% in poor storage conditions (Lyschinski et al., 2002). The estimation of DML in storage is difficult as it depends on the weather conditions, moisture content, condition in which biomass is stored and also the duration of storage. In addition, a portion of biomass is lost by machines depending on the sensitivity of biomass to breakage and the efficiency of the machines to process biomass. Another issue related to the harvest season is the dependency of the biomass harvest schedule on the harvest window of grain. Grain is the primary product, and thus, harvest window is scheduled when grain reaches maturity and moisture is optimal (Hess et al., 2009). Farmers prefer to complete the harvest of a mature crop as quickly as possible in time for preparing the land for the following cropping season or to minimize the potential of work stoppage due to cold, humid, or freezing conditions. Thus, the commencement of harvest season and its duration may not be optimal to the harvest and collection of residues. Lack of optimality of the harvest window for agricultural residues could negatively impact the quality and quantity of harvested and collected biomass. To mitigate the negative impacts such as high moisture content, extra operations may be required in the supply chain. These extra operations, such as biomass drying, add up to the complexity of the system and also increase supply costs. Other dominant cellulosic materials such as switchgrass do not have such harvest-related problems. Switchgrass is a perennial grass species with a longer harvest season. For example, Switchgrass is harvested from June to February in the Southeastern US (Ravula, 2007). Thus, sufficient switchgrass must be only stored during three non-harvest months. One of the primary sources of uncertainty in the supply chain is the biomass yield. The net yield depends on the grain yield, straw to grain ratio, soil conservation rate and competitive markets such as livestock feeding and bedding (Sokhansanj et al., 2006a). Grain yield varies based on climate, water, soil type, pests and fertilizer application. In addition, straw to grain ratio varies with location, cultural practices, species, and grain varieties (Boyden, 2001). The amount of residues left on the field required for wind and water erosion control depends on soil texture, field slope and tillage practice. This amount could be variable from 30% to 75% of produced residues in the Prairies (Sokhansanj et al., 2006a). 7  Using agricultural residues for livestock feeding and bedding is another parameter that increases the uncertainty in the availability of biomass for ethanol production. Agricultural residues have been used for raising livestock for many years in the Canadian Prairies. Often, due to a higher return on investment, cellulosic ethanol plants have to use the remaining biomass after the consideration of livestock requirements. The net available crop residues on the Prairies after the deduction of livestock feeding has been studied by Sokhansanj et al. (2006b). They showed that the amount of annual cereal straw available on the Prairies for industrial use ranges between 2.3 Mt and 27.6 Mt with the average of 15 Mt. Frequent crop rotation in the Prairies also affects the availability of biomass in the long term. Farmers typically include several crops in rotation. The size of land under each kind of crop (wheat versus barley versus canola, etc.) depends upon a number of factors such as market price, yield expectation, and rotational concerns. As such, it is difficult to know with certainty how much of a crop a farmer may grow in each year. Thus, during the lifetime of the ethanol plant, it would be indefinite which farm will grow which crop and what portion of the land will be assigned to the cultivation of a crop. In addition, farmers may decide to leave the entire produced biomass on the field for soil conservation purposes in some years. Therefore, farmers may change their attitudes toward participating in the biomass supply during the business life of the ethanol plant making the biomass procurement a difficult task in the supply chain. The unfavorable and uncertain weather conditions such as low temperature, sporadic precipitation and frequent droughts in the Canadian Prairies impact the supply chain. The weather conditions affect the availability and performance of machines. The logistics operations, in particular field operations, may be slowed down or shut down due to the unfavorable weather conditions. Another impact of the climatic conditions is on the quantity and quality of biomass delivered to the ethanol plant. Harsh climatic conditions may intensify the dry matter loss and chemical breakdown of the biomass structure and composition. In addition to the specific characteristics of the Prairies, there are other factors that turn the supply chain of crop residues into a complex system. These factors include the different operational window of the logistics operations and also the business model of the supply chain. The conversion facility at the ethanol plant operates 24/7, while the operations upstream of the plant such as field operations and transportation usually have 12 to 14 working hours in a day. Moreover, the transportation system may operate 5/6 days a week. The different operational 8  window between logistics operations necessitates at-plant storage. Type, capacity and inventory of at-plant storage should be determined so that the sufficient biomass is delivered to the conversion facility during the off-shift hours of operations at the upstream of the plant. These storage-related decisions have great impacts on the quality, quantity and cost of delivered biomass to the plant. The next complexity of the supply chain is related to its business model. The on-time delivery of the right amount of low-cost biomass to the ethanol plant requires interaction and coordination among the supply actors involved in the supply chain. These actors encompass farmers, hauling contractors and the ethanol plant. Farmers are usually responsible for harvesting and collecting biomass and then transporting it to the storage locations. Thereafter, the stored biomass is hauled to the plant by hauling contractors. Finally, the ethanol plant receives the delivered biomass and converts it to the final product. Farmers and hauling contractors have incompatible preferences on the number and location of storage sites within the supply area as these would affect their total operating cost (Resop et al., 2011). Another differing view exists between farmers and the ethanol plant regarding the supply contracts. The ethanol plant prefers long-term contracts with farmers to secure the availability of biomass for the conversion facility during its economic life. In contrast, farmers are able to sell their products in a spot market and may have little interest in long-term contracts (Cundiff et al., 2009a). The length of contract could be a more serious issue for annual crops such as cereal straw. For a perennial grass such as switchgrass, the contract with switchgrass growers could be as long as 10-15 years, equivalent to the average life of a stand of switchgrass before reseeding (Judd, 2011). Figure 1-4 illustrates the sources of the complexity and uncertainty in the agricultural biomass supply chain. It is noted that the varying parameters in the agricultural biomass supply chain such as yield, weather condition and harvest season are uncontrollable, thus, planning and scheduling of logistics operations becomes a challenge. In summary, the logistics challenge is how to provide a secure and steady daily feedstock delivery to a commercial-sized ethanol plant at the minimum cost with respect to the dynamics and stochastic nature of the supply chain and also the region-based constraints.  9  Grain yield Straw to grain ratio  Variability in the availability of biomass for ethanol production  Weather conditions  Short and variable harvest season  Other stochastic parameters such as machine breakdown and efficiency  Soil conservation rate Competitive markets  Stochastic parameters Crop rotation and annual crops Complex resource and storage management  Agricultural biomass supply chain Biomass characteristics Dry matter loss and microbial degradation  Structural complexity Distributed natural resource  Low-energy content  Low-bulk density  Different operational window of upstream and downstream operations Farmer participation rate Dependency of biomass harvest schedule on the grain harvest window Incompatible preference between farmers and hauling contractors on the number and location of storage Incompatible preference between farmers and the ethanol plant on the length of supply contracts  Figure 1-4: Sources of complexity and uncertainty in an agricultural biomass supply chain  10  A thorough review of the relevant literature reveals some studies have considered a static analysis of the supply chain to estimate the available feedstock for bioenergy production and the logistics costs (Clegg and Noble (1987); Allen et al. (1998); Noon and Daly (1996) and Graham et al. (1996)). The developed static models incorporated many assumptions and did not capture the dynamics of the supply chain. Dynamic models have been developed to estimate the amount of biomass delivered to the biorefinery plants and also the associated logistics costs subject to uncertainties such as delays in harvest due to weather, moisture and machine breakdown (Nilsson (1999a); Nilsson and Hansson (2001); Sokhansanj et al. (2006); Sokhansanj et al. (2008b)). However, these models were not able to find the optimal logistics solutions to minimize the delivery cost. Several studies have found the optimal solutions for selection of farms to contract, location and number of storage sites and also the location of the biorefinery plants to minimize the total supply cost (Ekşioğlu et al. (2009); Judd et al. (2010); Zhu et al. (2011); Zhu and Yao (2011) and Judd et al. (2012)). The developed optimization models are mainly at the strategic and tactical levels and thus, the stochastic nature and the part of the complexity and dynamics of the supply chain were neglected in the developed optimization models. Therefore, the optimal solutions do not guarantee the daily fulfillment of the feedstock demand. Finally, a combination of these modeling tools has been used to increase their capabilities to evaluate the biomass supply chain (Freppaz et al. (2004), Ravula (2007) and Berruto and Busato (2008)). However, the literature lacks an integrated modeling approach to explore and analyze a highly constrained agricultural biomass supply chain similar to the one in the Canadian Prairies at both tactical and operational planning levels. The tactical planning level concerns the design of the supply chain to meet the annual biomass demand at the minimum delivery cost. In the operational level, the logistics operations are scheduled on a daily basis to meet the daily demand. This calls for an integration of the optimization approach with the dynamic modeling of biomass flow. It is noted that part of the portfolio of the solutions to improve the performance of the supply chain is associated with the development of more advanced and flexible equipment and technologies. These developments are outside the boundaries of this study. Thus, no future equipment and technologies are included in this study. The focus of this study is on the integration of design and scheduling of the supply chain to meet the feedstock demand of a 11  commercial-scale cellulosic ethanol at the minimum cost possible. Note that, it is assumed that the location of the ethanol plant and its daily biomass demand is predetermined. It is also assumed that the delivered biomass meets conversion process quality specifications and thus biomass quality is not considered in modeling and analysis of the supply chain.  1.3 Research objectives and contributions 1.3.1 Research objectives The aim of this research is to design and schedule a highly constrained agricultural biomass supply chain to meet the daily biomass demand of a commercial-sized cellulosic ethanol plant. The specific objectives of this study are as follows: 1. To develop a new simulation model that would seek out feedstock from a blend of available biomass in the area surrounding the cellulosic ethanol plant. The developed simulation model incorporates time-dependency, stochastic parameters and regional constraints in the scheduling of the logistics operations to meet the daily biomass demand. 2. To develop a new optimization model to determine the design of the supply system. The design includes the selection of farms, the location of storage sites and also the assignment of the selected farms to the storage sites. It also determines the flow of biomass between farms, storage sites and the plant. The prescribed design and flow ensures the annual fulfillment of the biomass demand at the minimum delivery cost. 3. To integrate the multi-biomass simulation model with the optimization model to ensure an uninterrupted flow of biomass to the ethanol plant at the minimum cost possible. The efficiency of the integrated model is evaluated by applying it to a real-life case study. A sensitivity analysis is conducted to measure the impact of changes in input parameters on the results of the integrated model.  1.3.2 Research contributions The contributions of this study consist of four parts. The first contribution is that this study provides a detailed source of input data for a commercial-scale bioenergy plant. Due to the  12  similarities in the biomass supply chain of different bioenergy products, the gathered data in this study can be used to model and analyze the biomass supply chain for other bioenergy products. This study can be used as a guide to design and schedule the supply chain for agricultural residues, mainly cereal straw which has received less attention compared to other types of cellulosic materials. This study considers both data and model details and provides meaningful information on a constrained supply chain in the Canadian Prairies. This information can be used by decision makers in practical cases. The next contribution is the modeling aspect of this study. Both the developed simulation and optimization models have new features compared to the ones in the literature. The simulation model provides a detailed understating and evaluation of the dynamic and stochastic biomass supply chain on a daily basis. Additionally, the optimization model finds the logistics solutions at the tactical level of the supply chain. The last contribution is on the integration of both developed simulation and optimization models. An iterative procedure is developed to make interaction between both models. The interaction results in feasible and consistent solutions at both tactical and operational levels. The effectiveness of the proposed integrated simulation/optimization model is clearly shown by applying it to a case study.  1.4 Case study A proposed cellulosic ethanol plant located in the Prince Albert region, in north central Saskatchewan was used as the case study in this research. The region is one of the richest agricultural areas in the province and there is an emphasis by the provincial and federal governments on developing the biofuel industry in the region. The total number of farms in the Prince Albert area is 9,647. These farms cover 1,005,572 ha of land. More than 20 different types of crops, such as wheat, oats, barley, dry field peas and canola, are grown in this region. Table 1-2 lists the crop type, the number of farms and the cultivated area in this region in the reference year of 2006. This information was provided by Agriculture and Agri-Food Canada (AAFC).  13  Table 1-2: Crop type, number of farms and cultivated area for each crop in Prince Albert Crop Type  Number  Seeded area  Number of  Seeded area  of farms  (ha)  farms  (ha)  1,629  260,012  Alfalfa  1,952  153,272  Oats  1,315  82,916  Dry Field Peas  447  49,247  Barley  1,302  128,561  Other field crops 4  64  7,791  Mixed grains  106  6,330  Forage seed  127  8,583  Corn2  11  175  Potatoes  26  375  Rye3  132  6,067  Mustard Seed  28  3,152  1,474  225,995  Canary Seed  31  1,616  Flaxseed  171  11,050  Buckwheat  3  24  Other Tame Hay  794  58,779  Triticale  35  1,627  Wheat  1  Canola  Crop Type  and Fodder crops 1  Includes spring, winter and durum wheat Includes grain corn and corn for silage 3 Fall rye and spring rye 4 Includes solin, safflower, coriander and other spices, etc 2  In 2008, the Government of Saskatchewan, Iogen Corporation and Domtar Corporation reached an agreement to set the stage for the potential redevelopment of the Prince Albert pulp mill site as a cellulosic ethanol plant. The pulp mill has been closed since April 2006. It is located 12 km east of Prince Albert. The plant will be developed by Iogen. The production capacity of the plant is more than 70 million liters (ML) of cellulosic ethanol per year. The conversion facility would process 750 tonnes (t) per day of wheat straw. To procure this amount of wheat straw, Iogen has considered a 160-km supply radius (Iogen Corp., 2009). The details of the case study are provided in Chapter 3. In this study, first the considered supply area by Iogen will be modeled and analyzed using the developed simulation model. Thereafter, the integrated simulation/optimization model is applied to enhance the considered supply area in terms of cost-efficiency and demand fulfillment. Although the developed models are applied to a single specific case study, they can be adapted and applied to other regions, agricultural residues, and bioenergy products.  14  It is noted that since the considered agricultural biomass in this study is wheat straw, the term "biomass" refers to "wheat straw" in the rest of the dissertation. In addition, "the plant" refers to "the cellulosic ethanol plant". The term "tonne", or shortly, "t" is used in the dissertation referring to the weight of biomass at the acceptable moisture content for ethanol production (less than 20% w.b.).  1.5 Organization of the dissertation In addition to the introduction chapter, this dissertation includes one chapter on the literature review, one chapter on the case study, two chapters on the developed simulation and optimization models and one chapter of conclusions, limitations of the study, and finally suggestions for future research. The previous studies on the literature are discussed in Chapter 2. They are categorized and reviewed based on the decision-making tools used to find the solutions for different strategic, tactical and operational problems in the biomass supply chain. The strengths and shortcomings of the relevant literature are highlighted in this chapter and an integrated approach is proposed to tackle the shortcomings. The integrated approach includes both simulation and optimization modeling which are discussed in the following chapters. Prior to discussing the developed simulation and optimization models, the details of the case study are explained in Chapter 3. The development of the new simulation model is presented in Chapter 4. The framework of the developed simulation model including the input data, simulation structure and output data are given in this chapter. The developed simulation model is applied to the case study. The discussion on the verification and validation of the simulation model is also provided here. Chapter 5 elaborates on the development of an optimization model to prescribe the design of the supply chain. The integration of the developed simulation in Chapter 4 and the developed optimization model is also discussed in this chapter. The integrated simulation/optimization model is applied to the same case study and the efficiency of the integrated model is shown. The last chapter of this study is assigned to the final conclusions, the strengths and limitations of the study. Finally, several suggestions are given for future research direction.  15  Chapter 2. Literature review 2.1 Synopsis The biomass supply chain has been modeled and analyzed in the literature to improve its performance in terms of biomass delivery and the total delivery cost. In this regard, a wide range from strategic to operational decisions have been made, such as the location and capacity of the conversion facility, location of storage sites, inventory and shipment planning, and timing of harvest. Different decision-making tools have been developed to find the solutions for these decisions. Some studies employed static methods including spreadsheets and GIS-based tools. A portion of the relevant studies have exploited the power of simulation modeling mainly for planning and scheduling the operations at the operational level. Another popular tool is optimization modeling mainly used to make optimal decisions at the tactical and strategic levels. A combination of these modeling tools has also been used in the literature. In this chapter, the major relevant studies have been reviewed and categorized based upon the developed decision-making tools.  2.2 Static modeling in the biomass supply chain The initial and simplest tools used in the literature to estimate the logistics costs and biomass delivery are spreadsheets. A number of static spreadsheet models have been developed to calculate the costs of using biomass for bioenergy production, for example by Clegg and Noble (1987), Brundin (1988), Floden (1994), Mitchell (1995), Allen et al. (1998), Mitchell (2000), and Sokhansanj and Turhollow (2002). The developed spreadsheet models have been used as decisions support systems. They have different components such as a user interface, a database, a component that analyzes the data and information and a results screen. The logistics operations such as harvesting, storage and transportation are defined as equations and linked through the programming codes. The spreadsheet models allow the user to develop and evaluate different logistics scenarios through "what if" type questions. GIS-based modeling is another initial tool used to assess the resource availability and estimate the logistics costs. Due to the capability of GIS-based models in terms of storing, managing and 16  displaying geospatial data and also providing significant spatial patterns (Noon and Daly, 1996), it has been exploited either as an independent analytical tool or as part of a decision support system (DSS). Noon (1993) and Noon and Daly (1996) developed a GIS-based decision support system called Biomass Resource Assessment Version One (BRAVO). BRAVO assesses the availability of woody biomass including mill residues, logging residues and short rotation woody crops and estimates the delivery cost of biomass to coal-fired plants in the Tennessee Valley Authority (TVA) region. The data inputs in the BRAVO system include a digital map of the road network, a digital map of state and county boundaries and a digital map of plant locations. Graham et al. (1997) applied the BRAVO model to 21 locations in the state of Tennessee to investigate the impact of the biomass demand and the farmer participation rate on the delivered costs. In addition, BRAVO was used to determine the locations of bioenergy plants (Graham et al., 1996). To find the location, the farm gate cost and transport cost were calculated based on spatial data for any specific location in the region under study and the location with minimum total biomass cost was selected as the optimum location. Husdal (2000) developed a raster-based GIS method to minimize the transportation costs by calculating the shortest paths in a network. A raster is a grid of square cells (pixels) where each cell contains a value representing information such as temperature (ESRI, 2009). Resop et al. (2011) developed a raster-based model to assess the potential production of switchgrass and locate satellite storage sites near Gretna and Keysville, Virginia where switchgrass can be grown in a large scale to feed commercial-scale bioenergy plants. Resop et al. (2011) considered several criteria for locating satellite storage sites in the regions under study including: 1) direct road access to ease the transportation of switchgrass from satellite storage locations to the conversion facilities, 2) switchgrass bales can be collected from selected surrounding fields within a 3.2-km radius around storage, 3) minimum 40 ha of switchgrass production available within a 3.2-km radius and, 4) level land areas with average slopes of less than 10% were considered in the process of locating satellite storage sites. Brownell and Liu (2010) developed a heuristics method written in Visual Basic (Microsoft Corp.) to determine the size and number of satellite storage sites. Given the supply area and the location and size of the bioenergy plant in the middle of the supply area, the model starts with an initial satellite storage scenario and changes the size and number of satellite storage sites 17  gradually until an optimal scenario is reached. The optimal scenario has the minimum total cost of field, satellite storage and transportation operations. The model was implemented for three different plant sizes: small (2000 tonnes/day), medium (5000 tonnes/day) and large (10000 tonnes/day). The small, medium and large-sized plants require 25, 64 and 105 satellite storage locations, respectively. As reported by Brownell and Liu (2010), it took two days to find the optimal solution. In addition to the computational time drawback, the authors did not elaborate on how their developed method finds the optimal solution while assuring that all of the possible scenarios have been considered. The focus of this study was on satellite storage. At-plant storage was not taken in to account and only the handling system was modeled. The details of some of the studies on the static modeling are given in Table 2-1. The developed static models are easy to develop and implement and can provide a basic understanding of the biomass supply chain, mainly in terms of the amount of produced biomass in a region and the cost of delivered biomass to a bioenergy plant. However, many assumptions were made in the static models, such as those on average yield, average distance between farms and bioenergy plants, and the fixed moisture content across the supply chain. In addition, the dynamic behavior of the supply system such as continuous changes in moisture content, inflow and outflow of biomass from storage and also the functional relationships among many of dependent and independent variables in the supply system were neglected as the static models could not capture the dynamics of the system. Therefore, the obtained results of the analysis cannot be considered as reliable information for decision makers. To improve the quality of the obtained results, the dynamic behavior of the biomass supply system must be taken into account. This requires the deployment of a decision-making approach which enables one to model the time-dependency of the supply chain.  18  Table 2-1: Literature review on the static modeling of biomass supply chain Study  Method  Biomass/Bioenergy  Objective  type  Case  Findings/important aspects  study/region  Mitchell (1995)  Spreadsheet  Short rotation woody crops  Estimation of logistics costs  N/A  - Better understanding of the viability of growing short rotation woody crops for industrial uses.  Allen et al. (1998)  Spreadsheet  Forest biomass, short rotation coppice, straw and miscanthus/power plant  Estimation of biomass delivery cost to the power stations  N/A  - Delivered cost for large rectangular baled straw would be lower than the cost of other considered biomass types.  Sokhansanj and Turhollow (2002)  Spreadsheet  Corn stover  Estimation of corn stover collection costs  Midwestern United  - Total collection cost for round baling systems were 9% less than the rectangular baling systems.  Noon (1993), Noon and Daly (1996)  GIS  Mill residues, logging residues and short rotation woody crops (SRWC)  Estimation of total purchase and transportation costs  Tennessee Valley Authority  - Mill residues were the most viable biomass fuel compared to two other types of forest biomass.  Graham et al., 1996  GIS  Energy crops  Location of bioenergy facilities  State of Alabama  - A conversion facility demanding 600,000 dry t/yr or 27 facilities each requiring 100,000 t with feedstock cost under $35/t could be established.  Resop et al. (2011)  Raster-based GIS  Switchgrass  Assessment of the potential production of feedstock and location of satellite storage sites  Gretna and Keysville, Virginia  - Potential production of switchgrass could supply a hypothetical bioenergy plant with average consumption of 24-43 t/h within 32km radius and 61-98 t/h within 48-km radius.  States  19  2.3 Dynamic modeling in the biomass supply chain Simulation has been widely used in the supply chain networks. This is due to the capability and flexibility of simulation in modeling and evaluating complex dynamic systems, while considering uncertainty and variability in the system (Almeder et al., 2009). Kleijnen (2005) divided the simulation tools used in the supply chain into four types: spreadsheet simulation, system dynamics, discrete-event simulation and business games. Among them, discrete-event simulation is the most powerful simulation tool to model complex stochastic systems (Almeder et al., 2009). Due to the time-dependency and stochasticity of the biomass supply chain, discreteevent simulation has received the most attention among researchers to model and analyze the biomass supply chain (Ebadian et al., 2011). This tool has been mainly used to model the logistics operations in the biomass supply chain at the operational level to estimate the amount of delivered biomass to the bioenergy plant and the associated logistics costs. The simulation modeling was initially used to schedule the farm operations such as selection of forage machinery on a dairy farm (Noel P. Russell, 1983), evaluation of technologies or management practices in forage systems on dairy farms (Savoie et al., 1985), scheduling of labors and equipment for wheat harvesting (Elderen, 1987), and planning of hay harvesting equipment (Axenbom, 1990). One of the initial simulation models developed to model and analyze the biomass logistics is the work of Mantovani and Gibson (1992). They developed a simulation model written in GASP IV (simulation language) to compare harvesting and handling systems for corn stover, hay, and wood residues for ethanol production. They incorporated historical weather data and farmers’ changing attitude towards harvesting biomass. Ten years of climatological and biomass production data were included in the model. The impact of weather variations and late harvest on biomass availability and equipment cost were discussed. One of the most applicable simulation frameworks developed for designing a biomass delivery system is the Straw HAndling Model (SHAM) (Nilsson, 1999a). SHAM was presented as a dynamic simulation model for analysis of various delivery alternatives only for straw. That is because SHAM was used for analysis of agricultural fields in Sweden where straw is regarded as the prominent renewable energy source. Straw has a considerable potential to meet the energy  20  demand in rural areas in Sweden, especially as fuel in district heating plants (Nilsson, 1999a). SHAM simulates the effects of climatic, geographical, and biological factors on the cost of delivering biomass such as field size, transport distances between storage sites and heating plants, and straw yield. Some of the required input data are defined by appropriate random variables, for example the areas of the fields, the start of the harvest season, the arrivals of fields to the simulation model and crop yields. SHAM has been applied to three main regions in Sweden including Svalov, Vara, and Enkoping, as in Nilsson (1999b) and Nilsson (2000). The obtained results of the application of SHAM in these regions showed that the harvesting operation highly depends on the climatic and geographical conditions such as the frequency and duration of precipitation, field size and fraction of the land area with harvestable straw. Another finding was that the employment of management strategies such as prolonging the harvest season, optimal number of balers and transporters, and storage locations could lead to significant cost reductions. Nilsson and Hansson (2001) modified SHAM to incorporate a new crop, reed canary grass (RCG). They evaluated using RCG as feedstock in district heating plants in addition to straw and oil. The obtained results revealed that the total delivered cost could be reduced by using a mix of straw and RCG in suitable proportions instead of solely using straw. This could reduce costs by 15-20%. They also concluded that although RCG is an expensive feedstock (more than three times the cost of straw), the cost savings due to better use of machines and storage space and less consumption of oil as the primary fuel make this crop an attractive complementary fuel. Another comprehensive simulation framework developed to represent various stages of biomass collection, processing, storage, and transport activities is the Integrated Biomass Supply Analysis and Logistics (IBSAL) model. This supply model was developed using the ExtendSim simulation platform available from Imaginethat Inc. (Sokhansanj et al., 2006b). Contrary to SHAM, which was mainly developed for analysis of logistics systems of cereal straw, IBSAL is capable of modeling and analyzing the supply logistics system for a variety of crop residues, such as cereal straw and corn stover and also grasses such as switchgrass. It also calculates different outputs including costs associated with each logistics operation, the quantity of delivered biomass to the conversion facilities, dry matter loss, energy input and carbon emissions and completion time for each operation.  21  IBSAL was applied for various biomass types and logistics scenarios as in Sokhansanj and Fenton (2006), Sokhansanj et al. (2006a), Kumar and Sokhansanj (2007), Sokhansanj et al. (2008a), Sokhansanj et al. (2008b), Sokhansanj and Hess (2009), Sokhansanj et al. (2009), Stephen (2008), and Stephen et al. (2010). The details of some of these studies are given in Table 2-2. SHAM and IBSAL are similar in context. However, the issue of dry matter loss is not highlighted in SHAM. IBSAL estimates the dry matter loss as biomass undergoes various operations. In IBSAL, different preprocessing modules such as grinding, pelletizing and briquetting have been developed to assess their effects on the logistics costs and quality of delivered feedstock. There are no similar discussions on the possible preprocessing in SHAM. SHAM considered breakdown of equipment as an activity in the supply chain to shut down the operations, while IBSAL included breakdowns in the machine operational efficiency. Other differences of IBSAL and SHAM include assignment of harvesting equipment to farms, stoppage/delay of operations due to weather conditions and also the length of the harvest season. In addition to SHAM and IBSAL, another simulation model was developed by the Idaho National Laboratory (INL), US Department of Energy (DOE) (Hess et al., 2009). Their simulation model estimated the supply cost of herbaceous lignocellulosic biomass for biofuel production. The considered lignocellulosic biomass types included corn stover, corn cob and switchgrass due to their potential and availability in the US. Three feedstock supply systems were considered in the work of Hess et al. (2009): conventional bale, pioneer uniform, and advanced uniform. The primary difference among these three supply systems is the location of preprocessing operations in the supply chain. In the conventional bale system, the preprocessing occurs in the biorefinery plant, while it is considered at centralized depots in the pioneer uniform supply system, and during harvest and collection in the advanced uniform system. Pushing the preprocessing operation towards the upstream of the supply system eases the handling of a variety of feedstock with the same equipment. This would pave the way for treating biomass as a commodity in the biofuel market.  22  Table 2-2: Literature review on the dynamic modeling of biomass supply chain Study  Simulation  Biomass/Bioenergy  platform  type  Nilsson (1999 a,b), Nilsson (2000)  Arena  Straw  Modeling and analyzing the straw fuel delivery systems  Svalov, Vara and Enkoping in Sweden  - Climatic and geographical factors highly impacted the straw harvesting system. - longer transport distances and lower straw yields were the main reasons for the higher costs of straw handling at Vara and Enkoping compared to Svalov.  (Nilsson and Hansson, 2001)  Arena  Straw and reed canary grass (RCG)/ theoretical heating plant  Evaluation of the impact of using a mix of straw and reed canary grass in district heating plants  Enkoping  - Using a mix of straw and reed canary grass would result in reduction in the total cost. - More saving could be achieved using a mix of straw and wood chips, and possibly a small portion of RCG.  Sokhansanj and Fenton (2006)  ExtendSim  Switchgrass and crop residues  Calculation of the cost of transporting biomass from roadside storage/ satellite depots to the Biorefinery  Canada  - Delivery cost mainly depends on the bulk density of the biomass, its moisture content, and the distance to be transported.  Kumar and Sokhansanj (2007)  ExtendSim  Switchgrass  Estimation of delivery costs, energy input and emitted CO2  N/A  - Cost of feedstock delivered to a biorefinery for loafing is 18% less than for baling. - Dry matter loss was around 3% to 4% in the delivery system.  Sokhansanj et al. (2008b)  ExtendSim  Straw  Comparison of harvest scenarios 1) large square bale 2) round bale 3) loaf 4) dried chops 5) wet chops  N/A  - The cheapest and most expensive scenarios were loafing ($17.1dt-1) and wet chopping ($59.8dt-1). - Sensitivity analysis showed 20% reduction in cost by 33% increase in yield.  Stephen et al.  ExtendSim  Crop residues  Investigating the impact of biomass availability on the delivered cost over the service life of a biorefinery.  Peace River, Alberta  - Agricultural regions such as Peace River are not a good choice of location for a biorefinery due to the annual variability in biomass availability.  (2010)  Objective  Case  Findings/important aspects  study/region  23  Hess et al. (2009) applied their model to a theoretical plant with a size of 800,000 ton/year and a feedstock supply radius of 80 km for stover and 105 km for switchgrass. The obtained results of the simulation showed that the conventional bale supply system is not able to achieve the 2012 DOE cost target of $34.7 dt-1. The average delivery cost ($/dt-1) of stover and switchgrass were estimated to be 55.4 and 49.6, respectively. The obtained results also showed that the pioneer uniform supply system design is not able to achieve DOE cost targets. For example, the total logistics costs for corn cob would be $68.9 dt-1. Hess et al. (2009) conducted a sensitivity analysis. The obtained results revealed that the improvement in equipment efficiency (shredder field speed, baler capacity and efficiency, and harvest window) and biomass properties (bulk density and moisture content) could result in a cost-efficient bale stover supply system. The results for advanced uniform have not been published yet. Similar to IBSAL and SHAM, the developed simulation by Hess et al. (2009) models and evaluates the details of the supply chain but the discussion on whether the model enables to plan and schedule the operations to meet the daily demand has not been provided. In addition, they assumed an equal distance distribution of farms throughout the supply radius. In addition to corn stover, straw and switchgrass as the main agricultural biomass considered for biofuel production, the supply chain of energy crops including miscanthus, reed canary grass, willow and hemp have been studied. Huisman (2003) developed a simulation model to find the minimum supply costs by selecting the best harvesting and storage system for each energy crop. The total cost encompasses harvest, transport, drying and storage costs. Although the framework of the simulation model was illustrated, detailed information on the simulation model and how the model assists in the selection of the optimal supply chain was not given in this work. In addition to the estimation of the logistics costs and delivered biomass, simulation modeling has been used to determine the location of satellite storage sites in the supply area. Cundiff et al. (2009) employed simulation modeling to determine the number of satellite storage locations in the supply system and also the hauling distance from production fields to satellite storage locations. Their simulation study revealed that the allowable hauling distance is 3.2 km (2 mi) based on the total operating time to haul bales from a 16-ha field to satellite storage. The simulation study was carried out within a 50 km (30 mi) radius of Gretna, located in South Central Virginia, a region that has a high potential for switchgrass production. 24  Simulation modeling has also been employed to study the supply chain of woody residues. Zhang et al. (2012) developed a simulation model using Arena Simulation Software with a graphical user interface for a biofuel supply chain. Key supply chain activities including biomass harvesting/processing, transportation, and on-site storage were considered in the model. The considered woody biomass included pulpwood and wood residues. The model used the delivered feedstock cost, energy consumption, and GHG emissions as system performance criteria. Zhang et al. (2012) verified the developed model through a series of simulation runs applied to several biofuel facilities in the Lower Peninsula of Michigan. The moisture content was assumed constant (50%) throughout the supply chain. Thus, biomass weight delivered from harvesting areas to the biofuel facility remained the same. The simulation model finds the most and least favorable locations and sizes of bioenergy plants. Modeling of agricultural crops such as corn grain is another area explored by simulation. Arinze et al. (2001) and Sokhansanj et al. (2003) used simulation to model the changes in quality of potash fertilizer and alfalfa cubes during storage and transport. Humphrey and Chu (1992) and Benock et al. (1981) developed a GASP IV-based simulation model to analyze harvesting, transportation, and drying of corn. Although simulation modeling is a powerful tool for detailed evaluation of the supply chain under different circumstances, it is not able to find the best decisions among alternative ones. Therefore, to achieve the optimal decisions on biomass inventory, shipped biomass or location and size of different facilities, the application of mathematical optimization modeling1 is inevitable (Berruto and Busato, 2008).  2.4 Optimization modeling in the biomass supply chain Mathematical optimization modeling has been broadly used in a variety of industrial and academic fields. However, its application to the agricultural biomass supply chain has been limited. Several researchers have applied the mathematical optimization modeling approach to find the optimal solutions for logistics problems such as monthly working schedule for different logistics operations, the amount of biomass shipped, processed and stored in the supply system, and also the location, number, type and capacity of bioenergy plants and storage sites. 1  Selection of the best alternative from a set of available alternatives with respect to some criteria (INFORMS Computing Society, 1996)  25  Cundiff et al. (1997) developed a linear programming (LP) model to design a biomass delivery system for switchgrass producers who provide feedstock for a central plant. The proposed optimization model determined a monthly shipment plan and capacity expansion schedule for each switchgrass producer based on monthly harvest and four different weather scenarios. A mixed-integer linear optimization model was developed by Nagel (2000) to find the most economical and ecological (based on CO2 emissions) energy supply structure for heating a small rural community. The main decision variable was whether they should build heating systems (individual energy supply), a heating plant, or a co-generation plant. The proposed model was used in the rural municipality of Brandenburg, Germany with 660 inhabitants. The results showed individual heating systems became an attractive option with decreasing prices for biomass. The produced CO2 could decrease up to 25% by increasing the use of biomass. In addition, biomass would not be an economical feedstock with decreasing prices for fossil fuels. Another application of optimization modeling can be seen in the work of Kaylen et al. (2000). They developed a mathematical programming model to analyze the economic feasibility of producing ethanol from various lignocellulosic biomass materials including agricultural residues, energy crops, wood processing and logging residues in Missouri. They analyzed the tradeoffs between scale economies and transportation costs, and found that the estimated net present value (NPV) of the plant would be maximized at a capacity of 4,360 tons per day, under the conservative assumption that only 10% of available biomass could be used in the plant. Tembo et al. (2003) proposed a mixed integer mathematical programming model for the lignocellulosic biomass-to-ethanol industry. The optimization model determined the most economical source of biomass, timing of harvest and storage, inventory management, biorefinery size, and biorefinery location. It also identified the breakeven price of ethanol for a gasificationfermentation process. The objective function of this model was to maximize the industry’s net present value. The optimization model was applied in Oklahoma State where a variety of potential lignocellulosic feedstock including plant residue, indigenous native prairies, and pastures are available. The primary finding of the study was that gasification-fermentation of lignocellulosic biomass to bioethanol may be more economical than fermentation of corn grain with the further development of the conversion process to handle multiple feedstocks and reduction in the capital and operating costs of the plant. 26  Due to the high magnitude of potential agricultural biomass in Thessaly, Greece, several studies have been conducted in this area to investigate and optimize different aspects of the biomass supply chain using mathematical programming. Tatsiopoulos and Tolis (2003) applied an LP model to optimize economic aspects of the logistics network by comparing various biomass collection and transportation systems, including the farmer transportation system and third party logistics (3PL) handling system. The model was applied to the cotton biomass cultivated in Thessaly. The study concluded that it would be cheaper if the farmers were engaged in the logistics network. Rentizelas et al. (2009) analyzed the impact of three different biomass storage methods on the total cost of the system. The storage methods included: 1) closed warehouse with biomass drying capability, using heated air (WD), 2) a covered storage facility of a pole-frame structure having a metal roof without any infrastructure for biomass drying (CND), and 3) an ambient storage of biomass, covered only with a plastic film (AS). The model was implemented for the case study of a municipality of Thessaly. The final results showed that simple and cheap storage solutions (AS) could be used for the cheapest biomass type available. Whereas, expensive storage solutions (WD) were a more attractive option for multi-agricultural biomass approach in order to reduce the storage space required. In addition, no drying was required if the arriving biomass to store had a low moisture content (e.g. 15–20%). Other work done by Rentizelas et al. (2009) was the development of a decision support system (DSS) for multi-biomass energy conversion applications. The DSS provides the investor with information on: 1) the best location to establish the biomass-to-energy facility, 2) the optimal relative size of the base-load combined heat and power (CHP) unit and the peak-load boiler, and 3) the amount and location of each locally available biomass type that should be used. A system-wide modeling of the whole bioenergy system was considered in the DSS to find the global optimal design. The objective function of the optimization model was to maximize the net present value of the investment for the project’s lifetime. The DSS was applied to a trigeneration application (electricity, heating and cooling) at a municipality of Thessaly. They concluded that the developed DSS provides visibility to the potential investor regarding the details of the optimum design of the facility and the fuel supply chain, as well as the sensitivity of the investment on a set of investment parameters such as inflation, electricity price and biomass cost. 27  To make both long-term decisions (supply chain design-related decisions) and medium and short-term decisions (logistics management decisions), Ekşioğlu et al. (2009) developed a mixinteger programming model. The objective of the mathematical model was to minimize the cost of delivering biofuel by coordinating long- and mid-term decisions and also logistics management of a biorefinery. The long-term decisions were comprised of the number, size and location of biorefineries. The mid- and short-term decisions included the amount of biomass shipped, processed and stored during a certain time period. The State of Mississippi was used as the testing ground of the proposed model. The obtained results revealed that transportation costs and biomass availability are the primary parameters affecting the supply chain-design decisions. In case of low biomass availability, smaller sized biorefineries become commercial. The results also indicated that the supply chain-design decisions were not impacted by changes in biomass collection costs and biomass processing costs. Zhu et al. (2011) proposed a mixed integer linear programming (MILP) model to find the optimal configuration of a biorefinery supply chain. Two types of biomass were transported in the supply chain: switchgrass and residue. Residue was the by-product of the conversion process which was recirculated to the soil. The MILP model prescribed the optimal solutions for both strategic decisions about the supply chain design and tactical decisions on the annual operation schedules. These decisions consisted of whether to establish a biorefinery and an infield/intermediate/at-plant warehouse and their capacities, scheduling the harvest operation for each switchgrass production field, the size of the harvesting team, and transportation flow of switchgrass and residues among the logistics network. The developed model was verified via numerical examples. Due to the large size of the developed MILP model, it was not suitable for solving the large supply area. Moreover, it was mentioned in the study that the monthly fixed costs of operating an intermediate warehouse was assumed to be $60,000. However, no discussion was given on what would happen if the warehouse was kept closed for a month. Moreover, the authors did not elaborate on the components of the fixed cost of the warehouses and their service lives. As an extension of their previous study, in addition to dedicated energy plants such as switchgrass, Zhu and Yao (2011) considered agricultural residues (e.g., corn stalk and wheat stalk) in their MILP model. Thus, the developed MILP model was a multi-commodity network flow model. The model solution covered the harvesting and storage of switchgrass, purchasing of 28  stalk and straw, delivering of biomass to biorefineries, processing of biomass in biorefineries, and handling of biomass residue. The numerical study showed that the use of multiple types of biomass could increase the supply of biomass, alleviate the seasonality caused by the nonharvesting season of switchgrass, smooth the biofuel production, and eventually increase the unit profit of biofuel. The model also addressed the dry mass loss of biomass in the supply network due to various operations, fungal degradation, fermentation and breakdown of carbohydrates. Similar to their previous research, the established model was a large-sized optimization model. One of the interesting recent studies is the work of Akgul et al. (2012). They built a multiobjective optimization framework for a hybrid first/second generation biofuel supply chain. The main purpose of the optimization model was to design the hybrid bioethanol supply chain considering economic and environmental objectives, simultaneously. The problem was formulated as a static multi-objective, mixed integer linear programming (MILP) model. In this work, the total daily cost of the supply chain was minimized where the total environmental impact of the supply chain must be less than or equal to the maximum allowed GHG emissions from the supply chain. The proposed model was applied to a case study of bioethanol production in the UK. The environmental impacts were evaluated based on a life cycle analysis (LCA) approach. In addition, the effect of considering carbon tax on the overall environmental emissions was investigated. The results showed that the second generation biofuel production technologies would outperform the first generation ones in terms of the potential GHG savings. In a few recent studies, mathematical optimization modeling has been used to manage satellite storage locations in the biomass supply chain. Judd et al. (2010) developed an integer programming model to determine the optimal number of satellite storage sites and their locations as well as the assignment of production fields to the selected satellite storage sites. In the developed model, the total cost of transporting biomass from farms to satellite storage plus the establishment costs of satellite storage were minimized. The number of satellite storage locations determined by the optimization model was much less than that in Resop et al. (2011) who established a raster-based approach to locate satellite storage sites. However, the developed optimization model did not consider transportation costs from satellite storage sites to the ethanol plant. This cost component would impact the location and number of satellite storage sites within the supply radius.  29  In a similar study, Judd et al. (2012) exploited an MILP model to find the optimal number and location of satellite storage sites. The main novelty of the model was the consideration of sharing loading equipment among satellite storage sites. Equipment sharing was compared with the fixed position equipment scenario in which each piece of loading equipment was assigned to only one satellite storage location. The equipment sharing resulted in significant cost saving (14.8%) compared to the fixed position equipment scenario. In addition, three handling systems at the satellite storage sites were compared. Kim et al. (2011) considered uncertainty in their developed optimization model to find the optimal design of the supply chain network. The decision variables included the number, size, and location of two bio-processing units, optimal flow of biomass from forest areas to the conversion units as well as the transportation of the final products to the market. Fourteen variable parameters were considered in the optimization model such as biomass availability, purchase price of biomass and transportation costs. A two-step procedure was developed to find the optimal solutions. The procedure was applied to the Southeastern region of the United States and the final results showed the efficiency of the developed model to mitigate the impact of the variation on the overall profit. Chen and Fan (2012) also considered supply and demand uncertainties in the bioethanol supply chain. A two-stage stochastic programming model was developed to optimize the supply chain for a waste-based ethanol production system. The application of the model to an ethanol plant in California showed that ethanol can be produced at a competitive price of $1.2 per gallon. Another finding was the importance of feedstock diversification to mitigate the annual feedstock fluctuation. The developed model is at yearly-aggregated level and is not appropriate for monthly planning. Most of the developed optimization models focus on strategic and tactical planning levels of the biomass supply chain. These models do not take time-dependency and stochasticity of the supply chain into account in order to find the optimal solutions in a reasonable time. As reported by Ekşioğlu et al. (2009) and Zhu et al. (2011), the developed optimization models might not be solvable for real-world cases. A few of them elaborated on the operational level. Consideration of operational decisions in the optimization modeling increases the computational times. Due to complexity and dynamics of real-world supply chain networks, application of only one method such as simulation or optimization does not guarantee an optimal supply chain network 30  (Almeder et al., 2009). Using the combination of these methods under one framework to achieve an optimal operational plan for supply chain networks has recently received attention among researchers, for instance: Truong and Azadivar (2003), Lee and Kim (2002) and Almeder et al. (2009). The combined simulation/optimization model would allow consideration of all decisionmaking levels including strategic, tactical and operational levels. In addition, the timedependency and uncertainty of the supply chain can be considered in the combined approach.  2.5 Combined modeling methods in the biomass supply chain The idea of using a combined simulation/optimization approach in the biomass supply chain is new. To the best of the author’s knowledge, the only research in the context of the application of a combined optimization/simulation in the biomass logistics system is the work of Berruto and Busato (2008). They presented a combined approach of simulation and linear programming models to optimize the flow of biomass from the field to a power plant. In the combined approach, first the simulation model projected the performance of a given logistics network under varying input parameters including resource availability, crop yield, field size and shape, and transport distance. The results of the simulation model assisted in developing a linear regression equation. This equation represented the cost per ton of dry matter as a function of the transportation distance for different yields. This equation was then used in the optimization model. Thereafter, the optimization model prescribed the best combination of equipment for each field distance and yield in order to minimize the logistics costs. The considered biomass supply chain was a simple supply network which included forage harvesting, shipping biomass to the plant and unloading to the silos. The combined model had a simple structure and there was no feedback from the optimization model to the simulation model. It is noted that DeMol et al. (1997) developed both simulation and optimization models but these models were not combined. The simulation model, called Biologics (BIOmass LOGIstics Computer Simulation), estimated the logistics costs and energy consumption for different logistics scenarios. The developed optimization was an MILP model to find the annual flow of multi-biomass at the minimum costs. Both models were applied in a proposed bioenergy plant in 31  the province of North-Holland, Netherlands. The finding of their research was that the results of the simulation and optimization models were comparable. However, the optimization model can be used to select the optimal supply network or the optimal mixture of biomass types (strategic level) while the simulation model should be used when the supply network is given and it provides more detailed results on the biomass logistics (tactical level). In addition to the combination of simulation and optimization modeling, some studies have used the combination of other approaches. Freppaz et al. (2004) developed a decision support system (DSS) that includes a GIS-based interface, a database and an optimization module. The optimization module determined the size of thermal and electricity plants and also the optimal flow of biomass to the plants. The GIS-based interface was used to plan both forests eligible for biomass collection and sites to locate the energy conversion plants. The parameters of the optimization module were also computed by the GIS-based interface. Ravula (2007) developed a raster-based model to locate satellite storage sites in two separate production regions the Southern Piedmont and Northern Piedmont regions of Virginia. Potential satellite storage locations were determined by converting spatial data into a binary raster where the value of one/zero implies the high/low concentration of biomass inside a cell. Thereafter, the total number of cells within a 3.2-km radius of each cell in the raster was calculated. The calculation showed where the areas of highest and lowest concentration of potential production were located to establish satellite storage. A 3.2-km limitation on procurement regions around each satellite storage location was introduced to reduce bale wagons’ travel distance due to their low speed (less than 32 km/h (20 mph)). Bale wagons were used to transport the bales on the field to the storage location. Given the location of satellite storage sites, the location of the processing plant was then determined. To this end, each cell in the considered raster became a potential location. Thereafter, the travel time from each satellite storage location to every other cell in the region was computed. Finally, the cumulative travel time to all satellite storage locations was calculated for each cell. The cell with the minimum cumulative travel time was selected as the optimal location for the processing plant. Ravula (2007) also developed a simulation model to schedule the shipment of  biomass from satellite storage locations to the processing plant. The simulation model was also used to check the sensitivity of the supply system to variations in land use rate, truck travel times, and plant processing times. 32  2.6 Discussion and conclusions Depending on the type of the supply chain decisions, several modeling approaches have been used including static modeling, simulation modeling, optimization modeling and their combination. Each of these modeling approaches has advantages and disadvantages compared to others. Based on the research objectives explained in Chapter 1, the primary decisions made in this research are to design and schedule the highly-constrained agricultural biomass supply chain in the Canadian Prairies. This supply chain should be designed and scheduled to meet the daily demand of a commercial-sized ethanol plant with the minimum delivery cost possible. The simulation modeling can be used to plan and schedule the supply chain as it enables us to consider the dynamics and high level of complexity and uncertainty of the system. However, the estimated costs by the simulation model are not necessarily the minimum ones. The minimum delivery costs can be achieved via the development of an optimization model. The optimization model makes it possible to find the optimal design of the supply chain in which the biomass demand is met at the lowest delivery costs. To avoid a complex and large-sized optimization model, all the uncertainties and complexity of the supply chain cannot be incorporated in the model. Thus, the optimal design cannot guarantee the fulfillment of the daily demand. Therefore, an integration of the modeling tools is required. The literature lacks an integrated modeling approach in which the dynamics and stochasticity of the supply chain have been considered in both design and scheduling of the biomass supply chain. In addition, the details of the supply chain such as farm management, multiple biomass, storage and transportation management, moisture content, delays, and the business model of the supply chain have not been taken into account to fulfill the daily demand of a commercial-scale ethanol plant. Due to the different size and location of farms and their produced biomass, they should be considered as separate suppliers in the supply chain instead of aggregating them at a county or municipality level. Analyzing the supply chain at the farm level provides accurate cost and delivery outputs. It also allows the feedstock manager to help the farms in using their equipment efficiently. In addition, each farmer may grow different crops in a year resulting in the production of multiple biomass types. The supply chain should be able to manage the flow of multi-biomass considering the specific characteristic of each biomass type. 33  The storage management plays a key role in meeting the daily demand year-round as the harvest season is short while the conversion facility operates 24/7/365. Biomass gradually enters and leaves storage during the year depending on the harvest schedule, daily biomass demand, weather conditions and also equipment availability. In addition, biomass is prone to degrade during storage. These factors affect the storage costs, the amount of dry matter loss (DML) and the associated costs, storage capacity and inventory. In the relevant literature, the calculation of storage capacities and the associated costs are either based on the optimization models as in Cundiff et al. (1997) or are based on the assumption that storage is filled once or twice per year as in Cundiff et al. (2009), Judd et al. (2012) and Judd et al. (2010). It was also assumed that biomass is available at the storage sites on day one of every year which could not be the case in practice. An accurate estimation of the capacity of storage locations and the associated costs requires tracking of the daily inflow and outflow of biomass in storage. In addition, the daily flow of biomass in storage sites provides an accurate estimation of the amount of DML since the amount of DML in storage depends on the duration of storage time. Thus, the flow of biomass in storage locations must be modeled on a daily basis while considering the time-dependency and uncertainty of the supply chain. Another storage-related issue is the initial inventory in the supply chain. The initial inventory is important to meet the feedstock demand at the beginning of harvest season when there is not sufficient biomass produced available in the system. Initial inventory has not been explored in the relevant literature known to the author. In addition, as stated by Cundiff et al. (2009a), the role of the at-plant storage to avoid any feedstock shortage in the conversion facility has not been fully studied. In addition to storage, the transportation system can be improved by considering the uncertainties in the supply system. To assure the fulfillment of the daily demand, three decisions must be made which influence the fleet utilization: the quantity of daily biomass delivered to the plant, the capacity of at-plant storage and the number of trucks in the supply system. These decisions must be made based on the different operational windows of logistics operation, dry matter loss in the system, and uncertainties such as machine breakdown and weather conditions. Other detailed analyses neglected in the literature are delay times and the continuous track of moisture content. Sokhansanj, Kumar, et al. (2006), Sokhansanj et al. (2008) Sokhansanj et al. 34  (2009) and Kumar and Sokhansanj (2007) considered weather-related delay times and moisture content in the supply chain. However, the delays due to the unavailability of machines were not incorporated in these studies. The consideration of these delays is of importance as they may cause bottleneck in the supply chain affecting the planning and scheduling of the machines in the supply chain to meet the daily demand. In addition, the estimation of the moisture content in these studies were done in a discrete fashion in that the moisture content of biomass in a specific day was estimated based on the last recorded moisture content of biomass and the weather data for the respective day. Thus, the time period between the time when the last moisture time was recorded and the respective day were not taken into account. The continuous update of moisture content provides insights for the supply managers whether to take extra actions to manage the moisture content in the system in the case that the moisture content is not within a safe level. In addition, a better estimation of the moisture content assists in better estimation of dry matter loss and the associated costs in the system. Another deficiency of the relevant literature relates to the business model of the agricultural biomass supply chain. Although the general structure of the business model has been discussed in the literature, the duration of contract with farmers and their participation rate have not been taken into account as an input parameter in the modeling and analysis of the supply chain. These parameters impact the number and location of farms in the supply area. In this study, an integrated simulation/optimization model is proposed which alleviates the above-mentioned deficiencies. Prior to explaining the developed integrated model, the details of the case study to which the integrated model was applied, are given in the next chapter.  35  Chapter 3. Case study 3.1 Synopsis In this chapter, the case study used in this research to evaluate the efficiency of the developed models, is presented. The case study is a proposed commercial-scale cellulosic ethanol plant located in Prince Albert, in north-central Saskatchewan. Several organizations were contacted to gather the data. Each organization provided specific types of data based on their area of specialty. The data were gathered by searching their websites and communication by email, phone calls and in-person meetings. The organizations and the sets of data they provided include:   Iogen Corporation: Ethanol plant data    Agriculture and Agri-Food Canada (AAFC): Weather data and supply area data    Saskatchewan Ministry of Agriculture: Farm data, crop data, and harvest schedule data    Idaho National Lab & Oak Ridge National Lab, American Society of Agricultural and Biological Engineers (ASABE): Equipment data and cost data    Statistics Canada: Supply area and farm data In addition to these sources, the studies in the literature were used to gather general data such  as breakdown and repair data for equipment, grain moisture content and dry matter loss data. The details of the gathered data are explained in this chapter.  3.2 Ethanol plant data Figure 3-1 shows the location of the proposed ethanol plant. The estimated annual ethanol production capacity of the plant is more than 70 ML (Iogen Corp., 2009). The conversion process is a combination of thermal, chemical and biochemical technologies. In addition to the cellulosic ethanol as the main product, the plant would generate green electricity from processed by-products and forestry waste. The service life of the plant is estimated to be 25 years. The cellulosic ethanol is produced from locally-grown cereal straw, mainly wheat straw. Three prevailing types of wheat straw in the region under study include spring wheat (Triticum aestivum), durum wheat (Triticum turgidum) and winter wheat (Triticum hybernum). To reach 36  an annual capacity of 70 ML, a daily delivery of 750 t of wheat straw is estimated. No discussion on the preference of one type of wheat straw to another type to deliver to the plant has been given by Iogen. Table 3-1 summarizes the characteristics of the case study.  Figure 3-1: Location of the cellulosic ethanol plant, Prince Albert, Saskatchewan (by permission from Iogen Corp.)  Table 3-1: Characteristics of the proposed cellulosic ethanol plant Characteristics Bioenergy product  Cellulosic ethanol  Location of plant  12 km east of Prince Albert, north central Saskatchewan  Annual production capacity  70 million liters (ML)  Life time  25 years  Feedstock  Wheat straw (winter, spring and durum wheat)  Daily demand  750 t  37  3.3 Supply area data To meet the daily straw demand for the conversion facility, a supply area with the radius of 160 km was considered by Iogen (Iogen Corp., 2009). Figure 3-2 depicts the availability of straw within a 160-km supply radius. The yellow polygon represents the supply area which is about 53,800 km2. As shown in this figure, no wheat straw is available north of Prince Albert. This supply area was produced by the Biomass Inventory Mapping and Analysis Tool (BIMAT) (Stumborg et al., 2008). BIMAT1 is an Internet-based Geographic Information System (GIS) developed by Agriculture and Agri-Food Canada (AAFC). It estimates the availability of different types of agricultural and forestry residues within a user-defined radius from a userselected geographic point location across Canada. The estimation is based on a gross crop yield and discounted for soil conservation and the use of biomass for animal feeding and bedding. However, BIMAT does not take into account biological and climate constraints on the availability of biomass. Thirty one rural municipalities are located inside 160-km supply radius. Appendix A presents the list of the rural municipalities, their land area, number of wheat farms and distances from the ethanol plant. The locations of rural municipalities in the supply area are shown in Appendix B. Almost 98% of wheat farmers grow spring wheat and the rest produce durum and winter wheat. BIMAT estimates that 1.5 Mt of wheat straw is available within the 160-km radius which is the net available biomass for ethanol production. In other words, this value is the total remaining wheat straw that can be collected and delivered to the ethanol plant, taken into account soil conservation and livestock requirements.  1  http://www4.agr.gc.ca/AAFC-AAC/display-afficher.do?id=1226509218872  38  Figure 3-2: 160-km supply area considered for the proposed cellulosic ethanol plant (by permission from Agriculture and Agri-Food Canada)  3.4 Farm data The most challenging part of the data gathering process was collecting farm data. This is due to the uncertainty in the number of farmers who are willing to participate in the supply system during the business life of the ethanol plant. The average number of wheat farms which can be potential feedstock providers is 3,055 including 2,985 spring wheat, 49 durum wheat and 21 winter wheat farms. These numbers would change with the change in the participation rate. More important, the portion of farmlands that farmers assign to each type of crop in each year is indefinite beforehand since farmers typically include several crops in rotation. Thus, during the lifetime of the ethanol plant, it would be uncertain which farm will grow wheat and what portion of the land is assigned to the cultivation of this crop. Another difficulty in gathering farm data was the location of farms in the supply area. There is no spatial data available on the location of each single farm in the supply area. None of the organizations could provide fine spatial data. The available information is not at the farm and rural municipality levels, but at the crop district level which is a larger aggregation of the smaller rural municipalities. Figure 3-3 shows a map of the crop districts within the province, as well as  39  the rural municipalities within each crop district. Prince Albert is in crop district 9AE, and is surrounded by crop districts 9AW, 8B, and 8A. Therefore, due to the uncertainty in the amount of wheat cultivated in the supply area in any given year and also the unavailability of fine spatial analysis, the farms were created randomly inside each rural municipality. Three criteria were used in this regard:   Total number of created farms in each municipality does not exceed the number of farms growing wheat in the respective municipality.    Total size of created farms in each municipality does not exceed the total land area under cultivation of wheat in the respective municipality.    Total produced biomass in all created farms does not exceed the total available biomass within the 160-km supply radius, estimated by BIMAT (1.5 Mt). To assure there is no overlap between the created farms, the origin in each municipality was  set as the center of the municipality. A coordinate was then assigned to the center of each created farm. For each new created farm, it was checked whether the new created farm has no overlap with the preceding created farms. If one of the first two criteria is met, no more farms will be created in the respective municipality. If the last criterion is met, the farm creation process terminates.  40  Figure 3-3: Crop districts and rural municipalities in the province of Saskatchewan (by permission from Agriculture and Agri-Food Canada) To assign size to each created farm, the historical data on the size of farms in Saskatchewan were used. Table 3-2 shows these ranges. The farm size ranges from 0.4 to 1,620 ha. Roughly 70% of farms have a size between 100 and 900 ha. To assign size to each created farm, a random number is created using a uniform distribution function based on the size ranges in Table 3-2. It is assumed all the created farms are square-shaped.  41  Table 3-2: Farm size range in Saskatchewan (Saskatchewan Agriculture and Food, 2006) Size range Number of farms  Minimum farm size (ha)  Maximum farm size (ha)  431  0.4  4  1254  4  28  1084  28  53  4447  53  73  732  73  97  4698  97  162  3542  162  227  3668  227  308  5448  308  453  5589  453  648  5103  648  907  2946  907  1166  1795  1,166  1,425  3592  1,425  1,620  Another parameter that must be determined is the contract length with farmers. Iogen suggested a standard production contracts with a length of 5-6 years signed with individual farmers (Altman et al., 2007). In this study, it is assumed that contract length with farmers is 5 years. During this period, the same farmers are involved in the procurement process. Thus, the location of farms does not change during this 5-year period; however, the yield and the size of farms assigned to grow wheat may change from year to year.  3.5 Crop data The yield data for wheat grain were gathered from the Saskatchewan Ministry of Agriculture for the period of 1980-2009 through their website1. The yield data were available at the municipality 1  http://www.agriculture.gov.sk.ca/Default.aspx?DN=5e3d0f74-ef7a-49f5-a975-f340e11fa394  42  level not at the farm level. The historical yield data for each crop and municipality were fitted over different distribution functions to find the most suitable distribution function. This was done by the distribution fitness feature of the Extendsim software. The results showed that grain yield fairly follows a normal distribution regardless of wheat type and region. However, the mean and standard deviation of the wheat yield depends on the crop type and the rural municipality for which the data were available. Thus, grain yield is described by a normal distribution based on the statistical analysis on the available data: (̅  (  ))  (3.1) where ̅  nd S(Yij) are the mean and standard deviation of the yield of wheat type i in rural  municipality j. The values of mean and standard deviation for each wheat type and rural municipality are presented in Appendix C. Wheat grain distributions for three different rural municipalities are depicted in Figure 3-4. The average grain yields for spring, durum and winter wheat are 1.97 t/ha, 2.13 t/ha and 2.35 t/ha, respectively. These values are comparable with the average value reported for wheat yield (1.89 t/ha) in Saskatchewan (Sokhansanj et al., 2006a). In addition to the grain yield, the following parameters need to be estimated to calculate the straw yield:   Bulk density of grain: it is assumed the bulk density of wheat grain for all varieties is 37 bu/t (60 lb/bu) (Manitoba Agriculture and Food, 2009)    Grain moisture content: the standard grain moisture content is assumed to be 12% (w.b.) (Sokhansanj et al., 2006a).  43  0.2 Probability  0.15 0.1 0.05 0  1.1  1.2  1.3  1.5  1.6  1.7  1.9  2.0  2.2  2.3  2.4  2.6  2.7  Spring wheat grain (t/ha)  a) Prince Albert municipality  Probability  0.20 0.15 0.10 0.05 0.00 0.7 0.8 0.9 1.1 1.2 1.3 1.5 1.6 1.7 1.9 2.0 2.2 2.3 2.4 2.6 2.7 2.8 3.0 3.1 3.2 Durum wheat grain (t/ha)  b) Willow Creek municipality  Probability  0.2 0.15 0.1  0.05 0 1.7  1.8  1.9  2.1  2.2  2.4  2.5  2.6  2.8  2.9  3.0  3.2  3.3  Winter wheat grain (t/ha)  c) Spiritwood municipality Figure 3-4: Distribution of wheat grain (spring wheat, winter wheat and durum) yield in three rural municipalities inside the 160-km supply radius  44    Biomass to grain ratio: There is a wide range in biomass to grain ratio in the literature for each crop. Sokhansanj et al., (2006b) reported a range of 0.75-1.6 for wheat in Canadian Prairies. In this study, the ratio reported by Saskatchewan Ministry of Agriculture is used. This ratio is assumed to be 1.6 (1.6 t of residue per t of wheat harvested). This ratio has also been used by Agriculture and Agri-Food Canada (2003). On average, the biomass yield for spring, durum and winter wheat equates to 3.11, 2.88 and  3.44 t/ha, respectively. The average estimated yield of wheat straw in the region under study is comparable with the estimated value for the yield of wheat straw in the province of Saskatchewan by Sokhansanj et al., (2006b). They estimated that the average straw yield would be 2.61 t/ha. After the estimation of straw yield and farm size, the gross available biomass in each farm is calculated. The following two parameters are then considered in order to estimate the net straw available at each farm for ethanol production:   Soil conservation rate: Different ranges have been recommended to be left for soil protection from wind and water erosion such as 30-50% of available biomass (Kline, 2000) or 50-70% (Lindstrom, et al., 1979). In this study, it is assumed the soil conservation rate changes from 30% to 70% with most likely rate of 50%. Thus, a triangular distribution function was used to express the conservation rate with 30%, 50%, and 70% parameters as minimum, most likely and maximum values.    Livestock requirement rate: BIMAT takes the livestock requirement into account to estimate the available biomass for industrial purposes. Based on the output of BIMAT, the livestock requirement rate is around 5% of produced biomass in the field. By considering that a portion of the available biomass must be used on the field for soil  conservation and for livestock requirement, the net available biomass for ethanol production was estimated for each farm. The net produced biomass in a farm ranges between 5 t to 4,510 t.  3.6 Weather data In order to incorporate the effect of weather conditions on the logistics activities, the daily weather data for the supply area were gathered from AAFC for the period 1990-2010. Weather data contain daily dry bulb temperature (°C), daily snowfall (mm), daily average relative 45  humidity (decimal), daily evaporation (mm), and daily rainfall (mm). These parameters affect the quality of biomass such as moisture content, dry matter loss and the performance of the equipment in the supply chain.  3.7 Harvest schedule data The Saskatchewan Ministry of Agriculture publishes historical data on weekly harvest progress of major crops in the province. These data are available from 2005 on their website1. The harvest progress reports show that the harvesting season for all three wheat types begins roughly in early August and lasts until mid-October, occasionally extending to late October or early November. The progress of the harvest season depends on the availability of machines, labour and local weather patterns (Sokhansanj et al., 2008b). However, the primary factor on the length of the harvest season is weather conditions (Judd, 2011). More than 60% of the crop is often harvested during September. Cool, rainy conditions in the northern areas slow down the harvest in the middle of September, but usually the return of warm, dry conditions by the end of the month allows the harvest to continue. Harvest operations end for the most part by the third week of October. The collection of straw starts immediately after the first field is harvested for grain. Among the available data for the harvest season, the shortest harvest season took place in 2006 which lasted for 30 days due to good weather conditions. In contrast, 2009 had the longest harvest season lasted for 90 days due to frequent periods of rain and an earlier than normal snowfall. Figure 3-5 shows the weekly progress of harvest season for three wheat crops. The graphs are based on the average of years 2005-2010. Winter wheat has the shortest harvest season, followed by durum and then spring wheat. As shown here, the harvest season is most likely interrupted and remains incomplete (the accumulative harvest percent is less than 100%). This is mainly due to the delays caused by unfavorable weather conditions. These delays cause the overlap of the harvest season and the next cropping season. Figure 3-6 illustrates the harvest percentage in each month. For all the wheat crops, most of the harvest takes place in September, followed by October and August.  1  http://www.agriculture.gov.sk.ca/crop-report  46  The high dependency of the harvest progress on the weather conditions complicates the machinery management to harvest, collect and stack enough biomass to meet the annual demand  Cumulative harvest percent  of the ethanol plant before the commencement of the following cropping season.  1 0.8 Spring Wheat  0.6  Durum Wheat Winter Wheat  0.4 0.2 0 1-Aug  1-Sep  1-Oct  1-Nov  Time  Figure 3-5: Weekly harvest progress for different wheat crops (Saskatchewan Ministry of Agriculture, 2011)  Winter Wheat  Durum Wheat 16.6%  1.6%  6.3%  Spring Wheat 12.4%  3.3%  10.7% August September October November  28.9%  74.4%  69.4%  69.3%  Figure 3-6: Monthly harvest percent for different wheat crops  47  3.8 Structure of the supply chain in the case study Figure 3-7 depicts the sequence of operations in the considered agricultural biomass supply chain for cellulosic ethanol production. The supply chain encompasses in-field drying, baling, in-field transportation, storing, loading, road transportation, truck weighing, unloading, at-plant storing and grinding. The operations at the conversion facility are part of the ethanol production system, and thus, they are not incorporated in the supply chain. In addition, operations associated with wheat production on the farm are not taken into account as part of the supply chain. Agricultural residues are a byproduct of grain harvesting and activities such as pre-harvest operations, fertilizing, mechanical weed control, seeding, and herbicides are performed on grain as the primary product. Thus, their associated costs are incorporated in the grain production costs.  Baling (square baler and tractor)  In-field drying  Loading bales onto trucks (self-propelled telehandler)  Unloading bales (self-propelled telehandler)  In-field transportation (self-propelled stinger)  Road transportation (flat-bed trailer and tractor)  Storing bales at the at-plant storage site  Storing bales at storage  Weighing trucks at the ethanol plant (truck scale)  Grinding (grinder in-feed system, hammer mill, dust collection system and surge bin  Figure 3-7: The modeled wheat straw supply chain  A square baler is used in the field to compress loose wheat straw to the large rectangular bales. The square baler is powered and pulled by a tractor. Thus, both baler and tractor are used in the baling operation. A self-propelled in-field transporter called stinger is used to move created bales from farmlands to a nearby storage site. Stinger also stacks the bales at storage. This equipment is used in the in-field transportation operation. The next self-propelled equipment is telehandler utilized to load stored bales on trucks. Each truck encompasses two pieces of equipment, a flat-bed trailer and a tractor to pull the trailer. Trailer and tractor are both used in the road transportation operation.  48  A truck scale is used at the gate of the ethanol plant to weigh the arriving trucks. The next equipment is a self-propelled telehandler to unload bales from trucks and place them at the atplant storage. In the last operation, four pieces of equipment are used in the grinding station to reduce the size of biomass and prepare it for the conversion process. First, the grinder in-feed system directs biomass into the grinder. The grinder then processes biomass into a bulk format for insertion into the conversion process. Dust and particulate emitted during the grinding operation are captured by the dust collection system. This system also helps meet regulatory emission standards. The last equipment, referred to as a surge bin, collects the ground biomass from the grinder and meters out biomass in an even flow to the conversion process (Hess et al., 2009). Note that all the above-mentioned equipment can be used to handle multiple types of biomass as long as the same logistics scenario, i.e. square bale, is considered. The characteristics and the cost data for each piece of equipment are provided in the following section. The details of the operations will be discussed in sections 4.3.  3.9 Equipment data In this study, the equipment data were gathered on the latest commercially available biomass handling and processing equipment. The data sources included reports from Oak Ridge National Lab (Sokhansanj et al. (2008), Sokhansanj and Turhollow (2002)) and Idaho National Lab (Hess et al., 2009). Specifications given for a piece of equipment are generic. No specific model is considered for a piece of equipment. The specifications of the equipment prices are given in Table 3-3 and Table 3-4. The selection of trailer is based on the legal dimensions and weight in the province of Saskatchewan. The legal dimensions for trucks and semi-tractor trailers is 2.6-m in width, 4.15m in height and 23-m in length (Saskatchewan Ministry of Highways and Infrastructure, 2011). In addition, the legal weight of truck with full trailer on different roads in the province of Saskatchewan ranges between 36,400 and 53,500 kg depending on number of trailer axles and road type (Saskatchewan Ministry of Highways and Infrastructure, 2011). Since large quantities of straw must be processed every day, especially during the short harvest season, machine breakdown is likely to occur. The distribution function that is  49  considered for maintenance in this study is the same as that in Nilsson (1999b). For field operations such as baling, the breakdown is assumed to occur at intervals described by an exponential distribution (exp(1.0 h)). It implies the machine works, on average, one hour between stops due to repair. The duration of each failure is described by exp(0.1 h) (Nilsson, 1999b). It implies that it takes 6 minutes to get a machine back into operation. For other operations that take place outside of the field such as loading, the breakdown interval times are described by exp(10 h) in the working state. The corresponding failure duration is described by exp(0.25 h) (Nilsson, 1999b).  50  Table 3-3: Specifications of equipment pieces (Hess et al., 2009, Sokhansanj et al. (2008), Sokhansanj and Turhollow (2002)) Equipment Baler  Power (kW)  Efficiency (%)a Rated capacity  Bulk density (kgm-3)b  Moisture Content (%w.b.)c  132  [70,80,90]  38 (bale/h)  128  20  Truck scale  -  [65,75,85]  15 (truck/h)  -  -  Grinder in-feed  8  100  14.6 (t/h)  -  -  Grinder  463  [75,85,95]  16.5 t/h  200  10  Dust collection  74  [80,90,100]  68 t/h  -  -  4  [80,90,100]  22.7 t/h  -  -  system  system Surge Bin a  Efficiency accounts for conditions that cause a machine to operate at less than its theoretical capacity such as refueling and unproductive travel. Efficiency is  described by a triangular distribution as it varies based on the work conditions (Hess et al., 2009) b  This column shows the bulk density of biomass after the respective operation.  c  Moisture content for baler shows the safe moisture content in order to commence the baling operation. For ginning operation, it shows the base moisture  content.  51  Table 3-4: Specifications of equipment pieces (Hess et al., 2009, Sokhansanj et al. (2008), Sokhansanj and Turhollow (2002)) Equipment  Power  Speed  Efficiency  Bales  Loading time  Loading  Unloading time  Unloading  (kW)  (km h-1)  (%)  per load  (min/load)  efficiency (%)  (min/load)  Efficiency (%)  Tractor to pull the baler  165  9.7  -  -  -  -  -  -  Self-propelled stinger  177  8-25a  [70,80,90]  8  2b  [65,75,85]  1  [75,85,95]  Telehandler  73  -  [70,80,90]  2  1  -  0.5  -  Tractor to pull trailer  330  80  -  -  -  -  -  -  -  -  [75,85,95]  30  3c  [65,75,85]  1.5c  [75,85,95]  12-m flatbed trailer a  In order to transport bales from fields to storage sites, stinger usually travels on both field (bad-conditioned road) and public roads (well-conditioned road). The speed of stinger on field roads and on public roads are assumed to be 8 kmh-1 and 25 kmh-1, respectively (Brummer et al., 2000). b The loading time is twofold: maneuvering the stinger into position and loading the bales. c Loading and unloading time for the trailer refers to the tie/untie, securing and inspection time of bales on the trailer. It is assumed all tractors to be four-wheel drive units, equipped with a cab.  52  There are two major agricultural equipment engineering-economic costing methodologies (Hess et al., 2009). These methodologies were developed by the American Society of Agricultural and Biological Engineers (ASABE) and the American Agricultural Economics Association (AAEA). Turhollow and Sokhansanj (2007) extended these methodologies and developed a standard costing methodology for biomass. The developed methodology by Turhollow and Sokhansanj (2007) was used in this study. Table 3-5 shows the cost data for all the machines and equipment for both custom and ownership cases. Custom rate represents the case in which the equipment is rented. For the ownership case, both variable and fixed costs are reported.  Table 3-5: Cost data of the machines and equipment (Turhollow and Sokhansanj, 2007) Equipment  Purchase cost  Custom rate  Variable cost  Fixed cost  ($)  ($ h-1)  ($ h-1)  ($ yr-1)  Baler  134,490  89.92  78.17  7,051.75  Tractor to pull baler  129,858  93.67  85.65  8,019.16  Self-propelled stinger  159,470  77.67  73.74  9,810.34  Self-propelled telehandler  79,635  49.2  47.25  3,895.5  Tractor to pull trailer  127,125  103.94  97.90  6,036.35  12-m flatbed trailer  60,180  24.66  23.39  3,169.86  Truck scale  74,456  6.13  5.5  3,308.52  Grinder in-feed system  63,669  4.18  3.58  2,829.18  Grinder  502,870  119.35  95.18  24,167.18  Dust collection system  64,938  72.45  69.18  3,269.68  Surge Bin  77,305  14.13  10.43  3,602.84  The purchase cost of the equipment, repair, maintenance, labor, fuel use, taxes and insurance are embedded in both custom and ownership costs. The rates and prices were projected for 2011 based on the consumer price indices of years 1997-2010. The indices were provided by Oak Ridge National Lab, US Department of Energy. The interest rate was assumed to be 6%. The charge for taxes, housing, and insurance were calculated as 2% of the purchase price.  53  It is noted for equipment and cost data, assumptions on the number of annual working hours, the service life of the equipment, field and road speeds, and field efficiencies were taken from ASAE EP 496.3 (ASABE, 2007a) and ASAE D497.5 (ASABE, 2007b). In this research, the supply chain is evaluated from a feedstock manager’s point of view who is responsible for coordinating the operations at the upstream of the plant and the at-plant operations. The upstream operations consist of activities from farms to the gate of the ethanol plant including harvesting and collecting, storing, handling and transporting. The at-plant operations are comprised of activities occurring at the ethanol plant including receiving and preprocessing operations. Given the supply radius and the location and capacity of the ethanol plant in the case study, the feedstock manager requires planning and scheduling the logistics operations in a way that an uninterrupted daily delivery of biomass throughout the year is assured. To this end, a new simulation model was developed and implemented into the case study. The details of the developed simulation are provided in the next chapter.  54  Chapter 4. Development of a new simulation model 4.1 Synopsis This chapter elaborates on the development of a new simulation model to plan and schedule the biomass supply chain for the proposed cellulosic ethanol plant. First, the framework of the developed simulation model is presented. Then, the modules of the simulation model and their connections are explained. The next section concerns the verification of the model. Thereafter, the model is applied to the case study and the results are provided. The last section of the chapter is dedicated to the validation, discussion and conclusions based on the obtained results of the developed simulation model.  4.2 Framework of the developed simulation model Figure 4-1 illustrates the schematic of the simulated biomass supply chain. As shown in the figure, biomass is not directly delivered to the plant from farms as trucks cannot access to the farmlands because they will cause soil damage. The collected biomass on the farms is first transported and stored at either roadside storage or satellite storage and then transported to the plant from these storage sites. No flow of biomass exists between the storage sites. In the roadside storage system, as its name implies, the collected biomass at each farm is transported and stored at the roadside of the respective farm. In contrast, satellite storage is not necessarily located at the roadside of the farms and it could be placed at any point between farms and the ethanol plant. Satellite storage usually contains biomass from more than one farm. Depending on their size, their distribution in the supply region and their proximity to each other, farmers store their collected biomass either at the roadside of their own farms or at satellite storage. In addition to the storage system, the simulated supply chain has specific features as follows:   Multi-biomass: due to the availability of multiple crops in the considered supply region, the supply chain should be able to handle multi-biomass. The different characteristics of the crops such as yield, harvest schedule and initial moisture content affect the planning and scheduling of the logistics operations.  55  Farm 1 (Crop 1,..,m)  Baling  In-field hauling  Farm 2 (Crop 1,..,m)  Baling  In-field hauling  Farm 3 (Crop 1,...,m)  Baling  In-field hauling  Farm 4 (Crop 1,...,m)  Baling  In-field hauling  Farm 5 (Crop 1,...,m)  Baling  In-field hauling  Farm 6 (Crop 1,...,m)  Baling  In-field hauling  Satellite storage  Loading  Roadside storage  Loading  Road transportation  Satellite storage  Unloading  At-plant storage  Grinding  Conversion process  Loading  Flow of Biomass Flow of Information  Farm n (Crop 1,...,m)  Baling  In-field hauling  Push System  Roadside storage  Loading  Pull System  Figure 4-1: Schematic of the simulated agricultural supply chain  56    Push/pull supply chain: As shown in Figure 4-1, the agricultural biomass supply chain is a push/pull system. The push part is comprised of all field operations. The agricultural practices dictate the timeliness of the field operations. Biomass on the field is available only during the harvest season and it must be collected and moved to the outside of the field as the farmers tend to prepare the land for the next cropping season after the termination of the harvest season. The rest of the supply system acts as a pull system; meaning that only when the downstream operation needs biomass, required biomass is pulled from the preceding operation. Therefore, the push part of the supply chain processes as much biomass as possible during the harvest season to assure sufficient biomass would be available to meet the annual biomass demand. The pull part only processes the amount of biomass required to fulfill the daily demand throughout the year.    Storage management: Storage sites including roadside storage and satellite storage provide a buffer between the push and pull parts of the supply system, and thus, they play a key role in creating a steady flow of biomass in the supply chain. The size of each storage site must be determined in a way that ensures sufficient biomass is stored to meet the daily demand yearround. In addition, at-plant storage should have enough capacity to feed the conversion facility on occasions such as off-shift hours of the transportation system and interruptions due to machine breakdown and weather conditions. Initial inventory of storage sites is another important factor to meet the demand during the harvest season due to the lack of availability of sufficient biomass at the start of the harvest season.    Information management: To coordinate push and pull parts of the supply chain, in addition to the flow of material, flow of information must be managed. This information includes daily delivery of biomass, delays in the system, machine availability and storage status. For example, once the daily demand of the conversion facility is met in a specific day and the capacity of at-plant storage is full, the transportation system must be informed not to transport more biomass as the plant does not require any more feedstock. This coordination would also avoid bottlenecks in the system.    Farm management: Due to the different size and location of farms and their produced biomass, they should be considered as separate entities in the supply chain instead of aggregating them at a county or municipality level. Analyzing the supply chain at the farm level provides accurate cost and delivery outputs and allows the feedstock manager to help 57  farmers in using their equipment more efficiently. For example, if the utilization rate of field machines is low for adjacent farms due to the small amounts of biomass produced on these farms, one piece of equipment can be shared between these farms. This would result in a higher utilization rate and lower operating costs. In addition, farm-level outputs help contract and pay farmers based on the amount of transported biomass from their farms to the plant.   Delays: Three types of delays take place in the supply chain including delay due to dry biomass, delay due to unfavorable weather conditions, and delay due to unavailability of machines (breakdown/busy). These delays affect the delivery of biomass to the plant. In addition to the above features, the constraints of the supply chain in the Canadian Prairies,  explained in Chapter 1, must be considered. The goal of the simulation model is to find solutions for the following logistics decisions:   Amounts of biomass to collect during the harvest season for the year-round delivery    Number of hectares of farmland to contract    Number of machines in each logistics operation    Capacities of roadside storage, satellite storage and at-plant storage sites    Initial inventory of storage sites at the beginning of harvest season    Number of daily delivered truckloads A new simulation model is developed to find solutions for these decisions. The structure of  the simulation model is illustrated in Figure 4-2. The model is comprised of two components: 1) the spreadsheet which is developed in Microsoft Excel and 2) the simulation model which is built using ExtendSim software. All the input and output data are entered and saved in the spreadsheet. The input data, as explained in Chapter 3, encompass the ethanol plant data, supply area data, farm data, crop data, weather data, harvest schedule data, equipment data and cost data. Macros written in Visual Basic for Application (VBA) send all the input data from Excel to ExtendSim. These data are placed in pre-established databases in ExtendSim. During the simulation run in ExtendSim, the outputs of the simulation model are recorded in a database. Upon the termination of the simulation run, the outputs are sent to the designated worksheets in the spreadsheet.  58  Cost data  Equipment data  Harvest schedule  Weather data Spreadsheet (Excel Worksheets)  Crop data  Farm data  Ethanol plant data  Supply area data  Visual Basic for Application (VBA)  The spreadsheet  Extendsim Input Database  The simulation model  Weighing module  Item creation module Baling module  Unloading module In-field hauling module Storage module  Information Management module (IM)  At-plant storage module  Loading module  Energy input and emissions  Moisture content  Daily delivery scheduling Extendsim Output Databse  Storage outputs  Supply area outputs  The simulation model The spreadsheet Spreadsheet (Excel Worksheets)  Economics Outputs  Delay time  Grinding module  Road transportation module  Figure 4-2: Structure of the simulation model 59  It is noted that the developed simulation model and the IBSAL model simulate the agricultural biomass supply chain with the primary focus on the crop conditions in North America. In addition, they are built using ExtendSim software. However, IBSAL has the following differences from the developed simulation model in this research, called IBSALMulti-Crop (IBSAL-MC):   IBSAL deals with only one biomass type in each simulation study.    In IBSAL, given the location of the bioenergy plant and the biomass producers, the annual amount of delivered biomass to the plant is estimated. However, IBSAL is not able to plan and schedule the supply chain based on the demand of the bioenergy plant. In addition, IBSAL does not manage the flow of information in the supply chain. For instance, if a storage site is full or a machine is busy at the downstream of the supply system, the machines at the upstream of the system continue processing biomass and push it toward downstream operations regardless of their status. Thus, each upstream logistics activity operates with no information from the following operations. In other words, IBSAL is a push-oriented simulation model.    In IBSAL, it is assumed that roadside and at-plant storage sites have unlimited capacity. Moreover, IBSAL only models roadside and at-plant storage and does not consider any intermediate storage such as satellite storage in the supply system.    IBSAL does not simulate farms individually in the supply system. The entire supply area is regarded as a single farmland. Then, this farmland is broken down into smaller lands (simulation item) based on the harvest schedule.    IBSAL does not explicitly take into account the breakdown and repair times for the machines utilized in the supply chain. The breakdown and repair processes are included in the machine operational efficiency not as an activity in the supply chain to shut down the operations.    The estimation of the moisture content in IBSAL is based on a discrete fashion in that the moisture content for a specific day is estimated based on the last recorded moisture content of biomass and the weather data for the respective day. The time period between the time when the last recorded moisture content was calculated and the respective day are not taken into account.  60    IBSAL considers the delay to dry and delay due to weather conditions but does not consider the delays due to the unavailability of machines.  4.3 Modules in the developed simulation model As shown in Figure 4-2, each operation is represented by a module. In addition to the developed modules for logistics operations, a module called Information Management (IM) module was developed. This module is the heart of the simulation model as it is responsible for integrating and coordinating all the modules in the biomass supply chain. The real-time information is sent from each module to IM. Upon the receipt of the information, the IM module analyzes the information and reacts. The reaction process usually involves sending information to other modules. The main purposes of the integration and coordination of the supply chain were to 1) fulfill the daily feedstock demand; 2) provide a smooth flow of biomass in the supply chain; 3) avoid bottlenecks in the supply chain; 4) utilize the equipment efficiently to reduce operating costs; and 5) quickly react to any disturbance and shock in the supply chain to keep the system robust. Three sources of delay in the supply chain are considered in the IM module: 1) moisture content (Delay-To-Dry); 2) weather condition (Weather-Delay); and 3) machine unavailability (Machine-Delay). The purpose of Delay-To-Dry is to prepare a stable feedstock. Stability prevents the selfheating of biomass during the densification process such as baling, and also reduces the amount of dry matter loss due to the biological activities while storing (Rentizelas et al., 2009b). In most agricultural biomass logistics scenarios, the cut biomass is left on the field to dry down to a safe moisture level at which it can be collected and densified into different sized packages. Densified biomass may remain in the field for several days prior to its removal from the field. During this time, overall dry matter of biomass would decrease as biomass is exposed to weather. The safe level of moisture content is 15−20% or less for all agricultural resources (Hess et al., 2009). In this study, it is assumed the safe moisture content for baling and storing is 20% (Sokhansanj et al., 2008a). It is noted that moisture content constantly changes depending upon climate conditions. Moisture content of biomass is updated continuously in the simulation model based on the daily weather data. 61  In addition to moisture content, Delay-To-Dry is estimated based on two other parameters including minimum days on field (MinDF) and maximum days on field (MaxDF). MinDF is the minimum number of days the residue is left on the field. After MinDF, the residue may be still kept on the field or baled depending on its moisture content. MaxDF is the maximum number of days the residue can be left on the field. After MaxDF, the residues are baled regardless of their moisture content. This parameter represents the situation in which the moisture content is higher than safe moisture content but residue has been laid on the field for several days and it needs to be removed from the field as the harvest season is short. For instance, Hess et al. (2009) considered seven days the maximum for field drying. MinDF and MaxDF parameters are defined by the user and can be neglected from the Delay-To-Dry logic. In this study, both MinDF and MaxDF are neglected due to low moisture content of wheat straw. Thus, only safe moisture content was considered as the source of Delay-To-Dry. Weather conditions are another source of delay. Weather-related parameters such as rain, snow, temperature, excessive soil moisture, and other conditions caused by weathering affect the performance of field machinery. Weather conditions may delay harvest and collection operations (Sokhansanj et al., 2008b). In this study, the weather delay is estimated based on the weather data including temperature and the precipitation rate. It is assumed the field operations would stop at temperatures below -20oC until weather gets warmer. Moreover, the operations are delayed by one or two hours for each millimeter (mm) of rain or snow precipitation (Sokhansanj et al., 2008b). The IM module calculates the amount of delay caused by the weather conditions and holds the simulation items until the weather conditions favor the work conditions on the field. Consideration of weather delay is important to assure the suitability of soil condition for machinery traffic. The last source of delay is machine unavailability. Before releasing a simulation item to the next module, IM module checks whether a machine is available to process the dry biomass. The unavailability of the machine can be due to two factors: 1) The machine is busy processing another item; or 2) The machine is broken down and needs to be repaired. Once the machine is idle or fixed, the item is released. Figure 4-3 depicts the delays and their connection in a flowchart developed in the IM module.  62  No  Is the minimum days to dry met?  Delay the item for one more day  Day=Day+1  Yes  Is weather favourable to proceed the operation?  No Delay the item until the weather permitted  Has the day advanced?  Day=Day+1  No  Yes  Is the maximum days to dry met?  No  Yes  Is the machine available?  Yes  Is moisture content of biomass less than the safe level?  No  Delay the item until its moisture reaches the safe level or the day advances  Delay the item until the machine is available  Yes  No  Yes  No  Has the day advanced?  Has the day advanced?  Yes  No Yes Start the operation  Figure 4-3: Delay logic flowchart in the simulation model  It is noted that the delay logic shown in Figure 4-3 is applied to the field operations. For other operations such as transportation and unloading, only the last condition regarding the availability of machinery is checked as moisture content is not a constraint for these operations. Weather conditions can also impact the loading operations taking place at storage sites- depending upon the type of storage regime. If the loading equipment travels on the ground, the performance of the loader will be affected by the weather conditions. On the other hand, covering the floor of the storage location with material such as crushed gravel pad eases equipment use and vehicle traffic. Thus, equipment can work even during high precipitation. The sequences of other modules in the simulation model are as follows:  63  1. Item creation module In this module, all the farms in the supply area are created at the beginning of the simulation run and their characteristics are assigned to them as attributes. The assigned attributes to each farm include farm size, produced biomass type and its biomass yield, distance from the plant, and the initial moisture content of biomass (harvest moisture content). Hess et al. (2009) reported the harvest moisture content of wheat straw could be in the range of 9-25% (w.b.). This range has been used in the simulation model to estimate the initial moisture content. A discrete uniform distribution function was used to generate the initial moisture content of straw at the beginning of the harvest season. The created farms are first held in a queue. Once the simulation time advances, farms are broken down into small piece of farmlands based on the harvest schedule. Each farmland has a specific area and biomass tonnage. 2. Baling module Once the moisture content of biomass is at a safe level for baling, the weather conditions are favourable for field operations and the baler is available, the simulation item (portion of a farm) is released to the baling module. In this module, loose biomass laid on the ground is compacted into 1.2 m×1.2 m×2.4 m (4 ft×4 ft×8 ft) large square bales. Then, the baler ties the bale with twine strings to maintain the integrity of the bale and drops it on the field. The cost of twine per bale is assumed to be $0.8 (Sokhansanj et al., 2008b). The density of the created bales is assumed to be 128 kgm-3 wet bulk (8 lbft-3) (Hess et al., 2009). This module also models a tractor utilized to pull the baler. The square baler is drawn behind a 165-kW tractor. It is noted the simulation item changes from farmlands to bales in this module. In this study, only the large square bale is modeled. The large square bale has distinct handling, transportation and storage footprint benefits over the round bale. Large square bales can be loaded two at a time and quickly stacked on trailers. They can also be picked up by stinger stacker. In contrast, round bales are handled once at a time. Hess et al. (2009) reported that a 16.5 m (53-ft) semi-tractor trailer with square bales can be loaded in less than 30 minutes (80 bales/hr) while loading the same trailer with round bales takes nearly one hour (40 bales/hr). Thus, it is not cost-efficient to ship round bales for long distances due to their handling difficulties. In addition, the square bale has been used for the commercial hay harvest, whereas  64  the round bale has been widely used by cattle farmers. Since the developed simulation model is applied to a commercial-scale ethanol plant, the large square bale was modeled here. 3. In-field hauling module Immediately or at a later time (depending on moisture content, and schedule and readiness of bale collection machine), bales are hauled to the storage sites by a self-propelled stinger stacker. A self-propelled stinger stacker enables the collection, transportation and stacking of bales at the storage sites. The stinger picks up 8 bales in the field, hauls them to storage and stacks them 4 bales high. Thus, the total operating time is the sum of loading, transporting and unloading (stacking) bales at storage. The loading and unloading of 8 bales takes 2 and 1 minutes, respectively. The transportation time depends on the location of storage and also the distribution of bales on the farm. It is assumed that bales are randomly distributed on the farm. To estimate the transportation time, a similar approach used by Cundiff et al. (2009b), was employed in this module. The module assigns a coordinate to each bale in the farm. Given the number of bales in each farmland from the baling module and the size of farmland, a coordinate is assigned to each single bale. The origin was set as the center of the storage site. For the first load, it is assumed the first bale is located at 0.16 km away from storage (Sokhansanj et al., 2008a). The module finds the second bale which has the minimum distance from the first bale, and the third one which has the minimum distance from the second one and so on. This process will continue until 8 bales are loaded on the stinger. Given the coordinates of the bales, Euclidean distance is used to estimate the distance of two adjacent bales with a winding factor of 1.2 (Kumar and Sokhansanj, 2007). Although the procedure to load bales on the stinger is not an optimal procedure, it reasonably reflects the impact of the distribution of bales on the farm on the transportation time spent by the stinger. The same process is applied for the subsequent loads until all of the bales are transported to storage or the harvest season ends. The field speed and highway speed of the stinger are assumed to be 8 kmh-1 (5 mph) and 25 kmh-1 (15 mph) (Brummer et al., 2000).  4. Storage module This module models the inflow and the outflow of biomass in both roadside and satellite storage sites. Stacked bales are held in storage if one of the following conditions occurs: 1) weather  65  conditions are unfavourable for loading operations; 2) there is no truck available at storage; 3) the loading equipment is broken down and needs to be repaired; or 4) the daily feedstock demand of the conversion facility is met and the capacity of at-plant storage is full. These conditions are continuously checked by the IM module. In addition to the type of storage system (roadside/satellite), the type of the storage regime and the stack configuration must be identified in this module. Table 4-1 lists different types of storage regimes, their associated construction cost and the average annual dry matter loss percentage. The dry matter loss percentage is based on the assumptions that the moisture content of stored biomass is 15% w.b. and biomass is stored for one year (Brummer et al., 2000). This moisture content is close to the moisture content of wheat straw. Table 4-1: Construction cost and average dry matter loss of different storage regime (Brummer et al., 2000) Storage regime  Construction cost  Average dry  ($m-2)  matter loss (%)  70.39-107.64  2  Open sides, pole frame building on crushed rock  53.82  4  Reusable tarp on crushed rock  1.471  7  Outside unprotected on crushed rock  2.70  15  Outside and unprotected on ground  0.00  25  Enclosed, pole frame building on crushed rock  1  The cost per m2 is only the cost for the tarp.  The large stack of bales increases the risk of fire, which would cause a biomass supply shock in the system. Fire could also spread to adjacent facilities and farms and cause a large loss of assets (Hess et al., 2009). To prevent fires, the stack configuration of storage is important. In this study, the fire codes recommended by the International Code Council (Hess et al., 2009) are taken into account. According to these fire codes, the size of each stack in storage should not exceed 91 t (100 tons). Similar to Hess et al. (2009), the large square bales are stacked 4 high and 5 wide. The length of the stack depends upon the density of bales. For example, bales are stacked 10 long if each bale weighs 0.5 tons. In addition, fire lanes are considered between stacks. The fire lanes between two adjacent stacks in a row and in a column are 3 m (10 ft) and 30.5 m (100 ft), respectively in open yards 66  and 6.1 (20 ft) and 30.5 m (100 ft), respectively in closed storage. For closed storage, a fire lane of 23 m (75 ft) is considered between the wall and the nearby stacks (Hess et al., 2009). Figure 4-4 shows a simple schematic of the stack configuration in a closed storage site.  Column Row  Bale stack (4 high, 5 wide, 10 long)  30.5 m  6.1 m  23 m Storage wall  Figure 4-4: Stack configuration in a closed storage site according to the fire codes  5. Loading module In this module, the stored bales in storage are loaded onto a semi-tractor trailer attached to a tractor truck. This module models both the loader and the trailer. The loading process is carried out by a telehandler. At each trip, a telehandler picks up two large bales and places them on a 12m (40-ft) flatbed trailer. This process continues until the trailer is fully loaded. The trailer is 3 m (102 in.) wide. The size and weight of the trailer comply with the legal dimensions and weight regulations in the province of Saskatchewan, as explained in section 3.9. The configuration of the trailer allows for a truckload of 30 bales with 5 rows of three-high (3.6 m) and two-wide bale stacks (2.4 m). On average, the loading operation takes 20 minutes for each empty trailer. The reason for modeling the trailer along with the telehandler in this module is to keep track of the  67  time that the trailer waits in order to be loaded. This time is considered in the associated costs and resource utilization rate. 6. Road transportation module This module models a truck with its attached trailer between the storage sites and the gate of the plant. A 330-kW tractor is used to carry the trailer. The truck with a semi-tractor trailer is modeled here since this transportation mode can travel to rural areas and has been widely used for moving square bale hay. 7. Weighing module In order to determine the net weight of loaded bales on a truck, each arriving truck to the plant is weighed twice by a truck scale, once loaded with bales and once after unloading bales (tare weight). This module models a scale unit above the ground with the capacity of 91 t (100 tons). 8. Unloading module A similar telehandler used in the loading module is also utilized here to unload the bales from the trailer and stack them at the at-plant storage site. On average, the unloading of each trailer takes 15 min. Once the telehandler unloads all the bales, the empty trailer must be dispatched to another roadside/satellite storage location to pick up a new load of bales. The dispatching method in the IBSAL-MC model is based on the shortest travel time first and the longest travel time next. Based on this method, the first empty truck is dispatched to the closest storage site to the plant and the next empty truck is sent to the farthest storage site from the plant. The IM module continuously updates the list of closest and farthest storage sites. Once the inventory of a storage site exhausts, it is removed from the list of storage sites to assign empty trucks to. This dispatching method creates a fair balance in the transportation times of trucks and reduces the maximum number of trucks and also the variation on the number of required trucks in the supply chain (Ravula, 2007). 9. At-plant storage module This module models storage located at the ethanol plant. This storage acts as a buffer to avoid the starvation of the conversion process in case of unforeseen supply shocks or during the off-shift hours of the transportation system. This storage should be a well-structured area that not only  68  makes it accessible 24/7 but also protects the stored bales from weather as the stored bales at this stage of the supply chain are value-added feedstock. The bales are released from at-plant storage and fed into the grinder by a telehandler. The release of bales is managed by the IM module.  10. Grinding module The last module of the simulation model is grinding. This module is comprised of four submodules: grinder in-feed system, grinder, dust collection system and surge bin. The grinder infeed system removes the bale strings and conveys the bale into the grinder. The grinder is a hammer mill with a 463-kW electric motor and a base throughput of 16.5 t per hour. The grinder grinds biomass into an average particle size of 2.5 mm (0.1 in.). Depending on the daily feedstock demand, several grinders may operate in parallel. The dust collection system filters and captures the dust created during the grinding process. Finally, the surge bin feeds the small particles of biomass into the conversion process. All the developed modules contain a block estimating the dry matter loss (DML). The estimated DML for each operation is the aggregate of both physical losses during an operation and chemical losses. Chemical DMLs take place at any stage of the supply chain in which the biomass waits to be processed, such as loose biomass left on the field to get dry or stored bales at storage. The exact amount of chemical DML as a function of time is not available. The development of such a function would require extensive field experiments. To estimate the percentage of the chemical DMLs, given the average annual DML percentage, the daily DML percentage was first calculated by dividing the annual percentage by 365. The final amount of DML was then calculated by multiplying the number of days the biomass remained in storage/queue by the daily DML percentage. The annual DML percentage for different types of storage regime is given in Table 4-1. It is assumed that if loose biomass is left on the field for one year, the whole biomass will be lost (100%). During a short time window, bales may be left on the field until they are transported to storage sites. As these bales rest on the ground, ground moisture which is pulled into the base of the bale by capillary action, can easily double the amount of spoiled material normally expected in outer layer losses (OMAF, 1998). Through a yearlong field experiment, Blunk et al. (2003) showed that the amount of DML for fully uncovered bales reaches 40–60%. In this study, it is 69  assumed that the maximum annual dry matter loss for a single bale left on the field is 50%. It is also assumed that although bales may lose dry matter, the integrity of the bale is maintained throughout the supply chain and there are no broken bales in the supply system.  4.4 Assumptions Although the supply radius and the location and capacity of the ethanol plant were predetermined by Iogen, the details of the supply chain were not available. Thus, to plan and schedule the supply chain by the developed simulation model, the following assumptions were made:   Each farmer assigns a portion of his farmland for a roadside storage site to store the collected bales. This roadside storage is not shared with other farmers. Thus, no satellite storage is considered.    Each farmer is responsible for the collection of available biomass on the field, and its transportation and stacking at roadside storage    Farmers store the collected biomass outside and on unprotected ground.    Each farmer either owns the required machinery to perform field operations or the adjacent farmers perform the harvesting and collection operations for the respective farmer.    One loader is dedicated to each roadside storage location to load biomass from storage onto the trailers by the hauling contractors. The loaders are not shared between the roadside storage sites. In addition to loading the bales, the loader can be used to do other chores on the farm such as moving rocks and removing bushes.    All trucks are initially located at the ethanol plant.    Trucks are dispatched to the roadside storage sites where there is sufficient biomass to fully load the trailer at each trip. Thus, each truck only travels between a specific storage site and the ethanol plant at each trip.    The price of laid biomass on the field is assumed to be fixed. Iogen offered a variable price in the range of $7-$15 per tonne of straw laid on the field (Altman et al., 2007). The Saskatchewan Soil Conservation Association estimated the value of straw would be in the range of $8-$21 per tonne of straw with the average value of around $15 (Saskatchewan soil conservation association, 2000). In this research, it is assumed the biomass price is $15t-1.  70    All the upstream operations including the field operations, loading and transportation operate 14 hours per day (6 a.m. to 8 p.m.). The field operations are scheduled for the whole week while the transportation system operates 6 days/week. The operations at the ethanol plant are scheduled 24/7. It is also assumed the conversion facility operates 350 days of the year and the remaining days of the year are assigned for the annual planned maintenance. Thus, the annual required biomass is 262,500 t (750×350).    At-plant storage is an open storage yard with a crushed-rock surface to ease equipment use and vehicle traffic    The farmer participation rate is 100%. The farmer participation rate is the percentage of available farmers inside the supply area who are willing to make contract with the ethanol plant and sell their biomass to the bioenergy plant. This rate ascertains the distribution of farmers inside this supply area.  4.5 Verification of the simulation model The verification of the IBSAL-MC model was performed to assure: 1) every single module operated as expected; and 2) all the modules were connected correctly and the whole model performed as expected. Upon the development of each single module, it was run separately and the outcomes of the respective module were closely observed to check for any possible errors. For example, once biomass moves through the baling operation, a portion of biomass is not picked up by the baler and remains on the field as dry matter loss (DML). The amount of biomass that arrives in the baling module must equal the sum of DML and the recovered biomass (baled biomass). This equation was checked for different input values. If the respective module had variable inputs, its variability was first neglected and the most likely value of the input parameter was first plugged into the module. Then, the module was run several times to make sure the same results were obtained. In addition, the input data were changed several times to determine how the outputs would change. For example, the number of balers was changed from 1 to 5 to observe if the baling module would react to the change properly. Finally, the extreme values were plugged into the module to examine how the module would react to the extreme changes. For example, the number of available trucks in the supply 71  chain was set to zero and 1000. Another example was a significant increase/decrease in the temperature and the amount of rainfall. Once the deterministic version of the module was debugged, similar changes in the input data, as explained above, were made in the module with variable inputs. Different sets of results were produced and checked to determine if they met expectations. Once every single module was checked and debugged, it was coupled with its preceding and following blocks and the same debugging process as in the single module was followed. This process continued until the whole IBSAL-MC simulation model was tested. To verify the whole model, only one item was first released into the model. This helped to check the arrival and the departure times of the item and also its movement between the modules. Then, the number of items was increased gradually and the waiting time of items in each queue was closely monitored. Releasing different numbers of items, from small to large, helped to examine whether the model responded properly. In addition, the number of items created in the model and the number of items left by the end of the simulation run were checked to see if any items remained in the simulation model and to evaluate whether it was expected or not. For more complex parts of the model such as the signal exchanges between the modules and equations, the debugging feature of the ExtendSim software was exploited. The debugging feature enables the user to run the model event by event. Once a new event occurs, the model pauses and allows the user to check the new values of an equation or changes in the attributes of the items. Another feature of the ExtendSim software is animation. Running the simulation model with animation helped to monitor the flow of items in the system. It helped to check for unusual behavior of the system while items were moving throughout the system.  4.6 Outputs of the simulation model for the case study The simulation model has a one-year time frame. To cover the planning horizon of 25 years, the service life of the ethanol plant, the model was run 25 times in each replication and the results were saved in the developed spreadsheet. The time step in the simulation model was set to one hour. To determine the number of simulation runs, the total supply cost was chosen as the  72  variable of interest. Then, the model was initially run 100 times. Taking the standard deviation of cost outputs (S), setting the error (ε=0.1) and confidence interval of 90% (α=0.1), the number of runs (R) was calculated by the following equation: (  ⁄  )  )  In total, the simulation model was run 825 times. The discussed results in this section are the annual averages. Since the case study is a proposed plant, there is no initial biomass inventory in the supply system for the first year and it takes some time for the conversion facility to ramp up to its targeted capacity. Thus, the results of the first year were excluded from the analysis and it was used as a warm-up period in IBSALMC. It is noted that the start of the simulation model is August, the commencement of the harvest season period. As explained in Chapter 3, the variable input data in the simulation model include location, size and yield of farms, distribution of bales on the farms, soil conservation rate, initial moisture content, machine breakdown and repair time, the equipment efficiency, weather data and harvest schedule. It is noted that the harvest schedule data were only available for 6 years while the weather data were collected for 21 years. Thus, the weather data for years 2005-2010 were used in the simulation model. In each simulation run, the weather data and the harvest schedule for a specific year were selected using a discrete uniform distribution from years 2005-2010. Due to the large number of farms within 160-km supply radius (3,055), the adjacent small farms (size<50 ha) with shared borders were aggregated into one larger farm with the total area of merged farms. The aggregation of small farms resulted in a set of 2,485 production fields. The aggregation of small farms increased the efficiency of modeling by decreasing the number of farms as the simulation items. It also helped capture the fact that owners of small farmlands are expected to rent hauling equipment from adjacent farms rather than haul the biomass themselves (Judd et al., 2012). On average, the computational time for each simulation run was 20 minutes. The number of created farms, pieces of farmland and bales were 2,485, 12,053 and 663,671, respectively. In each simulation run, the farms, as initial simulation items, were broken down into several farmlands based on the harvest schedule. Then, the available biomass in each farmland was converted to bales.  73  4.6.1 Flow of biomass in the supply chain As mentioned in section 3.3, the total annually produced biomass in a 160-km supply radius and 100% farmer participation rate would be around 1.5 Mt. The historical harvest schedule data in Prince Albert shows that the total produced biomass is not necessarily collected and a portion of produced biomass may be left on the fields due to the long harvest season and its overlap with the cropping season (Saskatchewan Ministry of Agriculture, 2011). In addition, a portion of biomass is lost while laid on the field to get dry due to microbial degradation and weather-related effects. Simulation results showed that, on average, 5% of the total produced biomass is not collected during the harvest season. Thus, the potential available biomass to be collected is around 1,441,000 t. The biomass left/lost on the field is desirable for soil health and conservation purposes. Table 4-2 provides the detailed flow of biomass in the supply chain.  Table 4-2: Annual recovered biomass and dry matter loss (DML) in the supply chain Operation  Recovered biomass (t)  DML (t)  Baling  326,098  32,610  In-field hauling  323,363  2,735  Roadside storing  302,720  20,643  Loading onto trucks  270,977  2,493  Road transportation  268,570  2,407  Truck weighing  268,570  0  Unloading trucks  265,760  2,810  At-plant storing  715,14  528  Grinding  262,500  2,400  As shown in Table 4-2, 326,098 t of biomass should be baled from farms to meet the annual biomass demand. These amounts of biomass are collected from 348,875 ha of farmlands in 648 farms (638 spring wheat farms and 10 durum wheat farms). Table 4-3 shows the distribution of farms required to contract within a 160-km radius. No farm is required to contract in the distance range of 115-160 km. Thus, the annual demand of the ethanol plant can be met within a 115-km radius with a farmer participation rate of 100%. There are no winter wheat farms within this 74  radius. The size of selected farms ranged between 40 ha and 1,610 ha. The average size of these farms was 560 ha.  Table 4-3: Distribution of contracted farms within 160-km radius 15-30 Number of farms  87  30-45 111  Distance range (km) 45-60 60-85 85-100 114  285  26  100-115 25  Total DML is 66,626 t. Figure 4-5 shows the contribution of each operation to the total DML in the supply chain. Baling and roadside storing operations are the primary contributors. The significant DMLs in both operations are expected. The harvest efficiency of the baling operation is affected by several factors such as the cut height of the combine, field conditions, the traffic of machinery in the field, and the weather conditions (Richey et al. (1982), Shinners and Binversie (2007) and Hess et al. (2009)). The combination of these factors may lead to matted and poorly windrowed biomass which is difficult to pick up. The DML during harvesting and collection operations could reach 60% of the total available biomass in the field (Hess et al., 2009). The reason for significant DML in roadside storage is due to the poor storage conditions. The storage regime is ambient storage with no protection resulting in an average annual DML percentage of 25%.  4.6.2 Daily delivery scheduling Due to the high capital and operating costs of cellulosic ethanol production, the shutdown of the conversion facility would be costly due to inventory stockout at the at-plant storage site. To avoid any feedstock shortage, the operations must be scheduled in a way that the daily biomass demand of the conversion process is met throughout the year. The primary decisions that must be made in the daily delivery scheduling include the quantity of daily delivered biomass to the plant, the capacity and inventory of storage sites and the number of machines in the supply system.  75  0.49  0.31  0.04  0.04  0.04  0.04  0.00  0.01  0.04  Figure 4-5: Contribution of logistics operations to the total DML in the supply chain  To fully meet the daily biomass demand, i.e. 750 t, the amount of daily delivered biomass must be more than the daily demand. The main reason for this is due to the different operational windows of the transportation system and the ethanol plant. The transportation system only operates 6 days/week, 14 h/day while the operations at the ethanol plant operate 24/7. Thus, the amount of daily delivered biomass throughout a week must be managed so that sufficient biomass is stored at the at-plant storage site to meet the demand during the off-shift hours of the transportation system. Another factor that affects the daily delivered biomass is dry matter loss during receiving and pre-processing operations. As shown in Figure 4-5, the at-plant operations contribute 9% of the total DML in the supply chain. Figure 4-6 illustrates the quantity of daily delivered biomass to the ethanol plant for a week. As shown in Figure 4-6, the daily biomass delivery is more than the daily biomass demand. The minimum, average and maximum daily deliveries are 830 t, 897 t and 970 t, respectively. On average, the annual delivered biomass to the gate of the ethanol plant is 268,570 t. Out of which, 262,500 t is fed to the conversion facility, 5,738 t is lost as DML during the receiving and preprocessing operations and the remaining biomass (332 t) is stored at the at-plant storage site at the end of the year. This inventory is used as the initial inventory for the following year. In the 76  last day of each week when there is no delivery, the inventory of at-plant storage is retrieved to meet the demand. This can be observed in Figure 4-7.  969  Delivered biomass (t)  1000 832  872  1  2  937  912  915  5  6  800 600 400 200 0  3 4 Simulation day  Figure 4-6: Daily delivered biomass to the ethanol plant in a week  1800  Biomass inventory (t)  1600 1400 1200 1000 800 600 400 200 0 1  29  57  85  113  141  169 197 225 Simulation day  253  281  309  337  Figure 4-7: Daily inventory of the at-plant storage site  77  Figure 4-7 shows the inventory of at-plant storage throughout the year. The maximum observed inventory during the year was 1,660 t. This implies that the storage capacity should be more than twice the daily demand to assure that sufficient biomass is stored. Thus, the capacity of at-plant storage is set to three times the daily demand, i.e. 2,250 t. This amount of biomass takes up 9,510 m2 (102,350 ft2) considering the international fire code discussed in section 4.3. This storage area accommodates 4,844 bales in 24 stacks of 91 t (100 ton). The average daily inventory of at-plant storage was 852 t. The next decision is the number of trucks required. A factor that affects this decision is bale mass. Although the dimension of the bale may remain the same over time, its mass is lost. The rectangular baler makes bales of 1.2×1.2×2.4 m with bulk density of 128 kg m-3. Thus, the mass of each created bale is around 0.45 t. This is the mass of a bale once made and left on the ground. At the end of the harvest season, all of the collected bales are placed at the roadside storage sites and will gradually be transported to the plant throughout the year. The bulk density of bales decreases since bales lose a portion of their mass as they remain in storage. Therefore, those bales that are transported earlier have a higher mass compared to the ones transported later in the year to the plant. This trend can be observed in Figure 4-8. Figure 4-8 shows a downward trend for almost the first half of the year and then a slow upward trend for the rest of the year, with the maximum number at the end of the year. The reason for the slight reduction in the number of truckloads at the first half of the year is due to the different sources of bales being transported at the beginning of the year. The primary source of delivered biomass at the beginning of the year is the initial inventory at the roadside storage sites. Since these bales have been stored for a long time, their masses have decreased. As the harvest season commences, the more recently created bales have a higher mass, thus, the number of truckloads dwindles. As the time elapses, the reduction in mass of stored bales causes the increase in the number of delivered truckloads. The number of truckloads decreases from a maximum of 79 in week 2 to a minimum of 64 in week 14 and then increases to a maximum of 90 in week 50. The average mass of each truckload in week 2, week 14 and week 50 are 12.2 t, 13 t and 10.7 t, respectively.  78  100 90  Number of daily truckloads  80 70 60 50 40 30  20 10 0 1  31  61  91  121  151  181  211  241  271  301  331  Simulation day  Figure 4-8: Number of daily truckloads delivered to the ethanol plant in a year  79  The next factor that impacts the number of trucks is the distribution of farms within the supply area. Figure 4-9 depicts the distribution of the annual number of delivered truckloads for different distance ranges. The total annual delivered truckloads are 22,070. As shown in this figure, most truckloads are transported from the 60-85 km distance range. Variation in the daily delivered truckloads and also uneven distribution of farms in the supply area complicates the transportation system. The simulation results revealed that the number of required trucks changes from 15 to 19. Table 4-4 shows the changes in the number of trucks utilized in the transportation system in three weeks of 2, 14 and 50.  Number of annual delivered truckloads  12000  11288  10000  8000  6000 3925  3791  4000 2767 2000  171  128  0  0  0  0 15-30  30-45  45-60  60-85  85-100  100-115 115-130 130-145 145-160  Distance range (km)  Figure 4-9: Annual delivered truckloads to the ethanol plant from different distance ranges  Table 4-4: Number of required trucks for three different weeks Week number  Monday Tuesday Wednesday Thursday Friday Saturday  Week 2  15  17  16  16  16  15  Week 14  16  15  15  16  15  16  Week 50  19  17  18  16  19  19  80  The number of machines in other operations was determined based on their capacity and the amount of biomass they should process. The capacity of a truck scale is 15 trucks per hour, thus, one truck scale was enough to weigh the daily arriving trucks. On average, the unloading of bales from each trailer takes around 15 min. Thus, one telehandler can unload the daily delivered bales in a 24-hour shift. In addition to unloading, a telehandler is used to load the stacked bales at the at-plant storage site to the grinding station. Thus, two pieces of telehandler were utilized at the ethanol plant. Finally, due to the hourly throughput of 16.5 t at the grinding operation, three grinders were utilized at the grinding station to meet the daily demand of the conversion process. For field operations, as explained before, one piece of equipment for baling and in-field transportation was considered for each farmer and one telehandler for each roadside storage site. The next decision in scheduling of the supply chain is the capacity of roadside storage sites and their inventory. The capacity of each storage site depends on the size of the respective farm, the amount of collected biomass, and the inflow and the outflow of biomass in storage. The results of the simulation showed that the capacity of all storage sites in the supply area would range from 15 t to 1,716 t with the average capacity of 540 t. The average storage time for each bale at roadside storage was 135 days. Since wheat is an annual crop and there is only one harvest season in a year, the roadside storage site is filled once. This fact can be observed in Figure 4-10. This figure shows the daily biomass inventory for the roadside storage site with the capacity of 540 t. This is the inventory at the end of the day. The initial inventory is consumed to meet the demand during the harvest season. With the progression of the harvest season, the inventory gradually increases and reaches the highest level at the end of the harvest season. At the end of the harvest season, there is no flow of biomass into storage and the inventory is consumed to meet the biomass demand for the rest of the year. For the illustrated storage site, the maximum observed inventory is 540 t. This amount of inventory requires a space of 1,765 m2 (19,000 ft2) considering the international fire code. This storage area accommodates 1,166 bales in 6 stacks of 91 t. Since each farmer owns a roadside storage site, the number of roadside storage sites in the supply area was 648. The simulation results showed that a total initial inventory of 29,250 t is required at roadside storage sites to meet the annual demand.  81  Biomass inventory (t)  600 500 400 300 200 100  0 1  29  57  85  113  141  169  197  225  253  281  309  337  Simulation Day Figure 4-10: Daily biomass inventory level of a roadside storage with 540 t capacity  Due to the specific characteristics of the biomass supply chain, the bottleneck took place at different stages of the supply chain. Due to the high frequency of daily biomass delivery and the limited capacity of loading operations, the waiting of trucks to be unloaded at the plant is inevitable. The average waiting time of each truck was 3.1 hours. In addition, the average number of trucks waiting in a queue to be unloaded was 7. The maximum observed waiting time was 4.1 hours with 9 trucks in the queue. Another bottleneck took place at the farms, where a baler waits, on average, one day for biomass to get dry enough to be baled. The maximum observed delay to dry was 3 days. The short period of delay to dry indicates the low moisture content of wheat straw during the harvest season. Figure 4-11 shows the range of changes in moisture content and its frequency in the supply chain. The moisture content in the supply chain changed from 0.07 to 0.2 (w.b.). Thus, the moisture content of biomass remains in a safe range in terms of aerobic stability. The highest moisture content was observed at the farmland and roadside storage and the lowest one at the atplant storage site.  82  0.50 0.45  Probability  0.40  0.30 0.23  0.20  0.10  0.06 0.01  0.01  0.01  0.01  0.01  0.07  0.08  0.09  0.1  0.11  0.02  0.04  0.04  0.04  0.04  0.04  0.13  0.14  0.15  0.16  0.17  0.00 0.12  0.18  0.19  0.2  Mositure content (w.b.)  Figure 4-11: Range of moisture content of biomass in the supply chain  4.6.3 Supply costs The supply costs are divided into three components based on the supply actors involved in the supply chain. These components encompass farm-gate costs, hauling costs and at-plant costs. Farmers bear farm-gate costs including baling, in-field hauling and storage. Hauling costs are incurred by the hauling contractors. These costs are associated with loading and road transportation operations. The last cost component is at-plant costs including receiving and preprocessing costs. These costs are incurred by the ethanol plant. In this section, the three cost components are first explained and then the explanation on total supply cost is given. It is noted that all cost figures incorporate DML costs. Figure 4-12 shows the range of farm-gate, hauling and at-plant costs.  83  Cost ($/t)  25 20 15  15.69  14.83  16.68  10 5  1.99  2.96  2.41  1.98  2.44  3.25  0 Baling  In-field hauling  Storing  a) Farm-gate costs  Cost ($/t)  25  21.56  22.73  24.31  20 15 10 5  2.78  2.35  2.04  0 Loading  Road transporting  b) Hauling costs  Cost ($/t)  25 20 15  11.08 11.24  11.48  10 5 0.70  0.90 1.00  2.14  2.57  2.84 0.62  0.75  0.84  0 Truck weighing Minimum Cost  Unloading  At-plant storing  Average Cost  Grindering  Maximum Cost  c) At-plant costs Figure 4-12: The components of the total supply cost  84  Table 4-5 lists the average costs of all the operations in the supply chain. The total delivered cost is $61.09 t-1. As expected, the most expensive operation is transportation which accounts for 37% of the total delivery cost. The next most costly operations are baling and grinding with contribution of 26% and 18% to the total delivery cost. Two machines including a baler and a tractor to pull the baler are utilized in the baling operation. In addition, grinding is an energyintensive process. For each operation, the DML costs are given. The storage costs represent the DML costs as the storage construction cost for roadside storage is zero. The high DML costs are due to the significant DML during the storage period as shown in Table 4-2. The baling operation and storage are the main contributors to DML costs. Although the total DML at the at-plant operations is 82% and 72% less than that of baling and roadside storing operations, the DMLrelated costs at at-plant operations are more than that of baling operation and almost equal to that of roadside storing. This is due to the higher value of lost biomass at the at-plant operations. The biomass delivered to the plant is a value-added feedstock which carries the costs of upstream operations.  85  Table 4-5: Average supply costs Logistics operation  cost ($/t) Operation  DML  Baling  14.44  1.25  In-field hauling  2.17  0.25  Roadside storing  0  2.44  Total farm-gate cost  16.61  3.94  Loading  2.05  0.30  Road transporting  22.42  0.31  Total hauling cost  24.47  0.61  Truck weighing  0.9  0  Loading  1.74  0.83  At-plant storing  0.14  0.61  Grinding  10.30  0.94  Total at-plant cost  13.08  2.38  Figure 4-13 depicts the distribution histogram of the total supply cost. The total supply cost would be in the range of $57.71 t-1 and $66.10 t-1 with 90% confidence interval. 0.20  Probability  0.18  Mean: 61.09 Standard deviation: 2.66  0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 56  57  58  59  60  61  62  63  64  65  66  67  68  69  70  Total supply cost ($/t)  Figure 4-13: Histogram of the total supply cost  86  4.6.4 Energy input and the associated emitted CO2 Figure 4-14 and Figure 4-15 show the amount of energy consumed during logistics operations to process biomass and the associated produced CO2, respectively. The total energy input is 943.11 MJt-1 and the total emitted CO2 is 71.23 kgCO2 t-1. Assume the energy density of ethanol is 21.1 MJ liter-1 (26.8 MJ/kg) (Thomas, 2000) and the ethanol yield is 340 liter t-1 (Iogen Corp., 2005). Then, 13% of produced ethanol is consumed in the biomass supply chain to deliver the required biomass to the conversion facility. It is noted that the energy input and the emitted CO2 were only estimated for the operations in the supply chain and the building of equipment was not considered in this study. The building of equipment requires a life cycle analysis (LCA). This is outside the boundary of this study. In addition, since CO2 is the most important greenhouse gas, only the emission of this gas was estimated.  500  471.82  Energy Input (MJ/t)  450 400 350 300  269.38  250 200 150 102.31  100 52.65  50  23.74  23.21  0  Baling  Infield transportation  Loading  Road transportation  Unloading  Grinding  Figure 4-14: Energy input of different operations in the supply chain  87  35 30.34  Emitted CO2 (kg/t)  30  27.07  25 20 15 10  7.02 3.59  5  1.63  1.58 0 Baling  Infield transportation  Loading  Road transportation  Unloading  Grinding  Figure 4-15: Emitted CO2 in different operations in the supply chain  4.7 Sensitivity analysis on farmer participation rate In the IBSAL-MC simulation model, it was assumed that all the farmers within 160-km supply radius are willing to participate in selling their produced biomass to the ethanol plant. Farmers may change their attitudes toward participating in the biomass supply during the business life of the ethanol plant. This results in changes in the farmer participation rate. To investigate the impact of the farmer participation rate on the performance of the supply chain, the IBSAL-MC model was run for two participation rates of 50% and 25%. The information on biomass availability at these two rates is available in BIMAT for the case study. Thus, they were selected for the sensitivity analysis. The results of the sensitivity analysis revealed that the daily biomass demand would be fully met year-round at 50% participation rate. In contrast, about 98% of the annual biomass demand would be met at 25% participation rate. The amount of available biomass within 160-km radius and at a 25% participation rate is 372,569 t. Out of this amount, 7% remains on the fields and cannot be collected due to the termination of the harvest season and 19% is lost in the supply chain. The total delivered biomass to the gate of the ethanol plant is 262,230 t. Out of this 88  amount, 256,630 t is fed to the conversion facility, 5,420 t is lost as DML in receiving and preprocessing operations and the remaining biomass (180 t) is stored at the at-plant storage site at the end of the year and used as the initial inventory for the following year. The biomass shortage occurs at the beginning of the harvest season. Figure 4-8 illustrates the daily delivered biomass to the ethanol plant during the harvest season. As shown in this figure, the delivered biomass on some days is less than 750 t. This shows the importance of the initial inventory to meet the daily demand in the agricultural biomass supply chain.  Delivered biomass (t)  1000  800  600  400  200  0  1  8  15  22 29 Simulation day  36  43  Figure 4-16: Daily delivered biomass to the ethanol plant during the harvest season (25% participation rate)  Although the daily demand is met at a 50% participation rate, the supply radius is greater than that for 100% rate. As shown in Figure 4-17, the feedstock is procured from farmers within the 160-km supply radius, while the daily demand can be met within the 115-km radius at 100% rate. The increase in the supply radius leads to the increase in number of trucks and their associated costs, energy inputs and the emitted CO2. Table 4-6 shows the impact of farmer participation rate on the transportation system.  89  Number of annual delivered truckloads  7000  5996 6000 5000 4000  3392 2900  2644  3000  1953  2000  1498  1757  1496  1000  477  0 15-30  30-45  45-60  60-85  85-100  100-115  115-130  130-145  145-160  Distance range (km)  Figure 4-17: Number of annual truckloads delivered to the ethanol plant from different distance ranges (50% participation rate)  Table 4-6: Impact of farmer participation rate on the transportation system Farmer participation rate 100%  Number of required trucks (Min/Max) 15-19  Operating cost ($t-1) 22.73  Energy input (MJt-1) 471.8  Emitted CO2 (kgt-1) 30.34  50%  21-26  30.66  642.1  44.03  25%  24-29  38.71  845  57.94  It is noted that the farmer participation rate does not significantly affect the field operations and the at-plant operations. The changes in farmer participation rate mainly affect the transportation system. However, if the field machines are shared between farms, the farmer participation rate may impact the field operations as the distribution of farms in the supply region and thereby the distance to move the equipment between farms will change.  90  4.8 Validation of the simulation model The validation of the IBSAL-MC model was one of the biggest challenges of this study. The IBSAL-MC model was applied to a proposed commercial-scale cellulosic ethanol plant. Thus, there were no data available from the real system in order to compare with the outputs of the developed model. The validation process even became more challenging since there was no commercial-sized cellulosic ethanol plant operating in the world to use for comparison. In order to validate the model, three sources were used: 1) the results of the feasibility study conducted by Iogen Corp.; 2) the obtained results of similar studies in relevant literature, and 3) the approval of the obtained results of the model by the experts. In terms of the number of contracted farms, Iogen anticipated that approximately 600 farms would be required to procure the required biomass for the ethanol plant (Canadian Biomass Magazine, 2009). The outputs of the simulation model showed that 630 and 648 farmers are required to meet the annual demand at 50% and 100% participation rates, respectively. It is noted that Iogen did not specify the farmer participation rate. In addition, Iogen projected an annual purchase of over $10 million of wheat straw from farmers within the supply area (Iogen Corp., 2009). The results of the simulation model predicted annual farm-gate costs of $11,644,290 including the price of laid biomass on the field, baling, in-field hauling and storage costs. The outcomes of the simulated biomass supply chain depends on biological, climatic and geographical factors (Nilsson, 2000) and thus, the outcomes are region-based. In addition, different assumptions on the input data and the structure of the supply chain in the literature make the comparison even more difficult. However, the results of the developed simulation model in this study are comparable with the ones in the relevant literature. Table 4-7 illustrates the logistics costs of the operations in the biomass supply chain estimated in different studies. As shown in Table 4-7, the cost outputs of the IBSAL-MC model are comparable with those of other studies. Our study supports the literature in that transportation, baling and grinding are the most significant contributors to the delivery costs. In addition, the variations in the operating costs of different operations shown in Figure 4-12 are comparable to the ones in Hess et al. (2009). It is noted that high transportation costs in this study are due to the low yield and bulk density of wheat straw resulting in the need for a large supply area to meet the biomass demand of the 91  ethanol plant. The majority of the studies in the literature modeled and analyzed corn stover and switchgrass with yields more than twice the average yield of wheat straw. In addition, cereal straw produce the lowest bale densities (Hess et al., 2009). Table 4-8 and Table 4-9 compare the estimated energy input and the associated produced CO2 in different studies, respectively. The estimated values of these two outputs for baling, in-field hauling, road transportation and grinding are comparable with those of previous studies. For loading and unloading operations, the results of this study are similar to those of Morey et al. (2010) but it is different from other listed studies. These differences are mainly due to the different specification of equipment such as horse power. The high DML of baling operations was also concluded by Hess et al. (2009). In this study, the amount of DML in baling operation is 10%. Hess et al. (2009) estimated a DML of 10% for switchgrass and 46% for corn stover in baling operation. In addition, Shinners et al. (2007) and Richey et al. (1982) reported significant losses ranging 20%-75% in collection operations including shredding, raking, and baling. The obtained results of the IBSAL-MC model showed that the moisture content of wheat straw in the entire supply chain is at a safe level which removes the need for extra activities to manage the moisture content. The same discussion has been provided by Hess et al. (2009). In addition to the comparison with the studies in the literature, the obtained results were checked and confirmed by the scientist from Agriculture and Agri-Food Canada (AAFC) and the Oak Ridge National laboratory.  92  Table 4-7: Comparison of logistics costs ($ t-1) in different studies Study  Logistics operations Baling  In-field  Storage  Loading  Transportation  hauling Hess et al. (2009)  9.77-  1.7-1.89  10.91a  Truck  Unloading  scale 1.17-  0.76-  1.54b  0.84  At-plant  Grinding  storage  11.09-13.37  -  1.44-1.61  0.37c  11.66  Morey et al. (2010)  21.16  5.51  3.24  -  30.75  -  2.08  -  13.49  Cundiff et al. (1997)  -  -  2.20  3  9.44  -  -  -  -  12.42  3.03  -  -  -  -  -  -  -  9.66  4.54  -  2.23  16.53  -  1.06  -  5.65  Judd et al. (2012)  -  3.35  -  1.66  -  -  0.36  -  -  Sokhansanj et al.  -  -  6.7  0.92  12.4  -  0.74  -  -  9.59  -  -  -  15.10-15.83  -  -  -  -  15.69  2.41  2.44  2.35  22.73  0.9  2.57  0.75  11.24  Sokhansanj et al. (2008) Sokhansanj and Fenton (2006)  (2008) Sokhansanj et al. (2009) This study a  DML costs are not included Sum of rental and DML costs c Sum of truck scale and storage costs b  93  Table 4-8: Comparison of energy input (MJt-1) in different studies Study  Baling  In-field hauling  Loading  Road  Unloading  Grinding  transportation Morey et al. (2010)  34.38  64.76  -  196.9  20.69  233.7  Sokhansanj et al. (2008)  168.4  14  -  -  -  -  133  80.3  94  791  208  96  Sokhansanj et al. (2006)  124.9  169.1  154.5  640.6  199.6  185.7  Sokhansanj et al. (2009)  -  -  -  374-407  -  267  Kumar and Sokhansanj (2007)  82.23  41.18  53.14  384.44  428.66  141.46  This study  102.31  52.65  23.21  471.82  23.74  269.38  Sokhansanj and Fenton (2006)  94  Table 4-9: Comparison of emitted CO2 (kg t-1) in different studies Study  Baling  In-field  Loading  hauling  Road  Unloading  Grinding  transportation  Morey et al. (2010)  2.62  4.93  -  13.1  1.58  17.19  Sokhansanj et al. (2008)  13.09  1.1  -  -  -  -  Sokhansanj and Fenton (2006)  10.4  6.5  7.4  33.8  16.2  7.5  Sokhansanj, Kumar, et al. (2006)*  2.65  3.59  12.05  49.88  15.59  14.45  -  -  -  -  -  -  Kumar and Sokhansanj (2007)  6.38  3.20  4.12  29.83  33.26  10.98  This study  7.02  3.59  1.58  30.34  1.63  27.07  Sokhansanj et al. (2009)  *The estimated emissions are carbon (C) not CO2  95  4.9 Discussion and conclusions In this chapter, a new simulation model, called the IBSAL-MC model, was developed to plan and schedule the biomass supply chain for the proposed cellulosic ethanol plant in Prince Albert, Saskatchewan. The developed simulation model considered the specific features and constraints of the agricultural biomass supply chain in the Canadian Prairies. In order to evaluate the fulfillment of the daily biomass demand, the model incorporated the details of the supply area such as location, size and yield of farms, distribution of bales on the fields, soil conservation rate, moisture content, machine breakdown and repair time, equipment efficiency, daily weather data, weekly harvest schedule, competitive market (livestock requirement), delays, length of contract with farmers, and farmer participation rate. This level of detail allows the model to provide accurate outputs. The model plans the field operations to process as much biomass as possible during the harvest season so that sufficient biomass would be available to meet the annual biomass demand (push planning). On the other hand, it plans the other operations to process the amount of biomass required to fulfill the daily demand (pull planning). The main findings of the simulation model for the case study are as follows:   The daily biomass demand was met within a 160-km supply radius and 100% farmer participation rate. On average, 27% of the farms (648) were required to contract. These farms were located in a 115-km radius and covered 348,875 ha. The amount of collected biomass from farms was 37% (92,208 t) more than the annual demand. The extra amount of biomass was either lost (26%) or used as the initial inventory (11%) for the next year.    Baling and roadside storing were the two major contributors to the DMLs with a total contribution of 80%.    To meet the daily biomass demand, an average daily delivery of 897 t should be scheduled.    The capacity of at-plant storage should be set to three times the daily demand. This storage area accommodates 4,844 bales in 24 stacks of 91 t and takes up 9,510 m2. The average daily inventory of at-plant storage was 852 t.    The capacity of storage sites ranged from 15 t to 1,716 t with the average capacity of 540 t.  96    The results of the simulation model revealed the importance of the initial inventory in the supply system to meet the biomass demand during the harvest season. Initial inventory of 29,250 t and 360 t was required at the roadside storage sites and at-plant storage, respectively.    Changes in mass of bales over time, the uneven distribution of farms in the supply region and other uncertainties such as machine breakdown led to the need for different numbers of trucks during the year. The number of required trucks ranged from 15 to 19.    The total annual delivered truckloads were 22,070. 51% of them were delivered from the 6085 km distance range to the ethanol plant.    The numbers of required equipment at the ethanol plant were one truck scale, 2 pieces of telehandler and 3 grinding stations.    The average waiting time of each truck in a queue to be unloaded was 3.1 hours. In addition, the average number of trucks in the queue was 7.    The moisture content of biomass remained in a safe range in terms of aerobic stability (720% w.b.). This implies that there is no need for extra activities to dry wheat straw or using expensive storage structures at roadside to protect biomass against weathering and detrimental microbial activity.    The average total supply cost was $61.09 t-1. The most expensive operations were transportation, baling and grinding accounting for 37%, 26% and 18% of the total delivery cost, respectively.    The total supply cost was in the range of $57.71 t-1 and $66.10 t-1 with 90% confidence interval.    DMLs in the supply chain contributed 11% to the total supply cost.    The total energy input and the total emitted CO2 were 943.1 MJt-1 and 71.2 kgCO2 t-1, respectively. In addition, 13% of produced ethanol would be consumed in the supply chain to deliver the required biomass to the conversion facility.    The results of the sensitivity analysis on the farmer participation rate showed that the farmer participation rate does not significantly affect the field operations and the at-plant operations. The changes in farmer participation rate mainly impacted the transportation system due to the change in the supply radius. For example, 50% decrease in the participation rate led to a 45 km increase in the supply radius. The transportation costs at 100% rate were 26% ($7.93 t-1) 97  and 41% ($15.98 t-1) less than those for 50% and 25% rates, respectively. In addition, the daily biomass demand was fully met year-round at 50% rate, while about 98% of the annual biomass demand was met at 25% rate. The obtained results reveal the efficiency of the developed model to evaluate different aspects of a dynamic and stochastic supply chain. The model simulates the daily inflow and outflow of biomass in the storage locations and finds the storage-related solutions including the size of the storage sites, their inventory, DMLs and costs. The model also finds the required initial inventory to meet the demand during the harvest season. In addition to the storage management, the simulation model provides a detailed analysis of the transportation system. Due to the high contribution of the transportation system in the total delivery cost, the management of the truck fleet is important. The model finds the solutions for the quantity of daily biomass delivered to the plant and the number of trucks in the supply system. The change in mass of bales due to the occurrence of DML was considered in the determination of the required trucks. To the best of the author’s knowledge, the matter of changes in bale mass and its impact on the number of trucks have not been discussed in the relevant literature. In spite of the capability of the simulation model to plan and schedule the supply chain to meet the daily demand, the outputs of the simulation model can be improved in several ways:   Although the simulation model enables one to estimate the supply costs, these costs are not necessarily the optimal ones. Complementing the developed simulation model with an optimization model can help to find the minimum supply costs possible.    The results of the simulation model showed that the 160-km supply radius accommodates a large number of farms distributed in the supply area and a percentage of these farms would be sufficient to meet the annual demand. The simulation model selected the farms randomly. An optimal selection of farmers based on their biomass production rates, their distance from the plant and also the annual biomass demand would result in a more cost-efficient biomass supply chain.    The results of the simulation showed that the average capacity of a roadside storage site is 540 t. Since telehandler loads an empty trailer in less than 30 minutes and each truckload contains around 13 t of biomass, the total utilization rate of this equipment would be low. To increase its utilization rate, satellite storage should be considered in the design of the supply 98  chain. Since each satellite storage location usually receives biomass from more than one farm, it contains more biomass compared to roadside storage. In addition, using satellite storage could result in fewer numbers of large storage sites compared to the roadside storage system making the flow of biomass more manageable. In order to take advantage of satellite storage, their optimal number and location should be determined in the supply area. The above-mentioned improvements relate to the optimal design of the supply area. Thus, a new optimization model is developed and integrated with the IBSAL-MC model to improve the cost-efficiency of the supply chain. The next chapter elaborates on the developed optimization model and the integrated approach.  99  Chapter 5. Development of an integrated simulation /optimization model 5.1 Synopsis In this chapter, a mathematical optimization model is developed to prescribe the design of the supply area and then it is integrated with the IBSAL-MC simulation model. The design of the supply area refers to the selection of farms to purchase biomass from, the number and location of storage sites as well as the assignment of the selected farms to the storage sites. In the integrated model, the IBSAL-MC simulation model exploits the prescribed design to manage the flow of biomass among farms, storage sites and the ethanol plant. On the other hand, the detailed outputs of the IBSAL-MC simulation model are used to update some of the input parameters of the optimization model and adjust the design. The next section explains the structure of the developed optimization model, its parameters, objective function, decision variables, and constraints. Section 5.3 elaborates on the proposed integrated simulation/optimization framework. In section 5.4, the integrated framework is applied to the proposed ethanol plant and the results of the integrated model are compared with those of the optimization model and simulation model. Thereafter, the sensitivity analysis is carried out on the input parameters in order to evaluate their impact on the supply design and the total supply cost. The last section provides the discussion and conclusions on the obtained results.  5.2 Structure of the optimization model The obtained results of the IBSAL-MC model showed that in order to improve the performance of the supply chain, a combination of farms should be selected. In addition, consideration of satellite storage would lead to a more manageable and efficient supply chain. In order to apply and investigate these two improvements in the supply chain, the following questions must be addressed: 1) which farmers does the ethanol plant need to contract; 2) how many storage sites need to be established and where should they be located; and 3) which farm should be assigned 100  to each storage site. The optimal solutions for these questions provide the optimal design of the supply area. A combination of roadside and satellite storage is considered in the optimal design. Thus, the second question includes the decision on whether to share an established storage site between farms (satellite storage) or to assign only one farm to storage (roadside storage). Roadside storage is the ideal storage system for farmers as they do not have to haul biomass to storage on a public road that may require specialized transport equipment. However, as concluded by the IBSAL-MC model, the roadside storage system would result in a large number of small storage sites. The distribution of small storage sites complicates the transportation system for hauling companies. In addition, loading operations at the roadside storage locations would be an inefficient logistics task. In this storage system, loading equipment has to move between many production fields or is dedicated to each farm. In either case, the utilization rate of the hauling equipment would be low. This is due to the frequent movement of equipment between farms or the equipment would be idle for a significant period of time during the year (Judd et al., 2010). In contrast, satellite storage usually contains biomass from more than one farm. Thus, fewer numbers of storage locations are required within the supply area. Moreover, higher availability of biomass at each satellite storage site makes the hauling operations more productive for hauling companies. However, such a storage system causes higher in-field hauling costs for farmers as the delivery of the collected biomass to satellite storage requires some transportation on roadways. The equipment used to haul biomass from farms to storage is a low-speed vehicle, thus, hauling times and operating costs would increase. Cundiff et al. (2009) argue that longer distances increase the operating time of the in-field hauling equipment resulting in reduction of its productivity to transport more biomass during the working hours. On the contrary, shorter distances decrease the amount of biomass accumulated at a given storage location, increasing the operating costs for the hauling contractors. Thus, farmers prefer the roadside storage system while hauling contractors prefer the satellite storage system. On the other hand, the ethanol plant is interested in low-cost delivery of biomass to the conversion facility. Therefore, the design of the supply area must be determined by considering both farm-gate costs and hauling costs. This consideration leads to optimal hauling distances from farms to storage locations and from storage locations to the plant. These distances affect the number and 101  location of storage sites and consequently the cost of delivering biomass to the conversion facility. To find the design, a mixed-integer linear programming (MILP) model was developed and solved using the AIMMS software package. The MILP model was solved by CPLEX 12.1 on an Intel Core 2 Duo E7300 2.66 GHz computer with 2 GB of RAM running Windows 7. Indices, input parameters and decision variables of the MILP model are defined as follows:  Indices i  Farm index (i=1,…,I)  k Biomass index (k=1,…,K) j  Storage index (j=1,…,J)  t  Time index (t=1,…, 5). The planning horizon is 5 years due to the five-year contract with farmers. The time period is one year.  Parameters A  Annual feedstock demand (t)  BaleMass  Mass of each bale (t/bale)  CAPbaler  Capacity of baler (t/h)  CAPsj  Maximum capacity of storage j (t)  Cbaler  Custom cost of baler ($/h)  Ctwine  Cost of twine per bale ($/bale)  Ctractor  Custom cost of tractor to pull baler ($/h)  Cstinger  Custom cost of stinger stacker ($/h)  Cstorage  Amortized establishment cost of storage ($/t-year)  CF_Loader  Annual fixed cost of telehandler ($-year)  CV_Loader  Variable cost of telehandler ($/h)  Ctrailer  Custom cost of trailer ($/h)  Ctruck  Custom cost of truck ($/h)  102  Parameters (Continued) DFSij  Distance between farm i and storage j (km). This distance is calculated from the center of farm to the center of storage  DSPj  Distance between storage j and the ethanol plant (km)  Ebaler  Efficiency of baler (%)  Estinger  Travel efficiency of stinger stacker (%)  Estinger_load  Loading efficiency of stinger stacker (%)  Estinger_unload  Unloading efficiency of stinger stacker (%)  Eloader  Efficiency of telehandler (%)  Etruck  Travel efficiency of truck (%)  Hbaler  Maximum available capacity of baler in a year (t)  Hloader  Maximum available capacity of telehandler in a year (t)  Hstinger  Maximum available capacity of stinger in a year (t)  LoadMassstinger  Mass of 8 bales loaded by stinger (t/load)  LoadMassloader  Mass of 2 bales loaded by telehandler (t/load)  LoadMasstruck  Mass of 30 bales loaded onto truck (t/load)  Lstinger  Loading time of each bale by stinger (h/bale)  Lloader  Loading time of a load of 2 bales onto truck (h/load)  Pk  Price of biomass k laid on the field after harvesting ($/t)  Sik  Size of farm i assigned to crop of biomass k (ha)  Speedstinger  Average speed of stinger (km/h)  Speedtruck  Speed of tractor-trailer truck (km/h)  Ustinger  Unloading time of each bale from stinger (h/bale)  Uloader  Unloading time a load of 2 bales from truck (h/load)  Yik  Average yield of biomass k in farm i (t/ha)  θdry_field θbaler  Average dry matter loss (DML) percentage once biomass is laid on the field to get dry (%) Average DML percentage of baling operation (%)  θstinger  Average DML percentage of in-field hauling operation (%)  θstorage  Average annual DML percentage of storage (%)  103  Parameters (Continued) θloader  Average DML percentage of loading operation (%)  θtruck  Average DML percentage of road transportation operation (%)  θplant  Average DML percentage of operations in the ethanol plant including unloading, at-plant storing and grinding (%)  Decision Variables Amount of biomass k collected from farm i and transported to storage j in period t (t)  Amount of biomass k transported from storage j to the ethanol plant in period t (t)  =  {  =  {  For the illustrative purposes, the annual flow of biomass in the supply chain is shown in Figure 5-1 in a mathematical form.  Objective Function Min Z=Minimize (Farm-gate costs + Hauling costs)  (5.1)  Farm-gate costs= Collection costs + In-field hauling costs + Storage costs  (5.2)  Hauling costs= Loading costs + Road transportation costs  (5.3)  The objective function (Eq. (5.1)) is to minimize the sum of farm-gate costs and hauling costs over the contract period, i.e. five years. The details of each of these cost components are shown in Eq. (5.2) and (5.3).  104  Farm i, biomass k Available biomass on the field  Storage j  Yik× Sik Field drying  Bikjt  Loading  θloader  In-field hauling  Road transporting  Roadside/ Rkjt satellite storage  θstinger  θbaler  θdry_field  Rkjt× (1-θloader)  Bikjt×(1-θbaler)  Baling  Bikjt×(1-θbaler) ×(1-θstinger)  Rkjt× (1-θloader) ×(1-θtruck)  Ethanol plant  θstorage  Rkjt× (1-θloader)× (1-θtruck)×(1-θplant)  Annual feedstock demand  θplant  θtruck  Amount of biomass delivered to the next operation Amount of dry matter loss in the respective operation  Figure 5-1: Mathematical symbols representing the annual flow of biomass in supply chain  Collection costs=Baling costs+ Baling DML costs  (5.4)  As shown in Eq. (5.4), collection costs are the sum of baling costs and the associated DML costs. [  ) )] (  ∑ ∑ ∑ ∑ [(  ))  )  (  )]  (5.5) Eq. (5.5) represents the total baling cost which is the sum of the costs associated with utilizing the baler and tractor to pull the baler and the cost of tying the bales with twine.  [  ]  ∑ ∑ ∑ ∑[(  )  ]  105  (5.6) As shown in Eq. (5.6), the baling DML costs are estimated based on the price of biomass laid on the field. As explained in section 4.3, it was assumed that the quality of biomass is at an acceptable level for all farms in the supply area, and thus, the price of laid biomass on the field is assumed to be fixed for each biomass type in all farms.  In-field hauling costs = Loading costs+ Transportation costs+ Unloading costs+ In-field hauling DML cost (5.7) The second component of the farm-gate costs is in-field hauling costs. This cost component is the sum of costs incurred in loading and unloading bales, transporting them to storage and also the associated DMLs.  [  ) ] )  ∑ ∑ ∑ ∑ [((  ((  )  )  (  )))  ]  (5.8) Given the estimation of the number of created bales in each farm and the loading and unloading time for each bale, the total loading and loading costs are calculated in Eq. (5.8).  [  ) ]  ∑ ∑ ∑ ∑ [((  )  (  )  ))  ]  (5.9) In-field transportation costs depend on the distance travelled to pick up eight bales from farms and haul them to storage. The in-field transportation costs are shown in (5.9). 106  [  )]  ∑ ∑ ∑ ∑ [(( ( (  ) )  )  )  ) )) )  (  )]  (5.10) The last cost component of the in-field hauling costs relates to DML. In this operation, biomass has higher value compared to the baling operation since biomass has already gone through the baling operation. Thus, the DML-related costs are estimated based on the value of biomass lost in this operation. In general, as biomass moves forward from upstream operations to downstream operations, the value of biomass increases based on the operations performed on biomass. This concept has been applied to DML-related costs in the rest of the DML-related equations.  Storage costs= Storage establishment costs + Storage DML costs  (5.11)  The last cost term of the farm-gate costs is the storage-related costs including establishment costs and DML costs.  ∑  [∑ ∑(  )  ))]  (5.12) Farmers are responsible for establishing storage sites and placing the collected biomass in these locations during the contract period. They must consider enough space at the roadside in case they need to expand the storage area when biomass production increases in a year. To deal with this issue, the maximum annual purchased biomass during the five-year contract is considered as the base to estimate the area required to establish storage. This issue is reflected in  107  Eq. (5.12). Since the optimization model is a tactical planning model, this assumption is accurate enough to estimate the establishment costs. To linearize Eq. (5.12), it is replaced with Eq. (5.13) and Eq. (5.14). Eq. (5.13) is used in the objective function and Eq. (5.14) is added to the list of constraints.  ∑̅  (5.13) ̅  [∑ ∑(  )  ))]  (5.14) It is assumed that the land costs are the same in all farms. Thus, the same amortized establishment cost (Cstorage) is used. It is noted that the amortized establishment costs depend on the storage regime. The costs for different storage regimes are given in Table 4-1.  [ )) )]  108  )  ∑ ∑ ∑ ∑ [((  ( )  ( (  ) (  ) )  ( ∑ ∑ ∑ ∑ [((  (  )  )  (  ))  )] )  )  )  ))  (( ( (  (  )  ) )  (  ) )  )  )]  (5.15) The second component of the storage costs is DML costs. These costs are estimated in Eq. (5.15). Sum of Eqs. (5.5), (5.6), (5.8), (5.9), (5.10), (5.12) and (5.15) make up the farm-gate costs. The second term of the objective function is the hauling costs. As shown in Eq. (5.3), it includes the loading costs and the road transportation costs.  Loading costs = Annual fixed cost + Variable cost + DML cost  (5.16)  It was assumed that one telehandler is dedicated to each storage site and there is no movement of loading equipment between storage sites. Thus, the purchasing cost of equipment should be reflected in the objective function. To this end, the loading costs are comprised of annual fixed cost, variable cost and the loading DML costs. Eqs. (5.17), (5.18), (5.19) show the details of the loading costs. The dedicated loading equipment has been also considered by Judd et al. (2010) and Judd et al. (2012).  109  ∑  (5.17) [  ]  )  ∑∑∑(  )  (5.18) [ [ [∑ ∑ ∑(  )] ] )  )  ∑ ∑ ∑ [((  (( ( ∑ ∑ ∑[(  ) )  ] ) ) )  (  )  ∑ ∑ ∑ ((  [  )]  (  ))  )  )  ]  ]  (5.19) Eq. (5.17) shows that the fixed cost is considered in the model only when the roadside of a farm is selected to establish a storage site (Zj=1). If the roadside of a farm is not selected (Zj=0), there is no biomass delivery to the respective roadside. Thus, all Bikjt would be set zero and the combination of constraints in the optimization model force Rkjt to get zero. Therefore, the variable loading cost in Eq. (5.18) would be zero as well. Eq. (5.19) represents the DML costs in loading operation.  110  [ )] )  [ ] )  ∑ ∑ ∑ ([(  (  )]  (  )  )  )  )  (5.20) The last cost component of the objective function is road transportation costs. Eq. (5.20) shows the transportation costs which are the sum of the trailer and tractor-trailed truck costs. The total operation time of a truck is the aggregation of the waiting times while being loaded at a storage site, unloaded at the plant and the transportation time.  [ [ [  )] ] ]  [  [∑ ∑ ∑(  )  )  (  )  )  ∑∑∑(  [∑ ∑ ∑ ((  ((  (  )  ) )  (  (  (  )  ))  ] )  )  (  (  )) )  [∑ ∑ ∑(  )  )  (  )  )  [∑ ∑ ∑  ]  )  )  ]  ]  (  )  ]  111  (5.21)  Eq. (5.21) represents the DML costs of road transportation operation. Biomass has a high value at this stage of the supply chain as it has gone through several operations. This is reflected in Eq. (5.21).  Constraints )  (∑ ∑  )  ))  (  (5.22) The first set of constraints, Eq. (5.22), is demand constraint. Sufficient biomass must be transported from storage sites to the ethanol plant to assure the annual feedstock demand is fulfilled taking DMLs into account in both upstream and downstream operations.  )  (∑ ∑  (  ))  (  )  ∑  (5.23) The set of constraints shown by Eq. (5.23) represents the biomass flow balance in roadside/satellite storage sites. The amount of annually released biomass from a storage site is equal to or less than the amount of annually arrived biomass.  (  )  (5.24) The set of constraints in Eq. (5.24) represents the supply constraint. The quantities of the purchased biomass from each farmer depend on the yield and the size of the farmland a farmer assigns to grow the respective crop.  112  ∑∑∑  (5.25) ∑∑∑  ) (5.26)  ∑ (5.27)  ̅ (5.28)  Eqs. (5.25), (5.26), (5.27) and (5.29) represent the resource capacity constraints for baler, stinger, telehandler and storage, respectively.  (5.29) Constraints in Eq. (5.29) show that a potential storage location is selected when at least one farm is assigned to it.  ∑  (5.30) Constraints in Eq. (5.30) represent whether each farm in the supply area is not selected or selected and assigned to only one storage location. ̅  113  (5.31) The last set of constraints in Eq. (5.31) represents sign restriction. Based on the objective function and the set of constraints, the goal of the optimization model is to find the optimal flow of biomass from farms to storage sites and from storage sites to the ethanol plant in a way that the annual biomass demand is met with the minimum delivery cost. The developed optimization model in this study is the abstract of the IBSAL-MC simulation model without considering the uncertainties such as weather data and harvest season in the supply chain to avoid a large-sized optimization model. No special consideration has been given in the optimization model from the viewpoint of equipment for different types of biomass since the same pieces of equipment are used to handle multiple types of biomass considered in this study. In case of the need for different equipment, the optimization model must be adjusted accordingly. It is noted that the optimization model was developed based upon the baling scenario. Consideration of different logistics scenarios such as baling and chopping in the optimization model also results in the changes in the model. As discussed in Chapter 2, a few optimization models were developed in the literature to find the optimal selection of farms and the optimal number and locations of storage sites including Judd et al. (2010), Judd et al. (2012), Zhu et al. (2011) and Zhu and Yao (2011). The developed model in this study has some advantages compared to the developed models in the literature. The primary advantage of the developed optimization model is the calculation of accurate cost figures by considering the price of biomass, the DML costs, the value of lost biomass at different stages of the supply chain and the details of the operating costs for machines and storage. In addition, the results of the IBSAL-MC model showed that the amount of DMLs is significant in the supply chain. Thus, the consideration of DMLs in the model results in more accurate numbers of selected farms and established storage sites. The developed models by Judd et al. (2010) and Judd et al. (2012) did not calculate the amount of multiple types of biomass delivered from farms to storages sites (Bikjt) and from storage sites to the plant (Rkjt). In addition, the collection costs (baling) were not incorporated in their models. These costs result in the selection of the minimum number of farms as with each new farmer in the supply chain, the ethanol plant has to pay for the biomass collection costs. Calculation of Bikjt, Rkjt and the associated operating costs provide details of the flow of biomass  114  in the supply chain and the total delivery cost for the decision makers at the tactical level. In addition, they did not consider the constraint on the capacity of storage. It is noted that Judd et al. (2010) did not consider transportation costs from satellite storage sites to the plant. This cost component would impact the location and number of satellite storage sites. In their next study, Judd et al (2012) found the optimal locations of storage sites with respect to transportation costs from farms to the storage sites, and also from storage sites to a bioprocessing plant. The developed models by Zhu et al. (2011) and Zhu and Yao (2011) considered the details of the supply chain but the models are large-sized optimization models and may not be solvable for real-life case studies. In contrast, the application of the developed optimization model in this research to a real-life case study shows that the model finds the optimal solutions in 2 minutes.  5.3 Integrated simulation/optimization model In addition to the design, the optimization model determines the annual flow of biomass from the selected farms to the established storage sites and from the storage sites to the plant. This information can be used by the simulation to schedule the supply chain to reduce the delivery costs. On the other hand, the developed optimization model is at the tactical planning level and does not consider the dynamics and stochastic nature of the supply chain while the IBSAL-MC model represents a dynamic, non-linear and stochastic model of the supply chain at the operational level. Thus, the design prescribed by the optimization model does not guarantee the fulfillment of the daily biomass demand throughout the year. In addition, the outputs of the simulation model can be used to estimate some of the input parameters of the optimization model and to adjust the supply chain design based on what happens on a daily basis. These parameters are listed in Table 5-1.  115  Table 5-1: Input parameters of the optimization model estimated by the simulation model Parameter  Parameter  Cstorage  θbaler  DFSij  θstinger  LoadMassloader  θstorage  LoadMasstruck  θloader  Speedstinger  θtruck  θdry_field  θplant  The IBSAL-MC model keeps track of the daily inflow and outflow of biomass at storage. Thus, the capacity of storage and the cost of storage per tonne of biomass can be estimated accurately by the IBSAL-MC model. For example, the storage cost (Cstorage) in the optimization model is estimated based on the assumption that each tonne remains in storage for one year. While, each tonne of biomass may remain from one day to several months in storage resulting in different storage cost per tonne. The accurate storage cost estimated by the IBSAL-MC model can be used in the optimization model. In addition, in the optimization model, the maximum arriving biomass to storage per year was used to estimate the total storage costs (Eq.(5.12)). The IBSAL-MC model is able to estimate the storage capacity based on the rate of arriving and releasing biomass from storage. This rate can be used in the optimization model to estimate the maximum space occupied by the baled biomass per year. Thus, Eqs.(5.12) and (5.13) should be replaced by Eqs.(5.32) and (5.33) , respectively:  ∑  [∑ ∑(  )  ))]  (5.32) ∑̅  116  (5.33)  To estimate  j,  the maximum observed biomass at storage j in a year is divided by the  maximum biomass produced at all the farms assigned to the respective storage (total arriving biomass). For example, if the maximum produced biomass is 100 t and the maximum observed biomass in the storage is 70 t, storage. Thus, the value of  j  j  equals 0.7. This reflects the space used to store biomass at  depends on the rate of arriving and releasing biomass from a  storage site. The next input parameter that can have its value estimated by the IBSAL-MC model is DFSij. The in-field transportation time depends on the location of bales in the field. To simplify this in the optimization model, DFSij was estimated based on the distance between the center of the farm i and the storage site j. This distance can be estimated by the IBSAL-MC model accurately as the location of bales in the fields is known. The obtained results of the simulation model showed that the mass of bales will change over time in storage resulting in the need for more bales to be transported to the ethanol plant to meet the demand. To reflect this in the optimization model, the parameters LoadMassloader and LoadMasstruck can be estimated by the simulation model. The speed of stinger depends on the road conditions. As explained in section 3.9, the speed of stinger on field roads and public roads was assumed to be 8 kmh-1 and 25 kmh-1, respectively (Brummer et al., 2000). The average speed of stinger was used in the optimization model (50% on field roads and 50% on public roads). This parameter can be better estimated by knowing the percentage of travel on field roads and public roads. The IBSAL-MC model can be used in this regard. As discussed before, the amount of DMLs in storage depends on the duration of storage. In addition, the amount of machine DMLs depends on the moisture content of biomass at the time of operation. Thus, all the DML-related parameters in the optimization model can be estimated by the IBSAL-MC simulation model. Finally, another output of the simulation model which can be used in the optimization model, is the waiting time of trucks at the ethanol plant to be unloaded. To this end, Eq. (5.20) should be changed to  117  (  )  )  ̅ (5.34)  ̅ represents the average waiting time of each truck at the plant. This parameter affects the  transportation costs. In summary, the integration of these two models can improve the outputs of both models and results in the consistency of their results. More importantly, the integration can result in the determination of the design of the supply chain in which the daily biomass demand of the conversion facility is met at the minimum supply costs possible. In this regard, a new framework is proposed, which is the integration of both the IBSAL-MC model and the developed optimization model. Figure 5-2 illustrates the general structure of the integrated simulation/optimization framework. As shown in Figure 5-2, the simulation model provides the accurate values of input parameters for the optimization model and the optimization model prescribes the design of the supply chain as a decision rule for the simulation model.  Estimation of input parameters  The simulation model (Extendsim software)  Input Data Output Data  Spreadsheet (Excel software)  Input Data  The optimization model (AIMMS software)  Output Data  Design of the supply chain and biomass flow  Figure 5-2: General structure of the integrated simulation/optimization model  To run the integrated model, an initial list of farmers must first be prepared. The same farmers and supply radius used in the IBSAL-MC model was used as the initial supply area, i.e. a supply area with a radius of 160 km. Given the initial supply area in which the locations of potential farms to contract are known, the next parameter to be identified is the list of potential locations for storage. There are two ways to identify the candidate locations for storage: 1) continuous approach: any point in the  118  supply area is considered as a potential location for storage. This results in a large-sized optimization model which requires significant time to solve it. Moreover, any point cannot be a good candidate as storage must have access to public roads; and 2) discrete approach: roadside storage in each farm is considered as the potential candidate (Judd et al., 2010). This would lead to a significant reduction in the candidate points and thereby a smaller-sized optimization model. It is noted that most farmers have an area considered to be roadside area which could be used as storage. These roadside storage sites usually have access to roads. In this study, it is assumed that each considered farm within the supply area is a potential candidate for locating a storage site in the supply area. The sequence of steps to find the design of the supply chain is as follows: Step 1: Run the IBSAL-MC simulation model for the roadside storage scenario To start the loop shown in Figure 5-2, the IBSAL-MC model is first run for the roadside storage scenario to assure sufficient biomass and farms are available in the supply area to meet the biomass demand. In case of unfulfilled demands, the farmer participation rate or the supply radius or a combination of them will be incrementally increased until the daily demand would be met year-round. In the roadside storage scenario, each farmer stores the collected biomass at the roadside of the farm. In addition, all farms within the supply radius are considered potential suppliers. Thus, the design of the supply chain is pre-determined. This scenario was modeled in Chapter 4. Step 2: Estimate the input parameters in Table 5-1 based on the outputs of the IBSAL-MC model in step 1. Step 3: Run the optimization model Given the values of the input parameters in Table 5-1, the optimization model is run. Step 4: Run the IBSAL-MC model based on the prescribed design of the supply system in step 3 In step 1, the IBSAL-MC model was run based on the roadside storage scenario. Step 3 prescribes a new design with the optimal selection of farms, locations of storage sites as well as the optimal flow of biomass between farms, storage sites and the plant. In this step, the IBSALMC model uses the new design as a decision rule to schedule the operations. Since the IBSALMC model is run based on a new design, the daily demand may not be met. In such a case, the  119  amount of unfulfilled demand will be added to the right-hand side of Eq.(5.22) (demand constraints) in the optimization model and step 3 will be rerun to justify the design. If the demand is fully met, the outputs of the IBSAL-MC model are used to update values of the input parameters of the optimization model in step 5. Step 5: Run the optimization model based on the latest outputs of the IBSAL-MC model. The optimization model is rerun based on the latest updated values of the input parameters in Table 5-1. Step 6: Termination criteria The number of iterations to run steps 3-5 depends on the case study. The distribution of farms, the number of potential locations for storage, and the magnitude of variations in the input parameters affect the number of iterations. Depending on the time, resource and budgetary limits, different termination criteria can be used to terminate the iterations. For example, a computational time limit can be used to terminate the integrated model. Another example is a fairly close proximity of the total supply costs in both simulation and optimization models. In this research, the termination criterion is "whether the designs in the last two consecutive optimization runs are the same". It means that if the list of the selected farms, the location of the storage sites and the assignment of the selected farms to the storage sites are the same, the search for the optimal design terminates (Go to step 7). Otherwise, the integrated model will be rerun (Go to step 4). The iterative procedure continues until no improvement can be made in the design of the supply chain. Step 7: Record the final results The outputs of the last run of the optimization model and the IBSAL-MC models are recorded in the spreadsheet. Figure 5-3 presents the flowchart of the integrated simulation/optimization framework showing the steps of the proposed framework.  120  Start  Run the IBSAL-MC simulation model for the roadside storage scenario  Increment the farmer participation No rate/the supply radius  Is the daily demand met year-round? Yes  Estimate the input parameters in Table 5-1 based on the outputs of the IBSALMC simulation model  Add the amount of unfulfilled demand to the right-hand side of Eq.(5.22)  Run the optimization model  Run the IBSAL-MC model based on the latest design of the supply chain  No  Is the daily demand met year-round? Yes  Run the optimization model based on the latest outputs of the IBSAL-MC model  Are the designs the same in the last two consecutive optimization runs ?  No  Yes Record the outputs of both the IBSAL-MC simulation model and the optimization model in the Excel spreadsheet  End  Figure 5-3: Flowchart of the integrated simulation/optimization model  121  5.4 Comparison of the integrated model with the simulation and optimization models To show the efficiency of the integrated model, it was applied to the same case study in Chapter 4 and the obtained results were compared with those of the IBSAL-MC simulation model and the optimization model. In order to compare the results, the simulation model in the integrated model was run the same times as the IBSAL-MC model in Chapter 4. The t distribution was used to test the hypothesis that the means of a specific output variable differ in two models with the confidence interval of 90% (α=0.1). The obtained results at a 100% participation rate are reported in this section. The final design was determined by the integrated model after 12 iterations. The total delivery costs in different iterations are shown in Table 5-2. Delivery costs include the sum of farm-gate and hauling costs. The last column in Table 5-2 represents the difference between the calculated costs by the simulation and optimization models in each iteration.  Table 5-2: Total delivery cost ($/t) in different iterations of the integrated model Iteration  Total delivery cost ($/t)-  Total delivery cost ($/t)-  Difference  Simulation output  Optimization output  (%)  1  45.62  57.12  0.20  2  43.66  55.77  0.22  3  43.28  54.04  0.20  4  42.85  52.21  0.18  5  42.85  50.09  0.14  6  41.89  49.09  0.15  7  42.56  48.68  0.13  8  42.67  48.10  0.11  9  42.31  47.32  0.11  10  41.11  44.29  0.07  11  42.14  46.13  0.09  12  42.03  46.27  0.09 122  The delivery costs were reduced by 8% and 19% from iteration 1 to iteration 12 in the simulation model and optimization model, respectively. In addition, the cost difference between the simulation model and the optimization model was reduced by 54% from iteration 1 to iteration 12 showing the convergence of both models. The delivery costs in the optimization model are higher than those for the simulation model. This is because the optimization model considers sufficient biomass in the supply chain so that the daily demand could be met during the years with unfavourable conditions for biomass delivery such as low biomass yield. In addition, the aggregated values of input parameters are considered in the optimization model such as biomass yield and the efficiency of equipment. On the other hand, the reported costs for the simulation model are the average delivery costs of different years (simulation runs). Figure 5-4 depicts the trend of the reduction in the delivery costs in both the simulation and optimization models. The lowest delivery cost was observed in iteration 10 but the daily demand was not met for 12 days. Thus, the integrated model was rerun again considering the amount of unfulfilled daily demand at the right-hand side of Eq (5.22), as explained in the iterative procedure. Finally, in iteration 11 and 12, the same supply design was obtained resulting in the termination of the iterative procedure in the integrated model.  123  60  Total delivery cost ($/t)  50  40  30  Simulation_Outputs  Daily demand is not fully met for 12 days in iteration 10  Optimization_Outputs  20  10  0 1  2  3  4  5  6  7  8  9  10  11  12  Iteration  Figure 5-4: Total delivery cost in different iterations of the integrated model  In the next sections, the details of the comparison of the integrated model with the optimization model and the IBSAL-MC simulation model are provided. In this regard, the results of the first iteration and the last iteration are compared. As shown in Figure 5-3, the executed simulation model in the first iteration is the IBSAL-MC simulation model in Chapter 4. Thus, the obtained results in Chapter 4 are compared with those for the last iteration.  5.4.1 Number of farms and storage sites The comparison of the results with those of the optimization model showed a significant change in the number of selected farms and established storage sites. These numbers were 587 and 308 in the first iteration and 551 and 278 in the last iteration. These differences reveal the improvement in the results of the optimization model in the integrated model. The average number of farms and storage locations for both the IBSAL-MC model and the integrated model are shown in Table 5-3. The numbers of farms and storage sites in the integrated model were 15% and 57% fewer than those for the IBSAL-MC model. The significant difference between the number of storage sites is due to the assumption in the IBSAL-MC model 124  that farmers collect and store biomass at the roadside of their own farmland. This significant reduction shows the advantage of using satellite storage in the supply area. Farms and storage sites in the integrated model are located in a 100-km supply radius while they are located in a 115-km supply radius in the IBSAL-MC model. Due to the high contribution of the transportation costs in the total delivery cost, the integrated model selects the farms that have shorter distances from the ethanol plant resulting in a smaller supply area. Fewer numbers of farms in the supply area led to the reduction in the amount of purchased biomass and thereby the total harvested area. This amount is 358,708 t and 334,544 t in the IBSAL-MC model and the integrated model, respectively. The harvested area was reduced by 13%.  Table 5-3: Comparison of the supply area Model The IBSAL-MC model The integrated model  Total harvested area (ha) 348,875  Number of farms 648  Number of storage sites 648  Supply radius (km) 115  305,235  551  278  100  The decrease in the number of farms and storage locations led to the reduction in the number of created items in the simulation model resulting in the reduction in the computational time of the simulation model by 8 minutes, as shown in Table 5-4. In the integrated model, only a portion of the farms are released in the simulation model while in the IBSAL-MC simulation model, all the farms within the supply radius were released into the model. In addition, the number of storage sites in the integrated model is fewer than that for the IBSAL-MC simulation model resulting in a more efficient and faster flow of biomass in the supply chain.  Table 5-4: Average number of created items and the computational time Model The IBSAL-MC model The integrated model  Created farmlands 12,053  Created bales 663,671  Computational time of the simulation model in each run (Min) 20  11,427  651,954  12  125  It is noted that the average number of continuous variables, binary variables and the constraints in the optimization model were 1,631,783, 271,962 and 1,903,229, respectively. The average time to solve the optimization model was 2 minutes. A smaller supply radius in the integrated model resulted in the reduction of the required number of trucks in the supply area. The range of required trucks was 14-17 in the integrated model and 15-19 in the IBSAL-MC model. In addition, the average size of storage sites was increased from 540 t in the IBSAL-MC model to 770 t in the integrated model. Figure 5-5 illustrates the minimum, average and maximum daily delivery in the IBSAL-MC model and the integrated model. There is not a significant difference between the daily biomass deliveries. However, the differences between the minimum, average and maximum deliveries in the integrated model are fewer than those for the IBSAL-MC model showing more efficient flow of biomass in the integrated model. This is due to fewer numbers of farms, fewer numbers of large satellite storage sites, and thus better management of trucks.  1000  970  Daily delivery (t)  950  950 897  900 850  891 845  830  800 750 The IBSAL-MC model Minimum daily delivery (t)  The integrated model  Average daily delivery (t)  Maximum daily delivery (t)  Figure 5-5: daily biomass delivery  5.4.2 Supply costs Table 5-5 shows the supply costs per tonne in both models. The primary difference between these two models can be observed in the in-field hauling and road transportation costs. The difference in the in-field hauling costs stems from the in-field hauling distances. Figure 5-6 shows the average and maximum observed hauling distances in both models. As expected, the 126  in-field hauling distances increased in the integrated model. In the IBSAL-MC model, the hauling distance is the distance from the bale pick-up points on the farm to the storage site at the roadside of the respective farm while the hauling distances in the integrated model also incorporate the public roadways. This led to a 136% ($3.28 t-1) increase in the in-field hauling costs in the integrated model compared to the IBSAL-MC model. The average distance of 3.9 km is comparable to that of Resop et al. (2011), Ravula (2007) and Morey et al. (2010). They considered a hauling distance of 3.2 km.  Table 5-5: Supply costs ($/t) Logistics operation  Cost ($/t) The IBSAL-MC model  The integrated model  Baling  15.69  15.74  In-field hauling  2.42  5.70  Roadside/satellite storing  2.44  2.10  Total farm-gate cost  20.55  23.54  Loading  2.35  2.02  Road transporting  22.73  16.42  Total hauling cost  25.08  18.44  Truck weighing  0.90  0.90  Unloading  2.57  2.47  At-plant storing  0.75  0.52  Grinding  11.24  11.41  Total at-plant cost  15.46  15.30  Total supply cost  61.09  57.28  127  In-field Hualing distance (km)  16  15.2  Average hauling distance  14 12  Maximum hauling distance  10 8 5.6  6  3.9  4  2  1.5  0 The IBSAL-MC model  The integrated model  Figure 5-6: Average and maximum hauling distance  Contrary to the in-field hauling operation, the integrated model outperforms the IBSAL-MC model in the hauling operation. The hauling costs in the integrated model are 28% ($6.64 t-1) less than those for the IBSAL-MC model due to the smaller supply radius and road distances. The average road distance in the integrated model and the IBSAL-MC model were 49 and 73 km, respectively. In spite of more biomass at each storage site in the integrated model compared to that for the IBSAL-MC model, the reduction in the loading costs is not significant. The ethanol plant pays the hauling companies based on the amount of loaded biomass at the storage sites. The amount of available biomass at the storage sites in both models is not sufficient to keep the equipment busy year-round. If the ethanol plant owned the loading equipment, the savings would be significant. The ownership costs in the loading operation were $7.84 t-1 and $11.32 t-1 in the integrated and the IBSAL-MC models, respectively. For other operations, working conditions are the same in both models resulting in slight cost differences. For example, the baling operation takes place in similar field conditions in both models. The same is true for at-plant operations. On average, the total supply cost per tonne in the integrated model is 6% less than that of the IBSAL-MC model. In addition, the amount of purchased biomass in the integrated model was 24,164 t less than that for the IBSAL-MC model resulting in a 1% reduction in the annual supply cost. If the ethanol plant were to own the loaders, the savings would be 12%. 128  Figure 5-7 depicts the distribution histogram of the total supply cost for the integrated model. The total supply cost would be in the range of $50.29 t-1 and $62.16 t-1 with 90% confidence interval. In addition to testing the difference between the means of the total supply cost in both models, the difference between their variances was also tested with the confidence interval of 90% (α=0.1). The result of the tests was the rejection of the equality of the means. However, the equality of the variances was not rejected.  0.12  Mean: 57.28 Standard deviation: 3.6  0.1 0.08 0.06  0.04 0.02 0 48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  Total supply cost ($/t)  Figure 5-7: Total supply cost ($/t) in the integrated model  5.4.3 Energy input and the associated emitted CO2 The average energy input and the emitted CO2 in both models are shown in Table 5-6. As expected, the energy input and the emitted CO2 in the in-field hauling operation in the integrated model are more than those of the IBSAL-MC model due to the greater in-field hauling distances. In contrast, they are less in the road transportation operation due to the shorter distances between storage sites and the ethanol plant in the integrated model. For the rest of the operations, there is not a significant difference between the models. Overall, the energy input and the emitted CO2 in 30 the integrated model are 19% and 12% less than those for the IBSAL-MC model.  129  Based on the total consumed energy in the integrated model, 11% of the produced ethanol in the plant would be consumed in the supply chain which is slightly different than that (13%) of the IBSAL-MC model. Table 5-6: Energy input and emitted CO2 Energy input (MJt-1) Operation  IBSAL-MC  The integrated  CO2 (kgt-1) IBSAL-MC  model  The integrated model  Baling  102.31  102.35  7.02  7.02  In-field hauling  52.65  89.57  3.59  6.84  Loading  23.21  22.67  1.58  1.55  Road transporting  471.82  254.38  30.34  18.76  Unloading  23.74  23.09  1.63  1.59  Grinding  269.38  268.83  27.07  27.11  Total  943.11  760.89  71.23  62.87  5.5 Sensitivity analysis The design of the supply chain is prescribed by the integrated model while considering the dynamics and stochasticity of the supply chain. Thus, the developed design should be robust against the changes in the input parameters of the supply system. To investigate this issue, the sensitivity analysis was carried out to measure the magnitude of the impact of the input parameters on the supply design and the total supply cost. In this study, ±10% range of variation from the base value (average value) was considered in the sensitivity analysis for all the input parameters. The reason for the selection of this range is twofold: 1) This range of variation occurs for all the considered input parameters based on their distribution functions, and 2) Consideration of the same range of variation makes it possible to compare the impact of input parameters on the performance of the system. Figure 5-8 shows the sensitivity of the design of the supply chain to the input parameters. The first finding of the sensitivity analysis was that only a few parameters impact the design of the supply chain. The most influential parameter is biomass yield. A ten percent decrease in the 130  biomass yield in the entire supply chain resulted in the contracting of 90 more farms and establishment of 17 more storage sites. The significant increase in the number of farms did not lead to the significant increase in number of storage sites. This is due to the close proximity of the farms in the supply area. The distance of 85% of farms from their respective storage sites in the prescribed design was smaller than 5 km. Another influential parameter on the number of contracted farms is DML which emphasizes the importance of DML management in the agricultural biomass supply chain.  Biomass Yield (t/ha) Cost of stinger ($/hr) Stinger efficiency (%) -10%  Bale bulk density (kg/m3)  10% Loading efficiency (%) Cost of loader ($/hr) -10%  -5%  0%  5%  10%  Change in number of storage sites  Biomass Yield (t/ha) Plant DMLs (%) -10%  Baling DMLs (%)  10%  Storage DMLs (%) -25%  -15%  -5%  5%  15%  25%  Change in number of selected farms  Figure 5-8: Sensitivity of the supply design to the input parameters  131  Ekşioğlu et al. (2009) also concluded that biomass availability is one of the primary parameters affecting the supply chain-design decisions. Their results also indicated that the supply chain-design decisions are not affected by changes in biomass collection costs and biomass processing costs. Figure 5-9 shows the key parameters that affect the total supply cost. Among the input parameters, the bale bulk density has the highest impact on the total supply cost. This impact is expected as the bale bulk density affects the efficiency of all hauling and storage operations in the supply chain, mainly the transportation operation. As discussed before, the bale bulk density impacts the number of delivered truckloads and thereby the number of trucks.  Bale bulk density (kg/m3) Custom cost of truck ($/hr) Stinger efficiency (%) Custom cost of stinger ($/hr) Truck efficiency (%) Biomass yield (t/ha) Storage DMLs (%) Grinder efficiency (%)  -10%  Custom cost of trailer ($/hr)  10%  Baling capacity (tonne/hr) Cost of grinder ($/hr) Baling efficiency (%) Loading efficiency (%) Baling DMLs (%) -10%  -8%  -6%  -4%  -2%  0%  2%  4%  6%  8%  10%  Change in the total supply cost  Figure 5-9: Sensitivity of the total supply cost to the input parameters  Figure 5-9 also shows that the transportation-related parameters (both in-field transportation and road transportation) including custom cost and efficiency of truck and stinger are the next set of parameters impacting the total supply cost. Among the downstream operations, efficiency and cost of grinding operations have the highest impact. 132  In terms of DMLs, the baling and storage operations have the highest impacts on the total supply cost compared to the other operations. Although the amount of DMLs in baling operation is around twice the DMLs in storage sites, the impact of storage DMLs on the total supply cost is more than four times the baling DMLs. Since biomass at the storage sites is a more value-added material, the loss of biomass at the storage sites has a higher impact on the total supply cost compared to the loose biomass laid on the field. As shown in Figure 5-8, biomass yield is the most influential parameter on the design of the supply chain. However, it has less impact on the total supply cost compared to the design. This is due to the close proximity of farms and storage sites. Thus, change in biomass yield does not lead to significant change in hauling distances and the associated costs. The obtained results of the sensitivity analysis on the supply costs are comparable to those of Hess et al. (2009) and Judd (2011). The annual biomass demand was not considered in the sensitivity analysis since it was assumed that the production capacity of the conversion facility at the ethanol plant is fixed during its business life. Similar to Chapter 4, the sensitivity of the integrated model to the farmer participation rate was evaluated. The integrated model was executed at 50% and 25% participation rates. The integrated model was infeasible at a 25% rate due to lack of sufficient biomass to meet the annual biomass demand. Due to the significant amount of DMLs in storage sites, the possibility of the fulfillment of biomass demand at a 25% rate was investigated by changing the storage regime. The storage regime was changed from outside storage unprotected on the ground to the outside storage on crushed rock. As shown in Table 4-1, the annual DMLs percentage in this storage regime is 15%. The results of the integrated model showed that the daily demand would be met at 25% rate with outside storage on crushed rock. The total DML in the storage sites was 15,984 t which is 20% less than that for the integrated model with outside storage unprotected on the ground. The percentage of saved biomass was used to meet the biomass demand. Table 5-7 shows the impact of the farmer participation rate on the outputs of the integrated model. There is not a significant difference between the number of farms and storage sites at different participation rates. However, the number of farms increases with the increase in the farmer participation rate. The optimization model preferably selects farms that are closer to the ethanol plant. The number of small farms closer to the plant increases with the increase in the participation rate resulting in more selection of farms to meet the demand. The number of farms 133  that are smaller 64 ha (160 ac) was 14, 20 and 42 at 25%, 50% and 100% participation rates. In addition, farmers at 50% and 25% rates are more distributed compared to 100% rate resulting in greater number of storage sites established to cover farms.  Table 5-7: Impact of farmer participation rate on the outputs of the integrated model Outputs  Farmer participation rate 100%  50%  25%  Number of selected farms  551  541  534  Number of established storage sites  278  285  292  Supply radius (km)  100  140  160  50.29-62.61  58.12-69.25  68.61-78.04  Total supply cost (90% confidence ($t-1))  5.6 Discussion and conclusions To improve the performance of the supply chain in the case study, an optimization model was developed and integrated with the IBSAL-MC model. The optimization model is a tactical decision-making model with a five-year time horizon. It prescribes the design of the supply chain and the annual flow of biomass from farms to storage sites and from storage sites to the plant. The IBSAL-MC simulation model exploits this design and biomass flow to plan and schedule the operations on a daily basis. On the other hand, the outputs of the simulation model are used as the inputs of the optimization model to adjust the design. Thereafter, the IBSAL-MC model is rerun based on the adjusted design. This iterative procedure continues until no improvement can be made in the design. The primary advantage of the proposed integrated model is to provide feasible and consistent solutions at both tactical and operational levels due to the connection between two levels. The feedback sent from the operational level to the tactical level would result in the adjustment of the design of the supply chain based on the dynamics and the stochasticity of the supply chain. On the other hand, the design and flow of biomass determined by the optimization model is used by the simulation model in order to deliver biomass at the lowest costs possible. The integrated model was applied to the case study and the final design was determined after 12 iterations. The 134  obtained results were compared with those for the IBSAL-MC simulation model and the optimization model. The comparison revealed the superiority of the integrated model over the optimization and simulation models. The main findings of the comparison are as follows:   The total delivery costs were reduced by 8% and 19% from iteration 1 to iteration 12 in the simulation model and optimization model, respectively. In addition, the cost difference between the simulation model and optimization model was reduced by 54% from iteration 1 to iteration 12.    The number of selected farms and the established storage sites was reduced by 6% and 10% in the first and the last iterations, respectively. This shows the improvement in the results of the optimization model in the integrated model.    The integrated model led to a 15% (97 farms) and 57% (370 storage sites) reduction in the number of selected farms and the established storage sites compared to the IBSAL-MC model, respectively. In addition, the computational time of the simulation time at each run was reduced by 40% (8 min).    The amount of biomass purchased from farmers in the integrated model was reduced by 7% (24,164 t) compared to the IBSAL-MC model.    The harvested area in the integrated model was reduced by 13% (43,640 ha) compared to that of the IBSAL-MC model.    The supply radius determined in the integrated model was 13% (15 km) smaller than that for the IBSAL-MC model. This led to the reduction in the number of maximum trucks from 19 in the IBSAL-MC model to 17 in the integrated model.    The average capacity of the storage sites in the integrated model was increased by 43% (230 t).    The average hauling distance between farms and storage sites was increased from 1.5 km in the IBSAL-MC model to 3.9 km in the integrated model. In contrast, the average road distance between storage sites and the plant was decreased from 73 km in the IBSAL-MC model to 49 km in the integrated model.    The average and standard deviation of the total supply cost in the integrated model were $57.28 t-1 and $3.6 t-1.    The integrated model resulted in 6% ($4 t-1) savings in the total supply cost. The savings could be as high as 12% ($7 t-1) if the ethanol plant were to own the loading equipment. 135    The energy input and the emitted CO2 in the integrated model were 19% (181.6 MJ t-1) and 12% (84 kg t-1) less than those for the IBSAL-MC model.    The amount of the ethanol consumed in the supply chain in the integrated model was 11% of the ethanol produced by the plant.    The results of the sensitivity analysis revealed that a few parameters affect the design of the supply chain. The most influential parameter was biomass yield.    Bale density and custom cost and efficiency of in-field and road transportation operations had the highest impact on the total supply cost compared to other parameters. All these parameters are equipment-related parameters rather than supply chain-related ones.    Daily biomass demand would be met year-round at a 25% farmer participation rate if a more protective storage regime such as outside storage on crushed rock is employed instead of outside storage unprotected on the ground. Under such conditions, 20% less biomass (4,000 t) would be lost which can used to meet the demand.    Change in the farmer participation rate would not significantly impact the number of selected farms and the storage sites as these numbers mainly depend on the annual biomass demand and distribution of farms. However, it significantly affected the supply radius and the total supply cost. The average supply costs at 50% and 25% rates are 12% ($7.1 t-1) and 28% ($16.31 t-1) more than those of a 100% participation rate. In summary, the integrated model led to the improvements in the performance of the supply  chain compared to the IBSAL-MC model. The improvements were due to a smaller supply radius, fewer numbers of farms, fewer numbers of larger satellite storage sites, and thus, more efficient of flow of biomass in the supply chain.  136  Chapter 6. Conclusions, strengths, limitations and future research 6.1 Conclusions The apparent large quantity of crop residues on the Canadian Prairies such as cereal straw can be used as feedstock for bioenergy production like that of cellulosic ethanol. However, there is only one demonstration plant in Canada exploiting cereal straw to produce cellulosic ethanol. One of the obstacles restricting the development of this emerging industry is the inefficiency in the biomass supply chain (Fales et al., 2007). Due to the dynamics and high stochastic nature of the supply chain, a steady supply of biomass during the lifetime of the ethanol plant is highly uncertain. This uncertainty and complexity increases in the Canadian Prairies due to factors such as frequent crop rotation, weather conditions, short and variable harvest seasons, wide distribution of farms and variable yields. The overall goal of this research was to design and schedule a highly constrained agricultural biomass supply chain to meet the daily biomass demand of a commercial-sized cellulosic ethanol plant at the minimum delivery cost possible. In Chapter 2, the decision-making tools developed in the literature to model and evaluate the biomass supply chain were reviewed. The review of literature revealed that it lacks an integrated modeling approach to explore and analyze a highly constrained agricultural biomass supply chain similar to the one in the Canadian Prairies, at both the tactical and operational planning levels. The tactical planning level concerns the design of the supply chain to meet the annual biomass demand at the minimum delivery cost. In the operational level, the logistics operations are scheduled on a daily basis to meet the daily demand of biomass. This calls for an integration of an optimization approach with the dynamic modeling of biomass flow. Therefore, an integrated simulation/optimization model was developed. Given the location and the annual biomass demand of the ethanol plant, the developed integrated model prescribes the design of the supply chain in a way that the daily demand is met with the minimum delivery cost. The design of the supply area refers to the selection of farms to purchase biomass from, the  137  number and location of storage sites as well as the assignment of the selected farms to the storage sites The framework of the developed simulation model was explained in Chapter 4. The developed simulation model is used to plan and schedule the logistics operations on a daily basis (operational planning) with respect to the constraints, time-dependency and stochasticity of the system. The developed optimization model is a mixed-integer linear programming model. It determines the design of the supply chain at the tactical level. It also determines the annual flow of biomass between farms, storage sites and the ethanol plant. The details of the developed optimization model were discussed in Chapter 5. In Chapter 5, an iterative procedure was also developed to integrate the simulation and optimization models. In this procedure, the design at the tactical level is used in the simulation model to manage the flow of biomass from farms to storage sites and from these storage sites to the plant. Additionally, the outputs of the simulation model are used as the inputs of the optimization model to adjust the supply design. The iterative procedure continues until no improvement can be made in the supply design. The connection between the simulation and optimization models leads to feasible and consistent solutions at both tactical and operational planning levels. The developed integrated model provides both model detail and data detail. The details of the supply chain such as moisture content, dry matter loss, delays, and machine breakdowns are incorporated in the simulation model. This model simulates the daily inflow and outflow of biomass in the storage locations and finds storage-related solutions including the size of the storage sites, their inventory, DMLs and costs. The model also finds the required initial inventory to meet the demand during the harvest season. In addition to the storage analysis, the simulation model provides a detailed analysis of the transportation system. The developed optimization model calculates accurate cost figures by considering the price of biomass, the DML costs, the value of lost biomass at different stages of the supply chain and the details of the operating costs for machines and storage. In terms of data detail, the data were gathered at the farm level including their number, location, size and yield. Other data included the soil conservation rate, livestock requirements, initial moisture content, daily weather data and the weekly harvest schedule. 138  The model and data details enable the feedstock manager to make informed decisions in practical cases. To show the efficiency of the integrated model, it was applied to a case study. The case study was a proposed ethanol plant located 12-km east of Prince Albert, Saskatchewan. The plant would produce more than 70 million liters (ML) of cellulosic ethanol per year. The daily biomass demand of the conversion facility would be 750 t of wheat straw. To procure this amount of wheat straw, a 160-km supply radius was considered by Iogen Corporation. The case study was presented in Chapter 3. The obtained results of the integrated model for the case study revealed that the daily biomass demand can be met year-round at 100% and 50% farmer participation rates. In contrast, 98% of daily demand would be met at a 25% rate. Using a more protective storage regime, such as outside storage on crushed rock instead of outside unprotected on the ground, can lead to the fulfillment of the daily demand at a 25% rate by reducing the amount of dry matter loss. The supply radius would be 100, 140 and 160 km at 100%, 50% and 25% farmer participation rates. Due to the smaller supply radius, the minimum supply cost would happen at a 100% farmer participation rate. The total supply cost was in the range of $50.29 t-1 and $62.16 t-1 with a 90% confidence interval. The average supply cost was $57.28 t-1 at 100% rate. The average supply cost at 50% and 25% rates was 12% ($7.1 t-1) and 28% ($16.31 t-1) more than that for the 100% rate. The number of farms and storage sites at the 100% rate were 551 and 278, respectively. These farms cover 305,235 ha and collect 334,544 t of wheat straw. The results of the sensitivity analysis revealed that a few parameters impact the design of the supply chain. The most influential parameter was biomass yield. This implies that the ethanol plant should negotiate with extra farms to provide biomass for the plant in case of biomass shortage due to the low yield in some years. The number of extra farms can be determined by the integrated model as the model incorporates the yield variability. In addition, bale bulk density and in-field and road transportation operations have the highest impacts on the total supply cost compared to other input parameters. All these parameters are equipment-related parameters rather than supply chain-related ones. Thus, improvement in the technology of agricultural and industrial equipment used in the supply chain would lead to significant reduction in the supply costs. For instance, more advanced baling equipment with 139  higher capability of picking up biomass from the ground would lead to the significant reduction in baling DMLs. In addition, the creation of denser bales would reduce the supply costs, as shown in the sensitivity analysis. The comparison of the results of the integrated model with both the optimization model and the simulation model showed the superiority of the integrated model. The numbers of selected farms and the established storage sites in the integrated model were reduced by 6% and 10% compared to the optimization model. Compared to the simulation model, the integrated model led to a reduction in number of farms (15%), number of storage sites (57%), amount of purchased biomass from farmers (7%), harvested area (13%), supply radius (13%), number of maximum trucks (2 trucks), supply costs (6-12%), energy input (19%), and emitted CO2 (12%). The obtained results in this study indicated that roadside storage is not an attractive option at the commercial scale. Having a smaller number of large satellite storage systems instead of a greater number of roadside storage sites improves the flow of biomass in the supply chain. However, the location of satellite storage facilities should be determined in a way that farmers do not incur high in-field hauling costs. The average hauling distance of 3.9 km was suggested by the integrated model. Another advantage of the satellite storage system is that this system facilitates the contract process as the ethanol plant will only contract farmers who are responsible for the establishment of storage locations rather than each individual farmer. A review of the relevant literature reveals that the discussion on whether to enhance the storage regime has been restricted to the tradeoff between the DMLs-related costs and the construction costs. The obtained results of this study showed that other benefits of the enhancement of the storage regime should also be considered. For example, the replacement of an outside and unprotected on-ground storage regime with an outside, unprotected on-crushedrock storage regime would reduce the DMLs by 20% (4000 t). The potential benefits of this reduction are: needing contracts with fewer farmers, having fewer storage sites to establish, needing a smaller supply radius and thereby creating a more efficient flow of biomass in the supply chain. Therefore, the impact of the change in the storage regime on the entire supply chain should be taken into account rather than considering only the tradeoff between the DMLsrelated costs and the construction costs.  140  Since the ethanol plant has to face different farmer participation rates during the business life of the conversion facility, the outputs of the developed models can assist the feedstock managers to evaluate the performance of the supply chain under different participation rates. The proposed integrated model provides insights to enhance other aspects of the agricultural biomass supply chain which are beyond the scope of this study. One of the aspects of the supply chain, which needs to be improved, is the business model of the supply chain. The obtained results of this study showed that the amount of DMLs in the supply chain could be more than 20% of the annual biomass demand. Thus, one of the issues that needs to be addressed in the business model is how the DMLs-related costs should be shared between the supply actors. For example, bales are usually stored at the roadside/satellite storage sites for a long period of time resulting in the occurrence of significant DMLs. If the ethanol plant pays the farmers only based on the amount of biomass transported from farmers, the farmers have to incur the DMLs-related costs. This can make the farmers unwilling to supply biomass in the long run. The cost associated with DMLs should be shared between the supply actors. This can be specified in both the farm-gate and hauling contracts by the ethanol plant. The daily analysis of the logistics operations by the integrated model can help farmers and hauling companies to increase the productivity of their equipment by sharing them between farms and storage sites. Farmers can rent the equipment from the adjacent farmers. This can be done once the farmers do not need the equipment due to termination of their harvest season or termination of field operations on their own lands. If the field operations are performed by custom harvesting companies, similar information can be shared with them. In addition, the loading equipment can be shared between storage sites by the hauling companies. Once the inventory of a storage site exhausts, the loading equipment can be moved to another storage site. Another aspect of the supply chain that can be improved is the location of the ethanol plant. Although the location of the plant was definite in this study and used as an input in the developed models, the outputs of the models can be used to adjust the location of the established plan, if possible. For instance, the obtained results showed that 51% and 27% of annual truckloads are delivered from 60-85 km distance range in 100% and 50% farmer participation rates, respectively. The establishment of the ethanol plant at this range would reduce the transportation costs.  141  One of the outputs of this study was the quantities of the emitted CO2 in the entire supply chain. This output can be used in a life cycle analysis (LCA) study to evaluate the environmental aspects of the development of biorefineries in local communities. The reasonable computational time to solve the integrated model for a commercial-scale ethanol plant makes it a practical model to apply in real-world case studies. It is noted that each of the simulation and optimization models can be used independently depending on the decisions that the feedstock manager intends to make. In summary, the following factors have the highest impact on the cost-efficiency of the agricultural biomass supply chain:   Farmer participation rate    Selection of farms and location of storage sites    Type of storage system including roadside and satellite storage    Type of storage regime such as outside unprotected on crushed rock    Capacity of storage sites and their inventory, especially initial inventory    Sharing field equipment between farms and loading equipment between storage sites    Improvement in baling operation    Improvement in the transportation system including selection of efficient loading and unloading operations, the transportation mode, and the dispatching method  6.2 Strengths and limitations of the study The primary strength of this study is the detailed investigation of the agricultural biomass supply chain for a commercial-scale biorefinery plant. The dynamics, stochastic nature, and constraints of the supply chain were taken into account. The consideration of these factors resulted in accurate cost figures and daily delivery schedule. The thorough investigation of the supply chain also identified areas for improvement of the daily delivery and potential cost reductions. In addition, it provides meaningful information to the decision makers in practical cases. The developed models can be applied to other bioenergy products such as heating and electricity as long as the same feedstock type (cereal straw) and supply scenario (conventional baling system) are considered. The changes in the feedstock type and/or the supply scenario call for the modifications in the developed models, mainly the IBSAL-MC simulation model. 142  However, there is no limitation on the application of the general structure of the proposed integrated approach on different feedstock types, supply scenarios and bioenergy products. The primary limitation of the study was the lack of the fine spatial data for the region under study. Thus, the locations and sizes of farms were created randomly. The availability of spatial data on farm or rural municipality levels would lead to more accurate outputs, especially those of the biomass delivery costs. Similar to the majority of the studies in the literature, the data used in the developed model include several assumptions, such as the percentage of the annual dry matter loss in storage, biomass to grain ratio, grain moisture content and standard density of grain. The values of these parameters depend on the geographical, biological and weather conditions in the region under study. The estimation of such parameters would require extensive field experiments. Although this study attempted to estimate the values of such input parameters from other regions and studies, the availability of such data from the region under study could improve the obtained results of the developed models. It is noted that the obtained results of the developed models for the agricultural biomass supply chain are region-based. Thus, the results such as the delivery costs and the amount of energy input and CO2 may change in other regions. The obtained results may also change with the change in assumptions and the structure of the supply chain. Another limitation relates to the integrated model. In this research, the termination criterion was to obtain the same design in the last two consecutive optimization runs. This termination criterion does not necessarily lead to the global optimal solution. In addition, the application of the integrated model with this criterion to other real-life cases may not result in the termination of the iterative procedure in a reasonable computational time. The characteristics of the supply area such as the distribution of the farms and the range of variations in the variable inputs affect the number of iterations to meet the termination criteria. Thus, this criterion may need to be adjusted based on the case study. Two other forms of termination criterion have been used in the literature (Fu (2002)): 1) termination after a maximum number of iterations due to the time limit, and 2) a pre-determined threshold level of improvement in the solutions. Almeder et al. (2009) and Lee and Kim (2002) used the latter termination criterion. Almeder et al. (2009) continued the iterative procedure in their developed integrated model until the 143  difference between solutions in subsequent iterations was small enough. The application of their model to several numerical examples showed the fast convergence of the objective values of the simulation and optimization models. However, they stated that the general convergence in all cases needs to be proved. Lee and Kim (2002) used a similar termination criterion as in Almeder et al. (2009). They stopped the search for final solutions when the rate of difference between solutions in the iterations was small enough (5%). They considered the final solution as the realistically optimal solution. Although Almeder et al. (2009) and Lee and Kim (2002) called the final solutions of their models a good approximation of optimal solutions and  realistically optimal solutions,  consideration of this stopping criterion does not guarantee that the final solutions are the global optimal ones. Almeder et al. (2009) concluded that for complex supply chain models, other methods such as metaheuristics can be considered to improve the quality of the final solutions. Truong and Azadivar (2003) and Azadivar and Tompkins (1999) integrated genetic algorithm with simulation and optimization modeling to enhance the convergence of the solutions in the integrated model. However, the employment of heuristics and metaheuristics in an integrated simulation/optimization model does not necessarily lead to finding the global optimal solution. For instance, the integrated model developed by Azadivar and Tompkins (1999) finds a satisfactory solution. The efficiency of their model was shown by applying it to three test problems. However, the efficiency of their model on finding the optimal solution was not generally assured. Similar to the previous research, the convergence of the simulation and optimization models in an integrated approach was not generally proved. It is noted that the case study was a proposed plant and there was no operating commercialscale cellulosic ethanol plant in the world. Thus, there was no real data to compare to the obtained results of the integrated model. The obtained results were compared with the feasibility study conducted by Iogen Corporation and similar studies in the literature. In addition, the results were approved by the experts. A sensitivity analysis was also carried out to evaluate sensitivity of the obtained results to the small changes in the input parameters.  144  6.3 Future research The proposed integrated approach should be applied to other real-life cases to find out if the iterative procedure terminates in a reasonable computational time. In addition, it should be tested with different termination criteria. Another direction for future research is to prove the convergence of the objective functions of simulation and optimization models in the integrated approach. In case of the availability of the GIS data for the region under study, the obtained results can be improved by considering the real location of farms instead of their random creation. In addition, distances between farms, storage sites and the ethanol plant can be estimated more accurately instead of using the center of farms and storage sites to estimate the travelling distances. In this study, the location and capacity of the ethanol plant were assumed to be predetermined. Thus, this study can be extended to find the optimal type, number, size and location of bioenergy plants. This extension leads to the full integration of strategic, tactical and operational decision-making levels under a hierarchical framework. The obtained results of this study showed that the average amount of available biomass at farms and storage sites is not enough to keep the equipment busy throughout the year. Thus, the developed models in this study can be extended by sharing the field equipment among farms and loading equipment among storage sites. Efficient sharing of equipment can provide more cost saving. Since the movement of loading equipment between storage sites impacts the truck dispatching, the loading assignment between storage sites and the truck dispatching rule should be optimized together to decrease the total hauling costs. The developed IBSAL-MC simulation model can be extended in a number of ways. The IBSAL-MC model schedules the supply chain for only one commercial-sized plant. In case of the establishment of several small or medium-sized plants in the region, the model should be extended to meet the demand of several plants. The next extension is the consideration of correlations between the input parameters. For example, the projection of the biomass yield based on the weather data, soil type and tillage practices would improve the quality of the input data and their analysis. In addition, the impacts of the weather conditions on the performance of  145  field operations can be modified by keeping track of the moisture balance in the soil to check the suitability of the soil for machinery traffic.  146  References Agriculture and Agri-Food Canada (2003). Tillage Practices that Reduce Soil Erosion. Accessed from http://www4.agr.gc.ca/AAFC-AAC/display-afficher.do?id=1187362338955&lang=eng in 2010. Akgul, O., Shah, N., and Papageorgiou, L.G. (2012). An optimisation framework for a hybrid first/second generation bioethanol supply chain. Computers & Chemical Engineering, 42, 101114.. Allen, J., Browne, M., Hunter, A., Boyd, J., and Palmer, H. (1998). Logistics management and costs of biomass fuel supply. International Journal of Physical Distribution & Logistics Management 28, 463–477. Almeder, C., Preusser, M., and Hartl, R. (2009). Simulation and optimization of supply chains: alternative or complementary approaches? OR Spectrum 31, 95–119. Altman, I.J., Sanders, D.R., and Boessen, C.R. (2007). Applying Transaction Cost Economics: A Note on Biomass Supply Chains. Journal of Agribusiness 25, 107-114. Arinze, E.A., Sokhansanj, S., Besant, R.W., Wood, H.C., and Schoenau, G.J. (2001). Effects of material and environmental conditions on caking and breakage of potash fertilizer products during storage, shipment and handling. The International Journal of Storing, Handling, and Processing Powder 13, 45–54. ASABE (2007a). Standards. ASAE EP496.3 FEB2006, agricultural machinery management, 356_361. ASABE (2007b). Standards. ASAE D497.5 FEB2006, agricultural machinery management data, 362_369. Axenbom, Å. (1990). A simulation model for planning of hay harvesting machinery systems and management. Rapport - Sveriges Lantbruksuniversitet, Institutionen För Lantbruksteknik (02830086). Azadivar, F., and Tompkins, G. (1999). Simulation optimization with qualitative variables and structural model changes: A genetic algorithm approach. European Journal of Operational Research 113, 169–182. Benock, G., Loewer, O.J., Bridges, J.T., and Loewer, D.H. (1981). grain flow restrictions in harvesting-delivery drying systems. Transactions of the ASAE 24, 1151–1161. Berruto, R., and Busato, P. (2008). System approach to biomass harvest operations: simulation modeling and linear programming for logistic design. In ASABE Annual International Meeting, Rhode Island, Paper Number: 084565.  147  Blunk, S.L., Yore, M.W., Summers, M.D., Lau, G.K., Tang, S.T., and Jenkins, B.M. (2003). Quality and property changes in rice straw during long term storage. ASAE Annual International Meeting, Nevada, USA. Boyden, A. (2001). Reliable data on sustainable wheat straw availability - straw production. Final Report Project Number 500E March 2001. ADF #20000117. Humboldt, SK: Prairie Agricultural Machinery Administration (PAMI). Brown, L.R. (2010). World on the edge: how to prevent environmental and economic collapse. Earth Policy Institute. Brownell, D.K., and Liu, J. (2010). Computer Model of Satellite Storage Locations for Herbaceous Biomass Handling. American Society of Agricultural and Biological Engineers, 2010 ASABE Annual International Meeting, Paper No: 1009058. Brummer, E.C., Burras, C.L., Duffy, M.D., and Moore, K.J. (2000). Switchgrass production in Iowa: economic analysis, soil suitability, and varietal performance. Iowa State University, Ames, Iowa. Brundin, S. (1988). Fuels from agriculture. Cost calculations of systems based on straw and energy grass. Department of Economics, Swedish University of Agricultural Sciences. Canadian Biomass Magazine (2009). Advanced biofuels. Accessed from http://www.canadianbiomassmagazine.ca/index.php?option=com_content&task=view&Itemid=1 32&id=1261 in 2009. Canadian Renewable Fuels Association (2010). Growing Beyond Oil Delivering Our Energy Future. Report Card on the Canadian Renewable Fuels Industry. Carolan, J.E., Joshi, S.V., and Dale, B.E. (2007). Technical and Financial Feasibility Analysis of Distributed Bioprocessing Using Regional Biomass Pre-Processing Centers. Journal of Agricultural & Food Industrial Organization 5, Article 10. Chen, C.-W., and Fan, Y. (2012). Bioethanol supply chain system planning under supply and demand uncertainties. Transportation Research Part E: Logistics and Transportation Review 48, 150–164. Clegg, J.M., and Noble, D.H. (1987). A cost model of straw conservation systems. Agricultural Engineer 42, 15–20. Cundiff, J.S., Dias, N., and Sherali, H.D. (1997). A linear programming approach for designing a herbaceous biomass delivery system. Bioresource Technology 59, 47–55. Cundiff, J.S., Fike, J.H., Parrish, D.J., and Alwang, J. (2009a). Logistic constraints in developing dedicated large-scale bioenergy systems in the Southeastern United States. Journal of Environmental Engineering 135, 1086–1096.  148  Cundiff, J.S., Grisso, R.D., and Judd, J. (2009b). Operations at satellite storage locations (SSL) to deliver round bales to a biorefinery plant. ASABE Paper No 95896. DeMol, R.M., Jogems, M.A.H., Van Beek, P., and Gigler, J.K. (1997). Simulation and optimization of the logistics of biomass fuel collection. Netherlands Journal of Agricultural Science 45, 219–228. Ebadian, M., Sowlati, T., Sokhansanj, S., Stumborg, M., and Townley-Smith, L. (2011). A new simulation model for multi-agricultural biomass logistics system in bioenergy production. Biosystems Engineering 110, 280–290. Ekşioğlu, S.D., Acharya, A., Leightley, L.E., and Arora, S. (2009). Analyzing the design and management of biomass-to-biorefinery supply chain. Computers & Industrial Engineering 57, 1342–1352. Elderen, E. van (1987). Scheduling farm operations: a simulation model. Institute of Agricultural Engineering, Wageningen, The Netherlands 218pp. ESRI (2009). Understanding raster data. Accessed from http://webhelp.esri.com/arcgisdesktop/9.3/index.cfm?TopicName=What%20is%20raster%20data %3F in 2011. Fales, S.L., Hess, J.R., Wilhelm, W., Erbach, D., Provine, W.D., Vogel, K.P., Peterson, T.A., and Runge, E.C.A. (2007). Convergence of Agriculture and Energy: II. Producing Cellulosic Biomass for Biofuels. Publications from USDA-ARS / UNL Faculty. Floden, S. (1994). Cost calculations of handling big straw bales. Swedish Institute of Agricultural Engineering. Freppaz, D., Minciardi, R., Robba, M., Rovatti, M., Sacile, R., and Taramasso, A. (2004). Optimizing forest biomass exploitation for energy supply at a regional level. Biomass and Bioenergy 26, 15–25. Fu, M.C. (2002). Feature Article: Optimization for simulation: Theory vs. Practice. INFORMS Journal on Computing 14, 192–215. Graham, R.L., Liu, W., Jager, H.I., English, B.C., Noon, C.E., and Daly, M.J. (1996). A regional-scale GIS-based modeling system for evaluating the potential costs and supplies of biomass from biomass crops. Proceedings of the 9th European Bioenergy Conference, Copenhagen (DK). Graham, R.L., Liu, W., Downing, M., Noon, C.E., Daly, M., and Moore, A. (1997). The effect of location and facility demand on the marginal cost of delivered wood chips from energy crops: A case study of the state of Tennessee. Biomass and Bioenergy 13, 117–123. Hess, J.R., Kenney, K.., Ovard, L.., Searcy, E.., and Wright, C.. (2009). Commodity-Scale Production of an Infrastructure-Compatible Bulk Solid from Herbaceous Lignocellulosic Biomass. US Department of Energy, Idaho National Lab (INL). 149  Huisman, W. (2003). Optimising harvesting and storage systems for energy crops in the Netherlands. In Proceedings of the International Conference on Crop Harvesting and Processing, p. 110. Humphrey, D., and Chu, J. (1992). Optimization of a corn-processing simulation model. In Proceedings of the 24th Conference on Winter Simulation, (New York, NY, USA: ACM), pp. 1349–1355. Husdal, J. (2000). How to make a straight line square - Network analysis in raster GIS. University of Leicester. INFORMS Computing Society (1996). The Nature of Mathematical Programming. Accessed from Http://glossary.computing.society.informs.org/index.php?page=nature.html in 2012. Iogen Corp. (2005). Cellulose Ethanol. Project Overview, Iogen Energy Saskatchewan Plant. Iogen Corp. (2009). Project Overview: Iogen Energy Saskatchewan Plant. Judd, J.D. (2011). Modeling and analysis of a feedstock logistics problem. Ph.D. thesis, Virginia Polytechnic Institute and State University. Judd, J., Sarin, S.C., Cundiff, J.S., and Grisso, R.D. (2010). An optimal storage and transportation system for a cellulosic ethanol bio-energy Plant. ASABE, St. Joseph, Mi 1009413. Judd, J.D., Sarin, S.C., and Cundiff, J.S. (2012). Design, modeling, and analysis of a feedstock logistics system. Bioresource Technology 103, 209–218. Kaylen, M., Van Dyne, D.L., Choi, Y.S., and Blase, M. (2000). Economic feasibility of producing ethanol from lignocellulosic feedstocks. Bioresource Technology 72, 19–32. Kim, J., Realff, M.J., and Lee, J.H. (2011). Optimal design and global sensitivity analysis of biomass supply chain networks for biofuels under uncertainty. Computers & Chemical Engineering 35, 1738–1751. Kleijnen, J.P.. (2005). Supply chain simulation tools and techniques: a survey. International Journal of Simulation and Process Modelling 1, 82–89. Klein, K.., and LeRoy, D.. (2007). The Biofuels Frenzy: What’s in it for Canadian Agriculture? The Annual Conference of Alberta Institute of Agrologists Banff, Alberta. Kline, R. (2000). Estimating crop residue cover for soil erosion control. Soil Fact Sheet, Order No. 641-220-1. Abbotsford, BC: Resource Management Branch, Ministry of Agriculture and Food. Kumar, A., and Sokhansanj, S. (2007). Switchgrass (Panicum vigratum, L.) delivery to a biorefinery using integrated biomass supply analysis and logistics (IBSAL) model. Bioresource Technology 98, 1033–1044.  150  Lee, Y.H., and Kim, S.H. (2002). Production–distribution planning in supply chain considering capacity constraints. Computers & Industrial Engineering 43, 169–190. Lindstrom, M.J., E. Skidmore, E.L., Gupta, S.C., and Onstad., C.A. (1979). Soil conservation limitations on removal of crop residues for energy production. Journal of Environmental Quality 8, 533–537. Lyschinski, D., Leduc, P., and Wasserman, J. (2002). Wheat straw quality changes during storage. Final Report Project N. 5100E. Humboldt, SK: Praierie Agricultural Machinery Institute PAMI. Manitoba Agriculture and Food (2009). Crop Price-Yield and Imperial-Metric Calculator. Mantovani, B., and Gibson, H. (1992). A simulation model for analysis of harvesting and transport costs for biomass based on geography, density and plant location, 253–280. Mitchell, C. (1995). New cultural treatments and yield optimisation. Biomass and Bioenergy 9, 11–34. Mitchell, C.. (2000). Development of decision support systems for bioenergy applications. Biomass and Bioenergy 18, 265–278. Morey, R.V., Kaliyan, N., Tiffany, D.G., and Schmidt, D.R. (2010). A corn stover supply logistics system. Agecon Search Institution. Nagel, J. (2000). Determination of an economic energy supply structure based on biomass using a mixed-integer linear optimization model. Ecological Engineering 16, Supplement 1, 91–102. Nilsson, D. (1999a). SHAM—a simulation model for designing straw fuel delivery systems. Part 1: model description. Biomass and Bioenergy 16, 25–38. Nilsson, D. (1999b). SHAM—a simulation model for designing straw fuel delivery systems. Part 2: model applications. Biomass and Bioenergy 16, 39–50. Nilsson, D. (2000). Dynamic Simulation of Straw Harvesting Systems: Influence of Climatic, Geographical and Biological Factors on Performance and Costs. Journal of Agricultural Engineering Research 76, 27–36. Nilsson, D., and Hansson, P.-A. (2001). Influence of various machinery combinations, fuel proportions and storage capacities on costs for co-handling of straw and reed canary grass to district heating plants. Biomass and Bioenergy 20, 247–260. Noel P. Russell, R.A.M. (1983). A stochastic simulation model for evaluating forage machinery performance. Agricultural Systems 39–63. Noon, C.E. (1993). TVA GIS-based biomass resource assessment. Vol. I. Proceedings of the First Biomass Conference of the Americas. National Renewable Energy Laboratory, Golden CO, 74-78. 151  Noon, C.E., and Daly, M.J. (1996). GIS-based biomass resource assessment with BRAVO. Biomass and Bioenergy 10, 101–109. OMAF (1998). Big bale hay storage. The Ontario Ministry of Agriculture and Food (OMAF), Accessed from Http://mbforagecouncil.mb.ca/resources/forage-grassland-manual/7a-harvestingfeeding-stored-forage-hay/75a-big-bale-hay-storage/ in 2010. Ravula, P.P. (2007). Design, simulation, analysis and optimization of transportation system for a biomass to ethanol conversion plant. Ph.D. thesis, Virginia Polytechnic Institute and State University. Rentizelas, A.A., Tatsiopoulos, I.P., and Tolis, A. (2009a). An optimization model for multibiomass tri-generation energy supply. Biomass and Bioenergy 33, 223–233. Rentizelas, A.A., Tolis, A.J., and Tatsiopoulos, I.P. (2009b). Logistics issues of biomass: The storage problem and the multi-biomass supply chain. Renewable and Sustainable Energy Reviews 13, 887–894. Resop, J.P., Cundiff, J.S., and Heatwole, C.D. (2011). Spatial analysis to site satellite storage locations for herbaceous biomass in the piedmont of the Southeast. Applied Engineering in Agriculture 27, 25–32. Richey, C.B., Liljedahl, J.B., and Lechtenberg, V.L. (1982). Corn stover harvest for energy production. Transactions of the ASAE 25, 834–844. Saskatchewan Agriculture and Food (2006). Census StatFact-2006: size of farms. Saskatchewan Ministry of Agriculture (2011). Weekly harvest progress of major crops in the province. Saskatchewan Ministry of Agriculture, Accessed from Http://www.agriculture.gov.sk.ca/crop-report in 2009. Saskatchewan Ministry of Highways and Infrastructure (2011). Typical load limits on highways - highways and infrastructure. Accessed from Http://www.highways.gov.sk.ca/load-limits/ in 2011. Saskatchewan soil conservation association (2000). Value of Cereal Straw. Saskatchewan Soil Conservation Association, Accessed from Saskatchewan Soil Conservation Association in 2010. Savoie, P., Parsch, L.D., Rotz, C.A., Brook, R.C., and Black, J.R. (1985). Simulation of forage harvest and conservation on dairy farms. Agricultural Systems 17, 117–131. Shinners, K.J., and Binversie, B.N. (2007). Fractional yield and moisture of corn stover biomass produced in the Northern US Corn Belt. Biomass and Bioenergy 31, 576–584. Shinners, K.J., Binversie, B.N., Muck, R.E., and Weimer, P.J. (2007). Comparison of wet and dry corn stover harvest and storage. Biomass and Bioenergy 31, 211–221.  152  Sokhansanj, S., and Fenton, J. (2006). Cost benefit of biomass supply and pre-processing. BIOCAP Canada Foundation, Research Integration Program. Sokhansanj, S., and Hess, J.R. (2009). Biomass supply logistics and infrastructure. Methods in Molecular Biology (Clifton, N.J.) 581, 1–25. Sokhansanj, S., and Turhollow, A.F. (2002). Baseline cost for corn stover collection. Applied Engineering in Agriculture 18, 525–532. Sokhansanj, S., Khoshtaghaza, H., Schoenau, G.J., Arinze, E.A., and Tabil, L.G. (2003). Heat and moisture transfer and quality changes in containerized alfalfa cubes during transport. Transactions of the ASAE 46, 423–432. Sokhansanj, S., Mani, S., Stumborg, M., Samson, R., and Fenton, J. (2006a). Production and distribution of cereal straw on the Canadian prairies. Canadian Biosystems Engineering 48, 3. Sokhansanj, S., Kumar, A., and Turhollow, A.F. (2006b). Development and implementation of integrated biomass supply analysis and logistics model (IBSAL). Biomass and Bioenergy 30, 838–847. Sokhansanj, S., Turholow, A.F., Stephen, J., Stumborg, M., Fenton, J., and Mani, S. (2008a). Analysis of five simulated straw harvest scenarios. Canadian Biosystems Engineering 50, 2.27– 2.35. Sokhansanj, S., Turhollow, A.F., and Wilkerson, E. (2008b). Development of the Integrated Biomass Supply Analysis and Logistics Model (IBSAL). Technical Memorandum ORNL/TM2006/57. Oak Ridge, TN: Oak Ridge National Laboratory. Sokhansanj, S., Mani, S., Turhollow, A., Kumar, A., Bransby, D., Lynd, L., and Laser, M. (2009). Large‐scale production, harvest and logistics of switchgrass (Panicum virgatum L.) – current technology and envisioning a mature technology. Biofuels, Bioproducts and Biorefining 3, 124–141. Statistics Canada (2009). Farm data and farm operator data tables. Statistics Canada, Accessed from Http://www.statcan.gc.ca/pub/95-629-x/2007000/4123856-eng.htm in 2010. Stephen, J.D. (2008). Biorefinery feedstock availability and price variability : case study of the Peace River region, Alberta. Ph.D. Thesis, The University of British Columbia. Stephen, J.D., Sokhansanj, S., Bi, X., Sowlati, T., Kloeck, T., Townley-Smith, L., and Stumborg, M.A. (2010). The impact of agricultural residue yield range on the delivered cost to a biorefinery in the Peace River region of Alberta, Canada. Biosystems Engineering 105, 298–305. Stumborg, M., Townley-Smith, L., Sidders, D., Lee, D., and Black, M. (2008). Biomass Inventory Mapping and Analysis Tool (BIMAT). Accessed from Http://www.biofuels.apec.org/pdfs/apec_200810_stumborg.pdf in 2009.  153  Tatsiopoulos, I.P., and Tolis, A.J. (2003). Economic aspects of the cotton-stalk biomass logistics and comparison of supply chain methods. Biomass and Bioenergy 24, 199–214. Tembo, G., Epplin, F.M., and Huhnke, R.L. (2003). Integrative Investment Appraisal of a Lignocellulosic Biomass-to-Ethanol Industry. Journal of Agricultural and Resource Economics 28. Thomas, G. (2000). Overview of storage development. US DOE Hydrogen Program 2000 Annual Review, Sandia National Laboratories Livermore, California. Truong, T.H., and Azadivar, F. (2003a). Simulation based optimization for supply chain configuration design. In Simulation Conference, 2003. Proceedings of the 2003 Winter, pp. 1268–1275 vol.2. Truong, T.H., and Azadivar, F. (2003b). Simulation based optimization for supply chain configuration design. In Simulation Conference, 2003. Proceedings of the 2003 Winter, pp. 1268–1275 vol.2. Turhollow, A.F., and Sokhansanj, S. (2007). Costs of harvesting, storing in a large pile, and transporting corn stover in a wet form. Applied Engineering in Agriculture 23, 439–448. Zhang, F., Johnson, D.M., and Johnson, M.A. (2012). Development of a simulation model of biomass supply chain for biofuel production. Renewable Energy 44, 380–391. Zhu, X., and Yao, Q. (2011). Logistics system design for biomass-to-bioenergy industry with multiple types of feedstocks. Bioresource Technology 102, 10936–10945. Zhu, X., Li, X., Yao, Q., and Chen, Y. (2011). Challenges and models in supporting logistics system design for dedicated-biomass-based bioenergy industry. Bioresource Technology 102, 1344–1351.  154  Appendices Appendix A . Rural municipalities in 160-km supply area (Statistics Canada, 2009) Rural municipality  Land area (km2) 1,184  Number of wheat farms (spring-durum-winter) 142-0-0  Cultivated area (ha) 108,681  Distance from ethanol plant (km) 13.9  Buckland  806  58-0-0  58,638  25.4  Birch Hills  556  87-0-0  55,530  34  Garden River  666  71-0-0  52,600  41.1  Duck Lake  1,170  45-0-0  68,447  47.5  Shell Brook  1,346  101-0-0  52,600  50.2  Kinistino  1,113  98-0-0  78,515  60.2  St. Louis  849  124-6-0  78,257  62.4  Invergordon  854  117-0-0  84,707  63  Paddock Wood  3,168  40-0-0  58,816  79.5  Leask  1,444  52-0-0  126,207  80.1  Fish Creek  601  78-1-0  41,822  81.7  Hoodoo  817  128-2-0  77,702  85.9  Rosthern  960  109-0-0  95,959  90  2,231  125-0-0  162,065  91.3  845  122-3-0  84,432  98.5  5,183  202-0-0  160,488  107  Prince Albert  Can Wood Fletts Springs Torch River  155  Appendix A. Rural municipalities in a 160-km supply area (Statistics Canada, 2009) (Continued) Rural municipality  Land area (km2) 775  Number of wheat farms (spring-durum-winter) 112-0-0  Cultivated area (ha) 73,655  Distance from ethanol plant (km) 114  Grant  804  110-8-5  79,832  123  Blaine Lake  803  63-4-0  50,502  130  Bayne  804  110-8-5  79,832  132  Willow Creek  845  104-0-0  76,203  134  Redberry  1,017  91-2-0  94,066  134  Aberdeen  675  82-5-0  58,460  135  Star City  841  121-0-7  81,392  136  Laird  734  104-4-0  73,406  139  Lake Lenore  726  61-1-2  71,220  142  Humboldt  810  136-5-2  81,020  147  Spiritwood  2,581  69-0-0  207,735  150  Nipa Win  896  116-0-0  85,842  157  Big River  22,561  7-0-0  53,205  160  Three Lakes  156  Appendix B . Location of rural municipalities in the supply area (by permission from Iogen Corp.)  157  Appendix C . Mean and standard deviation of wheat grain Rural municipality  Spring wheat-grain yield (bu/ac) Mean  Durum wheat-grain yield (bu/ac) Mean  Winter wheat-grain yield (bu/ac)  Prince Albert  28.7  Standard deviation 7.09  Mean  -  Standard deviation -  -  Standard deviation -  Buckland  31.64  7.13  -  -  -  -  Birch Hills  35.13  8.33  -  -  -  -  Garden River  31.09  7.1  -  -  -  -  Duck Lake  30.92  8.27  -  -  -  -  Shell Brook  34.86  7.53  -  -  -  -  Kinistino  33.51  8.61  -  -  -  -  St. Louis  -  -  -  -  -  -  Invergordon  34.79  9.87  31.62  12.9  38.45  6.87  Paddock Wood  32.2  8.02  -  -  -  -  Leask  30.13  7.75  -  -  -  -  Fish Creek  32.17  7.06  -  -  -  -  Hoodoo  29.14  7.09  25.42  6.28  -  -  Rosthern  30.75  7.5  30.06  9.31  -  -  Can Wood  33.77  8.03  -  -  32.58  7.01  Fletts Springs  32.97  9.16  -  -  29.88  7.8  Torch River  36.34  10.34  32.61  9.73  -  -  Three Lakes  31.76  7.22  -  -  -  -  158  Appendix C. Mean and standard deviation of wheat grain (Continued) Rural municipality  Spring wheat-grain yield (bu/ac)  Grant  Mean  Durum wheat-grain yield (bu/ac) Mean  31.02  Standard deviation 7.46  Blaine Lake  27.34  Bayne  Winter wheat-grain yield (bu/ac) Mean  -  Standard deviation -  -  Standard deviation -  7.24  28.44  8.8  -  -  30.33  8.23  25.45  7.63  -  -  Willow Creek  29.07  8.1  29.97  10.56  39.22  7.94  Redberry  33.83  7.7  -  -  -  -  Aberdeen  27.01  9.21  25.71  8  -  -  Star City  29.57  7.28  30.24  9.02  -  -  Laird  34.86  9.66  -  -  -  -  Lake Lenore  34.8  9.38  36.06  7.42  -  -  Humboldt  32.05  7.44  26.18  6.16  31.81  7.33  Spiritwood  29.96  7.91  30.56  10.46  37.95  6.22  Nipa Win  28.14  6.9  -  -  -  -  Big River  33  8.27  -  -  -  -  159  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0073710/manifest

Comment

Related Items