Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Electricity use analysis of existing and planned university buildings, and opportunities for life cycle… Glaser, Leonard 2016

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

Item Metadata


24-ubc_2017_february_glaser_leonard.pdf [ 2.16MB ]
JSON: 24-1.0340482.json
JSON-LD: 24-1.0340482-ld.json
RDF/XML (Pretty): 24-1.0340482-rdf.xml
RDF/JSON: 24-1.0340482-rdf.json
Turtle: 24-1.0340482-turtle.txt
N-Triples: 24-1.0340482-rdf-ntriples.txt
Original Record: 24-1.0340482-source.json
Full Text

Full Text

Electricity Use Analysis of Existing and Planned University Buildings, and Opportunities for Life Cycle Costing by  Leonard Glaser  BSc, The University of Tübingen, 2012  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Resource Management and Environmental Studies)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  December 2016  © Leonard Glaser, 2016 ii Abstract Increasing environmental awareness has initiated a change in building design and efficiency in order to reduce the large amount of energy associated with this industry. Life cycle cost (LCC) analysis is a decision-making tool to evaluate the economic long-term benefits of different design options compared to the building’s basic design. LCC analysis can motivate decision-makers to reallocate building budgets towards higher initial capital costs if the long-term operational savings balance higher upfront expenses. However, LCC needs well-calibrated predictive modeling for such savings to be realized.  ‘Bottom up’ Energy modeling software has been used to evaluate savings associated with different building designs. Although these models require a large amount of building specific information, their predictions are often far off from the actual energy use. An alternative proposed in this thesis is to use ‘top down’ models that predict energy consumption using aggregate building characteristics such as size, age, type and occupancy.  We have developed a ‘top-down’ model for electricity use in buildings based on daily electricity consumption data of 48 research buildings at the University of British Columbia (UBC).  The model is a set of linear regressions analyzed with MATLAB. Our model requires only a few simple, aggregate inputs in order to make electricity use predictions. These compare favorably to the more complex LEED energy tested models for ten UBC research buildings. Thus, the ‘top down’ models are an additional, useful tool for energy planning and design. The effort to collect data for such models is also small compared to the ‘bottom-up’ alternative.  iii Preface This thesis is original, unpublished, independent work by the author, Leonard Glaser. I (Leonard Glaser) developed the ‘top down’ model that’s is discussed in this thesis. My supervisor Milind Kandlikar and I framed the project and all strategies together. Some ideas for this project were discussed and developed in cooperation with Harvard and Stanford Universities.       I received funding from the Bridge Program that was sponsored by the Canadian Institutes of Health Research (CIHR). iv Table of Contents Abstract .......................................................................................................................................... ii	  Preface .......................................................................................................................................... iii	  Table of Contents .......................................................................................................................... iv	  List of Tables ................................................................................................................................ vii	  List of Figures ............................................................................................................................... ix	  List of Abbreviations ..................................................................................................................... x	  Glossary ......................................................................................................................................... xi	  Acknowledgements .................................................................................................................... xiii	  Dedication ..................................................................................................................................... xv	  Chapter 1: Introduction ................................................................................................................ 1	  1.1	   Background ......................................................................................................................... 1	  1.2	   Life Cycle Cost Analysis .................................................................................................... 2	  1.2.1	   Data Collection – The Limitation of Life Cycle Cost Analysis .................................. 6	  1.3	   Electricity Use in Research Buildings ................................................................................ 8	  1.4	   The Energy Utilization Intensity (EUI) .............................................................................. 9	  1.5	   Research Contribution and Objectives ............................................................................. 11	  1.6	   Thesis Overview ............................................................................................................... 12	  Chapter 2: The Electricity Use of Research Buildings at UBC ............................................... 14	  2.1	   Building Types ................................................................................................................. 14	  2.1.1	   Classroom .................................................................................................................. 16	  2.1.2	   Laboratory ................................................................................................................. 17	  v 2.1.3	   Office ......................................................................................................................... 17	  2.1.4	   Library ....................................................................................................................... 18	  2.1.5	   Other Spaces .............................................................................................................. 18	  2.2	   Seasonal Outdoor Temperature in Vancouver .................................................................. 18	  2.2.1	   Occupancy Data ......................................................................................................... 20	  2.3	   Methods ............................................................................................................................ 20	  2.4	   Results .............................................................................................................................. 22	  2.4.1	   Seasonal Electricity Use and Occupancy Variation .................................................. 22	  2.4.2	   Daily Electricity Use and Occupancy Fluctuation .................................................... 29	  2.5	   Discussion ......................................................................................................................... 33	  Chapter 3: Linear Regression Analysis of Daily Electricity Consumption ............................ 35	  3.1	   Linear Regression Analysis .............................................................................................. 36	  3.2	   Data ................................................................................................................................... 37	  3.3	   Results .............................................................................................................................. 37	  3.3.1	   ‘Top Down’ Model: Building Parameters ................................................................. 38	  3.3.2	   ‘Top Down’ Model: Reallocated Other Spaces ......................................................... 43	  3.3.3	   ‘Top Down’ Model: Weekday & Weekend Separately ............................................. 46	  3.4	   Discussion ......................................................................................................................... 49	  Chapter 4: Our Results Compared to LEED ‘Bottom Up’ Models ........................................ 50	  4.1	   Data ................................................................................................................................... 51	  4.2	   Methods ............................................................................................................................ 52	  4.3	   Results .............................................................................................................................. 53	  4.4	   Discussion ......................................................................................................................... 58	  vi Chapter 5: Conclusion ................................................................................................................. 60	  5.1	   The Quality of Our Results ............................................................................................... 60	  5.2	   What Assumptions Were Made ........................................................................................ 64	  5.3	   Nature of Complexity of Input Data ................................................................................. 66	  5.4	   Role of the ‘Bottom Up’ Model in Building Design ........................................................ 67	  Bibliography ................................................................................................................................. 69	  Appendices ................................................................................................................................... 74	  Appendix A - The Building Specific Parameters for all 48 Research Buildings are Summarized in the Table Below. .............................................................................................. 74	  Appendix B - All Equations in the Table Below were Used for the ‘Top Down’ Models in Chapter 4. The Estimates are Different from those Described in Chapter 3 because the Building Sample is Reduced. ..................................................................................................... 76	  vii List of Tables Table 2.1: Total gross floor area (m2) accumulated for all 48 research buildings in this study. The total area is subdivided into the 5 main floor areas of research buildings. .................................... 15	  Table 2.2: Sample size of research buildings and number of buildings that represent each of the categories Classroom, Laboratory, Office, and Library. ............................................................... 21	  Table 3.1: Summary of the ‘top down’ model stages. Each equation has added parameters to increase the accuracy of the outcome. The variables in the same order as they appear are: GF: Gross floor area; A: Building age; C: Classroom; L: Laboratory; O: Office; LIB: Library; OT: Other Spaces; F: Number of floors; T: Outdoor Temperature; OO: Occupancy. ......................... 38	  Table 3.2: Summary of the ‘top down’ model with reallocated floor areas. The category “Other Spaces” is reallocated to the other floor area types according to their percentage of the total gross floor area. The variables in the same order as they appear are: CR: Classroom reallocated; LR: Laboratory reallocated; OR: Office reallocated; LIBR: Library reallocated; A: Building age; F: Number of floors; T: Outdoor Temperature; OO: Occupancy. ..................................................... 44	  Table 3.3: Electricity use on weekdays (EWD) and electricity use on weekends (EWE) are analyzed separately. The linear regressions have the following variables: C: Classroom; L: Laboratory; O: Office; LIB: Library; OT: Other Spaces; A: Building age; F: Number of floors; T: Outdoor Temperature; OWD: Occupancy on weekdays; OWE: Occupancy on weekends. .... 47	  Table 4.1: List of the only ten LEED modeled UBC campus buildings, their building category, and the modeled annual electrical Energy use. .............................................................................. 51	  Table 4.2: Summary of the ‘top down’ model results compared to the LEED ‘bottom up’ models. The five equations from Chapter 3 with the highest accuracy were used to estimate the ten viii buildings’ annual electricity consumption. The variables in the same order as they appear are: GF: Gross floor area; A: Building age; C: Classroom; L: Laboratory; O: Office; LIB: Library; OT: Other Spaces; F: Number of floors; T: Outdoor Temperature; OO: Occupancy; CR: Classroom reallocated; LR: Laboratory reallocated; OR: Office reallocated; LIBR: Library reallocated. All results in kWh/year. The equations can be found in Appendix B. ...................... 57	  ix List of Figures Figure 1.1:The electricity consumption of a typical Laboratory building on UBC campus is consumed by the HVAC (ventilation), ultra low temperature freezers (Other), Lighting, and Lab Equipment (UBC Green Labs, n.d.). ............................................................................................... 9	  Figure 1.2: Comparison of the electrical Energy Use Intensity (EUIel) for 48 research buildings on UBC Campus. ........................................................................................................................... 10	  Figure 2.1:  Seasonal monthly average temperature for Vancouver in the year of 2014. ............. 19	  Figure 2.2: Seasonal occupancy (green circles) and electricity consumption (blue diamonds) fluctuation of Classroom (A), Laboratory (B), Office (C), and Library (D) buildings. The solid lines and filled symbols represent the weekday behavior, and the dashed lines and unfilled symbols the weekend. The error bars show the standard error. .................................................... 28	  Figure 2.3: Electrical Energy Use Intensity (EUIel) in watt-hours per square-meter (blue diamonds) and occupancy in people per day (green circles) in February 2014. The curves show the average value of all buildings that were grouped as Classroom (A), Laboratory (B), Office (C), and Library (D) buildings according to their dominant floor use area. The error bars show the standard error. .......................................................................................................................... 32	  Figure 4.1: The red line is the 100% mark, which represents the actual electricity consumption of each building. The blue diamonds and the violet squares show the results of top-down model 1 and 2 respectively, relative to the actual electricity use of the building. The green triangles show the results of the ‘bottom up’ model relative to the actual electricity use. .................................... 56	  x List of Abbreviations CAPEX Capital Expenditure ECM Energy Conservation Measure EUI Energy Utilization Intensity EUIel Electrical Energy Utilization Intensity EWS Energy & Water Services GWh Gigawatt hour = 1,000,000,000 watt hours HVAC  Heating, Ventilating, and Air Conditioning  kWh Kilowatt hour = 1,000 watt hours LCC Life Cycle Cost LEED Leadership in Energy and Environmental DesignMT&R Monitoring, Targeting and Reporting UBC University of British Columbia ULT Ultra Low Temperature Wh Watt-hour xi Glossary Discounting – An economic methodology to translate inflated future costs back into current dollars. It is used to compare the net present value of inflated or escalated costs that are accumulated over a given time period for different products. Energy Conservation Measure – A project or any type of technical improvement that is implemented to reduce the overall energy consumption in a building.  Energy modeling software – A computer based tool that estimates buildings’ energy consumption according to a wide range of building specific parameters. Energy+ and eQuest are among the most commonly known modeling tools. Escalation – Similar to inflation, the escalation rate describes the increasing costs of a product or service over time. However, escalation follows other rules than inflation and is specific to a product or service. Green Building - A specially designed facility to reduce the environmental impact by using resources more efficiently and by selecting building materials that are considered to cause fewer harm to the environment. Heating, ventilating and air conditioning system (HVAC) – A major part of the mechanical equipment and often referred as the building’s lungs. The HVAC provides heated or cooled air to the place through the installed duct system. The air gets also filtered and mixed with fresh outdoor air.  Inflation – An economic term that describes the decreasing value of money over time. The inflation rate is usually correlated with the performance of a country’s economy.  xii Initial Capital Cost  - All costs for planning & design, material, grounds, permissions, labor, and any other construction costs until the building is operational are summarized as the Initial Capital Cost. LCC analysis – An economic decision making tool to evaluate the total cost of ownership under consideration of all affiliated costs over the building’s entire lifetime. It is used to choose the best economically performing building among different options during early stages of the planning & design phase. Life Cycle Cost (LCC) – A summary of all accruing costs from the planning and design phase of a building to the deposal at the end of its lifetime.  Simple Payback – An easy approach to calculate the time period it takes for a project’s savings to payback the project’s initial cost.   Value for Money – An old term that was invented in the UK when LCC analysis was in its infancy. It was developed to investigate the building option that offers the highest quality and lowest overall LCC.  xiii Acknowledgements The continual support of many people has allowed me to succeed in graduate school, and has made this thesis possible.  I would especially like to thank Milind for his support and invaluable guidance throughout my time at UBC. I want to recognize our uncountable and very productive meetings that always had great results. Thank you also to my co-supervisor Steve who contributed to my interest in life cycle costing in very early stages of my research.  I want to extend sincere appreciation to the prestigious Bridge Program that provided funding and an excellent education to me. The fellowship allowed me to extend my knowledge in the interdisciplinary cooperation between engineering and medicine in order to overcome public health related issues.     I would like to express my deepest gratitude to Jeff and the Department of Energy and Water Services for giving me the opportunity to gain valuable work experience during my studies, and for providing a very harmonic, enjoyable, and welcoming work environment. I hope we will still be working together in the future. I also want to thank Stefan Storey and Sensible Building Sciences for their generous provision of occupancy data that has added much value to this thesis. I would especially like to recognize Joel McKellar from Harvard University, as well as Laura Goldstein, Scott Gloud, and David Cuffy from Stanford University for all interesting discussions we had, and for taking their time to answer my endless questions.  xiv My time with the IRES would not have been as enjoyable without being surrounded by a great community. My gratitude extends to all students and staff for all they have done to support me, and for the good memories we share. Thank you to my incredible friends for all great moments we have had, in particular when work had occupied the majority of my time. You have truly helped me to preserve my positive attitude even after long days.  Finally, thank you to my family for encouraging me to live my dreams although living my dreams means that we cannot see each other as often as we would like to. Thank you for unconditionally supporting me throughout all adventures life has taken me. xv Dedication To my nieces Amelie and Helena. May their passion for science grow as strong as mine. 1 Chapter 1: Introduction 1.1 Background  The building sector is a major energy consumer accounting for 30-40% of all primary energy worldwide.  Facilities are also responsible for 40-50% of the global greenhouse gas emissions (CIWMB, 2000). Significant evidence of anthropogenic induced global warming, ozone layer depletion, and other fatal environmental interventions have raised international awareness with impacts on the building industry (Capehart, Turner, & Kennedy, 2012; Kneifel, 2010). Some cities have launched multiple Green Building initiatives to reduce their overall carbon footprint. The City of Vancouver, one of the global leaders in sustainability, promotes and supports the Green Building industry in order to meet the city’s carbon target and to become “the greenest city” in the world (City of Vancouver, 2012). Meanwhile, energy conservation and sustainable building operation has become a requirement for building developers and owners. The City of Vancouver requires all new buildings to meet the Leadership in Engineering and Environmental Design (LEED) guidelines. The buildings have to prove their efficiency during a one year test run before being awarded with the LEED (City of Vancouver, 2012, 2014). LEED is a North American rating system that awards buildings according to their environmental performance; the certification is recognized in over 150 countries (Canada Green Building Council, 2016).   Increasing energy prices and inevitable constraints are one reason why facilities are no longer designed to simply serve their purposes for the lowest construction cost (Kneifel, 2010). However, the industry has experienced a major planning and design change, not only because of increased environmental awareness and new building designs. Clients expect facilities that are efficient over their entire lifetime, and this efficiency is directly correlated with the operation 2 cost (Al-Hajj & Horner, 1998; Boussabaine & Kirkham, 2004b). The design with the lowest initial capital cost1 is no longer the most attractive choice amongst developers because it may not be the best business strategy when compared to the building option that uses the lowest amount of energy. Furthermore, upgraded heating, ventilating, and air conditioning (HVAC) systems, advanced building designs, and higher material quality provide an improved indoor-environmental performance (Boussabaine & Kirkham, 2004a) .  Alongside rising environmental and resource concerns, cost has remained an important and decision-making factor for developers and owners, creating an increasing demand for cost analysis and performance evaluating tools (Al-Hajj & Horner, 1998; Kneifel, 2010; Wübbenhorst, 1986). A concept that was originally developed in the UK to evaluate the “value for money” of buildings (and other products) in the early seventies has been established to a widely acknowledged decision-making tool (Dell’Isola, Alphonse J.; Kirk, 2003). The term “value for money” summarizes the two main benefits of Life Cycle Cost (LCC) analysis: obtaining the highest quality and the lowest overall LCC for a given budget (Dell’Isola, Alphonse J.; Kirk, 2003). 1.2 Life Cycle Cost Analysis Every item has an economic life cycle that begins with its purchase and ends with its disposal. The item has to be maintained, eventually repaired, broken parts replaced, and it might require electricity or some other form of fuel to do whatever it is designed for (Dell’Isola, Alphonse J.; Kirk, 2003).  1 The initial capital cost of a building is the product of all construction costs, ground costs, planning and design costs, and regulatory fees.  3 Life cycle cost analysis for buildings is an economic decision-making tool to evaluate the total cost of ownership over its entire life, taking into consideration initial capital cost, all operation costs, energy & water costs, maintenance costs, replacement costs, financing costs, and eventual disposal costs2 (Cole & Sterner, 2000). Anticipated expenses are documented, scheduled, and inflated over the expected lifetime of the building using inflation and escalation rates. Inflation expresses the decreasing value of money over time, giving the reason why costs for most goods and labor increase continuously. The rate is usually correlated with the performance of an economy but can also be company or organization specific (Dell’Isola, Alphonse J.; Kirk, 2003). Escalation is a different form of inflation because it is specific to a product or a service. Both rates are used for economical cost predictions and to investigate price developments of products and services in the future (Boussabaine & Kirkham, 1998; Dell’Isola, Alphonse J.; Kirk, 2003; Finance, 2013). Life cycle cost analysis has been developed to screen different building designs during early stages of a new construction project, large renovations, and capital expenditure (CAPEX) (Dell’Isola, Alphonse J.; Kirk, 2003).  The LCC of each option is discounted and the resulting net present value is used to economically compare different building design options (Cole & Sterner, 2000). Discounting translates inflated and escalated future costs back into current dollars. The discounted future cost is called the net present value. It is much easier to compare net present values than future costs, particularly when items have different life times (Boussabaine & Kirkham, 1998; Capehart et al., 2012; Dell’Isola, Alphonse J.; Kirk, 2003).    Despite higher initial capital costs owing to better mechanical equipment, improved building envelope insulation, or other Green Building strategies, the total LCC can be significantly lower 2 Disposal costs can reduce the overall LCC if the building has still a value at the end of its life cycle. 4 due to drastically reduced energy costs and carbon tax penalties (Dell’Isola, Alphonse J.; Kirk, 2003; Wong, Perera, & Eames, 2010). The total operation costs of a building accounts for 55% of the total LCC over a period of 40 years (Dell’Isola, Alphonse J.; Kirk, 2003; Liu, Benjamin Y. H.; Richard C., 1963)(Flanagan and Norman, 1989). Another study from Sweden has shown that the budget invested in producing new buildings in Sweden is lower than the total operation costs that is spent on existing buildings (Sterner, 2000). New building designs developed to increase efficiency can be operated with 80% less energy than normal buildings. Energy efficiency has become one of the most cost-reducing strategies in the manufacturing industry (Capehart et al., 2012). Furthermore, the design of Green Buildings is usually more robust, which is directly correlated to lower operation and maintenance costs (Cole & Sterner, 2000). More robust designs are appreciated by developers who expect the best value for their investment (Perera, Morton, & Perfrement, 2009). Qualitatively higher building designs offer better indoor-environments and contribute to a higher occupant satisfaction. That often supports improved productivity as well as reduced absenteeism. Although this argument can be used to advertise green buildings because increased productivity reduces the ongoing costs of any office or work place, it is usually not taken into account for LCC calculations (Al-Hajj & Horner, 1998; Cole & Sterner, 2000; Dell’Isola, Alphonse J.; Kirk, 2003).  Unlike many products with short life expectancies, buildings are durable and built to last for decades. Hence, buildings should be especially designed with long-term cost considerations (Morrissey & Horne, 2011). Life cycle cost analysis is much more accurate in providing long-term cost predictions and effectiveness than other economic models (Boussabaine & Kirkham, 2004a). It helps investors to understand that the building option with the lowest LCC is the smart 5 choice in the long term, and where savings are expected. Furthermore, the model ensures owners and designers actually consider ongoing costs and provides a monetary argument for smarter initial investments and design attention (Dell’Isola, Alphonse J.; Kirk, 2003). Thus, the commitment to building options with higher initial capital cost will only occur if its understood that higher rent and lower operation costs outweigh the additional expenses (Aye, Bamford, Charters, & Robinson, 2000). Although the LCC does not consider rent, it is already assumed to be five times the initial capital cost by the end of the building’s life time (Fu, Kaya, & Aouad, 2007). This has been shown by Aye at al., who compared different options for improving an existing commercial office building, ranging from “do nothing” to “buy and build” a brand new building. Their ‘do nothing’ comparison has shown that the higher cost for energy involved with this option offsets the initial capital cost of a the ‘new building’ option (Aye et al., 2000).  Another case study by Morrissey and Horne, in which life cycle costing was used for several residential buildings in Melbourne, demonstrated that energy savings due to thermally improved buildings outweigh the higher construction costs after the first years of occupancy (Morrissey & Horne, 2011).   In Canada, the C-2000 program made the use of LCC analysis mandatory (Larsson & Clark, 2000). Since 2001, most government departments in the UK are required to consider LCC analysis outcomes for all new building and building refurbishment projects (Mohamed, A. El-Haram; Sasa, Marenjak; Malcolm, 2002). Although LCC analyses are based on many estimates and their results are preliminary and approximate, they provide important information for planning and design by making explicit the building’s capital and ongoing costs(Cole & Sterner, 2000). In the absence of LCC analysis the 6 initial capital costs remains the only criterion dictating the economic preference of competing building design options (Cole & Sterner, 2000). Life cycle cost analysis is often suggested as a valuable alternative to simple payback calculation, which is a poor approach for appraising the total cost of a long-term building project. This is due to the factors ignored by the simple payback model, including the time cost of money and potential for risk reduction or fiscal returns of better-performing design alternatives (Chalifoux, 2006). The right questions have to be asked to get the right answers. Applying the simple payback method to evaluate the cost of building projects can be compared to inexperienced stock market investors who ask how soon they will get their money back. All stock market experts know that the only thing that matters is how well the money is working for the investor over the investment period (Chalifoux, 2006).  1.2.1 Data Collection – The Limitation of Life Cycle Cost Analysis “It is hard to make predictions, especially about the future.” [Niels Bohr]  Economic performance and future cost evaluating tools are struggling with that fact that the future is only predictable to a certain degree. Energy cost assumptions, operational attention, and maintenance costs of novel design and construction approaches can require significant analysis that is not commonly performed.  A constantly fluctuating economy and continuously changing market prices make escalation and inflation rate forecasts uncertain and difficult (Boussabaine & Kirkham, 1998; Risk & Responses, 2004). Many design teams have as yet failed to visualize the impact of their own decisions on operation costs (Aouad, Bakis, Amaratunga, & Sun, 2002). Thus, according to most authors, the main reason why LCC analysis has not found as many 7 proponents as one would expect, due to all previously listed advantages, is a lack of data availability and the accompanied effort of its collection. Data collection and analysis is a time intensive procedure, which in combination with the additional financial burden is often enough to outweigh the long term economic benefits the tool advocates (Boussabaine & Kirkham, 2004b; Neely & Neathammer, 1991; Wong et al., 2010). Nevertheless, as a data intensive methodology, access to data is a requirement for all LCC analyses because the quality of its outcomes relies on the quality of the inputs. It has been argued that better accessibility and data quality will continuously improve the reliability and confidence of LCC results while simultaneously reducing the time and effort that is required for the analysis (Boussabaine & Kirkham, 2004a; Cole & Sterner, 2000).  However, there is not much information provided on where to obtain the right data source and determine “how good is good enough?”. Most data is collected from existing buildings and their individual energy, maintenance, and other operational performances. Energy Monitoring, Targeting and Reporting (MT&R) is a common strategy to investigate facility’s energy performance by analyzing historical data (Capehart et al., 2012). Other data such as energy consumption can be modeled specifically for a building with the use of complicated energy modeling software. Building energy simulations are time intensive, computer based procedures to predict the future energy use of a building. These predictions are a result of hundreds or even thousands of differential equations (Chalifoux, 2006).    In summary, the accessibility to reliable databases seems to be the major obstacle that has hindered LCC analyses from becoming a successful and more appreciated decision-making tool. Only few databases have been developed on very small scales (Boussabaine & Kirkham, 2004b; Neely & Neathammer, 1991; Wong et al., 2010).  8 1.3 Electricity Use in Research Buildings Electricity conservation has great potential for cost savings because electricity accounts for 50% of the end-use energy and represents two-thirds of the overall energy costs to operate a commercial building (Capehart et al., 2012). The University of British Columbia consumes 293 gigawatt hours (GWh) per year and has saved over 33.6 GWh annually through energy conservation projects since 2014 (UBC Energy & Water Services, 2016). The largest portion of electricity is commonly used by the HVAC system that provides heating, cooling and ventilation to the spaces. The HVAC system is complex mechanical equipment part that combines ducts, motors, controls, fans, heat exchangers, chillers for cooling, and boilers heating. Apart from maintaining a comfortable room temperature, the HVAC system also filters particulates from the outdoor air and mixes it with indoor air before it gets heated or chilled and finally circulated through the entire building (Capehart et al., 2012). Not all buildings have HVAC systems installed but some facilities require high ventilation rates without any interruptions. The building specific ventilation requirements and the role of buildings with laboratories are discussed in Chapter 2. Lighting and plug load do not account for as much electricity as the HVAC system but can also add a significant portion to the building’s total energy use. Lighting can be zoned or timed that only occupied areas are illuminated and only for a certain amount of time per day. However, most buildings have only one or two light switches that control the entire floor. Plug load is the combined electricity demand of all computers, projectors, monitors, kitchen appliances, and any other electricity consuming equipment (Capehart et al., 2012). Some campus buildings also have very energy intensive research equipment installed such as ultra low temperature (ULT) freezers, lab lacers, chillers, and vacuum (UBC Green Labs, n.d.).   9 Figure 1.1 shows the electricity consuming areas in a typical laboratory building on UBC campus. Ventilation represents the HVC system and Other is a summary of large equipment that is permanently in use such as ultra low temperature freezers (UBC Green Labs, n.d.).   Figure 1.1:The electricity consumption of a typical Laboratory building on UBC campus is consumed by the HVAC (ventilation), ultra low temperature freezers (Other), Lighting, and Lab Equipment (UBC Green Labs, n.d.). 1.4 The Energy Utilization Intensity (EUI) Residential buildings tend to be more similar in the way they are used by their occupants (ASHRAE Standing Standard Project Committee 100, 2015). However, no research building is Ven%la%on	  48%	  Ligh%ng	  9%	  Lab	  Equipment	  16%	  Other	  27%	  10 like the other, which is perfectly mirrored by each building’s annual energy consumption as shown in  Figure 1.2.Figure 1.2: Comparison of the electrical Energy Use Intensity (EUIel) for 48 research buildings on UBC Campus. This measure expressed as kWh/m2 is a common metric for comparing or benchmarking different types of buildings, and is referred as Energy Utilization Intensity (EUI). A building’s EUI is the sum of the total annual energy use (Thermal and Electrical, both in kilowatt hours (kWh)) which is then divided by the building’s gross floor area (Capehart et al., 2012). This metric kWh/m2/yr is useful in helping to identify outlier buildings and energy conservation opportunities. As this study only investigates electricity consumption the electrical EUI (EUIel) is used instead. 	  -­‐	  	  	  	  	  100	  	  	  200	  	  	  300	  	  	  400	  	  	  500	  	  	  600	  	  	  700	  	  	  800	  	  Chemistry	  D	  Block,	  Centre	  Wing	  Michael	  Smith	  Laboratories	  Chemistry	  A	  Block,	  Chemistry	  Physics	  Building	  The	  Leonard	  S.	  Klinck	  Building	  Biomedical	  Research	  Centre	  Life	  Sciences	  Centre	  Civil	  and	  Mechanical	  Engineering	  Laboratories	  Earth	  and	  Ocean	  Sciences	  	  -­‐	  Main	  Brimacombe	  Walter	  C.	  Koerner	  Library	  David	  Strangway	  Food,	  Nutri%on	  and	  Health	  Building	  Ins%tute	  for	  Compu%ng,	  Informa%on	  and	  Cogni%ve	  Chemical	  &	  Biological	  Engineering	  Building	  David	  Lam	  Learning	  Centre	  School	  of	  Popula%on	  &	  Public	  Health	  Forest	  Sciences	  Centre	  D.H.	  Copp	  Jack	  Bell	  Mathema%cs	  Building	  J.	  B.	  Macdonald	  Building	  I.K.	  Barber	  Learning	  Centre	  Lower	  Mall	  Research	  Sta%on	  Pulp	  and	  Paper	  Centre	  Frank	  Forward	  Building	  Douglas	  Kenny	  Building	  Woodward	  Library	  Medical	  Sciences	  Block	  C	  Asian	  Centre	  H.	  R.	  Macmillan	  Building	  Hennings	  Building	  Friedman	  Building	  Macleod	  Building	  Civil	  and	  Mechanical	  Engineering	  Building	  Buchanan	  Building	  Block	  D,	  E	  Aqua%c	  Ecosystems	  Research	  Laboratory	  Antrophology	  and	  Sociology	  Buchanan	  Building	  Block	  A,	  B,	  C	  Frederic	  Lasserre	  Building	  West	  Mall	  Swing	  Space	  Building	  Neville	  Scarfe	  Building	  P.A	  Woordward	  Instruc%onal	  Resources	  Centre	  Wesbrook	  Building	  Hebb	  Building	  Geography	  Building	  Buchanan	  Tower	  C.	  K.	  Choi	  Building	  for	  The	  Ins%tute	  of	  Asian	  EUI el	  [kWh/m2 ]	  11 The 48 buildings were selected according to their ability to represent a range of different research building types. As such they must either have classroom spaces, laboratory spaces, academic office spaces, library spaces, or a mixed combination of some or all floor types. Unfortunately, data availability was the limiting factor for our building selection because an actual occupancy database for all UBC buildings has yet to be established. The origin and the process of the data collection for this study is discusses in Chapter 2.  1.5 Research Contribution and Objectives  The literature review has proven that LCC analysis has great potential; however, its success is limited due to data availability and collection. Some developers bridge this issue by dispensing with LCC analysis and instead choosing the lowest capital cost project without consideration of any long term operational expenses. Other developers who are more committed to using LCC analysis often rely on energy modeling software that gives results that are far off the building’s actual energy use due to many assumptions. In particular, electricity savings related to better building design and more efficient mechanical equipment can offset higher capital costs associated with the better building option. Hence, electricity use predications have to be more precise than the current results from energy modeling software.   This study contributes to the current research stage of LCC analysis and building energy evaluation by collecting a database for and investigating the electricity use of 48 research buildings. The analyses are conducted on a building type specific scale because of the large variety of EUIel among the 48 buildings. The facilities are allocated to one of the four categories, Classroom, Laboratory, Office, and Library, in order to investigate the drivers of seasonal electricity use.    12 Furthermore, we develop a model based on linear regression analyses and the collected database to make electricity use predications without using complicated and often inaccurate energy modeling software. The model developed in this study could be part of an LCC calculator for research buildings if further research contributes similar models for all other operational cost categories. The model in its stage as presented in this study delivers reverence values to be compared with energy modeling software results. This helps energy modelers reconsider their inputs in case the discrepancy of both results is unexplainable.  Furthermore, this study contributes to understanding where electricity is being used in research facilities, how the overall electricity consumption is determined by the building type (Classroom, Laboratory, Office, Library), and other building specific parameters.    1.6 Thesis Overview Chapter 1 has summarized the current debate on LCC analysis with some success stories from other countries, and the limitations of the tool. We have shown through our literature research that LCC analysis is a promising decision-making tool, but it is likely to be missed out on without further development towards more precise estimations.  In Chapter 2 we establish an electricity use database for 48 existing UBC research buildings. The Chapter highlights the differences between research buildings and why it is helpful to group them according to their main purpose before analyzing their electricity use. We discuss why some facilities have extremely high EUIel and which parameters are correlated with higher electricity use. The seasonal analysis gives an insight on how electricity use is correlated with Outdoor Temperature but also with Occupancy fluctuations according to UBC’s academic year.  13 In Chapter 3 we develop a ‘top down’ model that analyzes our 48 buildings’ daily electricity consumption in relation with simple infrastructural and operational parameters. The parameters are added one by one in order to investigate which variable increase the linear regressions’ accuracy, and is hence, an important contributor to a research building’s EUIel.  Chapter 4 compares our ‘top down’ model’s results with ten LEED modeled UBC research buildings, and the facilities’ actual annual electricity consumption in 2014. This chapter adds value to our analysis because it discloses how accurate the ‘top down’ model’s predictions are for ten buildings that are not included in our database. We also discuss how to apply the ‘top down’ model in order to get the best results out of it, and how the results defer to the LEED ‘bottom up’ model.    Chapter 5 concludes our analysis and summarizes the benefits and limitations. We end this thesis by suggesting which role the ‘top down’ model in building design could have also under consideration of incorporating the model with the LEED ‘bottom up’ model.   14 Chapter 2: The Electricity Use of Research Buildings at UBC The University of British Columbia has its main campus in Vancouver where more than 50,000 students are enrolled. The 100-year old university has been continuously growing to become the largest university in the province of British Columbia and one of the largest in Canada. All faculties together contain more than 100 research facilities with a wide range of different uses and operational requirements (The University of British Columbia, 2016). Even some of the very first buildings are still being used as research institutions. Meanwhile, some of the other buildings referenced were recently added to the campus. This widely diverse selection of research buildings covers a multitude of different mechanical systems, lighting retrofits, renovations, energy conservation projects, building materials, and building designs.   A detailed building list and further information can be found in the Appendix A. This chapter investigates the drivers for electricity consumption of different research building types. First, we discuss how to make electricity consumption of buildings with different floor area size comparable, and how to categorize a research building according to its main purpose. The building categories are discussed individually and each category’s specific electricity use behavior is explained together with reasons why some buildings are more electricity intensive than others. 2.1 Building Types Buildings are unique structures that require a specific amount of energy due to a long list of parameters and reasons. As previously mentioned, the buildings on UBC’s Vancouver campus are diverse, and each has its own characteristics. These characteristics together are responsible 15 for the specific amount of energy that is required make a building operational. The UBC department of Infrastructure Development has a database of more than 500 buildings that contains information about the size, age, floors, room space allocation and much more. The majority of research buildings have large Laboratory areas, open-office3 student areas, smaller offices and meeting rooms. In addition to Laboratory and Office areas, there are buildings with large Classroom or Library spaces but less frequent and smaller in overall gross floor area on a campus wide scale. These four categories cover all useful research areas of any campus building. The portion of each floor category is building specific but not all four categories have to be represented in every facility. It is common that some research buildings have only one or two different floor area types. Apart from the useful research area, there is one more floor space category, which is present in all buildings, and shall be referred to as “Other Spaces” for this thesis.  The total floor area in square-meters of each floor space category is shown in Table 2.1 as the sum of all 48 research buildings in the data set. Table 2.1: Total gross floor area (m2) accumulated for all 48 research buildings in this study. The total area is subdivided into the 5 main floor areas of research buildings. Total Floor Area Classroom Laboratory Office Library Other Spaces427,192 m2 30,650 m2 102,519 m2 88,132 m2 22,775 m2 183,117 m2 3 Open-office areas are shared spaces that are commonly used by are large group of people. These offices do not have zoned areas for lighting or HVAC, and are operated as one large office. This means that lights are either on or off, the temperature is the same everywhere no matter if parts of the open-office are not being used. Open-offices are not treated as a separate category in this study and belong to the building category Office.  16 Classroom, Laboratory, Office, and Library, are assumed to have their specific electricity demand. Other Spaces is not treated as an additional useful space category and reallocated to all other floor categories as described in 2.1.5. The ratio and size of each category that is represented in a building determines the facility’s total electricity. The following paragraphs explain the main drivers of electricity use of each floor category. To prove our assumption we summarize buildings according to their dominating floor category and investigate the seasonal electricity use fluctuation in correlation with Occupancy and Outdoor Temperature.  2.1.1 Classroom “Classroom” is a building category dominated by classroom spaces. This building category has a relatively low EUI and energy use in these buildings is highly correlated with occupancy (ASHRAE Standing Standard Project Committee 100, 2015). Classroom buildings host usually large groups of people for a certain amount of time during the day. The majority of energy consumed in these buildings is due to the Heating, Ventilating and Air Conditioning (HVAC) system that provides heating, cooling and ventilation to the spaces. Modern HVAC systems can be scheduled to avoid excessive ventilation when the spaces are unoccupied. Some systems even have occupancy sensors and measure the exact amount of air movement that is required (Capehart et al., 2012).  Among our 48 research buildings, all Classroom spaces together account for only 7% of the campus wide total gross floor area. There are only three buildings in our sample that can be categorized as classroom buildings but there are more buildings that have at least a small portion of classroom space area.  17 2.1.2 Laboratory Buildings that are dominated by Laboratory spaces are expected to have very high EUIs. Unlike other buildings, laboratories typically operate 24/7 and require high air change rates with 100% outdoor air (ASHRAE Standards Committee, 2007). This requirement means that heating, cooling, and air movement equipment must operate 24/7. The fresh outdoor air is either heated or chilled to meet the desired room temperature. Because of this, energy consumption of Laboratory buildings tends to be poorly correlated with building occupancy. Laboratory is the best represented building category in our 48 buildings set and dominates with 25 buildings or 24% of the total floor area.  2.1.3 Office Office buildings usually have much lower EUIs because lighting and HVAC systems can be scheduled to turn off at night when not in use. The ventilation requirements differ from laboratory buildings because indoor air can be recirculated, and a smaller amount of fresh outdoor air is needed to maintain sufficient ventilation. This reduces the electricity consumption dramatically. Although almost every building has at least some Office floor area, only 15 facilities in our set are Office buildings according to our definition, and account for 21% of the campus wide total floor area.   18  2.1.4 Library Libraries at UBC generally have longer operational times, and can be occupied 24/7 during some parts of the academic year, such as exam time4. There are only five amongst our 48 buildings which have library area as the dominant section of the building’s total floor area. All buildings’ library spaces account for only 5% of the 48 buildings’ total floor area.   2.1.5 Other Spaces Other Spaces is not an actual building category in this analysis but it is important to mention because it is the dominant floor type in many buildings. Other Spaces include all floor areas that are necessary for building operations such as mechanical rooms, as well as floor areas that are commonly shared by all building inhabitants such as washrooms, stairwells, elevators, and atriums. Other spaces can have a significant portion of a building’s total gross floor area although they do not contribute to any actual research space. In our 48-building example Other Spaces account for 43% of the total floor area. As Other Spaces are “non-useful” floor areas in terms of their contribution to research spaces, they are not part of the seasonal electricity fluctuation investigation.       2.2 Seasonal Outdoor Temperature in Vancouver Weather can have much influence on a building’s EUIel, especially in summer when chillers are required to cool the air that is provided to the space (Capehart et al., 2012). The outdoor temperature is measured in degree Celsius [°C]. It is recorded as daily averages and is assumed                                                 4 Library operation hours are taken from UBC’s library website: (September 16, 2016)  19 to be uniform across campus. Hence, temperature is not a building specific parameter. Temperature data is provided UBC Energy & Water Services. Vancouver has a mild, tempered climate with warm summers and mild winters. The monthly average temperature for the year 2014 is shown in Figure 2.1. Outdoor temperature is an important parameter that can influence a building’s EUIel as shown in the following analysis.  Figure 2.1:  Seasonal monthly average temperature for Vancouver in the year of 20145. 5Temperature data is provided by Energy & Water Services. 	  4.41	  	  	  2.52	  	  	  6.90	  	  	  9.85	  	  	  14.01	  	  	  15.72	  	  	  18.95	  	   	  19.23	  	  	  16.05	  	  	  12.93	  	  	  5.93	  	  	  5.05	  	  	  -­‐	  	  	  5	  	  	  10	  	  	  15	  	  	  20	  	  	  25	  	  Jan	   Feb	   Mar	   Apr	   May	   Jun	   Jul	   Aug	   Sep	   Oct	   Nov	   Dec	  Temperature	  [°C]	  20  2.2.1 Occupancy Data Occupants influence a building’s electricity consumption in either an active or passive form. An active form that increases the electrical load is when more people use more areas in a building. Accordingly, more lights are turned on for the time the room is occupied. Another effect is the increased plug load when more office equipment such as computers is being used. A passive form of occupant influence on the electrical load is additional heat rejected by every person, lamp or office equipment, that has to be removed by the HVAC system, requiring more electricity.  Occupancy data was provided by Stefan Storey and Lisa Shiozaki. The occupancy is measured by counting the Wi-Fi connections within the building. It is expected that most building inhabitants have at least one device that is permanently connected6. Users with two devices such as a phone and a laptop, or tablet, etc. are expected to balance the amount of people with no Wi-Fi connecting device. The collected data is recorded for each building as daily in the year of 2014.  2.3 Methods This chapter focuses on the relationship between electricity consumption of 48 research buildings and two different seasonally varying parameters: outdoor temperature and occupancy.  Most of the recently constructed buildings on campus are high-end laboratories and require mechanical ventilation as previously discussed in the building category introduction. Although a                                                 6 The information is taken from Sensitive Building Sciences and the Sustainable Buildings Sciences Program website: (September 15, 2016);  (September 15, 2016) 21  great deal of effort has been but in developing energy efficient ventilation units, the HVAC still demands the largest portion of the building’s overall electricity use (Capehart et al., 2012).  The following subsections will investigate how the four different building categories, Classroom, Laboratory, Office, and Library, correlate in different ways with seasonal and occupancy changes through a year. We utilize each building’s monthly EUIel in 2014. Again, the EUI takes off the weight of a building’s floor area size in order to make the energy use of different buildings with different gross floor areas comparable. The 48 facilities are grouped according to their dominate floor area category as explained in the previous paragraphs. The number of buildings per category varies as not all building types are represented on campus equally as summarized in Table 2.2.   Table 2.2: Sample size of research buildings and number of buildings that represent each of the categories Classroom, Laboratory, Office, and Library. Total Sample Classroom Laboratory Office Library 48 3 25 15 5   We analyzed each building’s daily electricity consumption for the calendar year 2014 and summarized separately the amount of electricity that was being consumed on weekdays, and the amount of electricity that was being consumed on weekends including statutory holidays in British Columbia. Following this strategy two data sets were generated for each month, one containing the total electricity use on all weekdays (kWhWD/m2/month) and the other containing the electricity use on all weekends/holidays (kWhWE/m2/month). In order to get one value per month for the weekday consumption and one value per month for the weekend consumption that represent the building category instead of each individual building, we calculated the monthly averages of all buildings within the same category. These values were used to create the figures 22  in the section below. Occupancy data was analyzed exactly the same way. Hence, each building category has four different measures for each month: average weekday electricity consumption, average weekend electricity consumption, average weekday occupancy, average weekend occupancy.  The second part of the analysis in this chapter investigates the daily average electricity consumption of each building category with a focus on weekday-weekend variation in both electricity use and occupancy. February was chosen as the most suitable month because of the one-week long reading break between February 17 and February 23, 2014.  This week represents parts of UBC’s academic year during which buildings have lower occupancy as most students are not on campus, and is similar to summer session. With 2.5°C, February had the lowest average temperature in 2014. Hence, most of the electricity consumption is due to the HVAC system to cycle heated air through the building. The heating process itself is done via boilers that burn natural gas or oil, therefore that process is not included in this calculation. The data was analyzed with the method described for the seasonal fluctuation, but instead of monthly averages we calculated daily averages for each building category. Each day has two measurement points, one for electricity use and one for occupancy. The unit for electricity consumption in the daily analysis is watt-hour per square-meter (Wh/m2), which is the equivalent to kWh/m2/1000.    2.4 Results 2.4.1 Seasonal Electricity Use and Occupancy Variation  The monthly average EUIel of all UBC Classroom buildings and the corresponding Occupancy is shown in Figure 2.2-A. The blue diamonds show the electricity consumption on weekdays (solid blue line) and on weekends (dashed blue line).  23  Classrooms have a fluctuating seasonal weekday EUIel that is strongly correlated with Occupancy. This correlation is even more significant than the impact of Outdoor Temperature. Usually, one expects electricity peak loads in summer when energy intensive cooling is required. This behavior is not present in the data displayed in Figure 2.2-A. The solid blue electricity curve mirrors the solid green occupancy curve and electricity peaks whenever occupancy is highest. Figure 2.2-A also shows that the electricity curve has an upper and a lower boundary meaning that it varies between 7 kWh/m2 and 5 kWh/m2 on weekdays. The EUIel never crosses these boundaries no matter how much Occupancy increases or drops. The weekday analysis further illustrates several key points about UBC’s academic year. The beginning of first term in September causes an Occupancy peaks at roughly 450 people per day which stays constant throughout the entire term until it drops in December due to the Christmas break. The second term starting in January has another Occupancy peak around 350 people per day on average. The Occupancy drop in February is due to the one-week reading break. It illustrates that there are either more students enrolled or more classes offered in the first term in fall, as an average of 100 people per day difference is significant and roughly a 22% drop.  In the summer semester, fewer classes are being offered and Occupancy levels are an average of 150 people per day. We would like to remind the reader at this point that none of the buildings on campus are 100% designates as a Classroom building and always have mixed use of other floor area types that might be used all year around such as offices. Summer is generally a quiet time on campus, as students, faculty, and staff take vacation days. This can be an explanation for the significant Occupancy drop in August.  Both weekend electricity use (dashed blue lines) and weekend Occupancy (dashed green lines) are multiple times lower than the corresponding data on weekdays. With almost no fluctuation at 24  2 kWh/m2, the EUIel on weekends is assumed to be close to the building’s baseline for electricity use, which is the amount of energy that is being consumed by the building independently of its occupancy and any other factors. Weekend Occupancy is between 10 and 50 people per day, and is closely related to other floor area types in the building as Classroom spaces are not expected to be occupied on weekends unless there are conferences and other events on campus.   The air exchange rate in Laboratory buildings is always very high to keep the lab environment free of particles and potential contaminants. The HVAC system accounts for the largest portion of electricity use followed by energy intensive high-end lab equipment such as lasers, helium compressors, and ultra low temperature freezers. The HVAC system and some lab equipment run independently, no matter if there are people in the building or not. Therefore, the EUIel and Occupancy curves are not correlated with each other as shown in Figure 2.2-B.   The seasonal variation of the EUIel on weekdays fluctuates between 14 and 18 kWh/m2. There is slightly more electricity consumed in summer although the number of occupants decreases from April onwards until the end of summer. The increased EUIel is a result of additional heat in summer that has to be removed from the building by the HVAC system. The Occupancy curve follows UBC’s academic year, and is most likely more controlled by other floor area types in the buildings than by Laboratory spaces, unless the labs are teaching labs that are occupied according to UBC’s academic year. The maximum number of occupants on weekdays is expected to be 180 people per day in the first term, and 150 people per day in the second term. Summer hosts only 75 people per day on average.  Laboratory spaces are usually operated 24/7, which explains why these buildings are also occupied on weekends, and the HVAC is still expected to run due to the high air exchange 25  requirement in labs. The weekend consumption is most likely very close to the building’s baseline, and therefore has almost no variation throughout the entire year. The energy intensive HVAC system explains why Laboratory spaces have a three-times higher baseline than Classroom buildings.     The electricity consumption of the average laboratory buildings does not cycle in large volumes between times of higher and lower occupancy, as show in Figure 2.2-B.   Office buildings tend to have lower EUIel because lighting and HVAC systems can be scheduled off at night when not in use. The ventilation requirements differ from lab buildings because indoor air can be recirculated and a smaller amount of fresh outdoor air is needed to maintain sufficient ventilation.  Office buildings use more electricity the more people are in the building because of increased use of lights, computers, and other office equipment by occupants.  However, many UBC offices are open floor concepts and occupied by students. These offices are unlikely to be zoned independently, meaning that no matter how many students work at the same time all lights will always be turned on and the entire space is ventilated. The EUIel in Figure 2.2-C shows some correlation with Occupancy, assuming that it never drops below 8.0 kWh/m2 on weekdays. The maximum EUIel is only 1 kWh/m2 above the minimum, which does not give the EUIel room to fluctuate for any reason. The electricity consumption follows the Occupancy curve between January and March but cycles faster and almost randomly thereafter. Other floor area types in the buildings might have a big impact on this behavior. Another example of this consideration is the manipulation of low energy use of Classroom buildings by Laboratory spaces that use a several magnitude higher amount of electricity.  26  The summer months are usually much quieter because fewer faculty and staff are on campus. However, this reduction due to decreased Occupancy could be balanced by the more electricity intensive cooling that is required.   In summary, the electricity use curve of Office buildings looks similar to that of Laboratory buildings, but the EUIel is much lower.   Occupancy of Library buildings has to most suitable curves to discuss the academic year at UBC. The academic year begins in September, and Occupancy as shown in Figure 2.2-D increases quickly until it peaks in December just before the winter term ends. Library buildings have one of their busiest times in December because it is the exam preparation time for most students. An almost identical behavior is repeated from January to March. During these periods, some Library buildings are even operated 24/7 and on weekends, explaining why the weekend Occupancy curve is essentially a copy of the weekday Occupancy curve but reduced to 200 occupants in winter, and 100 occupants in summer on average. Libraries have seasonally changing hours of operation, which is correlated with the number of occupants per day. The EUIel does not seem to be closely correlated with occupancy according to Figure 2.2-D. Library buildings operate very consistently throughout the year. This means the HVAC system supplies ventilation to the building and all lights are on no matter how many people are in the building. Often, Library buildings have only one light switch for the entire floor and zoned lighting does not exist. Although occupancy increases between January and February, the EUI drops by more than 1 kWh/m2/day. One could assume that more students are taking advantage of the reading break and actually studying at libraries. As libraries have reduced hours of operation during the reading break, the HVAC, if scheduled, operates fewer hours too, and the 27 building’s EUIel decreases although more occupants (but within less hours per day) are using the facility. The EUIel curve on weekdays shows no fluctuation, but has its maximum in July during one of the hottest months in Vancouver when the cooling load is highest. The drop directly afterwards in August could be maintenance related or a data recording error. There are always a lot of energy conservation projects on campus that sometimes show significant impacts on a building’s energy use which would also be reflected in the data.    28 Figure 2.2: Seasonal occupancy (green circles) and electricity consumption (blue diamonds) fluctuation of Classroom (A), Laboratory (B), Office (C), and Library (D) buildings. The solid lines and filled symbols represent the weekday behavior, and the dashed lines and unfilled symbols the weekend. The error bars show the standard error.A	   B	  C	   D	  29 2.4.2 Daily Electricity Use and Occupancy Fluctuation The daily EUIel (blue diamonds) and occupancy (green circles) fluctuation is shown in Figure 2.3. Unlike all other analyses the daily EUIel has the unit watt-hours per square-meter (Wh/m2) that equals kWh/m2/1000. The average Classroom building on UBC campus cycles between 8 Wh/m2 on weekends and 13 Wh/m2 on weekdays as shown in Figure 2.3-A. The room temperature in the building cannot fall below a certain point even on weekends when the buildings are almost unoccupied. This is especially important in winter in order to prevent damage to the building due to cold temperatures. Occupancy and EUIel are directly correlated and peak and drop at exactly the same time, even copying the minor fluctuations from one weekday to the next. Occupancy varies between zero occupants on weekends and peaks of 450 occupants per day on weekdays. Electricity consumption during the reading break week is much lower (ca. 10 Wh/m2) and Occupancy is almost null. The Figure shows once again that Classroom buildings have other floor area types that are occupied when no students are on campus, as it is the case in February during the reading break. The blue electricity use curve in Figure 2.3-B is slightly lower on weekends when Laboratory buildings have very few occupants. However, 23 Wh/m2 on a weekend day is still a fairly high amount of energy when compared to all other building categories. This higher volume is due to several factors, including: the HVAC system is never completely off in Laboratory buildings, even running at reduced capacity on weekends; most labs are still operating during the reading break, with the exception of teaching labs that are unoccupied and mirrored by the green 30 occupancy curve that drops from 170 people per day to 80 people per day; and electricity use is almost entirely unaffected by occupancy decreases, remaining constant around 26 kWh/m2 during periods of high use. Figure 2.3-B supports the previous statement that electricity consumption of Laboratory buildings is not correlated with Occupancy and is a function of an almost constant air exchange rate.   The results discussed for Figure 2.2-B are supported by the daily analysis in Figure 2.3-B. Again, the EUIel curve looks much like that one previously discussed for Laboratory buildings, however, the EUIel is much lower and cycles between 13 Wh/m2 on weekends and 18 Wh/m2 on weekdays.  A lot of student offices are unoccupied during the reading break, and therefore Occupancy accounting for all other offices is only around 70 people per day, which is about 100 people short compared to regular weeks. The electricity curve of Office Buildings in Figure 2.3-C follows Occupancy and drops by roughly 2 Wh/m2 during the reading break. But again, unless the HVAC is controlled by Occupancy levels, it still operates during the reading break no matter how occupied the student or any other office is. Hence, any electricity reduction during the reading break is a result of fewer computers and other office equipment that is being used. Although not very significant in Figure 2.3-C, electricity consumption and Occupancy are much more correlated than it is the case for laboratory buildings.   Library buildings have a most significant EUIel difference between weekdays and weekends when compared to all other building categories. The solid blue line in Figure 2.3-D goes as low as 8 Wh/m2 although occupancy is still fairly high with more than 100 people per day on 31 average. The reason is the reduced operation time on weekends, and the HVAC system that shuts off sooner than on weekdays just like the previously discussed relationship between reduced operation time, occupancy, and EUIel during the reading break. The maximum average Occupancy of Library buildings on UBC campus reaches almost 420 people per day on regular weekdays. This is the highest Occupancy among the four building categories. The curves are similar to Classroom buildings with the difference that Library buildings have an earlier Occupancy drop towards the end of each week. Occupancy on weekends during regular weeks is as high as Occupancy on weekdays during the reading break. The EUIel is almost unaffected by the reduced Occupancy during the reading break, which supports the assumption that the EUIel is more a function of the HVAC system than Occupancy. A lot of libraries on UBC campus have reduced operation hours on Fridays, hence shorter HVAC operation time, fewer occupants and less electricity consumption. The lowest occupancy level is on the weekend just before the reading break but increases slowly throughout the week.    32 Figure 2.3: Electrical Energy Use Intensity (EUIel) in watt-hours per square-meter (blue diamonds) and occupancy in people per day (green circles) in February 2014. The curves show the average value of all buildings that were grouped as Classroom (A), Laboratory (B), Office (C), and Library (D) buildings according to their dominant floor use area. The error bars show the standard error. A	   BC	   D	  33 2.5 Discussion The preliminary electricity analysis in this chapter has given insight how different research building categories consume electricity on a seasonal and daily scale. The average research building can be allocated to one of the four categories, Classroom, Laboratory, Office, and Library, according to its main use that is designated by the dominating floor area. The average EUIel and Occupancy has building category specifically shaped curve. Classroom, Office, and Library buildings show moderate to considerable correlation between electricity use and Occupancy because most building internal systems, such as HVAC and lighting, respond to Occupancy changes or are scheduled off when the building is not occupied. Laboratory buildings, however, do not have the same correlation as the largest portion of the electricity use accounts for the HVAC system that runs independently of how occupied the building is. This statement is supported by both monthly and daily EUIel fluctuation analyses. Seasonal Outdoor Temperature changes have some impacts on the EUIel because additional electricity is required in summer to cool the air before it is distributed throughout the building. All building categories use less electricity on weekends and Occupancy is significantly lower. Building Occupancy also follows seasonal changes, which is in most cases well correlated with UBC’s academic year. Occupancy is highest during the winter months, and lowest during summer sessions.  The weakness of this analysis is that none of the building categories is able to represent any of the buildings accurately because all research facilities are mixed use. Any building categorized as Office building according to its dominant floor area might also have Classroom, Laboratory, and Library spaces. Most significantly, if Classroom, Offices, and Library buildings have Laboratory spaces the EUIel of the Laboratory part will corrupt the data and the shape of all other 34 building categories’ curve as its EUIel is many times larger than the EUIel of Classroom, Offices, and Library buildings. It is not indicated at what percentage each building category is mixed with other floor area types.  Thus, it is still obvious that the four different building types have individual curves with category specific patterns, which supports our assumption that the main purpose of each building is a criterion for higher and lower EUIel.  Although occupants increase the electrical load of a building, the EUIel data can only reflect occupancy fluctuations of buildings that have a low EUIel, and a scheduled HVAC system that does not constantly run at full capacity, as it does for Laboratory buildings. 35 Chapter 3: Linear Regression Analysis of Daily Electricity Consumption The model described in this chapter is a so-called ‘top down’ model. ‘Top down’ models use simple parameters such as Occupancy, Temperature, and Gross Floor Area to develop a relationship between those building specific parameters and electricity consumption.  The analysis is done on a single building scale, and the daily electricity consumption for the entire year of 2014. ‘Top down’ models require a lot of data from a wide range of different facilities in order to make electricity use predictions for future buildings based on a few inputs.  Energy use predictions as they are commonly used for LCC analyses are developed by so-called ‘bottom up’ models. ‘Bottom up’ models are developed for each building individually and require a lot of building specific information such as the glazing or envelope type, the mechanical system’s specifications, and the materials that are being used. The building parameters can reach high complexity and require many assumptions. The more assumptions are being made the higher is the likelihood for errors to appear at some point of the analysis.  The ‘top down’ model presented in this study focuses on simple building parameters that rely on measurable facts, and do not require any assumptions in order to make simple and user-friendly electricity use predictions.  Chapter 2 has already investigated the seasonal energy consumption of Classroom, Laboratory, Office, and Library buildings in separate categories.  In Chapter 3, the buildings are analyzed individually according to their specific ration of Classroom, Laboratory, Office, and Library spaces but all 48 buildings together create the data set for the regression analysis. More building parameters are added for each analysis step while 36 the complexity increases continuously in order to improve the correlation between building parameters and electricity use.  3.1 Linear Regression Analysis The ‘top down’ model in this chapter is based on liner regressions that were analyzed by the numerical computing software MATLAB. The total gross floor area (GF) serves to investigate simple relations between building and electricity consumption. More information, such as the building age (A) and number of floors (F) was added as the analyses got more complex. Adding parameters to the equation is expected to improve the accuracy of the ‘top down’ model by increasing the coefficient of determination (r2), which is a measure that indicates how close the actual data to the fitted regression line is. MATLAB’s “Robust Fit” function is activated to ignore all outliers that are not representative for the rest of the data set. Outliers are a result of data that was measured or recorded incorrectly due to technical failure of the electric meter when the latter is in process of being maintained or replaced. The “Robust Fit” function is an important consideration when dealing with data that is recorded in 15-minute-intervals for an entire year.   After the first simple analysis, the gross floor area was replaced by the five floor allocation variables Classroom (C), Laboratory (L), Office (O), Library (LIB), and Other Spaces (OT) to evaluate the relationship between each building’s specific floor type ratio and the daily electricity consumption.  The next analysis step included more additional parameters again, such as the Building Age, the Number of Floors, the Outdoor Temperature (T) and finally the daily Occupancy (OO). 37 This analysis is based on a data set of 365 entries per building, one every day, with daily entrees for electricity use, outdoor temperature, and occupancy. The number of observations of all 48 buildings combined equals 17,520 in total.  3.2 Data The same data as described in Chapter 2 was used for the ‘top down’ model. The unit for all floor area sizes is square-meters [m2]. The parameters Age, Number of Floors, Classroom area size, Laboratory area size, Office area size, Library area size, and Other Spaces area size are building specific but constant over time. The study does not consider any floor area allocation changes that might have been made during renovations throughout the year. The building Age in “years” [yr] is the time difference between the construction year and the reference year 2014. It is also assumed to be constant because the period in which this analysis was done investigates the electricity consumption in one and the same calendar year. The Number of Floors includes all basement floors if applicable. A building list summarizing all mentioned parameters can be found in Appendix A. Outdoor Temperature [°C] and Occupancy [people/day] have already been discussed in Chapter 2.  3.3 Results The results of our ‘top down’ model are grouped in three different tables. The first table summarizes the results that were achieved in order to increase the accuracy of the equations by adding more building parameters. The second table is a summary of the results based on reallocation Other Spaces a described in detail in Chapter 2.  38 The reallocation process is supposed to increase the weight and significance of the other floor areas Classroom, Laboratory, Office, and Library. Finally, the third table shows a different approach in which weekday electricity use and weekend electricity use are analyzed separately. 3.3.1 ‘Top Down’ Model: Building Parameters  A summary of different ‘top down’ model stages with increasing accuracy is shown in Table .31. The variables are explained in the introducing section 3.2 of this chapter. The results of each equation are explained in detail below the table. Table 3.1: Summary of the ‘top down’ model stages. Each equation has added parameters to increase the accuracy of the outcome. The variables in the same order as they appear are: GF: Gross floor area; A: Building age; C: Classroom; L: Laboratory; O: Office; LIB: Library; OT: Other Spaces; F: Number of floors; T: Outdoor Temperature; OO: Occupancy.  Eq. 1: 𝑇𝐸 = 1+ 𝜉 ∙ 𝐺𝐹 Estimate Unit R-squaredIntercept 63.099 kWh 0.743 GF 𝜉 = 0.37612** kWh/m2  Eq. 2: 𝑇𝐸 =   1+ 𝜉 ∙ 𝐺𝐹 + 𝜃 ∙ 𝐴 Estimate Unit R-squaredIntercept cept2837.9 kWh 0.774 GF 𝜉 = 0.3171** kWh/m2 A 𝜃 = -54.429** kWh/year   Eq. 3: 𝑇𝐸 =   1+ 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 Estimate Unit R-squaredIntercept C600.43 kWh 0.832 C 𝛼 = -0.40458** kWh/m2 L 𝛽 = 0.95093** kWh/m2 O 𝛿 = -0.039709** kWh/m2 LIB 𝜀 = 0.4522** kWh/m2 OT 𝜁 = 0.39318** kWh/m2   Eq. 4: 𝑇𝐸 =   1+ 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 39 Estimate Unit R-squaredIntercept 3004.8 kWh 0.858 C 𝛼 = -0.23117** kWh/m2 L 𝛽 = 0.85965** kWh/m2 O 𝛿 = 0.10656** kWh/m2 LIB 𝜀 = 0.34638** kWh/m2 OT 𝜁 = 0.31853** kWh/m2 A 𝜃 = -49.989** kWh/year F 𝜗 = -88.468** kWh/floor T 𝜅 = 14.665** kWh/°C   Eq. 5: 𝑇𝐸 =   1+ 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 + 𝜆 ∙ 𝑂𝑂 Estimate Unit R-squaredIntercept 2654.1 kWh 0.868 C 𝛼 = -0.42212** kWh/m2 L 𝛽 = 0.8166** kWh/m2 O 𝛿 = 0.081731** kWh/m2 LIB 𝜀 = 0.22806** kWh/m2 OT 𝜁 = 0.31147** kWh/m2 A 𝜃 = -45.49** kWh/year F 𝜗 = -82.95** kWh/floor T 𝜅 = 23.866** kWh/°C OO 𝜆 = 2.9968** kWh/person Even the simplest relationship between building parameters and electricity consumption already exhibits a fairly accurate result. Equation 1 shows how dominant the building’s Gross Floor area in correlation with the total electricity consumption is. Summarized in words, the result of Equation 1 states that bigger buildings use more electricity than smaller buildings, which is not an unexpected result. A variable is considered to be significant if the pValue is 0.05 and below. Therefore, both intercept and Gross Floor area are significant, with the latter accounting for 0.38 kWh/m2/day. Increasing a building’s Gross Floor area by one square-meter adds 0.38 kWh to its daily electricity use. This result is not building category specific and could be applied to any research building on UBC campus. The coefficient of determination (r-squared) shows that 74.3% of the measure points approximate with the regression line. Again, the analysis excludes 40 any outliers that are not representative of the majority of the data. This result is unexpectedly precise for an equation that creates a relationship between one variable (GF) and electricity consumption.  Equation 2 takes the building Age into account, which results in a slightly better r-squared of 77.4%. The intercept and both variables, total Gross Floor area and building Age, are considered to be significant. The estimate for the building Age variable is negative (-54.429 kWh/m2/day), which means that older buildings use less electricity than newer buildings. This result might not account for regular buildings, as most modern buildings are much more efficient than those that were built decades ago due to better insulation, glazing, and mechanical equipment. However, on UBC campus where most recently built buildings are high-end research laboratories that consume much more energy than old Office buildings without HVAC systems, a negative estimate for building Age is appropriate.  Equation 3 is a new approach to address a building’s electricity consumption because we divided the total gross floor area into Classroom area, Laboratory area, Office area, Library area, and Other Spaces area. The r-squared shows a positive reaction and reaches 83.2%, which is the best result of any analysis so far. Laboratory has the highest significance followed by Library, Other Spaces, the intercept, Classroom, and finally Office areas. We have shown in Chapter 1 of that Classroom and Office have a much smaller EUIel than all other floor types. That explains why both variables have negative estimates, as they are supposed to balance the excess electricity consumption of Laboratory, Library and Other Spaces that are in every building at a certain ratio. This equalization prevents the overestimation of the EUIel of buildings that have 41 majorly Office and Classroom spaces. Since none of the buildings is either a 100% Classroom or a 100% Office building, every research building will have positive total EUIel estimate.    Equation 4 has another complexity level added to it. Unlike the building’s floor area, which does not change throughout the year of investigation, other parameters have daily records just like the electricity consumption itself. One of them is the Outdoor Temperature and this parameter is added to Equation 4. The building Age, which had already proven its significance in Equation 2, and the Number of Floors are also considered in Equation 4. The equation achieved an accuracy of 85.8%, giving evidence that a research building’s EUIel is significantly determined our chosen parameters. The intercept, Laboratory, and the building Age are of equally high significance. Laboratories contribute 0.86 kWh/m2/day to each building’s total electricity consumption.  The Age is negative again, just like Classroom and Number of Floors, which translates into buildings with more floors use less electricity than buildings with fewer floors. One might be surprised by this result because buildings with more floors are expected to have more floor area and use more electricity according to Equation 1. However, the floor area is already taken into account by the five floor categories, Classroom, Laboratory, Office, Library, and Other spaces. The negative estimate for Number of Floors means that the building with a certain Gross Floor area distributed over several floors uses lesser electricity than a building with the same Gross Floor area distributed over one floor. A lot of more-story-buildings on campus have large atriums and open areas. That is also the reason why Other Spaces can account for a great portion of some buildings. Atriums have not the same ventilation requirements as all other “useful” areas in the building. Adding “non-useful” spaces to a building, which is often the case for more story buildings, decreases the EUIel.   42 Library and Other Spaces contribute with approximately 0.34 kWh/m2/day each, and have almost the same significance. The Outdoor Temperature in Equation 4 is the variable for seasonal changes. It is positive because some building categories use more electricity on hot days due to cooling and air ventilation. However, its significance is not as high as most other parameters because a lot of building types, such as Laboratory, have very high air exchange rates that are not correlated with the Outdoor Temperature. The seasonal fluctuating EUIel of different research building types has been discussed in detail in Chapter 2.  Equation 5 is the final result, showing the relationship between electricity use and all parameters that are considered in this study. The r-squared reaches the overall maximum of 86.8% giving evidence that almost 90% of the 17,520 observations approximate to the linear regression line. This result could be considered the maximum possible correlation, considering that adding more variables to Equation 3 (83.2%) and Equation 4 (85.8%) has not shown any significant improvements in the r-squared. The estimates for the variables show only minor differences compared to those discussed for Equation 4.  Occupancy is the second most significant parameter and shows that each occupant adds 3.00 kWh/day to any building’s EUIel. Although electricity consumption in Laboratory buildings and Occupancy are not correlated, as investigated in Chapter 2, Classroom, Office, and Library areas have already shown in the previous chapter the significant impact of Occupancy on the EUIel. This result is also supported by the results in this analysis. The estimate for Laboratory is almost unchanged after adding the occupancy parameter and drops negligibly from 0.86 kWh/m2/day (Equation 4) to 0.82 kWh/m2/day (Equation 5). The same difference is much more significant for the Library 43 parameter that is estimated to be 0.35 kWh/m2/day in Equation 4 and only 0.23 kWh/m2/day when occupancy is considered in Equation 5.  In summary, the data shows that Occupancy is an important factor for Library and Classroom areas that shifts significant weight away from “energy added per square-meter” and towards “energy added per occupant”. Similar to Laboratory areas, Other Spaces also has no significantly different estimates when the Occupancy variable is added. It can be argued that occupants add no additional electrical load to atriums, stairwells, mechanical rooms, etc. that have always the same light usage or ventilation rates independent of how many occupants are within the area.  Equation 4 and 5 have delivered greater understanding of what the electricity demanding building parameters are. Both equations are highly accurate and almost 90% of all 17,520 observations approximate to the linear regression line, which makes them two eligible ‘top down’ models to be tested against LEED ‘bottom up’ model results in order to proof which model predicts the building EUIel more precisely.  3.3.2 ‘Top Down’ Model: Reallocated Other Spaces It is questionable whether Other Spaces should be an individual category and treated as such, although it does not contribute to any useful research area. As previously discussed, Other Spaces is an invented floor category that summarizes all spaces in a building that are not Classroom, Laboratory, Office, and Library areas. This category is highly diverse and includes elevators, stairwells, mechanical rooms, atriums, hallways, washrooms, and sometimes cafes and other areas used for social purposes. It is obvious that buildings cannot be operated and inhabited without these rooms. However, Other Spaces do not have the same functionality as all other research areas do and therefore they do not contribute to the “useful” floor area. In the following 44 analysis, we reallocated Other Spaces and distributed the floor area to Classroom, Laboratory, Office and Library spaces according to the percentage of each floor area type compared to the total Gross Floor area. For example, if 5,000 m2 Other Spaces floor area were to be reallocated to a building that is composed of 10% Classroom, 50% Laboratory, 40% Office and 0% Library (total gross floor - Other spaces), the reallocated floor areas would be CR = 10% Classroom + 0.1 x 5,000 m2 Other Spaces, LR = 50% Laboratory + 0.5 x 5,000 m2 Other Spaces, OR = 40% Office  + 0.4 x 5,000 m2 Other Spaces, and LIBR = 0% Library + 0 x 5,000 m2 Other Spaces. Thereallocation makes it possible to get the building’s actual EUIel, which would be much higher if Other Spaces were simply subtracted from the total gross floor area.   Table 3.2: Summary of the ‘top down’ model with reallocated floor areas. The category “Other Spaces” is reallocated to the other floor area types according to their percentage of the total gross floor area. The variables in the same order as they appear are: CR: Classroom reallocated; LR: Laboratory reallocated; OR: Office reallocated; LIBR: Library reallocated; A: Building age; F: Number of floors; T: Outdoor Temperature; OO: Occupancy.  Eq. 7: 𝑇𝐸 =   1+ 𝛼 ∙ 𝐶! + 𝛽 ∙ 𝐿! + 𝛿 ∙ 𝑂! + 𝜀 ∙ 𝐿𝐼𝐵! Estimate Unit R-squaredIntercept C649.54 kWh 0.828 CR 𝛼 = -0.037232** kWh/m2 LR 𝛽 = 0.68542** kWh/m2 OR 𝛿 = 0.14561** kWh/m2 LIBR 𝜀 = 0.43699** kWh/m2  Eq. 8: 𝑇𝐸 =   1+ 𝛼 ∙ 𝐶! + 𝛽 ∙ 𝐿! + 𝛿 ∙ 𝑂! + 𝜀 ∙ 𝐿𝐼𝐵! + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 Estimate Unit R-squaredIntercept 3075.9 kWh 0.856 CR 𝛼 = 0.024946** kWh/m2 LR 𝛽 = 0.59928** kWh/m2 OR 𝛿 = 0.19577** kWh/m2 LIBR 𝜀 = 0.33653** kWh/m2 A 𝜃 = -49.875** kWh/year F 𝜗 = -92.065** kWh/floor T 𝜅 = 14.808** kWh/°C 45 Eq. 9: 𝑇𝐸 =   1+ 𝛼 ∙ 𝐶! + 𝛽 ∙ 𝐿! + 𝛿 ∙ 𝑂! + 𝜀 ∙ 𝐿𝐼𝐵! + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 + 𝜆 ∙ 𝑂𝑂 Estimate Unit R-squaredIntercept 2712.5 kWh 0.868 CR 𝛼 = -0.08438** kWh/m2 LR 𝛽 = 0.57479** kWh/m2 OR 𝛿 = 0.18439** kWh/m2 LIBR 𝜀 = 0.25871** kWh/m2 A 𝜃 = -45.987** kWh/year F 𝜗 = -84.826** kWh/floor T 𝜅 = 24.166** kWh/°C OO 𝜆 = 3.0376** kWh/person Equation 7 is the equivalent to Equation 3 in the previous discussion but with reallocated floor areas. The only parameter that changes significantly, and is therefore worth mentioning, is the estimate for Office areas. The negative variable in Equation 3 changes to 0.15 kWh/m2/day and is significant. The estimate for Classroom areas is still negative and the only category that loses significance as all other parameters gain or remain as significant as they were before Other Spaces were reallocated. Equation 8 is the equivalent to Equation 4 but with reallocated floor areas. The equation has only positive estimates for all floor area categories. The reallocation of Other Spaces shifts significant weight towards Classroom and Office areas, while the estimates for Laboratory and Library decrease slightly. It seems that the weight of each parameter in Equation 8 is more equally distributed than in all previous equations. Equation 8 will be part of the final analysis and tested against some selected buildings’ actual and the ‘bottom up’ model’s electricity consumptions.  46 Equation 9 shows very similar results to its equivalent Equation 5 that includes Other Spaces. The reallocation process does not seem to have any impact on the variables in this equation. The r-squared of all 3 equations discussed in this paragraph are similar to the correspondingequations that include Other Space in the previous analysis. In summary, the reallocation of Other Spaces to all other floor area types it is not as relevant as expected. The variables change only slightly, which is not expected to have a significant impact on the final EUIel estimate. Equation 8 benefits from the reallocation because it has no negative estimates, which makes it easier to interpret the variables.  3.3.3 ‘Top Down’ Model: Weekday & Weekend Separately The following analysis investigates the electricity consumption on weekdays (EWD) and the electricity consumption on weekends (EWE) separately. This analysis is based on the same electricity data set that was being used for the seasonal fluctuation analysis in Chapter 2. The difference is that instead of using daily data, the set is divided into two separate data groups with one containing the monthly electricity use on weekdays and the other containing the monthly electricity use on weekends. Occupancy data is analyzed exactly the same way in order to create a relationship between weekday occupancy (OWD) and electricity consumption on weekdays, as well as weekend occupancy, (OWE) and electricity consumption on weekends. The daily temperature is replaced by monthly averages. The 48 research buildings and all other parameters remain the same. Accordingly, the generated data set, compared the previously discussed regressions, is reduced to 576 observations.  Unlike all other ‘top down’ models, the variables 47 are not added one by one because only the correlation of electricity use and all parameters together is of interest. Table 3.3: Electricity use on weekdays (EWD) and electricity use on weekends (EWE) are analyzed separately. The linear regressions have the following variables: C: Classroom; L: Laboratory; O: Office; LIB: Library; OT: Other Spaces; A: Building age; F: Number of floors; T: Outdoor Temperature; OWD: Occupancy on weekdays; OWE: Occupancy on weekends.    Eq. 9: 𝐸𝑊𝐷 =   1 + 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 + 𝜇∙ 𝑂𝑊𝐷Estimate Unit R-squaredIntercept 60177 kWh 0.801 C 𝛼 = -5.6474 kWh/m2 L 𝛽 = 20.678 kWh/m2 O 𝛿 = 4.1692 kWh/m2 LIB 𝜀 = 6.5511 kWh/m2 OT 𝜁 = 3.9542 kWh/m2 A 𝜃 = -1071.7 kWh/year F 𝜗 = -1902.9 kWh/floor T 𝜅 = 694.58 kWh/°C OWD 𝜇 = 51.262 kWh/person    Eq. 10: 𝐸𝑊𝐸 =   1 + 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 + 𝜈 ∙ 𝑂𝑊𝐸 Estimate Unit R-squaredIntercept 24030 kWh 0.783 C 𝛼 = -2.8877 kWh/m2 L 𝛽 = 6.7567 kWh/m2 O 𝛿 = 0.08834 kWh/m2 LIB 𝜀 = 2.187 kWh/m2 OT 𝜁 = 3.2192 kWh/m2 A 𝜃 = -425.99 kWh/year F 𝜗 = -554.73 kWh/floor T 𝜅 = 109.82 kWh/°C OWE 𝜈 = 12.173 kWh/person Equation 9 is the linear robust regression investigating the relationship between the electricity consumption on weekdays and the nine variables as shown in Table 3.3. Equation 10 is the same model for electricity consumption on weekends. The estimates for Equation 9 are about 20 times 48 higher and Equation 10 and are 10 times higher compared to previously discussed ‘top down’ models. This is because the weekday/weekend data is the accumulated EUIel for approximately 20 weekday-days and respectively 10 weekend-days (holidays in British Columbia included). This number of weekday-days to number of weekend-days relationship is month specific and requires a day-by-day analysis in order to guarantee that all matching weekday EUIel-Occupancy-pairs and all weekend EUIel-Occupancy-pairs were grouped and summarized correctly.    Equation 9 has a negative estimate for Classroom, which is not significant (pValue > 0.05). Laboratory spaces account for 20.7 kWh/m2/month, which is about 20 times more than what was shown in all previous equations for Laboratory spaces. Office, Number of Floors, and Temperature are also parameters that have no significance in this equation. Library and Other spaces account for 6.6 kWh/m2/month and 4.0 kWh/m2/month respectively. The building Age parameter is negative and corrects the electricity consumption by -1071.7 kWh/m2/month. Weekday occupancy has a positive estimate and adds 51.262 kWh/occupant/month to the total EUIel.      Equation 10 has lower estimates because the electricity consumption is lower on weekends as discussed in Chapter 2. Classrooms are significant but still negative. Office, Number of Floors, and Temperature are not significant again. The Occupancy levels on weekends are very low and accordingly the variable is a much lower significance in the weekend model compared to Equation 9. The highest estimate is Laboratory spaces 6.8 kWh/m2/month but still 70% lower than on weekdays.  49 3.4 Discussion Adding more parameters to the ‘top down’ model has successfully resulted in a continuously improving accuracy shown by the increasing r-squared value. This improvement, however, slows down after the first added parameters and does not seem exceed a higher r-squared than 0.86 for the given variables. Most equations have positive and negative parameters that either add or subtract electricity to the building’s total consumption. However, the floor allocations are not supposed to have negative estimates, as positive floor space should always result in positive energy consumption.  Equation 1 that has already shown the positive correlation between electricity consumption and Gross Floor area. The parameters for Classroom and Office area result in negative estimates in most equations but it is not expected to cause a negative overall electricity consumption for any building on UBC campus. All research buildings are mixed-use and even dedicated Classroom or Office buildings have a large percentage of other floor types that, together with all positive parameters such as temperature, occupancy, etc., are able to account for a positive overall electricity consumption. The ‘top down’ model does not consider all possible building parameters that might be important for estimating electricity consumptions. Other variables could contribute to further increase in the accuracy of the model, but added complexity to this analysis is not desired as the ‘top down’ model’s simplicity is its strength and main goal of this study. All building parameters were carefully chosen according to available data, and in order to represent a large group of research buildings. The regression analyses have delivered a set of equations that are accurate but also easily applicable for any research building. How close the ‘top down’ models’ predictions to the actual electricity use are is discussed in the next chapter. 50 Chapter 4: Our Results Compared to LEED ‘Bottom Up’ Models The Leadership in Energy and Environmental Design organization has been addressed in the introductory part of this study. Energy modeling software has been developed to predict future buildings’ energy intensity during the design process of a construction project. However, ‘bottom up’ models exist for only very few UBC buildings. As previously mentioned, ‘bottom up’ models are very time intensive because they require a significant amount of detailed information, and therefore they are often avoided. Apart from technical specifications for boiler, chiller, fan coils, and other mechanical equipment, detailed information about the building’s envelope, glazing, structure, lighting, plug loads, etc. has to be provided and applied to the software tool. The full list of required input data can be found online in the latest version of the LEED Canada Supplementary Energy Modeling Guidelines7. In Chapter 4 we compare the ‘bottom up’ model’s electricity estimates of ten UBC research buildings to the predications of our own ‘top down’ models. Meanwhile, the ten buildings have been commissioned, and actual electricity data has been recorded in the same manner as for all other buildings in this study.  7 The Canada Green Building Council provides all information regarding the LEED energy modeling requirements: (September 18, 2016) 51 4.1 Data Energy & Water Services collects UBC’s utility data, and is responsible for all energy conservation projects in order to operate buildings on campus as efficiently as possible. Unfortunately, only ten UBC buildings have ever been modeled according to the ‘bottom up’ model.  These models are used to design buildings that meet all LEED criteria in order to get LEED certified. The models estimate the energy savings and the capital cost the capital cost of each energy conservation measure (ECM) (Capehart et al., 2012). Oftentimes, several ECMs are bundled to support the design team, which usually decides on the bundle that has the shortest payback period. Hence, the selection of buildings suitable for the comparison with our developed ‘top down’ model is limited. Among the modeled buildings is one Classroom, four Office, and five Laboratory buildings, with some buildings also containing Library areas. Table 4.1 summarizes the collection of ‘bottom up’ models of UBC research buildings, and their estimation for electricity consumption.  Table 4.1: List of the only ten LEED modeled UBC campus buildings, their building category, and the modeled annual electrical Energy use. Code	   Building	  Name	   Category	   Electricity	  B1	  B2	  B3	  B4	  B5	  Aquatic	  Eco	  Research	  Laboratory8	  Allard	  Hall9	  Biology	  West	  &	  South	  Building10	  Buchanan	  Block	  A,	  B,	  C,	  D,	  E11	  Chemistry	  D	  Building12	  Office	  Office	  Laboratory	  Classroom	  Laboratory	  332,082	  kWh	  1,659,391	  kWh	  6,401,133	  kWh	  1,250,622	  kWh	  657,329	  kWh	  8Information provided by UBC Energy & Water Services  9Energy Report: (Allard Hall - eQUEST Plant and Water-Side HVAC Summary Report, 2009) 10Energy Report: (Shaun Martin Consulting, 2011) 11Energy Report: (Stantec Consulting Ltd., 2010c) 52 Code	   Building	  Name	   Category	   Electricity	  B6	  B7	  B8	  B9	  B10	  Centre	  for	  Interactive	  Research	  on	  Sustainability13	  Earth	  Sciences	  Building14	  Friedman	  Building15	  Life	  Sciences	  Centre16	  Pharmaceutical	  Sciences	  Building17	  Office	  Office	  Laboratory	  Laboratory	  Laboratory	  432,472	  kWh	  4,296,746	  kWh	  310,568	  kWh	  7,311,051	  kWh	  9,177,464	  kWh	  4.2 Methods Five of the buildings in Table 4.1 (B1, B4, B5, B8, B9) are among the 48 research buildings that have been selected for this study. In Chapter 2 and 3 Buchanan Block A, B, C, D, E account for two separate buildings (Buchanan A, B, C & Buchanan D, E), unlike in this chapter as the five building blocks were modeled together. Therefore, B4 is considered to be one building in this chapter. The 48 research facilities in Chapter 2 and 3 were reduced by those buildings that are in common with Table 4.1. This makes the evaluation of the ‘top down’ model results and ‘bottom up’ models better comparable. The analyses in Chapter 3 were repeated with the remaining 42 buildings. The reduced sample did not change the linear regressions significantly, and are hence not discussed again. The results can be found in Appendix B.  Occupancy data was not available for the five additional buildings in Table 4.1, and had to be estimated according to the average occupancy per square-meter of each building category 12Energy Report: (Stantec Consulting Ltd., 2010d) 13Energy Report: (Stantec Consulting Ltd., 2010a) 14Energy Report: (Stantec Consulting Ltd., 2011) 15Information provided by UBC Energy & Water Services 16Energy Report: (GF Shymko & Associates Inc, n.d.) 17Energy Report: (Stantec Consulting Ltd., 2010b) 53 (Classroom, Laboratory, Office, Library).  The hereby gained values were multiplied by each building’s gross floor area fraction of the matching floor area type. Hence, the estimated Occupancy of all five buildings is both building size and building category specific.  4.3 Results The results of this analysis are summarized in Table 4.2 with the total annual electricity consumption calculated by the corresponding ‘top down’ model in the first row of each column. Building B1 is an Office building with an annual electricity consumption of 499,479 kWh. All ‘top down’ models except from one overestimate the buildings electricity use by more than 150%. However, ‘top down’ model 5 (TDM-5) is able to predict almost the exact actual value. Not even the ‘bottom up’ model gets as close, and estimates 332,082 kWh which is -34% below the actual electricity use. The reason why the ‘top down’ models overestimate the actual electricity use is because building B1 has almost 10% lab space (at reallocated “Other Spaces” conditions) according to UBC Infrastructure Development. The laboratories are rarely used and cannot be compared to high-end, energy intensive laboratories. The relatively large portion of Laboratory spaces add too much electricity to building’s overall use.   The next building is a mixed-use of Office and Library (B2) with an annual electricity use of 2,209,067 kWh. The ‘top down’ models 1, 2 and 4 (TDM-1, TDM-2, TDM-4) make very close predictions and are only 8% short compared to the ‘bottom up’ model that again is off by -25%. Building B3 is hard to capture by our ‘top down’ models because a significant portion of the building’s thermal energy use has shifted to electricity due to installed heat pumps. The operation of heat pumps is very electricity intensive but none of the ‘top down’ models has a parameter that could account for this specific technical equipment. ‘top down’ model 3 (TDM-3) 54 underestimates the actual electricity use of 10,801,161 kWh/year by only 17%. ‘Bottom up’ models consider a large variety of mechanical equipment and allows heat pumps to be added to the energy models. Thus, the ‘bottom up’ model result underestimates the actual electricity consumption by -41%.  Building B4 is a Classroom building that uses 1,634,282 kWh/year electricity annually. TDM-3 is the best approach but still overestimates the electricity by 38%. No other ‘top down’ model makes any more accurate predictions. The ‘bottom up’ model is only off by -23% and the best approach for this building example.  The Laboratory building B5 has one of the highest EUIel on campus, accounting for 5,525,286 kWh/year. The most accurate result is again estimated by TDM-3, which predicts a result only -23% below the actual consumption. The ‘bottom up’ model is not able to make an accurate prediction for building B5 and underestimates the actual result by -88%.  Building B6 is a Green Building, and one of the most sustainable research facilities on UBC campus. Although LEED Platinum certified and considered as such in the energy model, the actual electricity use differs from the LEED ‘bottom up’ model by -55%.  Office building B7 uses 5,060,175 kWh/year and was best modeled by TDM-1 and TDM-2, although both ‘top down’ models underestimate the actual use by roughly -40%. The ‘bottom up’ model is more accurate and only -15% below the actual electricity use. B8 is a Laboratory facility that consumes 604,102 kWh/year. The three models TDM-1, TDM-2, and TDM-4 overestimate the actual electricity use by roughly 30% but are still more precise than the ‘bottom up’ model that underestimates the actual use by almost -50%. No model was able to predict the annual electricity of Building B9. The high-end laboratory uses 20,763,958 kWh/year, which is the highest single building electricity use on campus. 55 Equation 3 and the ‘bottom up’ model are similarly accurate in estimating the Laboratory building’s (B10) annual electricity consumption. TDM-3 is 36% above, and the ‘bottom up’ model is 33% below the building’s actual electricity use of 13,768,578 kWh.  A summary of the results discussed in this chapter is shown in Figure 4.1. The estimated energy use of ‘top down’ model 1 relative to each building’s actual electricity use is represented by the blue diamonds. The red line is the mark where actual (measured) and estimated electricity use would match 100%. This applies also to ‘top down’ model 2 and the ‘bottom up’ model’s results, which are represented by the violet squares and the green triangles respectively. All measure points above the red line overestimated, and all measure points below the red line underestimated the actual electricity consumptions. Some model results were to not accurate enough, and are not represented in Figure 4.1.  56 Figure 4.1: The red line is the 100% mark, which represents the actual electricity consumption of each building. The blue diamonds and the violet squares show the results of top-down model 1 and 2 respectively, relative to the actual electricity use of the building. The green triangles show the results of the ‘bottom up’ model relative to the actual electricity use.   	  -­‐	  	  	  	  0.2	  	  	  0.4	  	  	  0.6	  	  	  0.8	  	  	  1.0	  	  	  1.2	  	  	  1.4	  	  	  1.6	  	  	  1.8	  	  	  2.0	  	  AERL	   Allard	  Hall	  BIO	  West	  &	  South	  BUCH	  A,B,C,D,E	  CHEM	  D	   CIRS	   ESB	   Friedman	   Life	  Sciences	  Pharmacy	  Top	  Down	  Model	  2	   Top	  Down	  Model	  3	   BoXom	  Up	  Model	  57 Table 4.2: Summary of the ‘top down’ model results compared to the LEED ‘bottom up’ models. The five equations from Chapter 3 with the highest accuracy were used to estimate the ten buildings’ annual electricity consumption. The variables in the same order as they appear are: GF: Gross floor area; A: Building age; C: Classroom; L: Laboratory; O: Office; LIB: Library; OT: Other Spaces; F: Number of floors; T: Outdoor Temperature; OO: Occupancy; CR: Classroom reallocated; LR: Laboratory reallocated; OR: Office reallocated; LIBR: Library reallocated. All results in kWh/year. The equations can be found in Appendix B. BLD	   Actual	  El.	  Use	  TDM-­‐1	  Equation	  11	  TDM-­‐2	  Equation	  12	  TDM-­‐3	  Equation	  14	  TDM-­‐4	  Equation	  13	  TDM-­‐5	  Equation	  15	   ‘Bottom	  Up’	  Model	  B1	   	  499,479	   	  1,338,686	   168%	   	  1,292,176	   159%	   	  2,620,613	   425%	   	  1,395,128	   179%	   	  505,612	   1%	   	  332,082	   -­‐34%	  B2	   	  2,209,067	   	  2,057,753	   -­‐7%	   	  2,036,507	   -­‐8%	   	  5,803,984	   163%	   	  2,042,146	   -­‐8%	   	  1,359,120	   -­‐38%	   	  1,659,391	   -­‐25%	  B3	   	  10,801,161	   	  2,156,692	   -­‐80%	   	  2,158,800	   -­‐80%	   	  8,992,813	   -­‐17%	   	  2,044,457	   -­‐81%	   	  2,628,350	   -­‐76%	   	  6,401,133	   -­‐41%	  B4	   	  1,634,282	   	  (174,623)	   -­‐111%	   	  4,045	   -­‐100%	   	  2,247,941	   38%	   	  117,231	   -­‐93%	   	  163,136	   -­‐90%	   	  1,250,622	   -­‐23%	  B5	   	  5,525,286	   	  272,354	   -­‐95%	   	  453,357	   -­‐92%	   	  4,259,758	   -­‐23%	   	  208,446	   -­‐96%	   	  1,503,593	   -­‐73%	   	  657,329	   -­‐88%	  B6	   	  951,686	   	  1,537,603	   62%	   	  1,545,970	   62%	   	  9,944,403	   945%	   	  3,059,485	   221%	   	  2,508,384	   164%	   	  432,472	   -­‐55%	  B7	   	  5,060,175	   	  3,056,614	   -­‐40%	   	  3,138,858	   -­‐38%	   	  8,992,813	   78%	   	  2,044,457	   -­‐60%	   	  2,628,350	   -­‐48%	   	  4,296,746	   -­‐15%	  B8	   	  605,102	   	  765,200	   26%	   	  783,897	   30%	   	  3,448,345	   470%	   	  767,072	   27%	   	  1,016,980	   68%	   	  310,568	   -­‐49%	  B9	   	  20,763,958	   	  10,206,513	   -­‐51%	   10,040,582	   -­‐52%	   	  39,135,553	   88%	   	  10,028,727	   -­‐52%	   	  10,652,835	   -­‐49%	   	  7,311,051	   -­‐65%	  B10	   	  13,768,578	   	  5,212,042	   -­‐62%	   	  5,124,992	   -­‐63%	   	  18,708,557	   36%	   	  5,301,102	   -­‐61%	   	  4,900,311	   -­‐64%	   	  9,177,464	   -­‐33%	  58 4.4 Discussion  The ‘top down’ model that has been developed in this study has the ability to estimate most of the ten research buildings’ annual electricity consumption more accurately than the LEED ‘bottom up’ model does.  This study has not considered any extremely electricity intensive mechanical equipment that is installed in some buildings in order to replace common heating and cooling technologies. Therefore, the ‘top down’ model should not be applied to buildings that have heat pumps or similar equipment, without adding a parameter to the linear regressions that is able to consider the additional energy load. Any additional parameter requires a set of buildings equipped with heat pumps that is big enough to be representative as another category in this analysis. The bigger the sample size the more accurate the predictions are.   After the comparison in Chapter 4 it is obvious that one single equation is not able to cover the full spectrum of research buildings and their unique energy consumptions. Some equations are more accurate for buildings that have low EUIel, such as Classroom and Office buildings. Other equations cover more energy intensive buildings. If the ‘top down’ model had more parameters to add and subtract electricity according to building specific characteristics, a single equation might be able to address a wider range of research buildings.   The scope of this study is the development of a ‘top down’ model that estimates the electricity consumption of research buildings based on very few building parameters. The simplicity of the model is its strength, and that allows its application to a wide range of research buildings on UBC campus.  59 Two of the previously discussed ‘top down’ models are chosen to be the final result of the developed model in this study. These two models combined are expected to make the most accurate electricity use predictions for UBC research buildings. The pair could either be ‘top down’ model 1 and 3 ore model 2 and 3 as both combinations cover the same amount of buildings in the discussed building sample. ‘Top down’ model 1 and 2 differ only in the additional parameter “Occupancy” that is added to model 2. Occupancy adds very little to the total annual electricity consumption, giving the reason why model 1 and 2 estimate almost the same amount of electricity use for all buildings in Table 4.2. ‘Top down’ model 3 has the most accurate predictions for four of the ten buildings and appears to be the best applicable model for electricity intensive facilities. Either model 1 or 2 is more suitable for Office and Classroom buildings, as well as Laboratory buildings that are not expected to have very high electricity requirements. The ‘top down’ model pair (1 AND 3 or 2 AND 3) covers seven out of ten buildings in Table 4.2 and either underestimated (four buildings) or overestimated (three buildings) the electricity use within an accuracy range of 7 to 40%.  The ‘bottom up’ model estimates the electricity use of two buildings more accurately than the linear regression model, although it is expected to be much more precise due to its level of complexity. The ten buildings used for the analysis in Chapter 4 do not account for a large enough sample size to conclude that the here presented model is able to compete ‘bottom up’ models Furthermore, the simplicity of the ‘top down’ model has its limitations, which are discussed in Chapter 5.     60 Chapter 5: Conclusion 5.1 The Quality of Our Results Although all research buildings are unique in design, functionality, and EUIel, each facility can be allocated to one of the four categories, Classroom, Laboratory, Office, and Library, according to its dominant floor area type. The building specific combination of up to all four of these categories provides a lot of information on how energy is being used throughout the year, and whether Occupancy and Outdoor Temperature have any effect on the amount of consumed electricity.  The reason why some building types use more electricity than others, and the strong correlation with specific ventilation requirements, has been discussed in Chapter 2. The preliminary results in this chapter do not allow for assumptions about a building’s EUIel because the electricity intensity can be quite different even amongst buildings within the same category. The Figures in Chapter 2 provide a range of expected energy use that is specific for the particular building category. The intention of Chapter 2 was not to predict how much electricity is being used by buildings due to their unique characteristics. Instead, the aim was to gain a better understanding of how buildings on campus consume electricity, why and when this electricity consumption varies, and what the main drivers for this behavior are. Understanding the drivers for energy use is necessary before choosing the parameters for a ‘top down’ model that estimates the annual electricity use of research buildings. Occupancy and Outdoor Temperature, for instance, show some correlation with electricity consumption of Office and Classroom buildings, and should be considered for these categories. However, both parameters might be neglected for other facilities because neither parameter has a significant impact on their overall electricity consumption due to 61 constant ventilation requirements and the energy intensive HVAC system that accounts for the largest portion of electricity use but is not correlated with Occupancy. The outcome that can be taken from Chapter 2 is that different building categories have different input data requirements in order to estimate their energy consumption. The input parameters have different values for each building category. In summary, the results in Chapter 2 have contributed to clarify prior assumptions regarding the relationship between seasonal electricity fluctuation, Occupancy, Outdoor Temperature, and building category. Chapter 3 has investigated the infrastructural and usage related parameter interactions with the EUIel on a different level of complexity. Adding more variables increases the ‘top down’ model’s accuracy, as proven by the continuously growing r-squared value that describes the fit between actual data and regression line.  The ‘top down’ models have highlighted the significance of each individual variable for energy predictions, and those parameters that add the most electricity use per unit18. The results of Chapter 2 and 3 are consistent, as the estimates of electricity use per square-meter in Chapter 3 show the same building category specific differences as the seasonal investigations in Chapter 2. The ‘top down’ models in Chapter 3 can be applied to calculate most research building’s EUIel without any modifications of the model itself, as long as the variables for the investigated building are known. However, the value of the developed regression model would have remained uncertain without the evaluation in Chapter 4.  18 The units in Chapter 3 vary between kWh/m2, kWh/year, kWh/number of floors, kWh/occupant, and kWh/°C according to the parameters, Floor Type, Building Age, Number of Floors, Occupancy, and Outdoor Temperature. 62 The evaluation of the ‘top down’ model in Chapter 4 has shown how meaningful the results are compared to a common methodologies that are used to predict the electricity intensity of buildings, as represented by the ‘bottom up’ models. Comparing not only the ‘top down’ model as well as the more complex ‘bottom up’ model to actual electricity data has highlighted the pros and cons of a well-established methodology in the building industry.  The conclusion of Chapter 4 is that a ‘top down’ model comprised of two equations as developed in Chapter 3 is able to estimate the electricity use of seven out of ten research buildings more precisely than the complex ‘bottom up’ models that require more input of data, time, and labor. Accordingly, we summarize that simple building parameters have the ability to create an accurate electricity use range that covers the annual consumption of most research buildings on UBC campus. The model could not estimate the electricity use of all facilities within a tolerance range of  ±40%, and in some cases all ‘top down’ models were off by over 100%. The reason for this failure has been discussed in previous chapters, and it has been explained that the ‘top down’ model does not have the ability to consider electricity intensive mechanical equipment that is not common for research facilities. The ‘bottom up’ models’ predictions consistently underestimate all facilities’ electricity consumptions, and are less precise than the ‘top down’ model for the majority of the ten buildings analyzed in Chapter 4. There could be consideration of the time and effort inherent in the ‘bottom up’ model in comparison to the accuracy of the model results. We also conclude that this study is in favor of extensive data collections in order to enable more accurate predictions and LCC analyses for future buildings. The ‘top down’ model in its current stage already has the ability to create an electricity use baseline for a range of research buildings, and this could be used as a check and balance for more precise energy models. The success of our model is based 63 on the high quality data that was collected for this study. The advantage associated with databases and high quality data inputs is a major outcome of this study. Our conclusion that the quality of future building’s energy predictions can be increased by data that is being collected from existing buildings, is an important addition to the current stage of research, and is in alignment with the assumptions made by many authors as discussed in the introduction. In summary, old existing buildings deliver valuable information that should be used to establish databases, which in combination with the projected building’s unique parameters make us able to forecast future energy demands and costs associated with new facilities. The simplicity of our ‘top down’ model particularly exhibits its benefits when the expected level of accuracy is in range of quick estimates and “back of the envelope” calculations, but it cannot replace more detailed and flexible energy modeling tools at this time. Some analyses require much more flexibility and a ‘bottom up’ model that can be easily adjusted to minor building design and mechanical equipment changes. Further, if minor changes in the energy model can cause dramatic fluctuations in the actual outcome of a building’s energy intensity, the use of extensive modeling may not be necessary as demonstrated by the variance in predictions made by ‘bottom up’ models for the ten research buildings in Table 4.2. Nevertheless, ‘bottom up’ models have the ability to address different options for the building envelope, glazing type, mechanical equipment, etc., and the associated energy savings potential. Again, this flexibility is a requirement for energy predictions of facility projects and cannot be realized with our ‘top down’ model.   We conclude that information collected from existing buildings is very useful in making predictions about future buildings. Reliable databases are the major contributor to making LCC analysis a powerful decision-making tool that has not truly been acknowledged yet.  64 The following paragraphs are dedicated to discuss more benefits and limitations of our ‘top down’ model in comparison with the ‘bottom up’ model.  5.2 What Assumptions Were Made Some assumptions have already been addressed in the Results sections of the previous chapters. Most of the data that has been collected for this study was precisely recorded and documented. The floor allocations are established and constantly updated by UBC’s Department of Infrastructure Development. The department uses architectural building drawings to measure the actual size of each floor area. The hereby generated data is provided to the provincial government for cost estimates and other purposes. The reallocation of the floor category Other Spaces according to the ratio between Classroom, Laboratory, Office, and Library floor area, is based on our own assumption and supported by the following example.  An imaginary two-story building that has only Classroom spaces on the first floor (accounting for 40% of the total gross floor area) and a mixed use of Office and Laboratory spaces (each 20% of the total gross floor area) on the second floor. The facility also has two washrooms on each floor (combined 20% “Other Spaces” of the total gross floor area). Our method considers the fraction of each floor type (40% Classroom, 20% Office, 20% Laboratory) instead of reallocating Other Spaces in equally large portions three ways. Since most first floor washroom users are assumed to be Classroom occupants it is justified that the washrooms are considered to be “part” of the Classroom facility. The same procedure is applied to all building areas that account for Other Spaces. The “equally reallocated” method would increase the significance of smaller area categories in the building because their percentage of total gross floor area would gain relatively more weight through the reallocation than larger areas. Moving the relative 65 weight of one floor category’s size to another could cause misinterpretations of the relationship between EUIel and floor allocation.    Outdoor temperature and electricity use is collected and documented by UBC Energy and Water services. Electricity meters can fail due to many reasons, causing a data recording gap until the meter is set back or replaced. This is a common reason for energy data gaps that can occur for a couple of hours or even days. It is important to fill the gaps for an uninterrupted daily regression analysis. This was done by replacing nulls with the calculated monthly averages of that particular month in which the recording failure occurred. The averages were calculated for weekdays and weekends separately, and the day specific nulls replaced accordingly. Considering that electricity is recorded in 15 minute-intervals, any inaccuracy due to this gap filling method for short time intervals is negligible.  Occupancy data is based on WI-FI device connections, and is therefore more an occupancy activity variable than an exact occupancy measure. The accuracy is in the order of 80% according to ©Sensitive Building Sciences who has developed the devices to track WI-FI connections. The data is already being used on UBC campus to adjust ventilation rates on a room specific scale based on occupancy levels to accomplish energy savings of 2-10% depending on the building type19. Building category specific occupancy averages were used for the five buildings in Chapter 4 that had no actual recorded Occupancy data. This approach might not 19 Information is gratefully provided by ©Sensible Building Sciences and Stefan Storey: (September 15, 2016) & (September 15, 2016). 66 represent the building’s actual occupancy. However, Occupancy has not proven to be a very important contributor to most facility’s electricity consumption as the parameter adds only little to the total energy use.      5.3 Nature of Complexity of Input Data We want to highlight again what has already been discussed in previous paragraphs. • The advantage of the methodology in this study lies in its simplicity.• Any research building’s electricity use can be estimated with the knowledge of only fiveparameters: Floor Allocation (m2-Classroom, m2-Laboratory, m2-Office, m2-Library),Occupancy, Number of Floors, Building Age20, and Outdoor Temperature.The benefit of low complexity on the one side is also the ‘top down’ model’s limitation on the other side. The five parameters might be easily collected for most research buildings but the flexibility of different facility designs, and especially the economic performance comparison of different building options is only possible to a limited extent. We can make assumptions how a projected building’s electricity use is affected by design changes and if the weight of one floor category is shifted towards another. We can also make suggestions if it is an energy advantage to have more, smaller floors or fewer but much bigger floors for the designed building type and layout. One could use the model to calculate if high Occupancy levels will have a significant impact on the building’s energy use or if Occupancy does not necessarily have to be addressed at all.  20Building Age can be ignored for future building projects if they are still in their design stage. 67 The chosen parameters of this ‘top down’ model in its current stage are not able to capture any energy conservation measures, green building design strategies, advanced insulation, glazing-type or mechanical efficiency. Additional parameters are necessary to consider these specific characteristics and to investigate their impact on the EUIel. The required variables would cost the model’s simplicity and require significantly more data to be collected. However, this would be an unavoidable requirement if our developed ‘top down’ model were to be used for LCC analysis or more precise building energy evaluations.  Common energy modeling software tools are based on ‘bottom up’ models, and have very detailed input data requirements, and parameters can be tuned and customized at any time according to building design changes or cost reallocations. Thus these tools do not use data from existing buildings but rely on a multitude of calculations according to the input information of the designed building. This complexity is required to support decision-makers by choosing the best building design with the lowest LCC amongst different options. Although the ‘top down’ model as developed in this study cannot replace more complex ‘bottom up’ models in its current stage, it can have an important role in building designs as concluded in the next paragraph. 5.4 Role of the ‘Bottom Up’ Model in Building Design Our ‘top down’ model has proven that it can compete with complex energy modeling software tools but it is not able to replace them. The ‘top down’ models have delivered electricity consumption predictions at even higher accuracy than the ‘bottom up’ energy models. Although the ‘top down’ models based on nine variables do not have the same complexity and flexibility as complex engineering tools, our model can provide planners and decision-makers with a simple way to estimate electricity use in buildings. The tool has great opportunities for further 68 development including the incorporation of more exact variables in order to address specific building design characteristics, mechanical equipment, and other efficiency increasing parameters that are known to affect the electricity consumption. If the ‘top down’ model is taken to a stage that allows decision-makers to incorporate building specific design variables in order to investigate the energy intensity and savings of different design options, the tool could be used as part of a LCC analysis calculator. This calculator would require similar analyses for all building cost categories that add to the building’s overall LCC as explained in the introduction of this study. This methodology is very data intensive and would require years of research and data collection. However, according to the quality of our electricity use estimates, an LCC calculator based on existing building data could be a promising alternative to common methodologies and potentially replace both the intensive engineering calculations as well as the ‘bottom up’ models. The here presented ‘top down’ model in its current stage has the ability to test and challenge the outcomes of more complex ‘bottom up’ models. It can provide an electricity use range or a reference value to encourage energy modelers to rethink their input information if the discrepancy between the ‘top down’ model and the ‘bottom up’ model is unexplainably high.  69 Bibliography Al-Hajj, A., & Horner, M. W. (1998). Modelling the running costs of buildings. Construction Management and Economics, 16(4), 459–470. Allard Hall - eQUEST Plant and Water-Side HVAC Summary Report. (2009). Aouad, G., Bakis, N., Amaratunga, R., & Sun, M. (2002). An integrated life cycle costing database: System proposal and methodology, (1997), 134–144. Retrieved from ASHRAE Standards Committee. (2007). ASHRAE STANDARD 62.1, Ventilation for Acceptable Indoor Air Quality (Vol. 2007). ASHRAE Standing Standard Project Committee 100. (2015). ASHRAE Standard 100, Energy Efficiency im Existing Buildings. Aye, L., Bamford, N., Charters, B., & Robinson, J. (2000). Environmentally sustainable development: a life-cycle costing approach for a commercial office building in Melbourne, Australia. Construction Management and Economics, 18(8), 927–934. Boussabaine, H. A., & Kirkham, R. J. (1998). Fundamentals of Whole Life-cycle Cost Analysis. Boussabaine, H. A., & Kirkham, R. J. (2004a). Whole Life-Cycle Costing: Risk and Risk Response. Wiley. Retrieved from Boussabaine, H. A., & Kirkham, R. J. (2004b). Whole Life-Cycle Costing Risk Management. Whole Life-Cycle Costing, 12. Retrieved from edb&AN=74882995&site=eds-live Canada Green Building Council. (2016). Canada Green Building Council. Retrieved May 5, 2016, from Capehart, B. L., Turner, W. C., & Kennedy, W. J. (2012). Guide to Energy Management (7th Edition). Fairmont Press, Inc. Retrieved from cathode fuel cell&sort_on=default&b-group-by=true&b-search-type=tech-reference&b-sort-on=default Chalifoux, A. (2006). Using Life Cycle Costing and Sensitivity Analysis to “Sell” Green Building Features. Journal of Green Building, 1(2), 39–48. City of Vancouver. (2012). Greenest City 2020 Action Plan. City of Vancouver. (2014). Green Buildings Policy for Rezonings. Retrieved from CIWMB. (2000). Design With Vision: A Technical Manual for Material Choices in Sustainable Construction. Cole, R. J., & Sterner, E. (2000). Reconciling theory and practice of life-cycle costing. Building Research & Information, 28(5–6), 368–375. Dell’Isola, Alphonse J.; Kirk, S. J. (2003). Life Cycle Costing for Facilities. Reed Construction Data. Finance, P. (2013). Treasury Board and Finance – Economic Trends – September 2013, (September), 1–4. 71 Fu, C., Kaya, S., & Aouad, M. K. G. (2007). The development of an IFC-based lifecycle costing prototype tool for building construction and maintenance: Integrating lifecycle costing to nD modelling. Construction Innovation: Information, Process, Management, 7(1), 85–98. GF Shymko & Associates Inc. (n.d.). UBC Life Sciences Centre - LEED EA Credit 1. Kneifel, J. (2010). Life-cycle carbon and cost analysis of energy efficiency measures in new commercial buildings. Energy and Buildings, 42(3), 333–340. Larsson, N., & Clark, J. (2000). Incremental costs within the design process for energy efficient buildings. Building Research & Information, 28(5–6), 413–418. Liu, Benjamin Y. H.; Richard C., J. (1963). A Rational Procedure for Predicting The Long-Term Average Performance of Flat-Plate Solar-Energy Collectors, 7(2). Mohamed, A. El-Haram; Sasa, Marenjak; Malcolm, W. H. (2002). Development of a generic framework for collecting whole life cost data for the building industry. Journal of Quality in Maintenance Engineering, 8(2), 144–151. Retrieved from Morrissey, J., & Horne, R. E. (2011). Life cycle cost implications of energy efficiency measures in new residential buildings. Energy and Buildings, 43(4), 915–924. Neely, E. S., & Neathammer, R. (1991). Life-Cycle Maintenance Costs by Facility use. J. Constr. Eng. Manage., 117(2), 310–320. Perera, O., Morton, B., & Perfrement, T. (2009). Life Cycle Costing: A Question of Value. 72 Risk, W. L., & Responses, R. (2004). Whole Life-Cycle Costing. Shaun Martin Consulting. (2011). UBC Biosciences Building Renovation. Stantec Consulting Ltd. (2010a). Centre for Interactive Research on Sustainability (CIRS) - High Performance Building Program Post Tender Energy Study. Stantec Consulting Ltd. (2010b). UBC-Pharmacy Building, BC Hydro Power Smart New Construction Program, Energy Study. Stantec Consulting Ltd. (2010c). UBC Buchanan Complex LEED – EAp2 and EAc1 ENERGY MODELING REPORT Prepared. Stantec Consulting Ltd. (2010d). UBC Chemistry Block D, EAp2 and EAc1 Documentation. Stantec Consulting Ltd. (2011). UBC - Earth System Science Building (ESSB) LEED – EAp2 and EAc1 Energy Modeling Report. Sterner, E. (2000). Life-cycle costing and its use in the Swedish building sector. Building Research & Information, 28(5–6), 387–393. The University of British Columbia. (2016). The University of British Columbia. Retrieved September 5, 2016, from UBC Energy & Water Services. (2016). UBC Energy & Water Services. Retrieved January 1, 2016, from UBC Green Labs. (n.d.). UBC Green Labs. Retrieved September 1, 2016, from Wong, I. L., Perera, S., & Eames, P. C. (2010). Goal directed life cycle costing as a method to evaluate the economic feasibility of office buildings with conventional and TI‐façades. Construction Management and Economics, 28(7), 715–735. 73 Wübbenhorst, K. (1986). Life cycle costing for construction projects. Long Range Planning, 19(4), 87–97. 74 Appendices Appendix A  - The building specific parameters for all 48 research buildings are summarized in the table below. Building	   Age	   #	  of	  Floors	  Electricity	  [kWh]	  Gross	  	  Floor	  [m2]	  Classroom	  [m2]	  Lab	  [m2]	  Office	  [m2]	  Library	  [m2]	  Other	  [m2]	  Occupancy	  Overall	  EUI	  [kWh/m2]	  Anthropology	  and	  Sociology	   64	   3	   	  495,925	  	   	  5,893	  	   	  461	  	   	  643	  	   	  2,272	  	   	  -­‐	  	  	   	  2,517	  	   	  34	  	   	  84.2	  	  Aquatic	  Ecosystems	  Research	  Laboratory	   9	   4	   	  493,749	  	   	  5,368	  	   	  325	  	   	  245	  	   	  2,281	  	   	  -­‐	  	  	   	  1,971	  	   	  48	  	   	  92.0	  	  Asian	  Centre	   39	   5	   	  560,115	  	   	  5,532	  	   	  -­‐	  	   	  151	  	   	  704	  	   	  2,095	  	   	  2,164	  	   	  23	  	   	  101.3	  	  Biomedical	  Research	  Centre	   27	   4	   	  1,731,951	  	   	  4,798	  	   	  -­‐	  	  	   	  1,926	  	   	  492	  	   	  -­‐	  	  	   	  1,814	  	   	  33	  	   	  361.0	  	  Brimacombe	   19	   7	   	  2,811,240	  	   	  9,864	  	   	  -­‐	  	  	   	  3,860	  	   	  705	  	   	  -­‐	  	  	   	  2,735	  	   	  42	  	   	  285.0	  	  Buchanan	  Building	  Block	  A,	  B,	  C	   56	   4	   	  918,624	  	   	  10,936	  	   	  3,292	  	   	  426	  	   	  1,923	  	   	  -­‐	  	  	   	  4,008	  	   	  322	  	   	  84.0	  	  Buchanan	  Building	  Block	  D,	  E	   54	   4	   	  670,596	  	   	  7,134	  	   	  1,908	  	   	  48	  	   	  1,673	  	   	  -­‐	  	  	   	  2,902	  	   	  126	  	   	  94.0	  	  Buchanan	  Tower	   42	   15	   	  514,944	  	   	  10,728	  	   	  318	  	   	  202	  	   	  5,005	  	   	  -­‐	  	  	   	  3,581	  	   	  60	  	   	  48.0	  	  C.	  K.	  Choi	  Building	  for	  The	  Institute	  of	  Asian	  Research	   18	   4	   	  116,480	  	   	  2,912	  	   	  33	  	   	  76	  	   	  1,140	  	   	  -­‐	  	  	   	  1,295	  	   	  23	  	   	  40.0	  	  Chemical	  &	  Biological	  Engineering	  Building	   9	   8	   	  3,030,480	  	   	  14,030	  	   	  768	  	   	  4,358	  	   	  1,911	  	   	  -­‐	  	  	   	  5,928	  	   	  119	  	   	  216.0	  	  Chemistry	  A	  Block,	  Chemistry	  Physics	  Building	   25	   6	   	  4,027,380	  	   	  7,805	  	   	  -­‐	  	  	   	  3,421	  	   	  1,226	  	   	  -­‐	  	  	   	  2,166	  	   	  30	  	   	  516.0	  	  Chemistry	  D	  Block,	  Centre	  Wing	   89	   5	   	  5,513,692	  	   	  7,274	  	   	  416	  	   	  2,637	  	   	  1,107	  	   	  -­‐	  	  	   	  1,835	  	   	  191	  	   	  758.0	  	  Civil	  and	  Mechanical	  Engineering	  Building	   38	   4	   	  889,295	  	   	  9,361	  	   	  568	  	   	  2,523	  	   	  2,780	  	   	  -­‐	  	  	   	  3,063	  	   	  136	  	   	  95.0	  	  Civil	  and	  Mechanical	  Engineering	  Laboratories	   43	   3	   	  1,354,405	  	   	  4,043	  	   	  44	  	   	  2,719	  	   	  551	  	   	  -­‐	  	  	   	  534	  	   	  21	  	   	  335.0	  	  D.H.	  Copp 45	   5	   	  1,334,535	  	   	  8,514	  	   	  137	  	   	  1,433	  	   	  928	  	   	  -­‐	  	  	   	  6,017	  	   	  26	  	   	  156.7	  	  David	  Lam	  Learning	  Centre	   22	   6	   	  1,350,200	  	   	  6,751	  	   	  109	  	   	  374	  	   	  849	  	   	  1,582	  	   	  2,330	  	   	  147	  	   	  200.0	  	  David	  Strangway	   9	   7	   	  2,914,705	  	   	  11,388	  	   	  -­‐	  	  	   	  3,136	  	   	  2,787	  	   	  -­‐	  	  	   	  5,465	  	   	  139	  	   	  256.0	  	  Douglas	  Kenny	  Building	   31	   5	   	  1,067,043	  	   	  9,613	  	   	  145	  	   	  3,655	  	   	  1,404	  	   	  -­‐	  	  	   	  3,941	  	   	  84	  	   	  111.0	  	  Earth	  and	  Ocean	  Sciences	  	  -­‐	  Main	   43	   5	   	  3,384,606	  	   	  10,780	  	   	  114	  	   	  4,959	  	   	  1,345	  	   	  -­‐	  	  	   	  2,207	  	   	  61	  	   	  314.0	  	  Food,	  Nutrition	  and	  Health	  Building	   32	   4	   	  1,401,070	  	   	  5,962	  	   	  461	  	   	  2,012	  	   	  983	  	   	  -­‐	  	  	   	  1,872	  	   	  60	  	   	  235.0	  	  Forest	  Sciences	  Centre	   16	   7	   	  4,414,060	  	   	  26,915	  	   	  1,378	  	   	  7,020	  	   	  3,855	  	   	  -­‐	  	  	   12,337	  	   	  234	  	   	  164.0	  	  Frank	  Forward	  Building	   46	   7	   	  961,360	  	   	  7,880	  	   	  449	  	   	  2,195	  	   	  1,645	  	   	  -­‐	  	  	   	  2,981	  	   	  74	  	   	  122.0	  	  Frederic	  Lasserre	  Building	   52	   5	   	  395,640	  	   	  4,710	  	   	  726	  	   	  1,073	  	   	  1,145	  	   	  -­‐	  	  	   	  1,346	  	   	  87	  	   	  84.0	  	  Friedman	  Building	   53	   5	   	  597,312	  	   	  6,222	  	   	  408	  	   	  1,676	  	   	  1,468	  	   	  -­‐	  	  	   	  2,060	  	   	  67	  	   	  96.0	  	  75 Building	   Age	   #	  of	  Floors	  Electricity	  [kWh]	  Gross	  	  Floor	  [m2]	  Classroom	  [m2]	  Lab	  [m2]	  Office	  [m2]	  Library	  [m2]	  Other	  [m2]	  Occupancy	  Overall	  EUI	  [kWh/m2]	  Geography	  Building	   89	   3	   	  270,725	  	   	  5,525	  	   	  962	  	   	  672	  	   	  2,080	  	   	  -­‐	  	  	   	  1,628	  	   	  68	  	   	  49.0	  	  H.	  R.	  Macmillan	  Building	   47	   5	   	  1,419,300	  	   	  14,193	  	   	  989	  	   	  3,963	  	   	  2,655	  	   	  -­‐	  	  	   	  5,672	  	   	  129	  	   	  100.0	  	  Hebb	  Building	   50	   7	   	  329,505	  	   	  5,991	  	   	  922	  	   	  2,572	  	   	  487	  	   	  -­‐	  	  	   	  2,035	  	   	  84	  	   	  55.0	  	  Hennings	  Building	   69	   4	   	  1,131,372	  	   	  11,413	  	   	  983	  	   	  3,875	  	   	  2,758	  	   	  -­‐	  	  	   	  3,192	  	   	  133	  	   	  99.1	  	  LIU	  Institute	  for	  Global	  Issues	   14	   4	   175,320	  	   	  1,729	  	   	  -­‐	   	  -­‐	  	  	   910.77	  	   	  -­‐	  	  	   592.68	  	   	  21	  	   101.8	  	  I.K.	  Barber	  Learning	  Centre 7	   6	   	  3,441,816	  	   	  24,893	  	   	  1,197	  	   	  211	  	   	  2,844	  	   	  10,063	  	   10,577	  	   	  687	  	   	  138.3	  	  Institute	  for	  Computing,	  Information	  and	  Cognitive	  Systems	  /	  Computer	  Science	   21	   5	   	  2,360,009	  	   	  10,583	  	   	  43	  	   	  2,673	  	   	  2,981	  	   	  -­‐	  	  	   	  3,915	  	   	  241	  	   	  223.0	  	  J.	  B.	  Macdonald	  Building	   47	   4	   	  1,062,560	  	   	  7,328	  	   	  150	  	   	  2,031	  	   	  2,063	  	   	  -­‐	  	  	   	  2,625	  	   	  89	  	   	  145.0	  	  Jack	  Bell	   22	   5	   	  385,088	  	   	  2,589	  	   	  380	  	   	  133	  	   	  1,027	  	   	  -­‐	  	  	   	  1,050	  	   	  26	  	   	  148.7	  	  Life	  Sciences	  Centre	   10	   8	   	  20,661,965	  	   	  59,596	  	   	  1,932	  	   20,568	  	   	  6,649	  	   	  -­‐	  	  	   25,675	  	   	  423	  	   	  346.7	  	  Lower	  Mall	  Research	  Station	   54	   5	   	  875,028	  	   	  6,629	  	   	  -­‐	  	  	   	  980	  	   	  1,217	  	   	  -­‐	  	  	   	  3,722	  	   	  23	  	   	  132.0	  	  Macleod	  Building	   51	   5	   	  705,120	  	   	  7,345	  	   	  799	  	   	  3,165	  	   	  626	  	   	  -­‐	  	  	   	  2,159	  	   	  131	  	   	  96.0	  	  Mathematics	  Building	   89	   3	   	  896,440	  	   	  6,140	  	   	  939	  	   	  23	  	   	  2,184	  	   	  -­‐	  	  	   	  2,646	  	   	  48	  	   	  146.0	  	  Medical	  Sciences	  Block	  C	   53	   6	   	  429,498	  	   	  4,014	  	   	  77	  	   	  1,526	  	   	  756	  	   	  -­‐	  	  	   	  1,332	  	   	  15	  	   	  107.0	  	  Michael	  Smith	  Laboratories	   10	   5	   	  5,600,384	  	   	  7,712	  	   	  230	  	   	  3,493	  	   	  1,299	  	   	  -­‐	  	  	   	  2,690	  	   	  60	  	   	  726.2	  	  Neville	  Scarfe	  Building	   49	   6	   	  1,356,740	  	   	  19,382	  	   	  1,689	  	   	  2,307	  	   	  5,044	  	   	  1,377	  	   	  7,699	  	   	  94	  	   	  70.0	  	  P.A	  Woordward	  Instructional	  ResourcesCentre	   42	   8	   	  843,430	  	   	  12,049	  	   	  1,724	  	   	  235	  	   	  2,755	  	   	  -­‐	  	  	   	  6,399	  	   	  220	  	   	  70.0	  	  Pulp	  and	  Paper	  Centre	   29	   4	   	  470,528	  	   	  3,676	  	   	  -­‐	  	  	   	  1,096	  	   	  833	  	   	  -­‐	  	  	   	  1,404	  	   	  34	  	   	  128.0	  	  School	  of	  Population	  &	  Public	  Health	   35	   5	   	  1,603,980	  	   	  8,442	  	   	  460	  	   	  492	  	   	  3,760	  	   	  -­‐	  	  	   	  3,062	  	   	  62	  	   	  190.0	  	  The	  Leonard	  S.	  Klinck	  Building	   57	   6	   	  4,813,280	  	   	  10,720	  	   	  1,085	  	   	  394	  	   	  4,351	  	   	  -­‐	  	   	  3,681	  	   	  160	  	   	  449.0	  	  Walter	  C.	  Koerner	  Library	   18	   7	   	  2,081,355	  	   	  7,303	  	   	  -­‐	  	  	   	  -­‐	  	  	   	  848	  	   	  2,105	  	   	  3,660	  	   	  217	  	   	  285.0	  	  Wesbrook	  Building	   65	   6	   	  688,224	  	   	  10,272	  	   	  538	  	   	  1,343	  	   	  612	  	   	  -­‐	  	  	   	  6,536	  	   	  84	  	   	  67.0	  	  West	  Mall	  Swing	  Space	  Building	   9	   5	   	  382,645	  	   	  4,907	  	   	  2,677	  	   	  -­‐	  	  	   	  125	  	   	  -­‐	  	  	   	  2,105	  	   	  120	  	   	  78.0	  	  Woodward	  Library	   50	   5	   	  839,916	  	   	  7,777	  	   	  -­‐	  	  	   	  -­‐	  	  	   	  -­‐	  	  	   	  5,553	  	   	  1,825	  	   	  97	  	   	  108.0	  	  76 Appendix B  - All Equations in the Table Below were Used for the ‘Top Down’ Models in Chapter 4. The Estimates are Different from those Described in Chapter 3 because the Building Sample is Reduced.  Eq. 11: 𝑇𝐸 =   1 + 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 Estimate Unit R-squaredIntercept 3295.5 kWh 0.705 C 𝛼 = -0.79283** kWh/m2 L 𝛽 = 0.81937** kWh/m2 O 𝛿 = 0.22707** kWh/m2 LIB 𝜀 = 0.33107** kWh/m2 OT 𝜁 = 0.35665** kWh/m2 A 𝜃 = -54.702** kWh/year F 𝜗 = -118.46** kWh/floor T 𝜅 = 15.841** kWh/°C  Eq. 12: 𝑇𝐸 =   1 + 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 + 𝜆 ∙ 𝑂𝑂 Estimate Unit R-squaredIntercept 3048.4 kWh 0.72 C 𝛼 = -0.89689** kWh/m2 L 𝛽 = 0.79441** kWh/m2 O 𝛿 = 0.1921** kWh/m2 LIB 𝜀 = 0.22744** kWh/m2 OT 𝜁 = 0.33463** kWh/m2 A 𝜃 = -51.573** kWh/year F 𝜗 = -110.81** kWh/floor T 𝜅 = 24.604** kWh/°C OO 𝜆 = 2.6657** kWh/person  Eq. 13: 𝑇𝐸 =   1 + 𝛼 ∙ 𝐶! + 𝛽 ∙ 𝐿! + 𝛿 ∙ 𝑂! + 𝜀 ∙ 𝐿𝐼𝐵! + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 Estimate Unit R-squaredIntercept 3442.7 kWh 0.685 CR 𝛼 = -0.22201** kWh/m2 LR 𝛽 = 0.58754** kWh/m2 OR 𝛿 = 0.27674** kWh/m2 LIBR 𝜀 = 0.33244** kWh/m2 A 𝜃 = -56.034** kWh/year F 𝜗 = -121.04** kWh/floor T 𝜅 = 16.262** kWh/°C 77   Eq. 14: 𝑇𝐸 =   1 + 𝛼 ∙ 𝐶! + 𝛽 ∙ 𝐿! + 𝛿 ∙ 𝑂! + 𝜀 ∙ 𝐿𝐼𝐵! + 𝜃 ∙ 𝐴 + 𝜗 ∙ 𝐹 + 𝜅 ∙ 𝑇 + 𝜆 ∙ 𝑂𝑂 Estimate Unit R-squaredIntercept 3136.1 kWh 0.705 CR 𝛼 = -0.29106** kWh/m2 LR 𝛽 = 0.56531** kWh/m2 OR 𝛿 = 0.25382** kWh/m2 LIBR 𝜀 = 0.25787** kWh/m2 A 𝜃 = -52.673** kWh/year F 𝜗 = -112.57** kWh/floor T 𝜅 = 25.243** kWh/°C OO 𝜆 = 2.7859** kWh/person  Eq. 15: 𝑇𝐸 =   1 + 𝛼 ∙ 𝐶 + 𝛽 ∙ 𝐿 + 𝛿 ∙ 𝑂 + 𝜀 ∙ 𝐿𝐼𝐵 + 𝜁 ∙ 𝑂𝑇 + 𝜅 ∙ 𝑇 + 𝜆 ∙ 𝑂𝑂 Estimate Unit R-squaredIntercept C332.44 kWh 0.61 C 𝛼 = -1.0703** kWh/m2 L 𝛽 = 0.94983** kWh/m2 O 𝛿 = -0.02600** kWh/m2 LIB 𝜀 = 0.30451** kWh/m2 OT 𝜁 = 0.37395** kWh/m2 T 𝜅 = 27.336** kWh/°C OO 𝜆 = 3.9561** kWh/person 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items