Open Collections

Water demand forecasting : a flexible approach

University of British Columbia

Water distribution systems (WDS) operators would benefit greatly from educated estimates of water demand to help with the water scarcity threatening many regions worldwide. A wide range of stochastic and deterministic techniques have been proposed to model water demands in urban WDS. Every WDS needs a fair estimate of its future state for both development and operational purposes. Statistical models like time series forecasting and regression analysis have been widely used in the field of water demand forecasting. Recently, so-called artificial intelligence techniques became increasingly popular among scholars due to their high accuracy in prediction as well as not being limited to certain statistical assumptions. This research introduced gene expression programming (GEP), and support vector regression (SVM) and well-known artificial neural networks (ANN) as supervised learning algorithms for predictive analytics of water demand. K-means clustering as an unsupervised learning algorithm was tested to group the data based on a suitable distance metrics. The performance of the developed models was improved through phase space reconstruction of the time series data using optimum lag time determined by average mutual information. Monthly long-term water demand data of the City of Kelowna district (CKD) was used as the main case study throughout the research. Due to unavailability of water demand at finer short-term resolutions in Kelowna, the water demand data in the City of Milan was also used as a second case study for developing a framework of studying different temporal resolutions in the short-term analysis of water demand. Some scholars believe that predictive models are often wrong given the significant uncertainty in the conditions of underlying complex engineering systems. A novel technique based on data augmentation (data cropping and distorted data), and information theory were used to propose a flexible range of water demand (358 ML for upper bound and 335 ML for lower bound) for City of Kelowna which anticipates a wide range of uncertainties WDS.

Text

2018-09-30

Attribution 4.0 International

http://hdl.handle.net/2429/67223

Civil Engineering

University of British Columbia

UBCO

http://creativecommons.org/licenses/by/4.0/