Short-Term Water Demand Prediction in Residential Complexes: Case Study in Columbia City, USA

Abstract

Shortage of freshwater resources and increasing water demand are significant challenges facing water utilities. Accordingly, reliable and accurate short-term prediction is a valuable tool to efficiently operate and manage an existing municipal water supply system. The present study aims to develop an accurate and easy to apply methodology to predict the water demand based on past water consumption data. The proposed methodology uses singular spectrum analysis (SSA) and a linear autoregressive (AR) model to forecast accurately the required water quantities in forthcoming years. The SSA is used to clean the signal of structure-less noise. Then the AR is used to describe the behaviour of the past water consumption data and then to forecast the daily expected water demand in a short-term period. The suggested methodology is validated using daily water consumption data from July 2007- December 2016 in Columbia City, USA, as inputs for the short-term model. The initial results show that the suggested methodology, SSA-AR, has the ability to predict water demand accurately and outperform an AR model.