A Study of Forecasting Error for Water Demand in Saitama City

Abstract

Forecasting water demand is very important for pure water reservoirs because the amount of water supply to water supply stations cannot be changed frequently according to the variation in demand. The aim of this study is to propose a forecasting method for water demand applicable all over Japan. In our previous studies, we proposed three forecasting methods and one correcting method for water demand using data from four Japanese prefectures. These forecasting methods adopted the following three processes: (1) multivariate regression analysis, (2) using multiple year data of the amount of daily water supply, and (3) correcting the obtained multivariate regression formula by single regression analysis. The results showed that the proposed forecasting methods ’precision was approximately 2%. However, the forecasting precision could not be improved further. Thus, this study focuses on correcting the method and forecasting error. In our previous studies, the correcting method that was proposed by using a single regression analysis consisted of the following three steps: (1) temporary forecasting values were calculated by obtained multiple regression formula using daily water supply data close to the forecasting day; (2) correcting coefficients were estimated by a single regression analysis using the relation between temporary forecasting values and actual amounts; and (3) the multiple regression formula was corrected using the obtained correcting coefficient. Thus, the proposed correcting method was very simple in our previous studies. In this paper, the factors of forecasting error are estimated based on the relationship between the forecasting and actual amounts. Then, the new correction method is examined using these factors.