AWWA JAW56653

A water utility's primary purpose is to provide safe drinking water to its customers, but it must also plan for its customers' future water needs. A critical aspect of this planning is predicting short-term (peak) water demands and optimizing the water supply system to meet these demands. Jain and Ormsbee evaluated eight models (four conventional and four artificial intelligence [AI]) to forecast short-term water demand. The models were developed and tested using daily water demand, daily maximum air temperature, and daily total rainfall. AI models outperformed the conventional models and predicted short-term water demands during both "normal" and drought conditions relatively accurately. These AI models can be run using commercial software, if the system operator is experienced in using models, or simple rule-based AI models can be easily developed and applied using historical demand, weather data, and a standard spreadsheet program. The authors encourage utility operators or managers to incorporate a short-term water-demand forecasting model into their predictions of future water needs. Includes 25 references, tables, figures.