Status and prospects for the use of algorithms and models for short-term water demand forecasting

Short-term water demand forecasting

Historical and predicted daily water demand of a supply area (simulation data)

The application of algorithms and mathematical models for the evaluation and prediction of information is becoming increasingly interesting for water management. The DVGW research project “short-term water demand forecasting” dealt with the identification of potential fields of application of short-term forecasts and evaluated the application of these methods on real water supply systems.

Short-term forecast models have long been used in the fields of electricity, gas, stock exchange, weather forecast, etc. The application in the field of water management is still new. However, this will continue due to the increasing automation of process flows ("Water 4.0"). For the water industry, automated data evaluation and the resulting optimization processes using intelligent algorithms are of high relevance. In this context, water demand analyses and models form the central variable for the control and management of plants and aggregates in the water industry. With the help of robust forecasts for water demand, adapted operating strategies can be developed and strategic decisions can be supported. For the application of demand forecasts in drinking water supply, simple statistical methods, models from the field of machine learning and time series analysis were investigated. In the 13 investigated application areas a very high forecast quality (deviations ≤ 5 %) for the prediction of the average and peak water demand was achieved with machine learning models. Simple statistical methods achieved forecast values with a mean deviation of ≤ 10%, whereas peak demand values could not be predicted. The final report serves as a guideline for action for the preparation, creation and evaluation of forecast models for hourly and daily water demand and is available on request. For further advice on the preparation and implementation of water demand forecasts, please contact us.