Short-term water demand prediction using stacking regression–based machine learning

Abstract

Forecasting of water demand and equitable allocation of local water resources are used to reduce and eliminate water shortages and waste. The key emphasis of this research article is to estimate water demand using the prediction model for the Peroorkada urban water distribution network. The characteristics, such as head, pressure, and base demand, related to the water demand were the features of the prediction model. The prediction model has been developed using python. The water distribution network consists of 99 nodes. The demand graph for a time interval of 6 h has been plotted and predicted for all the nodes, and 24-h interval demand has been plotted for vulnerable nodes, which were determined by the sensor placement toolkit. This study included 13 machine learning algorithms, including three hybrid/stacked regression techniques. The least absolute shrinkage and selection operator-based stacking regressor model performs the best at demand prediction. Single prediction models were outperformed by stacking regressor models.