Water Demand Forecasting Using Machine Learning and Time Series Algorithms

nowadays, most of the water distribution networks are still managing their operation using the Instantaneous demand. This means that the machinery's use is determined by the immediate need for water. The network's water reservoirs are packed pumps that start to work when the level of water exceeds a given minimum threshold and stops when it reaches the peak level. Establish a water management strategy focused on predicting future demand is reducing the cost of capture, storage, processing, and distribution. In this paper, we present a comparative study for water demand forecasting using support vector linear regression and AutoRegressive Integrated Moving Average (ARIMA). The study has been carried out on the state of Kuwait daily water consumption. The result shows that ARIMA has MAPE (1.8) and RMSE (9.4) while support vector linear regression has MAPE (0.52) and RMSE (2.59) which indicates the deviation of the forecasted water demand versus the actual water consumption.