Water demand time series generation for distribution network modeling and water demand forecasting

Knowledge of water consumption can help water distribution network modeling. Consumption has been studied based on its stochastic and spatial random features. Considering the difficulty to have access to real-demand data, this work presents three methods to generate synthetically water demand time series from observed data, thus producing final demands in different ways. Each method generates a time series that considers either temporal consumption trends or jointly temporal trends and climatic influence. A random forest algorithm is applied to obtain the relevance of each climatic variable. This study uses water demand and climatic data of various Brazilian cities to extract temporal patterns. The final synthetically generated data can be used as input data for water network models, to feed the methods used according to the objectives of each study or project.