Comparing Predictive Machine Learning Models for Short- and Long-Term Urban Water Demand Forecasting in Milan, Italy

Urban water demand forecasting is essential for supply network optimization and management. In this case study, we comparatively investigate different state-of-the-art predictive models on short- (1 day-ahead) long-term (7 urban (UWD) the city of Milan, Italy. The contribution paper two-fold. First, compare performance time series machine learning daily UWD. tested include Autoregressive Integrated Moving Average (ARIMA) models, Artificial Neural Networks (ANN), Support Vector Regression (SVR), Light Gradient Boosting Machine (LightGBM), Long Short-Term Memory (LSTM) networks. Second, whether coupling a Wavelet Data-Driven Forecasting Framework (WDDFF) with these further improves capacity. Results show that, in general, WDDFF can improve model performance. LSTM coupled wavelet decomposition technique achieve high levels accuracy R2 larger than 0.9 both UWD forecasts. LightGBM efficiently reduce number predictors potential to forecast select important features hydrology resources field.