Daily multivariate forecasting of water demand in a touristic island with the use of artificial neural network and adaptive neuro-fuzzy inference system

Abstract

Water demand forecast has emerged as an imperative component of intelligent Internet and Communication Technologies based methodologies of water management. The need of increased time resolution of forecast in order to implement such methodologies is driving stakeholders to long for new more specialized forecast approaches that will take into account the special drivers of water demand in each case study. Advanced techniques have the ability to overcome the nonlinearity issues commonly met when investigating the complex relationship of water demand and weather, socioeconomic and other variables. In this article we present two approaches, an Artificial Neural Network and an Adaptive Neuro-Fuzzy Inference System, for forecasting a Mediterranean touristic resort daily water demand based on weather variables, tourism and leakage. Both models seem to have an adequate response, though ANFIS can more smoothly catch winter non-touristic water demand profile.