Water Distribution Network Modeling Based on NARX

Abstract This paper proposes a NARX (Nonlinear Auto-Regressive with Exogenous Inputs) model for the water distribution network real-time prediction and control. The model estimates the time-variable nodal demand equivalently by exploiting the real-time and historical operating data and establishes a functional relationship between the major variables among the network. In addition, a training scheme with a combination of offline training and online training, and corresponding algorithm are proposed. And the NARX model is established for a real distribution network. The results demonstrate the model is applicative and satisfactory, and it shows good tracking and predicting performance.