On-line Hydraulic State Estimation in Urban Water Networks Using Reduced Models

Abstract

A Predictor-Corrector (PC) approach for on-line forecasting of water usage in an urban water system is presented and demonstrated. The M5 Model-Trees algorithm is used to predict water demands and Genetic Algorithms (GAs) are used to correct (i.e., calibrate according to on-line pressure and flow rate measurements) these predicted values in real-time. The PC loop repeats itself at each subsequent time-step with the forecasting model inputs being the corrected outputs of previous iterations, thus improving the model performances over time. To meet the computational efficiency requirements of real-time hydraulic state estimation, the urban network model which is comprised of over ten thousand pipelines and nodes is reduced using a water system aggregation technique. The reduced model, which resembles the original system's hydraulic performances with high accuracy, simplifies the computation of the PC loop and facilitates the implementation of the on-line model. The developed methodology is tested against the real input data of an urban water distribution system comprised of approximately 12500 nodes and 15000 pipes.