State Estimation of Municipal Water Supply Network Based on BP Neural Network and Genetic Algorithm

Abstract

Nowadays, it is necessary to simulate the comprehensive operating state of water network based on less monitoring points, which is of great significance to operation optimization and leak detection. In this paper, first the existing state simulation models of water network are briefly introduced, and on this basis a new nonlinear dynamic method with better fault tolerance is put forward. Then, a specific model is constructed, namely GA is first used to optimize the BP network's initial weights, and then BP network is available for completing the final training algorithm. Finally, taking the water network of Tianjin Port Free Trade Zone as an example, the unknown pressure values of other nodes are estimated by the known information on the various monitoring points from SCADA system. By the results, the samples whose absolute value of relative error less than 5% are about 85% of the total, which shows the feasibility of the model.