Study On Leakage Detection Of Urban Water Distribution System

The leakage of urban water distribution system not only will result in waste of water and energy resource, but also may lead to the hazards of public environmental. In view of the present situation of serious leakage in urban water distribution system, it is a great significant to study the problem of leakage in depth and explore the effective method of detecting leakage actively, which can utilize the water resources effectively and reduce the hazards to environment. What's more, it can cut down the expenditure of operation and management so as to enhance the benefits of water supply enterprises in economic and social through detecting leakage. Taking into account the funding of investment, it is generally uneconomic to monitoring of all the nodes in a large-scale water distribution system. Consequently, it can only be detected the leakage by the scant number of monitoring nodes. Combine with the feature of resolving non-linear problems by back propagation (BP) neural network and genetic algorithm, the applications of the BP neural network and genetic algorithm in leakage detection of unban water distribution systems are developed.Combine with the function of complex non-linear mapping, the method of predicting the pressure variation of non-monitoring nodes that after the burst by the monitoring pressure variation with back propagation (BP) neural network is introduced. Then it is possible to detect the leakage combine with the contour map. The analysis of a case shows that it is possible to pre-detect the water leakage with major error by the BP method.Aimed at the problem that back propagation neural network is probe to local optimum , the genetic algorithm is used to optimize the back propagation neural network. Then the optimized GA-BP neural network is applied to detect the leakage of water distribution system. The analysis of a case shows that it is possible to pre-detect the water leakage with minor error by the optimized GA-BP method.It is concluded that the genetic algorithm is able to avoid the BP neural network fall into local minimum by the comparison of the both methods. As a result, the effect of pressure variation prediction as well as leakage pre-detection by GA-BP neural network is much excellent than BP neural network.