| There is a relative shortage of water resources in China,and there are a large number of leakage of urban water distribution,which not only brings economic losses to society,but also seriously affects the normal life of residents.The leakage prediction of water supply network can sense the leakage event in advance and effectively control the amount of leakage,which is of great significance to the work of water resources in our country.This thesis combines the DMA technology to carry out the following work in the leakage detection of water distribution.(1)In this thesis,the commonly used prediction methods of leakage in water distribution are summarized,and through experimental comparison,it is found that most of the prediction models directly through single fitting can’t meet the requirements of prediction accuracy.Furthermore,this thesis proposes a fusion prediction model in water distribution based on residual correction,which judge a leakage event by predicting water flow data,and improve the accuracy of a single prediction model.(2)Through in-depth research,this thesis finds that different algorithms have their own advantages on different data,but the corresponding disadvantages are also obvious.Therefore,based on the residual correction fusion prediction model,a multi-fusion neural network prediction model in water distribution based on residual correction and weight distribution is designed.Before building the weight distribution network,the residual correction of the predicted value of each model is performed,and the weight parameter learning is performed through the BP neural network.The final prediction result is the fusion of the residual correction value and the weight coefficient,so as to integrate the advantages of multiple models.The prediction accuracy of the model is further improved and the engineering availability is improved.(3)Finally,this thesis designs and implements an intelligent pipe network leakage monitoring system based on the multi-fusion neural network prediction algorithm designed in this thesis,which can well apply the algorithm designed in this thesis,and realize the real-time monitoring of the state of water supply pipe network.leakage event detection and timely notification,as well as the basic functions of various platform services. |