Study On State Estimation Of Urban Water Supply Network Based On Regression Tree

The urban water supply network system is a dynamic network system with large scale,complex structure and strong disturbance.To carry out prediction modeling can better master its running rules and improve the ability of daily scheduling,operation optimization and emergency treatment.The main content of this project is to establish a hydraulic model of urban water supply network and estimate the running state of urban water supply network.First,based on the actual data of the experimental pipe network in a city of central plain area water works and the data of the experimental pipe network built by the water plant,the topological structure of the experimental pipe network is established by combining the field survey data,and the hydraulic model of the experimental pipe network is established.The hydraulic model of the experimental pipe network is as follows: the model is consistent with the actual running state of the pipe network,the average absolute error of the simulated pressure value and the measured value is 0.205 m,and the built hydraulic model has a higher precision.Secondly,for large cities,there are few monitoring points of the water supply network,and the state variables are limited,which can not fully reflect the status of the operation of the pipeline network.A region of central plain area is selected as the research object.There are 12 monitoring points in this area.Based on the actual operation data of the region,two forecasting models based on regression tree are established.CART regression tree and model tree are used to construct two forked tree of training data recursively.The prediction model is obtained and the pressure values of RTU11 and RTU12 in this area are predicted.Finally,the model experiment in this area is as follows: the error between the predicted pressure and the measured pressure of the CART regression tree model for RTU11 and RTU12 is less than 5%,and the error of the prediction pressure and the measured pressure of the model tree to RTU11 and RTU12 is about 10%.Through the experiment,the prediction error of the CART regression tree model is low and can predict the single or multiple monitoring points.The prediction effect is good,and it has a certain indication effect on the estimation of the running state of the pipe network.