The Study Of Forecasting Model Of Urban Pipe Network Based On BP Neural Network And Its Application

Urban drainage is an important urban infrastructure construction for the urban sewage, industrial waste water, and some precipitation to collect, transport, purify, re-use and emit. With the accelerating process of urbanization, urban sewage emissions also increased, resulting in increased flooding, increased runoff pollution, urban water shortages and other problems, urban drainage faces enormous challenges. The current simple level control of pump station exists lag characteristics, if large amounts of sewage flowed into the sewage pipe network in a short time, can easily cause sewage spills. Therefore, the study on prediction models of establishing a sewage pumping station for the safe drainage operation has great significance. But influenced by weather, urban population, pipe leaking and other complex factors, the urban sewage flow is uncertainty and serious nonlinear. Moreover with the testing equipment limitations and the lack of basic information, it makes the physical modeling more difficult.This paper utilizes mass operating data from the SCADA system of Hangzhou drainage Co.Ltd., studies the urban drainage network prediction modeling technique based on neural network , and establishes the entire drainage model for the whole sewage line. Through the model prediction of each pump station monitoring and control using fuzzy algorithm gives the optimal strategy and change a simple single-level control station, the sewage overflow is minimized. Through monitoring and controlling each pump station with the established model, further optimizing operation using fuzzy algorithm, it changes the current simple level control of pumping station, and may makes the sewage overflow minimized. Here summarized the main research work as follows.1. To establish prediction model of cascade pumping stationsThrough the analysis for cascade sewage pumping stations operation mechanism and correlation of the sewage pumping station, we confirm the main factors which influence the front pool level, determine the delay time by calculating the correlation between the upstream and downstream, and make those factors and the delay time as the input of neural network model. Learning and training by large historical data, the various forecasting model with different predictable period based on neural network are established. Then the further validation test and comparison for those models show that the modeling technique has high accuracy and good generalization ability.2. To establish prediction model of convergence pumping stations Because the confluence pumping station has multiple upstreams, it makes the correlation between the upstream stations and the downstream station weaken, uncertainty increased, and can not obtain the delay time by using cross correlation calculation. To solve the problem, through the gray relational analysis and calculation for the confluence pumping station and its multiple upstream pumping stations, we determine the delay time according to the calculation of the rate of delay values at the various time points. Combined with the convergence analysis of real cases, we determine the input for neural networks, train the model by historian data. The validate test shows the confluence model has high accuracy with lateral flow in less than 50%.3.The study on urban drainage simulation and prediction system based on BP neural network and its applicationCombined the above two modeling techniques, we began the engineering application research. First of all, we designed and implemented a prediction system prototype for urban drainage based on neural network: discussed the software design, the key implementation technologies and the UI effect respectively, realized the BP neural network algorithm, configurable pumping station prediction model and real-time simulation functions, so that the software prototype can be intuitive presented to the user. Moreover, we optimize the current simple level control of urban drainage with fuzzy control technology: analyzed the shortcomings of traditional control method, proposed the optimization rules to improved control, and finished the detail design of fuzzy controller. Finally we took Wulin-Gate station as a test example to verify the effectiveness of Fuzzy Optimal Control.