Research On Water Demand F Orecast And Optimal Pressure Operation Of Water Distribution System In Hefei City

With the continuous growth of population and quick urbanization,the scale and complexity of urban water distribution systems are increasing,resulting in many problems such as high energy consumption,pipeline leakage and water quality risks.This poses significant challenges to the traditional water system operation strategies.To this end,the scientific coordination of water quantity and pressure in water distribution systems to achieve optimal scheduling is important to ensure the stable system operation and also an inevitable requirement for improving the socio-economic efficiency of the water sector.This master thesis aims to conduct research on the technologies for optimizing the scheduling of urban water distribution systems,involving water demand prediction and pressure operation optimization,based on a real water distribution system in Hefei,China.Accurate water demand prediction is a prerequisite for optimal scheduling of water distribution systems.In this study,a long short-term memory(LSTM)based recurrent neural network is applied to urban water demand prediction,where the operating mechanism and characteristics of this prediction model are explored.In addition,the performance of the LSTM is compared with traditional moving average integrated autoregression(ARIMA),support vector regression(SVR)and random forest(RF)models.As the pressure operation is the key sector of a water distribution system,this thesis analyzes the pressure data of the Hefei water distribution system and the historical working conditions of the pumps,based on the result of water demand prediction,thereby identifying the practical solutions to energy saving and consumption reduction with a scientific scheduling scheme.The main observations and conclusions of this research include(1)This research shows that the LSTM model can capture the intrinsic correlation mechanism of the data and is therefore well suited for urban water demand prediction,especially for short-term(1-hour and 15-minute resolution)water demand and DMA water demand prediction.According to the results of the water demand prediction in Hefei city,the LSTM model can provide a more accurate prediction of water consumption in a short period of time than ARIMA,SVR and RF models,with the advantage being more noticeable in predicting urban water demand with data in multiple consecutive periods and DMA water demand with high uncertainty.Besides,the LSTM model does not require the inclusion of external variables(such as temperature and rainfall)in the forecasting process,which is a major advantage in practical application,as the collection of external variables with high time-resolution is often time-consuming and costly.The LSTM model has been applied in the Hefei water distribution system for 6 months and has been running steadily,with an average prediction accuracy around 2.2%for the daily demands evaluated by a third party,satisfying the practical engineering requirement;the 15-minute and 1-hour LSTM prediction model provides technical reservation for the establishment of the real-time global hydraulic model of Hefei and leakage supervision.(2)Based on the water demand and pressure data,this study conducts a systematic analysis of the pressure operation scheduling of the water distribution system in Hefei.The result shows that there were unreasonable spatiotemporal operations in the system,where the water demand during nighttime hours is usually low.Meanwhile,this study has proposed a specific technical solution for the nighttime pressure control of the Hefei water distribution system,which has been gradually implemented in this system so far.In accordance with the need of the water distribution company in Hefei,field testing and energy analysis were carried out on the operation of the pumping stations of the water distribution system in Hefei.It is found that the pump efficiency is basically within a reasonable range,while it can still be further improved.Then this study designs a pump optimization scheduling module,in order to determine the optimal pump combination and frequency setting scheme based on any given pressure and flow requirements,which provides technical support for the optimization of the pumping stations in Hefei.