Intelligent Collaborative Optimal Scheduling For Water Intake-supply Pump Stations In Water Purification Plant

Water purification plants are major electric energy consumers.Electric energy consumption in water intake-supply pump stations account for up to 80% of which in water purification plants.The electricity produced by clean energy is so limited that traditional fuels must be used to produce electricity stably,which has led to global climate change and energy crisis.Therefore,water intakesupply pump stations are imminent to be optimally scheduled for electric energy saving and carbon dioxide reduction while meeting water supply demand and security constraints.Thus,an intelligent collaborative optimal scheduling method for water intake pump group,clean-water reservoir,and water supply pump group is proposed.Firstly,the long short-term memory network(LSTM)algorithm is applied in flow prediction of water supply pump station because of its learning advantages in processing time series data.Secondly,clean-water reservoir is used to buffers the flow changes of water intake pump station and water supply pump station based on the strategy of “lowpeak storage water at low water consumption period,replenish water at high water consumption period”.Thirdly,the parallel working pumps form a configuration if only these pumps deliver water simultaneously.Non-linear relationship models of the characteristics of working pump configurations are established to improve the accuracy of models.In the end,intelligent collaborative optimal scheduling objectives are designed,and the optimal scheduling plan in the next 24 hours is obtained based on the dynamic programming algorithm.The proposed method adaptively updates data and optimal scheduling plan every hour to improve the accuracy of collaborative optimal scheduling.The experimental results of long-term operation shows that it can effectively save 8.11% electric energy compared to the results of previous manual scheduling.The main work is as follows:(1)The flow prediction model of water supply pump station is established based on the LSTM algorithm and engineering scheduling data of the past 4 years.The problem of large prediction deviation of special time nodes is solved.The prediction model is corrected by feeding back the prediction results to improve the accuracy of the model prediction.Then the predictive flow of water supply pump station in the next 24 hours is obtained.(2)The flow of water intake pump station is planned based on the predictive flow of water supply pump station,the adjustment and storage capacity of the clean-water reservoir,seasonal and diurnal variation regular.Then the planned flow of water intake pump station in the next 24 hours is obtained.(3)The dynamics of working pump configurations in water intake-supply pump stations is not equivalent to the linear combination of the dynamics of each working pump due to upgrades and repairs over years.Thus,it is necessary to accurately acquire the dynamics of working pump configurations.Especially for working pump configurations containing variable frequency drive(VFD)pumps,the non-linear relationship between the frequency of VFD pump and water flow/electric energy consumption/main pipe pressure of the working pump configurations is obtained.For working pump configurations containing only constant frequency pumps,the non-linear relationship between main pipe pressure and water flow/electric energy consumption of the working pump configurations is obtained.(4)Based on the above three achievements,intelligent collaborative optimal scheduling objectives are established to minimize the electric energy consumption and the number of pump switch while meeting flow constraint of water intake-supply pump group,main pipe pressure constraint and water level constraint of clean-water reservoir.And then optimal scheduling plan of water intake-supply pump groups,namely working pump configurations at each time,can be obtained based on dynamic programming algorithm.