Research And Application Of Optimal Operation Of Long-distance Multi-source Water Delivery System

As an important part of the urban water supply system,the degree of optimal operation of the water delivery system is directly related to the normal operation and production of water plants.At present,the operation and management of pumping stations in China is in a state of experience scheduling.Although the scheduling scheme can ensure the normal water transmission,it lacks theoretical and technical support,and the energy consumption of pumping station is greater.Therefore,in this paper,the daily water supply forecast,macro-hydraulic model and pumping station optimization model are carried out to enhance the rationality and economy of the water delivery system scheduling scheme.First of all,a suitable prediction model is selected to predict the daily supply of water delivery system.Based on long short-term memory networks model and grey model,the relative errors of daily water supply are 3.65% and 4.39% respectively.Based on RBF neural network and BP neural network,an explanatory prediction model is established.The influence factors of water quantity are screened by grey correlation analysis and taken into account in the prediction model.Because of its special hidden layer structure,RBF neural network avoids the weakness that BP neural network is prone to fall into local optimization.The average error percentage is 2.98%,and the prediction accuracy is higher than that of BP network.Based on the validity,the long short-term memory networks model is weighted with the RBF neural network model,and the combined prediction model is obtained,and the average prediction error is 2.47%.The above models are evaluated with prediction results,and the combined prediction model is superior to the single prediction model in terms of stability and prediction accuracy.Secondly,based on the BP neural network and the historical operating condition data of the water delivery system,with the flow of each pumping station and the total flow of the system as input variables while the pressure of control point as output variable,the macro-hydraulic model of pressure measuring point is established.The average relative error between the simulated value and the measured value is 3.82%.The macro-model of the whole water delivery system was trained and fitted with the control point pressure and total system flow as input variables,and the outlet flow and pressure of each pumping station as output variables.Input the control point pressure simulation value and the whole system water quantity prediction value,gained the simulation results of the two pumping stations each outlet flow and pressure.Compared with the measured value,the simulation error is basically controlled within 5%,which meets the requirements of pumping station optimization scheduling accuracy.Finally,based on the water quantity prediction results and macro-model simulation results,the outlet constraint conditions of pumping station are obtained.Based on the characteristic curve fitting of pump,the minimum energy consumption of pumping station is taken as the objective function,the ratio of start-stop and speed regulation of pump is taken as the decision variable,and the optimal mathematical model of pumping station is established with the constraint conditions of outlet pressure,flow rate,threshold of speed regulation ratio and high efficiency interval of pump.By using the improved genetic algorithm to solve the model,the pump combination scheme with the lowest energy consumption and the frequency regulation parameters of the speed regulating pump are obtained.The total power of the A pumping station dispatching scheme is reduced by 6.44%on average and the 100-meter head of 1000 tons of water was reduced by 6.06% on average compared with the actual operation scheme after optimization,which achieves the purpose of energy saving and optimal operation of the pumping station.