Study Of Power Optimal Scheduling Of Water Distribution System Based On Differential Evolution And Bp Neural Network

With the continuous expansion of the city, water distribution system as an important guarantee develops to a large and complex system, in the current dual requirements of stable water distribution system and energy saving, it is very necessary to study on the optimal scheduling of water distribution system. The optimal scheduling of water distribution system research is divided into three parts: the forecast of water consumption, water distribution network model and optimal scheduling decision.This dissertation studies the prediction of water consumption firstly, through understanding the common methods, and analyzes the time series variation of water consumption. Using BP neural network method to build a water demand forecast model, it is a three-tier network structure, in which the hidden layer nodes are identified to be 9 through repeated comparisons; using seven days’ water consumption data as training sample for the model to predict water consumption of the 24 hours next day. The results show that the model has a certain accuracy and applicability, prediction results can be used to optimize the scheduling decisions.Then this dissertation analyzes the method how to establish the microscopic and macroscopic model in order to research on the water distribution network. Than using the method of BP neural network to build the pressure points of water supply pipe network macro model,which contains the network structure of three layers and the hidden layer node is determined as 7. The water pumping station and pipeline network total traffic flow is established to be the input while the measuring point pressure is the output index. The model is verified by the data of T city. The result shows that model can be applied to optimize the scheduling decision.Finally, the optimal scheduling model is to be established in view of the energy saving demand of current water distribution system according to the operation characteristics of the water distribution system. Using differential evolution algorithm to solve the model, this dissertation uses real coding to represent the decision variables to be the individual in the population evolution. It uses penalty function method to transform constraint conditions, than constructs the fitness function on the basic of combination by the way of addition and multiplication form. The standard differential strategy is used for mutation operation. The study verifies the validity of the model and the differential evolution algorithm through the water distribution system in T city. The optimization scheduling solution has good energy saving effect compared with the genetic algorithm scheduling results.