Research Of The Method Of Optimization Of Water Distribution In Irrigation District Gate Based On ANN

With the increasing scarcity of water resources, the water-saving irrigation of irrigationdistricts in China is increasingly concerned. However, the present reform of watersaving infrastructure in irrigation area of China is mainly to solve the leakage of canalsystem from the point of engineering. The extensive management mode in northeastarea and human-experienced water supply mode lead to a serious waste of irrigationwater resources. Therefore, the optimization of water allocation in irrigation district isstudied and it can make the water saving irrigation come true and have vital significancefor saving water resources.In this paper, the optimization of water allocation among the sluice gates in theprocess of irrigation water distribution scheduling is taken as the research content.Considering that the complex relationship of mutual influence of the canal intake gatesin the running process, the optimization of water allocation models based on neuralnetwork theory is researched. By means of a comparative study of various techniques ofneural network method and drawing lessons from the experience of the application ofneural network technology in water conservancy engineering field, the BP neuralnetwork method and the support vector machine method are applied to the study ofoptimization of water allocation model of gates. In accordance with the irrigation waterrequirement, the opening, flow and water level of sluice gates at the same level weretaken as the main control parameters and the nearest approximation of practicalirrigation water supply to water demand was taken as the goal. The optimization ofwater allocation models were trained and validated by test based on a lot of test dataunder laboratory conditions.Test results are shown as follows:(1) the optimization of gate water allocation modelwas established by the methods of BP neural network and the support vector machine,the error between actual amount of water supply and water demand is small. It canachieve the purpose of saving water and the application of the method has positivesignificance to the irrigation area;(2) the two kinds of optimization of water allocationmodel based neural network are both reliable, the average error between actual amountof water supply and water demand from the model based on the support vector machineis4.46%, which has higher precision and better stability compared with BP neuralnetwork optimization model. So it is valuable to be applied and researched and it isworth in-depth researching in the study of optimization of gate water allocation in thefuture.