Power prediction for centrifugal pumps at shut off condition based on BP neural network

At present, the power of centrifugal pumps at shut off condition can not be obtained by theory computation. The structure of the BP artificial neural network and its application situation in energy performance prediction of centrifugal pumps were introduced in detail. Based on BP artificial neural network, the characteristic prediction model is established to predict power of centrifugal pumps at shut off condition. The input mode of the BP network prediction model is presented and the number of middle layer is fixed by many tests. The characteristic data of 46 centrifugal pumps at shut off condition are used to train the network model, and the data of the other 3 centrifugal pumps are used to test the network model. The weight of each layer is also presented. The study fruits show that the prediction results of the model agree well with the experiment results. The average prediction discrepancy of the network is 4 percent, the minimum prediction discrepancy is 3.35 percent, and the maximal prediction discrepancy is 4.51 percent. The prediction precision of the BP network model can meet the engineering practical requirement.