Optimization of Oil Pumping Decision Model Based on Radial Basis Function Neural Network

Abstract

Aiming at the problems of low production of single oil well, high energy consumption and production cost of pumping units in ultra-low permeability oilfields, the oil pumping decision model based on RBF neural network was studied to optimize the current intermittent pumping system in oil fields. In paper, the influencing factors of the pumping decision model of the pumping unit were analyzed firstly. Then a three-layer RBF neural network was created, and a dynamic adjustment algorithm for node center of network hidden layer was proposed, and a weight adaptive training algorithm was studied, in which the output error was satisfied through multiple iteration. Finally, the model simulation experiment was carried in Matlab, predicting the motor speed, threshold speed and stop time. With 3000 training samples, when the error was set to 0.0001, the RBF neural network achieved convergence after learning for 300 times. Compared with the network output when the error was set at 0.005, the predicted values of motor speed, threshold motor speed and stop time are closer to the actual values when 100 samples were tested. The simulation results has showed that it is reasonable and feasible to optimize oil pumping decision model of the pumping unit through RBF neural network.