【摘要】 针对特低渗透油田单井产量低、抽油机能耗大及开采成本高等问题,研究基于径向基函数神经网络的抽油决策优化模型,以便对目前油田采用的间抽制度进行优化,实现高效节能开采。分析了一种基于开关磁阻电机的抽油机自动控制系统结构,建立三层径向基函数神经网络,提出网络隐含层节点中心动态自适应调整算法。研究了网络输出层权重自适应训练算法,用于对电机转速、阈值转速和停抽时间进行预测。采用Matlab进行仿真实验,在训练样本量为2 000、误差设置为0.000 1时径向基神经网络学习300次后达到收敛,并且相比于误差设置为0.000 5时的网络输出,在对100个测试样本进行测试时,电机转速、电机阈值转速和停抽时间预测值均更接近实际值。仿真结果表明,采用径向基函数神经网络优化抽油决策是合理可行的。更多还原

【Abstract】 In order to solve the problems of low single-well production, high energy consumption and high production cost in ultra-low permeability oilfield, the optimizing pumping decision-making model of pumping unit based on radial basis function neural network is established, so as to optimize the intermittent pumping system used in oilfields at present and realize efficient energy-saving production of crude oil.The structure of an automatic control system of pumping unit based on switched reluctance motor is analyzed, a three-layer radial basis function neural network is established, and a dynamic adaptive adjustment algorithm for the node center of the hidden layer of the network is proposed.The weight adaptive training algorithm of network output layer is studied to predict motor speed, threshold speed and stop pumping time.Matlab is used for simulation experiments.When the training sample size is 2 000 and the error is set to 0.000 1,the RBF neural network reaches convergence after 300 times of learning.Compared with the network output when the error is set to 0.000 5,the predicted values of motor speed, motor threshold speed and shutdown time of testing 100 test samples are closer to the actual values.The simulation results show that to use radial basis function neural network for realizing the optimal pumping decision of pumping unit is reasonable and feasible.更多还原