几种人工神经网络在潜油螺杆泵转速预测中的比较

Comparison of Several ANN Models on Predicting of ESPCP Speed

在潜油螺杆泵的使用中,转速的选择与控制十分重要,同时转速的预测能够有效地避免泵效下降、延长泵的使用寿命。目前在分析各因素对螺杆泵转速影响的基础上,采用人工神经网络方法建模并进行预测其后续转速已成为研究的热点。在对螺杆泵转速影响因素的分析基础上,采用BP网络、RBF网络和Elman网络分别对转速进行实例研究并观察其误差。结果表明,三种典型的神经网络模型均能够很好地描述任意温度、原油粘度和泵端压差等工况条件下的螺杆泵转速变化特性,实现螺杆泵转速的预测。RBF神经网络训练速度最快且误差值最小,Elman神经网络和BP神经网络的预测误差接近,但前者的训练误差要比后者平滑。

The choice and control of speed is very important in the use of electrical submersible-motor-driven Progressive Cavity Pumping(ESPCP),while the forecast of speed can efficiently avoid pump efficiency inefficient and extend the service life.Based on analyzing the influence of Progressive Cavity Pump(PCP)speed including temperature,oil viscosity and pumpend pressure difference;it has become a hot topic of modeling and predicting its follow-up speed by artificial neural network.It uses BP network,RBF network and Elman network to study on the speed prediction and observe the error.The results show that these three typical neural networks have a good description of PCP speed characteristics in any working conditions,and realize the PCP speed prediction.The RBF neural network training speed is the fastest and the error value is the minimum,and the Elman neural network and BP neural network prediction error is the most close,but the former training error is smoother than the latter.