基于卷积神经网络的油井供液程度定量分析方法及应用

Quantitative Analysis Method and Application of Oil Well Fluid Supply Degree Based on Convolutional Neural Network

供液不足是抽油井最常见的工况之一,实现供液不足程度的定量化、自动化、智能化分析,对于提升油井运行效率、降低设备磨损和故障风险意义重大。以油井实时监测的示功图图像作为供液不足程度分析的主体,提出了一种基于卷积神经网络的油井供液不足程度定量分析方法,设计了包含4层卷积层、4层池化层和3层全连接层的卷积神经网络模型。采用反向传播算法,以示功图样本集为输入对卷积神经网络模型进行反复训练。结果显示,训练完成的卷积神经网络模型能够高效、准确识别供液不足程度,准确率达98.58%。选取某油田3口油井,通过所建立的油井供液不足程度量化分析方法监测油井供液情况,在此基础上进行远程动态调频生产,实现了抽油机冲速与油井供液程度的合理匹配,在保证日产油量的基础上有效减少了电能浪费。

Insufficient liquid supply is one of the most common working conditions in rod pumped wells.Realizing the quantitative,automated and intelligent analysis of the degree of insufficient liquid supply is of great significance for improving the operating efficiency of oil wells and reducing equipment wear and failure risks.Taking the indicator diagram image of the real-time monitoring of oil wells as the main body of the analysis of the insufficient liquid supply,a quantitative analysis method of the insuffi-cient liquid supply of oil wells based on a convolutional neural network is proposed.A convolutional neural network model including 4 layers of convolution layers,4 layers of pooling layers,and 3 layers of fully connected layers is designed.Using the back-propagation algorithm,the convolutional neural network model is repeatedly trained with the indicator diagram sample set as the input.The results show that the trained convolutional neural network model can efficiently and accurately identify the degree of insufficient fluid supply,with an accuracy rate of 98.58%.Three oil wells in an oilfield are selected,through the established quantitative analysis method for insufficient liquid supply of oil wells,the liquid supply of oil wells is monitored,and on this basis,remote dynamic frequency modulation production is carried out,and a reasonable match between the pumping unit strokes and the degree of liquid supply of oil wells is realized,the oil production is guaranteed,and the waste of electric energy is reduced effi-ciently.