基于卷积神经网络的在线泵效计算方法研究与应用

Research and Application of Online Pump Efficiency Calculation Method Based on Convolutional Neural Network

油井泵效的计算是油田生产管理中的一项重要工作,针对传统泵效计算方法计算复杂,速度慢的问题,提出一种基于卷积神经网络的泵效计算方法。该方法通过卷积操作从地面示功图中提取特征图,并融合油井参数和抽油机参数,挖掘多源参数与泵效之间的非线性关系。此外,将井下泵示功图作为先验知识引入,有助于泵效计算精度的提升。应用测试结果表明,基于卷积神经网络的泵效计算方法精度高、速度快,可支持多井在线计算,符合油田数字化的需求。

The calculation of oil well pump efficiency is an important work in oilfield production management. Aiming at the problem of complex and slow calculation by using the traditional method, a pump efficiency calculation method based on the convolutional neural network is proposed. The method extracts the feature chart from the ground indicator diagram by the convolution operation, with integrating parameters of the well and pumping unit to excavate the nonlinear relationship between the multi-source parameters and pump efficiency. In addition, the introduction of downhole pump indicator diagram as the prior-in-use knowledge is helpful to improve the accuracy of pump efficiency calculation. The practical application results show the pump efficiency calculation method based on the convolution neural network has high accuracy and fast speed, which can support multi-well online calculation, and satisfy the needs of oilfield digitization.