基于BP神经网络识别的电泵井实时电流故障诊断方法与应用

Application of BP Neural Network Identification in Real-time Current Failure Diagnosis of ESP Wells

电泵井的运行电流数据是判断电泵井故障原因的主要依据。以往电泵井故障诊断主要依据技术人员依据描绘的电流卡片形态进行诊断,诊断结果的可靠性一方面依据技术人员工作经验的积累,另一方面难以实现大批量快速电泵井诊断。现海上油田可采集与传输时电泵井的实时电流数据,大量实时电流数据可为电泵井故障诊断提供数据支持。本文提出基于BP神经网络识别的电泵井实时电流故障诊断:首先利用现场实测电流样本和人工模拟的电流卡卡片形态,按照一定规则进行样本特征值提取建立样本库,其次对实时电流曲线进行模式识别,提取样本特征值,得到其权重值和阀值。最后将实时电流曲线特征值与权值矩阵相似度计算,相似程度可作为判断电泵井故障类型的依据。实践证明该方法可快速、高效、批量用于全油田的电泵井故障诊断。

The operating current data is the troubleshooting basis for ESP wells. In the past, the diagnosis of ESP wells was mainly based on the technician's diagnosis based on the current card. On the one hand, the reliability of the diagnostic results is based on the accumulation of technical experience of the technicians. on the other hand, it is difficult to realize the diagnosis of large-scale rapid for ESP wells. Real-time current data of electric pump wells that can be collected and transmitted in offshore oil fields,a large amount of real-time current data provides data support for ESP wells failure diagnosis. This paper proposes real-time current failure diagnosis of ESP wells based on BP neural network identification: Firstly, use the field measured current samples and artificially simulated current card card, Sample eigenvalue extraction according to certain rules to establish a sample library. Secondly, the real-time current curve is pattern-recognized, extracted the sample feature value, and obtain the weight value and threshold. Finally, the similarity between the real-time current curve eigenvalue and the weight matrix is calculated, the degree of similarity can be used as the basis for judging the type of electric pump well failure. Practice has proved that this method can quickly, efficiently and batch for failure diagnosis of ESP wells in oilfields.