基于改进EfficientNet的抽油机井故障分级评价方法  

Fault Classification Evaluation Method of Pumping Unit Wells Based on Improved EfficientNet

作　　者：李翔宇 张帅 高宪文[2] 袁春华 LI Xiangyu;ZHANG Shuai;GAO Xianwen;YUAN Chunhua(Shenyang Ligong University,Shenyang 110159,China;Northeastern University,Shenyang 110819,China)

机构地区：[1]沈阳理工大学自动化与电气工程学院,沈阳110159 [2]东北大学信息科学与工程学院,沈阳110819

出　　处：《沈阳理工大学学报》2025年第3期1-6,共6页Journal of Shenyang Ligong University

基　　金：国家自然科学基金项目(62173073);辽宁省教育厅高等学校基本科研项目(LJKMZ20220618,JYTMS20230215)。

摘　　要：针对抽油机井电参数的特征提取及故障分级评价的准确性问题,提出使用相对位置矩阵对抽油机井电参时序数据进行图像编码增强抽油机井电参数特征并生成抽油机井故障数据集的方法,改进EfficientNet模型对抽油机井故障进行更加精确的分级评价。该方法对抽油机井电参数进一步处理并凸显其关键特征,利用图像编码和深度学习技术提高抽油机井故障分级方法的准确性和效率。通过对典型故障的分级实验以及与多种模型性能的对比实验表明,该方法在抽油机井故障分级评价中具有较高的准确性。The method is proposed to use a relative position matrix to encode the time-series data of electrical parameters of pumping unit wells into images,enhancing the feature extraction of these parameters,and generating a fault dataset for the pumping unit wells.The improved EfficientNet model is utilized for more accurate fault-level evaluation of the pumping unit wells.This approach further processes the electrical parameter data of the pumping unit wells and highlights its key features,leveraging image encoding and deep learning technologies to increase the accuracy and efficiency of fault-level evaluation in pumping unit wells.The results of experiments on typical faults and performance comparison experiments with various models indicate that this method exhibits high accuracy in fault-level evaluation of pumping unit wells.

关 键 词：抽油机井 故障评价 相对位置矩阵 图像编码 EfficientNet 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]