针对油井长时程基于深度残差收缩网络的模型故障诊断

Model Fault Diagnosis for Long-term Oil Wells Based on the Deep Residual Shrinkage Network

目前,大多数的故障检测都是针对故障发生时的这一段时间来进行检测的。当这种方法被用于检测多种故障类型时,其准确性往往会下降。针对上述问题的多故障、长时间序列的油井电参数信号,提出了一种基于深度残差收缩网络(DRSN)模型的故障诊断方法。首先,将采集到的油井长时间序列的电参数信号,按照一定尺寸将其矩阵化。其次,通过将深度残差收缩网络模型应用于故障诊断中,首先是将残差项加入到CNN中解决深度网络的模型退化问题,再通过软阈值化进行样本降噪。最后,为了验证所提方法的有效性,将采集到油井时间序列的数据用于改模型中用于故障诊断。实验结果表明:通过验证该文所提的方法有效性和可行性,表明该诊断方法在油井的故障诊断中有较好的表现和远大前景。

At present,most fault detection is aimed at the period of time when the fault occurs,and when this method is used to detect a variety of fault types,their accuracy tends to decrease.A fault diagnosis method based on the deep residual shrinkage network(DRSN)model is proposed for the multi-fault and long-time sequenced electrical parameter signals of oil wells of above problems.First of all,the collected long-time sequenced electrical parameter signals of oil wells are matriculated according to a certain size.Secondly,the deep residual shrinkage network model is applied to fault diagnosis,that is,the residual term is first added to the CNN to solve the model degradation problem of the deep network,and then the sample noise reduction is carried out through soft thresholding.Finally,in order to verify the effectiveness of the proposed method,the collected time series data of oil wells is used in the modified model for fault diagnosis.Experimental results show that the verification of the effectiveness and feasibility of the method proposed in this paper indicates that the diagnostic method has good performance and broad prospects in the fault diagnosis of oil wells.