【目的】高压隔离开关在户外运行时工作环境恶劣，易受到外力、自然老化和高温潮湿等因素的影响而发生一系列故障，影响电网的健康正常运行。针对隔离开关的机械故障诊断，本文提出了一种基于牛顿拉夫逊优化算法(NRBO)优化特征模态分解(FMD)和支持向量机(SVM)的隔离开关故障诊断方法。【方法】首先，利用NRBO优化FMD的三个参数，得到最优的参数组合，用基于NRBO优化的FMD(NRFMD)对试验采集到的隔离开关振动信号进行分解处理，得到最优本征模态分量；其次，使用精细复合多尺度波动散布熵(RCMFDE)对本征模态分量进行特征提取，得到一个高维的特征矩阵；最后，通过核主成分分析对高维特征矩阵进行降维处理，并输入基于NRBO优化的SVM(NRSVM)模型对隔离开关的故障进行诊断。【结果】对于某220 kV隔离开关进行故障模拟试验，采集四种工况下的隔离开关振动信号，将本文所提故障诊断方法与其他常用诊断方法进行对比。试验结果表明，在不同的机械故障情况下，本文方法对隔离开关的故障分类精度可达到98.33%，具有较高的识别精度，识别准确率高于其他常用算法。【结论】本文使用的NRFMD可以忽略机械信号的周期性和脉冲性，具有较好的鲁棒性;RCMFDE可以更好地提取模态分量的特征。综上所述，本文提出的NRFMD-RCMFDE-NRSVM算法对隔离开关故障诊断具有良好的适用性，为后续针对隔离开关故障的研究提供新的思路。

[Objective] High voltage isolation switches operate in harsh outdoor environments and are susceptible to external forces, natural aging, high temperatures, humidity, and other factors, which can lead to a series of faults and affect the healthy and normal operation of the power grid. This paper proposes a fault diagnosis method for isolation switches based on the Newton-Raphson-based optimizer (NRBO) improved feature mode decomposition (FMD) and support vector machine (SVM). [Methods] Firstly, NRBO was used to optimize the three parameters of FMD, and the optimal parameter combination was obtained. The vibration signals of the isolation switch collected in the experiment were decomposed by the FMD based on NRBO optimization (NRFMD), and the optimal intrinsic mode components were obtained. Secondly, the refined composite multiscale fluctuation dispersion entropy (RCMFDE) was used to extract the intrinsic mode components and obtain a high-dimensional feature matrix. Finally, the kernel principal component analysis was used to reduce the dimension of the high-dimensional feature matrix, and the SVM based on NRBO optimization (NRSVM) model was applied to diagnose the fault of the isolation switch. [Results] The fault simulation experiments were carried out for a 220 kV isolation switch, and the vibration signals of the isolation switch under four working states were collected. The fault diagnosis method proposed in this paper was compared with other commonly used diagnosis methods. The results showed that under different mechanical fault conditions, this method could achieve a fault classification accuracy of 98.33% for isolation switches, demonstrating high recognition accuracy, outperforming other commonly used algorithms. [Conclusion] The NRFMD used in this paper can ignore the periodicity and impulse of mechanical signals, exhibiting good robustness. RCMFDE can better extract the features of mode components. In summary, the proposed NRFMD-RCMFDE-NRSVM algorithm has good applicability for fault diagnosis of isolation switches, providing new insights for future research on isolation switch faults.

国网江苏省电力有限公司重点科技项目(J2024047)