基于谱峭度分析和粒子群Kmeans算法的高压断路器故障诊断研究

Study on Fault Diagnosis of High Voltage Circirt Breaker Based on Spectral Kurtosis Analysis and Particle Swarm Optimization Kmeans Clustering Algorithm

为满足电网对高压断路器高效诊断要求,提出一种谱峭度分析和粒子群K均值算法(PSO-Kmeans)相结合的故障诊断方法。该方法首先对正常状态和故障状态振动信号进行快速Kurtogram谱峭度分析,得到谱峭度指标最大中心频率和相应频率分辨率,据此设计带通滤波器对信号进行去噪;对去噪后的信号进行小波分解,提取小波包能量熵作为特征量;进一步采用PSO-Kmeans对特征量进行聚类分析。实验结果表明:改进谱峭度分析法弥补了传统带通滤波器参数确定的不足,提升去噪效果;去噪与PSO-Kmeans算法相结合的诊断方法克服了传统Kmeans易受初始聚类中心影响的缺点,聚类效果良好且精度高于传统算法,证实该方法适用于高精度高压断路器机械故障诊断。

In order to meet the power grid requirement of high precision circuit breaker fault diagnosis, an improved spectral kurtosis analysis combined with PSO Kmeans algorithm is proposed. Firstly, the Kurtogram spectrum analysis is performed for normal and fault vibration signals, the central and the corresponding frequencies are obtained, the band pass filter is designed;The de-noising signal is decomposed by wavelet, the wavelet packet energy entropy is extracted;PS0-Kmeans is used to cluster the feature quantity. It is proved that, the effectiveness of designed filter is verified;The employed PSO-Kmeans algorithm overcomes the shortcoming of traditional Kmeans, achieves good clustering results and improves the accuracy of fault diagnosis. The method is suitable for high-precision mechanical fault diagnosis of high voltage circuit breaker.