基于混沌理论与蜉蝣优化K-means算法的变压器绕组松动故障特征分析方法

Analysis Method of Transformer Winding Looseness Fault Based on Chaos Theory and K-means Algorithm Optimized by Mayfly Optimization Algorithm

为了更加准确有效地对变压器绕组状态进行分析,提出了一种基于混沌理论与蜉蝣优化K-means算法的变压器绕组松动故障特征分析方法。首先,运用C-C法重构变压器振动信号的相空间,分析变压器振动信号的混沌特性,得到关联维数、Kolmogorov熵作为混沌特征。然后,将蜉蝣优化算法引入K-means聚类分析中,对高维相空间轨迹的簇中心选取进行优化,得到相轨迹的簇中心矩之和、矢径偏移,并作为几何特征。实验结果表明:变压器振动信号的最大Lyapunov指数均大于0,适用于混沌特性分析;由变压器振动信号计算出的混沌特征能够表征变压器绕组的松紧程度;同时,经蜉蝣优化的K-means算法得到的簇中心能够作为特征点提取整个相空间轨迹的几何特征,也能够区分绕组的松动故障;将两种特征结合能够实现变压器绕组状态的准确监测,从而为变压器绕组在线检修提供了一种理论依据。

In order to analyze the state of transformer windings more accurately and effectively,a method for analyzing the characteristics of transformer winding looseness fault based on the chaos theory and K-means algorithm optimized by mayfly optimization algorithm is proposed.Firstly,the phase space of transformer vibration signal is reconstructed using C-C method,and the chaotic characteristics of transformer vibration signal are analyzed.The correlation dimension and Kolmogorov entropy are obtained and taken as the chaotic features.Then,the mayfly optimization algorithm is intro-duced into K-means clustering analysis to optimize the cluster center selection of high-dimensional phase space trajectory,and the sum of cluster center distance of phase trajectory and vector offset are obtained and taken as geometric features.The experimental results show that the maximum Lyapunov exponents of transformer vibration signals are all greater than 0,which is suitable for the analysis of chaotic characteristics.The chaos features calculated by transformer vibration sig-nals can characterize the states of transformer windings.At the same time,the cluster centers obtained by the mayfly optimized K-means algorithm can be used as feature points to extract the geometric features of the whole phase space tra-jectory,which distinguish whether the winding looseness fault occurs.The combination of the two features can realize the accurate monitoring of transformer winding states,thus providing a theoretical basis for the online maintenance of trans-former windings.