A power quality disturbance classification method based on improved Shapelet method

The extensive grid connection of new energy and nonlinear power electronic devices has made power quality disturbance (PQD) problems more frequent, seriously affecting the stable operation of the power grid system. In response to the real-time response requirements of the research model of this problem, this study proposed an improved Shapelet method and applied it to the classification of PQDs. First, the concept of subsequence blocks was proposed, and the diversity of Shapelet was enhanced by multiple subsequences of multiple length ranges. In order to solve the problem of high time complexity of searching subsequence blocks, the length range of subsequence blocks was determined by the multi-scale extreme point peak distance method. This method uses the Black Kite Algorithm (BKA) to optimize the parameters of the Variable Mode Decomposition (VMD), decomposes the PQD signal into multiple modal components, and then screens out the disturbance components through permutation and combination entropy and calculates the average peak distance of the extreme points; secondly, a multiple loss function is used to optimize the quality of the selected subsequence blocks through the similarity loss and distance loss between subsequence blocks; finally, the K-means weight initialization method is used to accelerate the convergence of the model. Experimental results show that this method has an accuracy rate of 98.63 % in identifying PQDs in 16 simulated environments, with an average time consumption of 0.141 ms for per data sample. On the measured real data, the recognition accuracy rate is 98.20 % with a time consumption of 0.08 ms for per data sample. This method can provide a good choice for real-time PQD analysis of power grid systems.