Novel Power Load Pattern Recognition Algorithm Based on Unbalanced Data Clustering

An improved fuzzy C-means clustering algorithm is proposed for the imbalance between different date types in the power load dataset and the uniformity effect of the cluster size faced by the fuzzy C-means clustering algorithm when dealing with the unbalanced dataset. Firstly, in the improved algorithm, the sum of the membership values of all samples belonging to a certain class is taken as the volume of the class, and the objective function of the traditional algorithm is normalized by its inverse ratio to overcome the problem of uniformity. Secondly, through the Lagrange multiplier method, a new membership matrix update formula is obtained. Finally, Dynamic Time Warping (DTV) was selected as the distance measure in order to solve the Euclidean distance cannot measure the shape similarity of time series. Experimental results show that the proposed algorithm can effectively identify the load pattern of holidays from the historical load data set.