Density Peak Clustering Algorithm Considering Time Series Characteristics for Electricity Load Pattern Analysis

To address the problems of Instability of clustering results, poor classification accuracy and inability to extract time series characteristics in electricity load curve clustering analysis, a density peak clustering algorithm (DPC) based on optimized derivative dynamic time warping (DDTW) is proposed. Firstly, calculate the estimated derivatives of the standardized load curve dataset and reconstruct the dataset to extract precise time series characteristics, then the reconstructed dataset is processed by dynamic time warping (DTW) to generate similarity distance matrix, and the new matrix is used to replace the Euclidean distance matrix of the original algorithm as the input data, and finally, DPC is used to cluster the new data on the basis of redefining the sample neighborhood. The OpenEI datasets and UCI datasets are used to analyze and verify the practicability and effectiveness of the improved algorithm.