Power Load Curve Clustering Based on IGWO-Kmedoids

This paper introduces an advanced clustering algorithm for power load curves, named IGWO-Kmedoids, addressing challenges faced by traditional clustering methods like k-means and DBSCAN in handling power system load data. These challenges include difficulty in parameter setting, slow clustering speeds, and suboptimal results. The proposed algorithm begins with data preprocessing, including outlier removal and filling missing values through interpolation, followed by standardization. It then employs Principal Component Analysis (PCA) for dimensionality reduction, easing the computational load for clustering. The core innovation lies in optimizing the K-medoids clustering algorithm with an Improved Grey Wolf Optimization (IGWO) approach, enhancing clustering accuracy and speed. This optimization significantly boosts the power system’s operational efficiency and the optimization of scheduling policies.