Comparison of Dimensionality Reduction Techniques for Clustering and Visualization of Load Profiles

Due to the recent deregulated electrical market and the increasing consumption of electrical energy, new possibilities have been opening for the electricity suppliers to formulate tariff packages and to improve the quality services to satisfy the daily demand. A key aspect to visualize and interpret a huge volume of data is to cluster customers according to their individual electrical load profile. For this purpose, it is proposed to apply dimensionality reduction techniques, namely, Principal Component Analysis (PCA), Isometric Feature Mapping (Isomap), Sammon Mapping, Locally Linear Embedding (LLE) and Stochastic Neighbor Embedding (SNE) to such data. The IEEE 30-bus system is used in this paper for evaluating these techniques. SNE performed the best when clustering the load profiles into the expected eight clusters.