A Load Classification Framework Based on VMD and Singular Value Energy Difference Spectrum

High load data dimension and insufficient sample characteristics challenge the load clustering accuracy. For this challenge, this paper proposes a load analysis method based on variational mode decomposition (VMD) and the energy difference spectrum of singular value (EDSSV). Firstly, the load data is decomposed by VMD algorithm. Simultaneously, the lowest frequency intrinsic function in the decomposition results is selected for EDSSV. The decomposition manifests load sample characteristics and reduces the load data dimension. Furthermore, the singular value is used to obtain an energy difference spectrum curve by EDSSV, which converts curve features to energy features and reduces the amount of data. The k-means clustering result of the original sample and the result of the energy difference spectrum curve are compared through compactness index. Compared with the k-means method, the result of clustering index CP with the proposed method is reduced by 0.0627, which shows that the clustering accuracy is improved. Also, the load data dimension is reduced by about 80%.