A Short-Term Power Load Prediction Algorithm of Based on Power Load Factor Deep Cluster Neural Network

Abstract

Accurate short-term prediction of power system load is of great importance to improvement of power system’s stability and electrical equipment’s safety, and a critical step for power load prediction is clustering of existing historical information of load. Since load itself is a time series data with high dimensionality, and meanwhile, load is influenced by meteorological factors and seasonal factors, as a result it’s difficult to establish simple linear relationships with the load and these factors, and consequently, the clustering quality obtained by traditional algorithms is low, which further affects quality of load prediction. Deep learning algorithm was introduced in this paper to construct a power load factor deep cluster neural network (PLDCNN) which consisted of multi-group neural network. Station operating data for consecutive 3 years were introduced in the simulation experiment. PLDCNN was compared with traditional clustering algorithms, and experiment indicated that PLDCNN could describe and classify power load information more accurately, and short-term load prediction of electric power system based on PLDCNN could also archive higher accuracy.