Short-Term Load Forecasting Method Based on Self-Organizing Feature Mapping Neural Network and GA-Least Square SVC Model

Accurate load forecasting is very important for economic and effective operation of whole power grid. To make collaborative optimization improvement for load orecasting set and the forecasting model, based on particle swarm optimization- self-organizing feature mapping (PSO-SOFM) and genetic algorithm-based least square support vector machine (GA-LSSVM) a short-term power load forecasting method was proposed. The adaptive weighted PSO was used to optimize the weight of SOFM neural network, and the optimized PSO-SOFM neural network was used to classify the sort processing the original load date to obtain multi groups of training sets. For each group of training set a least square support vector machine forecasting model was established and its key parameters were optimized by GA, finally a GA-LSSVM forecasting model was obtained. Finally, performing load forecasting by existed load data, the effectiveness and accuracy of the proposed method are verified.