Short-term load forecasting using multiple support vector machines based on fuzzy clustering

According to the future of power load, a load forecasting method of multiple support vector machine based on fuzzy clustering is proposed. Data type, weather and temperature factors are considered in the model. Load data are classified using fuzzy clustering. Each class was modeled using support vector machines which best fitted the special class. The method was simulated utilizing the load data of Shan Dong electrical company from 2005 to 2007. The simulation result showed our method can improve the forecasting accuracy.