A COMBINED MODEL FOR SHORT TERM LOAD FORECASTING BASED ON MAXIMUM ENTROPY PRINCIPLE

As uncertainty exists in power system load demand, information theory is introduced to deal with the uncertainty in load forecasting. In this paper, a novel combined short time forecasting model based on maximum entropy principle is proposed. Taking the forecasting values and the historical forecasting error distributions produced by all the adopted individual load forecasting models as constraints, the new combined model can give out a probability distribution of forecasting value based on maximum entropy principle. The background, theory and implementation details of the new combined model are also presented. Application results in a real-life utility show that overfitting problem arising in the traditional combined models is overcomed by the proposed model, and the forecasting precision is improved by the proposed model when load demand is more stochastic.