A Novel Method for Hourly Electricity Demand Forecasting

Short-term load forecasting has been playing an increasingly important role in electric power systems. Effective forecasting of the future electricity demand, however, is difficult in view of the complicated effects on load by a variety of factors. A hybrid method based support vector regression (SVR) with meteorological factors and electricity price is proposed to address such problem. First, the input used in this paper, are specific ratio value combinations of each characteristic parameter affecting the hourly electricity load. Second, SVR is used to analyze and develop a load forecasting model. Third, an improved adaptive genetic algorithm (IAGA) is utilized to optimize the specific ratio value combinations of each characteristic parameter, penalty factor C and Gaussian kernel function σ to accurately establish a forecasting model. The experimental results show that the proposed method can obtain better forecasting performance in comparison with other standard and state-of-the-art methods.