The mid and long-term load forecast method based on synthesis best fitting forecasting model

It is very important for power system planning and market strategy development to forecast mid and long-term load. Building the mathematic model of the historical data of the forecast object is the pith of the load forecast. However, forecasting accuracy is a challenge when applying both classical load forecast methods or heuristic methods individually. In order to solve this problem, this paper proposes a new hybrid method with three separate load models, i.e. a grey model GM(1,1), an exponential smoothing model and an unitary nonlinear regression model base on historical data. Annealed with a integrated optimal fitting approach using genetic algorithm (GA) technique, three coefficients are obtained, including wl of grey model GM(1,1) model, w2 of exponential smoothing model, and w3 of unitary nonlinear regression. Case study is with the Chinese national electricity consumption data from 1990-1999. The proposed method shows very good midterm and long-term forecast accuracy.