基于协整-格兰杰因果检验和季节分解的中期负荷预测

In recent years, with the transformation of national economy, great changes have taken place in the economic structure of China. The prediction based on the historical data of electric power load will cause great error. In order to solve the problem which traditional load forecasting method is not enough for economic and meteorological factors, a forecasting method for medium-term load is proposed. This method can consider the influence of economy, climate and other factors. First, using seasonal decomposition, the monthly electricity consumption of history is decomposed into long-term and cycle component, seasonal component and irregular component, and the relationship between economic factors and long-term trend and cyclic components of electricity consumption is analyzed by cointegration test and Granger causality test in econometrics. The key indexes to influence the prediction of electric quantity is determined. Each component is predicted by support vector machine (SVM) based on electricity, meteorology and economic data, and the monthly total quantity of electricity is predicted. Finally, the effectiveness and feasibility of the method are illustrated by an example.