The invention relates to a modeling method for medium and long term power load forecasting. The modeling method comprises the following steps: dividing power demand in multiple layers according to industry and trade, extracting history data to form a forecasting matrix, and constructing an initial load trend model; extracting influence factors and analyzing the degree of correlation of forecasts, and determining the important influence factors of each forecast; forming a forecasting method base; extracting maximum load Pmax, annual average load and annual minimum load Pmin as key load indexes, and analyzing the characteristics and rules appearing in Pmax and Pmin and a forecasting method thereof; selecting the forecasting method to forecast the forecasts in a combined mode, carrying out curve fitting on forecasting results and the history data to obtain the development trend of the forecasts, and constructing a load trend model; amending the parameters of the load trend model and the weight of the forecasting method according to load influence factors and influence weights, and establishing a comprehensive load forecasting model. The modeling method has great advantages for improving the reliability of the power load forecasting and analyzing the inherent law of power load.