Research On Mid-long Term Power Load Forecasting Based On Nested Compositions

Power load forecasting plays a very important and positive role in power sector, and is an important part of power system planning.The accuracy of load forecasting not only have influence on optimizing the formulation of plans of economic power, but also on the power grid planning and scheduling reasonable operational arrangements influential economy, and on reducing the reserve capacity, the regularity of arrangement influential unit maintenance work, in order to save energy, saving electricity costs and improving the national economy and making contributions. Therefore, in order to improve the accuracy of load forecasting, to make a reasonable load forecasting, developing the new method is very urgent.Based on the actual power usage of Sichuan region in this paper,the consumption of the area were merged and predicted by linear regression method, polynomial prediction method and the improved gray prediction, then choose the prediction model which is more suitable for the area,and then use them as the basis of a combination of model prediction.Then variance- covariance combination model, and so on weighted average combination model and the minimum variance portfolio model were used to merge and predict the consumption of the area.The test results indicate that a combination forecasting model is superior to the single method of the power load forecast on the Sichuan region.Since different combination models for different focus,this paper presents the concept of nested combination forecasting method to further take advantage of the combination models and to obtain a more accurate predictions.The results showed that nested combination forecasting method is scientific and feasible,and it can obviously improve the power load forecasting accuracy and can reduce forecast error for the Sichuan region.Nested combination model is used to predict the next five years power loads of Sichuan.To verify the applicability of the nested combination forecasting model,this article randomly selected three regions of Yunnan, Guizhou and Tibet, and power load forecasting is predicted by nested composition prediction model.It concluded that the theory derived method can be applied to Yunnan and Guizhou, but not suitable for Tibet,which indicates that this method also has limitations.It can not be applied the object whose base models have the same numbers(Far or with negative) of relative error of predicted values in the same year.