| The weekly and monthly load forecasting plays an important role in the power system planning and operation. The purpose of this article is to discuss how we can forecast weekly and monthly load accurately.This article introduces the grey forecast technology and the artificial neural network forecast technology in detail. By analyzing the advantages and disadvantages and the complementary quality of them we can summarize five integrated models of load forecasting based on grey theory and artificial neural network, they are synthetic parallel connection model, synthetic series connection model, neural network system with auxiliary grey system model, grey system with auxiliary neural network system model and compensated model.According to the requirements of monthly and weekly load forecasting in the HeNan Load Forecasting Project, the synthetic series connection model is chosen for modeling. The synthetic series connection model utilizes the strong non-linear projection capability of neural network to produce more accurate forecasting data form the output of different grey forecasting models. By accepting or rejecting the historical data in different ways we can establish different grey models which have different results. The forecast result is fluctuant in some range but is not a value. It is difficult to find out the model which has the best result in this interval because different stage has different model which has the best accuracy, and it has risk to fix the grey model because the factors which affect the changes of load is random. Therefore we can utilize the strong non-linear projection capability of neural network to carry out optimum combination value form the grey results in the forecast interval.The weekly and monthly load forecasting experiment in HeNan power grid have testified obviously that the forecast accuracy of the synthetic series connection model is better than the results come form use the gray model and...
|
PDF Full Text Request