Research Of Load Forecasting Method Based On Grey Model

There are many factors, such as politics, economic, major accidents and so on,which will affect the stable operation of power system. Unstable voltage and current in the power grid will lead to the fluctuation of the load, then cause the abnormal operation of the power grid system and electrical equipment. Therefore, in the power system, the effective estimate of the power grid load is a urgent need of the power system for a reasonable economic dispatch, reducing production costs, to prevent a large area of power failure or power grid collapse in the power grid. Power system load forecasting is the basis of power system for stable operation, and it is a special analysis method to predict the future load through the research and analysis of historical load. In this paper,we study the short term forecasting of power system load, and analyze how to build the forecast model with higher accuracy and faster calculating speed.The work done in this paper is as follows:1)In this paper, the gray model of GM(1,1) based on the historical load data of the power network is used to predict the load value of local power network, proposing dynamic information model in the modeling process.2)Three methods are used to test the accuracy of forecasting results: the relative error test, the posteriori difference test and the correlation test method.3)Meanwhile, three methods are proposed to improve the model prediction accuracy.The first kind of method is to improve the optimization of grey model GM(1,1)in the process of constructing the model, such as, optimization of background value for GM(1,1), optimization of grey derivative for GM(1,1), reasonable selection of initial conditions for GM(1,1). The second kind of method is the application of grey model GM(1,1) and other forecasting model combined into a combinatorial forecasting model,for example, the grey model respectively with the least squares prediction method,exponential smoothing forecasting method, artificial neural network model is combined into a combinatorial optimization model in this paper, by using the way with making up the deficiencies of a single model to improve the prediction accuracy, expecting to achieve good data processing and forecasting effect. The third kind of method is to analyze and forecast the load forecasting model based on RBF neural network for the GM(1,1) improved with residual error correction. Through the final analysis of the results about the model to predict the load of the grid is found that the three methods to improve the accuracy of the model are very reliable, and have achieved good results.