Research On Electric Power Demand Forecasting And Pre-warning Of Regional Power Grid

Mid-term and long-term electric power demand forecast is the basis of electric power system planning and construction. The accuracy of forecast has an impact on the rationality of investment, layout of Grid and operation directly.At the very first, the pre-handling methods concerned of history data during mid-term and long-term electric power demand forecasting are elaborated in this paper. The methods on how to deal with the bad data in the power system are also summarized in the paper. Besides, the different methods of normalizing the history data are analyzed, and one of them is employed by the paper. The paper discusses the method on how to solve the vacancy of the original data set, a method called RBF is proposed by the paper to solve the problem.The method of mid-term and long-term electric power demand forecast of area grid is discussed in the paper. Besides, the feed-forward ANN which is used in the electric power demand forecast of area grid is researched in terms of the basic principal, learning algorithm and improved algorithm based on digital optimal algorithm. Besides, the paper set electric power demand forecast as research object, from the view of decreasing the input vector of ANN, and re-constructs the input vector with the method of PCA (Principle Component Analysis). As a result, the input vector is reduced effectively, and the accuracy of forecast is raised. What's more, the mid-term and long-term electric power demand forecast is done with an example based on gray theory and ANN.Thirdly, the pre-warning of mid-term and long-term electric power is deeply discussed in the paper, including the basic theory of pre-warning, how to set the index system, and the release of Five-Color pre-warning, and so on.It is about the development of mid-term and long-term electric power demand forecast prototype system. It is finished by VC6.0 and ACCESS on the basis of the method and theory of this paper.Last, the whole work of the paper is summarized. And the method of data pretreatment, the technique of forecast and pre-warning, the method of developing the software system are outlook.