Research Of Extended Short-term Load Forecasting Based On Cloud Computing

With the development of the modernization of power system, especially the putting forward of the concept of strong smart grid. The intelligent power system become the inevitable development trend. Power market is a very important step in the process, so the traditional load forecasting have been turning to the market demand forecasting. This is the basis of setting real-time price of electricity market, and the key point of improving economic and social benefits. So the authorities have put forward their own strategies to cope with the change. Power generation unit arises rolling power plan at the historic moment. The plan calls for real-time changes generation plan according to the actual load.Traditional short-term load forecasting method is generally at a particular point in one day, using historical load before that day to forecast the next day load values throughout the day. This approach obviously unable to meet the requirements of rolling power generation. So someone puts forward the concept of extended short-term load forecasting. This method introduced the revision forecast, and used the forecasting results to forecast the load values of next day. So this method can improve the prediction accuracy. Based on the analysis of the method, this article points out two shortcomings, one is the curse of dimensionality,the other is the too long retraining time, and put forward the solutions.This dissertation solves the two shortcomings of extended short-term load forecasting by introducing cloud computing. And innovate the method by combining the two characteristics of it. Design a cloud computing model with one master node and multiple child nodes. Establish the Cloud-SVM model and ECloud-SVM model. Finally verify the effectiveness of proposed method by using the grid’s operating data.