Design And Realization Of Mid-short Term Load Forecasting System

Power system load forecasting is an important task for power dispatch department. Not only accurate load forecasting is the foundation of power system reliability, economic operation and planning, but also it will influence the power supply plan and the power system operation mode. So it is necessary to develop the software about load forecasting to meet power market.Load forecasting models play an important role in the power system load forecasting. This thesis first summarizes the common forecasting methods, discussing the theory of regression model, trend-extrapolate model and grey model etc. Then the models were built and the predicted result was compared and analyzed. To overcome the limitation of the traditional load forecasting method, a new load forecasting system basing on radial basis Gaussian kernel function (RBF) neural network is proposed in this paper. Genetic algorithm adopting the real coding, crossover probability and mutation probability was applied to optimize the parameters of the neural network, and a faster convergence rate was reached. Theoretical analysis and simulations prove that this load forecasting model is more practical and has more precision than the traditional one.On the basis of fully study of the Load forecasting models, a system software on the Visual Basic and SQL platform which was meet the electric power market demand was designed. Especially, the features such as the hiberarchy in software design, system basic services, the structure of system database and the interface are described. Applying the software in the month electricity forecasting show good results. The forecast process and the results indicate that this software is easy to use, precise in forecasting and it can meet the electric power enterprise better.