Research Of The Maximum Load Forecastmodel Based On Combined Model

The power load forecasting is the core part of the power grid planning, the current grid development faced many problems, for example, the changes in the external environment, the risk of grid operation increasing and so on. In order to make a further solve these problems, people have to have a purpose and planning to do a good job of load forecasting, providing a solid basis for power grid planning.In the load forecasting of power system, there were a lot of different power system load forecasting methods, including all prediction model based on load system, simplifying the simulated from different angles. In the simulation of power system, the development trends of load variables and parameters that with certain limitations were unexpectedly complexity and uncertainties because of the influences of external conditions. So, the use of a single model for electric power load forecasting can not achieve the results which were satisfactory predictions in all cases. Therefore, choosing the comprehensive prediction models had became one of the main ways of power system load forecasting.Based on the analysis of power load forecasting content and characteristic, a varieties of models were taken to forecast the load. The paper analyzed neural network model and regression models predictions most widely used nowadays, respectively, which included basic models, principles, characteristics, calculation methods and so on. Obtained correlation coefficient, the fitted values and the relative error of each load forecasting. Discarded those less effective fitting model. A high precision fitting model were chosen.The article first analyzed the maximum load and six related factors effected power load collected from 1900 to 2014 of Fujin region, got a final prediction by overlaid two optimization models though BP neural network. Secondly, did fitting operation to single model of regression model with the use of existing data. By compared the relative error, chosen the superior one to establish portfolio optimization model with BP neural network, forecast power load of Fujin area. Compared new established model with BP neural network model, verified the accuracy of the improved load forecasting model was higher than single BP neural network model. Illustrate that even the prediction method with a larger prediction error, if it contains independent information of systems when it combined with a forecasting method with a smaller prediction error then it is entirely possible to increase prediction performance of the system. The improved model had good practical significance and use prospects.