Research On Combined Forecasting Model And Its Application Based On Support Vector Machine

Prediction has been widely used in the various fields of life, There are many common used prediction methods, and support vector machine(SVM) is a novel method based on Statistical Learning Theory, It has properties of good generalization ability and global optimal solution, and demonstrated has a great advantage in solving the small size problem and non-linear problem. However, due to the standard SVM solving quadratic programming problem and has larger computing, while it has the problem of the uncertain parameters and the low precision of the single forecasting model, this paper has studied this problems, and it mainly concentrates on the following contents:Firstly, the research significance of prediction and commonly single prediction method were described, then learned the short-term load time series is non-linear, non-stationary, and empirical mode decomposition(EMD) is an adaptive nonlinear processing method, and the historical data is decomposed into a series of relatively stable component of the sequence by the EMD, due to the EMD may exist mode mixing phenomenon, therefore on this basis, the time series can decomposed by the ensemble empirical mode decomposition(EEMD).Secondly, in order to solve the problem of SVM larger computing, using a least squares support vector machine(LSSVM) simplifies the calculation, then a combined forecasting model based on EEMD and the LSSVM is proposed, and applies this forecasting model used into short-term load forecasting. firstly, the historical data is decomposed by the EEMD, and then the appropriate forecasting model is established for each component of the sequence. The parameters of the LSSVM are optimized through the Bayesian evidence framework, finally, the results of each component forecasting are superimposed to obtain the final forecasting result, and the simulation results show that the combined forecasting model has achieved better than a single model and EEMD more suitable for non-stationary data processing.Finally, although LSSVM simplifies the calculations to some extent, but in the LSSVM case, the sparseness and robustness may lose, based on this, weighted least squares support vector machine(WLSSVM) is used. then identify the Chaos of short-term load data, and use the phase space reconstruction method of chaos theory for prediction, then a combined forecasting model based on chaos theory EEMD-WLSSVM is proposed. firstly, the historical data is decomposed by the EEMD, then calculate the delay time and embedding dimension and phase space reconstruction to establish WLSSVM forecasting model for each component, finally, the results of each component forecasting are superimposed to obtain the final forecasting result, and the simulation results show that the combined forecasting model has achieved better than a single mode, verify the validity of this combined model.