Power Load Forecasting Based On Empirical Mode Decomposition And Regroup Improved SVR

Power load forecasting is the basis of the safe and stable operation of the power system,and it is also an important basis for the power department to make the power supply plan.As a typical non-stationary time series,it is difficult to predict the power load sequence stably,accurately and effectively.In addition,with the deepening of the market-oriented reform of the power industry,people have higher requirements on the accuracy of power load prediction.Therefore,the study of power load forecasting has important theoretical and application value.Based on the Empirical Mode Decomposition and Regroup improved Support Vector Regression(EMDRISVR),a new prediction method based on Empirical Mode Decomposition and Regroup improved Support Vector Regression(EMDRISVR)is proposed in this paper.Firstly,the empirical mode decomposition method is used to decompose the power load sequence into several sub-component sequences,which are then classified into three sub-sequences: high-frequency mode,low-frequency mode and remainder.Then,an improved support vector regression method based on the optimization parameters of grid search method is proposed to predict the three sub-sequences respectively,and the final prediction results are obtained by adding them together.Finally,a simulation experiment was designed based on the power load data of Richmond,Virginia,USA,and the predictive performance of the model was evaluated with three specific indicators such as mean absolute percentage error,root-mean-square error,and directional change statistics.The experimental results show that,compared with the traditional support vector regression method,the improved support vector regression prediction method based on empirical mode decomposition reduces the mean absolute percentage error(MAPE)by about 0.59%,and the statistical performance of the direction change is improved by about 9percentage points.To sum up,based on empirical mode decomposition and classification recombination,the power load time series is expanded into sub-series from different perspectives,which enhances the inherent regularity of sub-series and reduces the complexity of prediction.In the future,the parameter and structure optimization ofpower load time series decomposition and recombination can be further studied,or the improvement of long and short term memory neural network,fuzzy inference prediction and other prediction methods can be studied under the framework of decomposition and recombination.