Short Time Load Prediction Based on Atomic Decomposition and Support Vector Machine

This paper proposes a combined forecasting method for power load,which is based on atomic decomposition and support vector machine(A-SVM) method.Firstly,the optimal path combined search strategy-based atomic decomposition is utilized to track and decompose the non-stationary load signal and generate the multiple atomic components and residual error components.Then SVM is applied to construct the mathematical model of the decomposed components.The component forecasting value is output by this model.Finally,each component forecasting value is added as the final load forecasting value of the next moment.The simulation is performed in the actual measured load data from a certain region of Zhejiang regional power grid and the results is compared with that from other two existing methods.It is verified that the proposed A-SVM algorithm can reduce the calculation consuming time to30.75 s.The root mean square error is reduced to 17.97%and the mean absolute error is reduced to 11.85%.Meanwhile,it is also verified that the proposed method is of good robustness and statistical significance and can play a reference role in the load forecasting of this regional power grid in the future.