Power load forecasting method based on demand response deviation correction

With the increasing deployment of demand response (DR), accurate short-term load forecasting (STLF) is playing an essential role in smart grid operation. This paper addressed the task of improving the accuracy of STLF using a two-stage approach, which adopts the results of traditional algorithms as baselines. The target DR deviation sequence is constructed by the mode decomposition and reorganization of the initial prediction deviation. In the above process, the deviation sequences caused by DR are obtained using dynamic mode decomposition (DMD), where the Hankel matrix is constructed to simplify the process. The final forecast accuracy is improved by su- perposing the results from traditional algorithms and the obtained deviation sequence. Multiple case studies on the data from DR pilot areas and comparison with existing models show that demand response deviation correction based on existing algorithms effectively improves power load forecasting accuracy and has good generalization.

Automated summary

This paper proposes a two-stage approach to improve the accuracy of short-term load forecasting (STLF) in the context of increasing demand response (DR). The approach utilizes traditional algorithms as baselines and constructs a target DR deviation sequence through mode decomposition and reorganization. The deviation sequences caused by DR are obtained using dynamic mode decomposition (DMD) with the assistance of a Hankel matrix. By superposing the results from traditional algorithms and the obtained deviation sequence, the final forecast accuracy is significantly improved. Multiple case studies on DR pilot areas and comparison with existing models demonstrate the effectiveness and generalization of demand response deviation correction based on existing algorithms.