Novel power system wind power hybrid forecasting based on secondary decomposition

In order to effectively improve the accuracy requirements of different forecasting tasks in the multi-timescale energy management of wind power novel power system to ensure the economic and stable operation of the system, a short-term forecasting method for wind power based on secondary decomposition and CNN-BiLSTM-Transformer is proposed. In this paper, the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) of wind power data is firstly used to obtain different components, and then the high-frequency component clustering sequences are subjected to variational modal decomposition (VMD), and then the secondary decomposition components are reconstructed by combining with feature selection, which are correspondingly inputted into the CNN-BiLSTM and Transformer to carry out the time-series modelling, and the final component forecasting results are cumulated. The case study demonstrates that the proposed combined model outperforms other models in both the day-ahead and intraday forecasting tasks.