A Markov Matrix Based Method for Prediction Direction in Wind Power Forecasting Attacks

Deep learning has a wide range of applications in the field of wind power prediction, but it is susceptible to attacks. The fast gradient sign method based on selected directions achieves better performance than methods relying on random direction determined by judging directions. However, this method still lacks sufficient interpretability and understanding of its underlying mechanisms. In this paper, we propose a method based on the Markov transition matrix to predict directions, which is comparable to the method using selected directions in terms of prediction effectiveness. Furthermore, this paper explains the reason behind such predictable outcomes, namely, that wind power data from the previous time step makes a significant and primary contribution to the wind power at the current time step.