Wind speed forecasting with correlation network pruning and augmentation: A two-phase deep learning method

To ensure the operational reliability of power systems, it is important for wind speed signal forecasting systems turbines to be efficient, accurate and stable. This paper proposes a two-phase deep learning structure with network augmentation pruning. By introducing cross-correlation quasi-convex optimization, fractional quadratic programming problem related convex optimization models are constructed generate augmented data this proposed internal network; by pruning weakly correlated convolution channels, redundant features its external reduced. Furthermore, closed-form solution model derived, which reduces computational complexity considerably from O(n*log(2N)) O(n). The approach has been extensively validated using real farm in China. results numerical experiments demonstrate that method achieves superior performance training flexibility, accuracy, stability, interpretability.