Total: 1
Accurate probabilistic forecasting of wind power is essential for maintaining grid stability and enabling efficient integration of renewable energy sources. Gaussian Process (GP) models offer a principled framework for quantifying uncertainty; however, conventional approaches rely on stationary kernels, which are inadequate for modeling the inherently non-stationary nature of wind speed and power output. We propose a non-stationary GP framework that incorporates the generalized spectral mixture (GSM) kernel, enabling the model to capture time-varying patterns and heteroscedastic behaviors in wind speed and wind power data. We evaluate the performance of the proposed model on real-world SCADA data across short\mbox{-,} medium-, and long-term forecasting horizons. Compared to standard radial basis function and spectral mixture kernels, the GSM-based model outperforms, particularly in short-term forecasts. These results highlight the necessity of modeling non-stationarity in wind power forecasting and demonstrate the practical value of non-stationary GP models in operational settings.