1 Department of Informatics and Mathematical Modeling, Technical University of Denmark2 Mathematical Statistics, Department of Informatics and Mathematical Modeling, Technical University of Denmark3 Department of Applied Mathematics and Computer Science, Technical University of Denmark4 Department of Electrical Engineering, Technical University of Denmark
This report concerns probabilistic forecasts for Nysted Offshore. Different approaches for issuing predictive densities are studied, discussed in details and compared. The results show that the spatial correction of the first order moments of the predictive densities improves the quality of the corresponding forecasts. The spatial correction of the higher order moments is shown to be unnecessary as does not bring any additional amelioration. The best performing of the studied models is based on the adaptive quantile regression using the spatially corrected point predictions as input. This model is shown to outperform the benchmark approach in termps of the CRPS score (accuracy measure) by 1.5%-8.29% depending on the considered prediction horizon.