Interpolation-Free Deep Learning for Meteorological Downscaling on Unaligned Grids Across Multiple Domains with Application to Wind Power

TL;DR

The results show that the learned grid alignment strategy performs as well as conventional pre-processing interpolation steps and that LR wind speed at multiple levels is sufficient as a predictor, enabling a more compact architecture, and suggest that extending to new spatial domains using transfer learning is promising.

Abstract

As climate change intensifies, the shift to cleaner energy sources becomes increasingly urgent. With wind energy production set to accelerate, reliable wind probabilistic forecasts are essential to ensure its efficient use. However, since numerical weather prediction models are computationally expensive, probabilistic forecasts are produced at resolutions too coarse to capture all mesoscale wind behaviors. Statistical downscaling, typically applied to enchance the resolution of climate model simulations, presents a viable solution with lower computational costs by learning a mapping from low-resolution (LR) variables to high-resolution (HR) meteorological variables. Leveraging deep learning, we evaluate a downscaling model based on a state-of-the-art U-Net architecture, applied to an ensemble member from a coarse-scale probabilistic forecast of wind velocity. The architecture is modified to incorporate (1) a learned grid alignment strategy to resolve LR-HR grid mismatches and (2) a processing module for multi-level atmospheric predictors. To extend the downscaling model's applicability from fixed spatial domains to the entire Canadian region, we assess a transfer learning approach. Our results show that the learned grid alignment strategy performs as well as conventional pre-processing interpolation steps and that LR wind speed at multiple levels is sufficient as a predictor, enabling a more compact architecture. Additionally, they suggest that extending to new spatial domains using transfer learning is promising, and that downscaled wind velocities demonstrate potential in improving the detection of wind power ramps, a critical phenomenon for wind energy.

Figures (77)

• Figure 1: Predictor and predictand grid shapes illustrated on one of the spatial domain, and the three configurations of domains used in this work. Each domain is represented by the predictand domain and marked by a red rectangle of size $120 \times 120$ km2 including the padding. In addition, each colored dot indicates a wind farm site.
• Figure 2: Legend for each kind of layer and operation used in the three modules that form the downscaling models' architecture.
• Figure 3: Architecture of the multi-modal predictor processing module for the interpolation strategy. Some predictors are not illustrated and are replaced by $...$. The missing predictors are multi-level $U$, multi-level $V$, multi-level $T$, see Fig. \ref{['fig:legend']} for the legend.
• Figure 4: Architecture of the multi-modal predictor processing module for the no-interpolation strategy. Some predictors are not illustrated and are replaced by $...$. The missing predictors are multi-level $U$, multi-level $V$, multi-level $T$, see Fig. \ref{['fig:legend']} for the legend.
• Figure 5: Residual block architecture. See Fig. \ref{['fig:legend']} for the legend.