- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper presents a foundation model for wind power forecasting. It combines fusion of multiple weather forecasts, time aligned patching, static site features, and historical weather and power observations. Experiments cover 1,012 wind power series from 33 datasets, including tests on unseen sites, unseen regions, probabilistic forecasting, and finer time resolutions.
Critical strengths
- The paper addresses a consequential problem on a large heterogeneous corpus , and its chronological splits and explicit cross-region holdout are more informative than random-window benchmarks.
- The paper maps its design to identifiable forecasting challenges, and ablations support the key choices.
Critical weaknesses
-
The paper does not separate its gains from pretraining scale and input access. W2P-FM is pretrained on 938 sites and fuses three NWP sources, while the baselines train on each dataset with a single source. The 12.3% improvement is therefore not attributable to the architecture alone.
-
The zero shot setting still uses each site's recent power and weather history at inference, as shown in Equation 1 and Section 4.2. It tests transfer without parameter fitting, not the cold start case motivating the paper. The preprocessing details in Appendix C.1 are unclear.
-
The paper reports no per seed variation or confidence intervals despite averaging over three seeds. This matters for small margins such as the 1.77% improvement for North America at 24 hours, as reported in Table 1. Geographic coverage is also uneven, with no North American sites in the holdout and no African sites in the corpus.
Questions for authors
-
Could you add a controlled comparison with matched pretraining sites, horizon mixture, and NWP sources against a strong pooled data baseline? If the baseline cannot natively accept multiple NWP products, could you evaluate a simple matched input fusion variant? This would help separate corpus scale, additional weather information, and architecture contributions.
-
What target site history is used at inference for held out sites? How do results change with shorter history or without Meteo Power Encoding? How are anomalous power readings identified, and can short gap interpolation be guaranteed not to use future observations or cross split boundaries?
-
Could you report per seed results or confidence intervals for the main full shot and zero shot comparisons, ideally with a paired site level analysis? Could you also clarify how the missing North American and African sites bound the global generalization claim?
7 comments
Rebuttal by Authors (Part 2 of 4)
Performance degrades consistently as the context is shortened, most at the 6-hour horizon. This reflects the strong short-term correlation of wind power: short-term forecasts rely more on recent observations, whereas long-term forecasts rely on the site's stable weather-to-power response, which a shortened context can still capture. Notably, even with one-quarter of the context, W2P-FM still substantially outperforms all full-context zero-shot baselines in Table 2 (e.g., MSE of 0.0260 vs. 0.0434 at 6h).
(2) Ablation of Meteo-Power Encoding. Extending the analysis in Figure 5(b), we retrain W2P-FM without MPE on the full pretraining corpus while retaining static context conditioning, using the same sites, partitions, configurations, and three random seeds.
| Setting | Horizon | MSE (W2P-FM) | MSE (w/o MPE) | Increase (%) | MAE (W2P-FM) | MAE (w/o MPE) | Increase (%) |
|---|---|---|---|---|---|---|---|
| Full-shot | 6h | 0.0189 | 0.0193 | 2.12 | 0.0928 | 0.0935 | 0.75 |
| Full-shot | 24h | 0.0250 | 0.0255 | 2.00 | 0.1072 | 0.1086 | 1.31 |
| Full-shot | 168h | 0.0441 | 0.0475 | 7.71 | 0.1486 | 0.1539 | 3.57 |
| Zero-shot | 6h | 0.0219 | 0.0230 | 5.02 | 0.1031 | 0.1062 | 3.01 |
| Zero-shot | 24h | 0.0294 | 0.0315 | 7.14 | 0.1197 | 0.1252 | 4.59 |
| Zero-shot | 168h | 0.0450 | 0.0525 | 16.67 | 0.1539 | 0.1673 | 8.71 |
Removing MPE consistently degrades performance, especially in cross-site zero-shot forecasting, indicating that unseen sites rely more on MPE to capture site-specific weather-to-power responses from historical observations. The gap also generally widens with horizon, consistent with the analysis above: long-term forecasts depend more on the stable response patterns that MPE summarizes from the full context, complementing the local aggregation of MSFFN.
W2 & Q2: Anomalous Power Readings and Short-Gap Interpolation
Regarding the preprocessing concerns in W2 and Q2, we have revised Appendix C.1 to clarify anomaly identification and interpolation boundaries. Interpolation and sample construction never cross the training, validation, or test boundaries, and short-gap interpolation has a negligible impact on the overall evaluation results.
- Invalid power observations, including missing or empty values (NaN), negative power readings, and periods explicitly marked as invalid operating conditions in the original records, are replaced with zeros.
- The complete power series of each site is first divided chronologically into training, validation, and test sets, and short-gap interpolation and sliding-window construction are then performed independently within each split. Consequently, interpolation and sample construction never cross the training, validation, or test boundaries.
- Since short-gap interpolation uses neighboring valid observations (Appendix C.1), one edge case may still arise within individual samples: an interpolated power value near the end of the historical context may be computed using a valid observation after the forecast origin. Such cases are rare: only about 0.02%, 0.005%, and 0.0007% of forecast target timestamps for the 6h, 24h, and 168h horizons, respectively, have their observed values used in such interpolation.
- Using the original W2P-FM and W2P-FM-NS checkpoints from Table 1 without retraining, we reset interpolated power values to zero only in the historical context (History) or only in the forecast targets (Targets), while retaining the same test samples and forecast target timestamps. Removing short-gap interpolation leads to only marginal changes in MAE and MSE, and the relative performance of W2P-FM and W2P-FM-NS remains unchanged across all forecasting horizons.
| Setting | Horizon | Model | MSE (Original) | MSE (History) | MSE (Targets) | MAE (Original) | MAE (History) | MAE (Targets) |
|---|---|---|---|---|---|---|---|---|
| Full-shot | 6h | W2P-FM-NS | 0.0210 | 0.0210 | 0.0212 | 0.0986 | 0.0987 | 0.0991 |
| Full-shot | 6h | W2P-FM | 0.0189 | 0.0189 | 0.0191 | 0.0928 | 0.0928 | 0.0933 |
| Full-shot | 24h | W2P-FM-NS | 0.0279 | 0.0279 | 0.0282 | 0.1144 | 0.1143 | 0.1148 |
| Full-shot | 24h | W2P-FM | 0.0250 | 0.0250 | 0.0252 | 0.1072 | 0.1072 | 0.1076 |
| Full-shot | 168h | W2P-FM-NS | 0.0469 | 0.0469 | 0.0472 | 0.1541 | 0.1541 | 0.1546 |
| Full-shot | 168h | W2P-FM | 0.0441 | 0.0441 | 0.0443 | 0.1486 | 0.1486 | 0.1491 |
| Zero-shot | 6h | W2P-FM-NS | 0.0228 | 0.0228 | 0.0229 | 0.1054 | 0.1054 | 0.1057 |
| Zero-shot | 6h | W2P-FM | 0.0219 | 0.0219 | 0.0220 | 0.1031 | 0.1032 | 0.1034 |
| Zero-shot | 24h | W2P-FM-NS | 0.0298 | 0.0298 | 0.0300 | 0.1213 | 0.1213 | 0.1216 |
| Zero-shot | 24h | W2P-FM | 0.0294 | 0.0294 | 0.0295 | 0.1197 | 0.1197 | 0.1199 |
| Zero-shot | 168h | W2P-FM-NS | 0.0454 | 0.0454 | 0.0456 | 0.1551 | 0.1551 | 0.1555 |
| Zero-shot | 168h | W2P-FM | 0.0450 | 0.0450 | 0.0452 | 0.1539 | 0.1539 | 0.1542 |
Rebuttal by Authors (Part 3 of 4)
W3 & Q3: Seed-Level Variation and Paired Site-Level Analysis
Following your suggestion, we report seed-level standard deviations and conduct paired site-level analyses with bootstrap confidence intervals, and find that the improvements of W2P-FM are consistent across seeds and sites: it outperforms the best-performing baseline on more than 81% of full-shot sites and 97.30%–100% of unseen sites, and all 95% confidence intervals of the mean improvement remain above zero. We have added the paired site-level analysis and bootstrap confidence intervals to Appendix F (Table 19).
(1) Seed-level variation. Seed-level standard deviations of all full-shot methods are reported in Appendix F (Tables 13–15). In the North American 24-hour setting you highlighted, W2P-FM achieves an MSE of 0.0277 ± 0.0004 vs. 0.0282 ± 0.0001 for the strongest baseline (LightGBM), with a 3.81% MAE improvement (0.1137 vs. 0.1182). The three-seed mean and standard deviation of W2P-FM and W2P-FM-NS are summarized below; zero-shot baselines use official pretrained checkpoints without additional training and thus have no seed variation.
| Setting | Horizon | W2P-FM (MSE) | W2P-FM (MAE) | W2P-FM-NS (MSE) | W2P-FM-NS (MAE) |
|---|---|---|---|---|---|
| Full-shot | 6h | 0.0189±0.0000 | 0.0928±0.0003 | 0.0210±0.0001 | 0.0986±0.0003 |
| Full-shot | 24h | 0.0250±0.0001 | 0.1072±0.0002 | 0.0279±0.0001 | 0.1144±0.0004 |
| Full-shot | 168h | 0.0441±0.0002 | 0.1486±0.0003 | 0.0469±0.0002 | 0.1541±0.0000 |
| Zero-shot | 6h | 0.0219±0.0001 | 0.1031±0.0002 | 0.0228±0.0000 | 0.1054±0.0000 |
| Zero-shot | 24h | 0.0294±0.0003 | 0.1197±0.0008 | 0.0298±0.0000 | 0.1213±0.0005 |
| Zero-shot | 168h | 0.0450±0.0004 | 0.1539±0.0007 | 0.0454±0.0003 | 0.1551±0.0008 |
(2) Paired site-level analysis. For each site and horizon, W2P-FM is compared against the best-performing baseline for that site and horizon. We report the site-level win rate and the mean and median relative improvements averaged over three random seeds. 95% confidence intervals of the mean improvement are estimated by paired bootstrap resampling over sites with 10,000 iterations.
| Setting | Horizon | Metric | Win Rate | Mean Imp. | Median Imp. | 95% CI |
|---|---|---|---|---|---|---|
| Full-shot | 6h | MAE | 779/938 (83.05%) | 7.29% | 8.04% | [5.90%, 8.37%] |
| Full-shot | 6h | MSE | 787/938 (83.90%) | 7.57% | 8.81% | [4.75%, 11.01%] |
| Full-shot | 24h | MAE | 790/938 (84.22%) | 6.37% | 7.31% | [4.94%, 7.54%] |
| Full-shot | 24h | MSE | 766/938 (81.66%) | 5.17% | 6.39% | [3.58%, 7.63%] |
| Full-shot | 168h | MAE | 826/938 (88.06%) | 8.10% | 8.02% | [7.11%, 9.05%] |
| Full-shot | 168h | MSE | 817/938 (87.10%) | 9.55% | 10.76% | [7.73%, 11.24%] |
| Zero-shot | 6h | MAE | 74/74 (100%) | 26.78% | 25.49% | [24.68%, 28.83%] |
| Zero-shot | 6h | MSE | 74/74 (100%) | 44.74% | 43.98% | [41.70%, 47.68%] |
| Zero-shot | 24h | MAE | 73/74 (98.65%) | 31.08% | 28.76% | [28.00%, 34.14%] |
| Zero-shot | 24h | MSE | 73/74 (98.65%) | 47.85% | 43.88% | [43.46%, 52.14%] |
| Zero-shot | 168h | MAE | 72/74 (97.30%) | 28.64% | 27.37% | [26.31%, 30.88%] |
| Zero-shot | 168h | MSE | 73/74 (98.65%) | 43.72% | 42.25% | [40.57%, 46.74%] |
Rebuttal by Authors (Part 1 of 4)
Dear Reviewer 3EP8,
We sincerely thank you for your thoughtful review. We are grateful that you acknowledge the informative evaluation on a large heterogeneous corpus and the challenge-driven design supported by ablations.
W1 & Q1: Controlled Comparison under Matched Data and NWP Inputs
Following your suggestion, we conduct controlled comparisons against strong pooled baselines with matched pretraining sites, horizon mixture, and NWP inputs, and find that W2P-FM-NS still achieves the lowest MAE and MSE across all three horizons, reducing average errors by 5.76%–6.56% relative to the strongest baseline. With corpus scale and weather information matched, these results demonstrate a clear advantage of its weather-to-power representation learning architecture. Moreover, comparing W2P-FM-NS under different NWP inputs shows that MeteoRouter effectively exploits the complementary information in multi-source NWP. The related experiments and analyses have been added to Appendix C.13.
(1) Controlled comparison experiment
(2) Contribution of additional weather information
These experiments also isolate the contribution of additional weather information. For W2P-FM-NS, the full-shot MSE decreases consistently from GFS to Uniform input and further to MeteoRouter-based fusion (W2P-FM-NS in Table 1). This indicates that multi-source NWP provides complementary information, which MeteoRouter exploits more effectively than uniform averaging, consistent with the corresponding ablation of W2P-FM (Table 8).
W2: Zero-Shot Setting
In our zero-shot setting, each target site's historical information is limited to the short context available immediately before each forecast origin (24 or 72 hours), without any fine-tuning on the target sites. This setting corresponds to the scenario motivated in the Introduction (Lines 44–47), where newly commissioned sites lack sufficient data for model training but can provide recent observations. It is also consistent with existing zero-shot forecasting studies, including WindFM (Fan et al., 2025) and Chronos-2 (Ansari et al., 2025). We have clarified this evaluation setting in Appendix C.1.
Q2: Shorter Historical Context and MPE Ablation
To address your questions, we conduct two additional experiments with shortened historical context and without MPE, and find that even with only one-quarter of the original context, W2P-FM still substantially outperforms all zero-shot baselines that use the full context, while removing MPE consistently degrades performance, particularly for unseen sites and longer horizons. We have added both analyses to Appendix C.14.
(1) Sensitivity to historical context length. Using the original pretrained checkpoints on the 74 cross-site zero-shot test sites without retraining, we shorten both historical power and meteorological observations to one-half and one-quarter of the original length. Shortened inputs are left-padded with zeros and marked as invalid positions, preserving the original input shape.