Abstract
dc:description.abstractAccurate PV power forecasting is essential for modern power system operations. While Transformers have shown strong potential for capturing temporal patterns, many existing models remain deterministic and overlook predictive uncertainty. This work proposes a unified Transformer-based framework for day-ahead PV forecasting. We first enhance deterministic performance through an encoder-only architecture, temporal embeddings, and nighttime zero-padding, yielding substantial accuracy gains across 11 PV sites. We then extend the model using quantile regression to produce multiple conditional quantiles, enabling empirical 90% and 95% prediction intervals without distributional assumptions. Compared to a traditional SARIMAX baseline, our Transformer-based models consistently achieve higher point accuracy and better-calibrated prediction intervals, offering a practical solution for uncertainty-aware solar forecasting.
Degree
thesis:*- Name thesis:degree_name
- M.S., Systems Engineering
- Level thesis:degree_level
- Master of Science
- Discipline thesis:degree_discipline
- Systems Engineering
- Grantor
- Cornell University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Li, Shuohan
- Chair dc:contributor.chair
-
- Gao, Huaizhu
- Committee member dc:contributor.committeemember
-
- Kowal, Daniel
Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Bibid
- 16938313
- Dc Identifier Other
-
ProQuest Submission ID: 12415
ProQuest Publication ID: 32000016