arXiv：cs.LG· Tomas Kaljevic, Ivan Arzola, Yu Zhang·· 4 小时前AI 评分35
Chronos-2 等时间序列基础模型在协变量不确定性下的负载预测基准测试
Benchmarking Time Series Foundation Models for Load Forecasting Under Covariate Uncertainty
AI 导读
研究对 4 个从头训练模型和 4 个时间序列基础模型（TSFM）在 3 个真实负载预测数据集上进行了基准测试，考察未来协变量信息可用性与质量的影响。Chronos-2 在协变量可用或预测准确时，零样本和微调设置下均达到 SOTA，但随着协变量预测噪声增大性能下降，TimesNet 在严重协变量不确定性下表现更稳健。
正文
Abstract:Accurate short-term load forecasting (STLF) is essential for the reliable and efficient operation of modern power systems. While time series foundation models (TSFMs) have recently demonstrated remarkable performance across a wide range of forecasting tasks, their effectiveness for STLF under realistic operational conditions remains largely unexplored. In this paper, we present a comprehensive benchmark of four trained-from-scratch (TFS) models and four TSFMs across three real-world load forecasting datasets under operational scenarios that differ in the availability and quality of future covariate information. Our results show that Chronos-2 consistently achieves state-of-the-art performance in both zero-shot and fine-tuned settings when future covariates are available or accurately forecast. However, its performance degrades as covariate forecasts become increasingly noisy, whereas TimesNet exhibits greater robustness under severe covariate uncertainty. These findings demonstrate the effectiveness of covariate-informed TSFMs for STLF while highlighting the critical role of robust covariate modeling in real-world forecasting applications.
| Comments: | 5 pages, 1 figure, 5 tables. Accepted to the 2027 IEEE PES Grid Edge Conference & Expo, Salt Lake City, UT, USA, 19-22 April 2027 |
| Subjects: | Machine Learning (cs.LG); Systems and Control (eess.SY) |
| Cite as: | arXiv:2610.07232 [cs.LG] |
| (or arXiv:2610.07232v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.07232 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tomas Kaljevic [view email]
[v1]
Mon, 5 Oct 2026 18:37:55 UTC (98 KB)
来源：arXiv：cs.LG · arxiv.org