Explainable Load Forecasting with Covariate-Informed Time Series Foundation Models
2026-04-30 • Machine Learning
Machine Learning
AI summaryⓘ
The authors explore special forecasting models called Time Series Foundation Models (TSFMs) for predicting energy use in power grids. They focus on making these complex models easier to understand by creating a new method to explain how the models make decisions, using a technique called SHAP. Their method takes advantage of the models' ability to handle different amounts and types of input data efficiently. Testing on real energy data showed the models made good predictions and used important factors like weather and calendar dates sensibly. This suggests TSFMs can be reliable and transparent tools for energy forecasting.
Time Series ForecastingFoundation ModelsSHAP (Shapley Additive Explanations)Load ForecastingPower GridsTransformer ModelsTemporal MaskingCovariate MaskingZero-shot LearningEnergy Systems
Authors
Matthias Hertel, Alexandra Nikoltchovska, Sebastian Pütz, Ralf Mikut, Benjamin Schäfer, Veit Hagenmeyer
Abstract
Time Series Foundation Models (TSFMs) have recently emerged as general-purpose forecasting models and show considerable potential for applications in energy systems. However, applications in critical infrastructure like power grids require transparency to ensure trust and reliability and cannot rely on pure black-box models. To enhance the transparency of TSFMs, we propose an efficient algorithm for computing Shapley Additive Explanations (SHAP) tailored to these models. The proposed approach leverages the flexibility of TSFMs with respect to input context length and provided covariates. This property enables efficient temporal and covariate masking (selectively withholding inputs), allowing for a scalable explanation of model predictions using SHAP. We evaluate two TSFMs - Chronos-2 and TabPFN-TS - on a day-ahead load forecasting task for a transmission system operator (TSO). In a zero-shot setting, both models achieve predictive performance competitive with a Transformer model trained specifically on multiple years of TSO data. The explanations obtained through our proposed approach align with established domain knowledge, particularly as the TSFMs appropriately use weather and calendar information for load prediction. Overall, we demonstrate that TSFMs can serve as transparent and reliable tools for operational energy forecasting.