Get Started
Home
Topics
Search
Library
Research questionHow can zero-shot LLMs predict weather-related distribution-grid outage severity without labeled data?Utilities need outage-severity forecasts from weather information even when labeled outage examples are unavailable. The difficulty is determining whether zero-shot language models can produce reliable predictions across forecast horizons and weather inputs.
AI
Evaluation & Benchmarks
Machine Learning
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.Evaluating Large Language Models for Forced Outage Risk Prediction: Benefits and Comparison to Machine LearningThe evidence covers four zero-shot LLMs and two supervised classifiers evaluated with six years of outage records and high-resolution weather data from a central Texas utility service area. Results are reported using macro-F1 and precision, with additional discussion of actionable reasoning and geographic scalability; they do not establish performance beyond this setting.research paper · Sep 2, 2026
Related questions
How can lightweight time-series forecasters learn accurate, specialized forecasts from scarce private data?How can power-grid operators rapidly classify N-k contingency severity for proactive security decisions?How can zero-shot building IoT forecasting adapt covariates to each building’s physical topology?How can we select the best probabilistic weather forecast model for a specific downstream decision?