Machine learning models benchmarked for electricity demand forecasting

A new benchmark study evaluated ten machine learning models for short-term electricity demand forecasting in New England, utilizing weather, calendar, and COVID-19 data. The research found that gradient-boosted tree models, specifically CatBoost and XGBoost, outperformed standalone neural network architectures like LSTMs and Transformers. Analysis indicated that historical demand data was the most significant predictor, while the inclusion of COVID-19 indicators showed signs of temporal validity decay. AI

IMPACT Provides a benchmark for applying machine learning to critical infrastructure forecasting, highlighting model performance differences.

RANK_REASON Academic paper presenting a machine learning benchmark. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Catboost
• COVID-19
• LightGBM
• long short-term memory
• New England ISO
• optuna
• random forest
• Reza Ghanavati
• Shap
• Transformer++
• XGBoost

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →