Parameter-Efficient Adaptation of Time Series Foundation Models for Electrical Load Forecasting

Key points

• Accurate electrical load forecasting supports power-system operation across countries, regions, cities, and buildings, but repeated development of target-specific models is costly.1
• Time series foundation models (TSFMs) provide reusable pretrained backbones, although multi-domain pretraining may not fully capture load-specific characteristics.1
• Pretrained foundation models outperform the implemented seasonal naive and XGBoost baselines across the evaluated aggregation levels.1
• Chronos-2 achieves lower pretrained NMAE than MOIRAI but benefits less from adaptation.1

Sources (1)

• [1]Parameter-Efficient Adaptation of Time Series Foundation Models for Electrical Load Forecasting
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 01:40 PM

  This paper compares pretrained inference, full fine-tuning, and low-rank adaptation (LoRA) in terms of forecasting accuracy and adaptation efficiency across four electrical load aggregation levels.

  Accurate electrical load forecasting supports power-system operation across countries, regions, cities, and buildings, but repeated development of target-specific models is costly.

Extractive summary: sentences quoted from the sources.