To main content

InterOpt

Deep integration between machine learning approaches and renewable energy optimization

Contact person

Four categories of ML applications to optimization models
Four categories of ML applications to optimization models

Project goals:

  • Develop knowledge enabling deep integration between ML approaches and the short-term hydro-dominant UC problem in a deregulated power system
  • Realize an ML-based optimization model that can enhance the computational efficiency of generating daily production schedules for Norwegian hydropower producers, accurately represent their physical watercourse, and support their bidding strategies in energy and capacity markets.

Background:

Traditional optimization methods repeatedly solve similar problems without accumulating experience. In contrast, machine learning (ML) efficiently gains experience from historical data and previous decisions. The fact that the same unit commitment (UC) problem is solved every day with only minor changes in input data is the perfect base for ML. Numerical studies have shown that ML can achieve speedups ranging from 2× to 260× in many large-scale power systems without any observed reduction in solution quality.

Explore research areas