ARTFEED — Contemporary Art Intelligence

Imitation Learning Optimizes Microgrid Energy Management

ai-technology · 2026-04-30

A new framework uses imitation learning to approximate mixed-integer Economic Model Predictive Control (EMPC) for microgrid energy management. The approach trains a neural network on offline trajectories from an expert EMPC, enabling fast real-time decisions without solving online optimization problems. This addresses scalability issues and variable solution times that hinder real-time deployment. The system manages fuel generators, renewables, energy storage, and curtailable loads.

Key facts

  • Imitation learning approximates mixed-integer Economic Model Predictive Control (EMPC).
  • Neural network trained on offline expert trajectories.
  • Enables fast real-time decision making.
  • Avoids solving online mixed-integer optimization problems.
  • Addresses scalability and variable solution times.
  • Manages fuel generators, renewables, energy storage, and curtailable loads.
  • Published on arXiv with ID 2510.20040.
  • Announcement type: replace-cross.

Entities

Institutions

  • arXiv

Sources