IEMS Virtual Microgrid 🔋

A Deep Reinforcement Learning agent that runs a factory's battery to cut its electricity bill — validated on held-out 2016 data against a perfect-foresight optimum.

Cost saving

32.9%

Gap to optimum

8.0%

Grid peak

preserved

Validation

5 seeds

▶ Improved agent

The headline result. Fine-grained control, day-ahead price look-ahead, Double + Dueling DQN. Full interactive day-by-day animation, LP benchmark, and a complete system explainer.

RECOMMENDED

Thesis-faithful agent

The baseline reproduction (3 actions, plain DQN, no look-ahead) used as the comparison point. Same dashboard, same animation, so you can see the difference the improvements make.

BASELINE

Each page is a self-contained interactive dashboard — press ▶ Play to watch the agent run the battery through any test day. Built with PyTorch + Plotly.