Renewable Energy Orchestrator — AI skill for Claude Code

Renewable Energy Orchestrator

An agentic system that autonomously operates a renewable energy portfolio: 5 solar farms, 3 wind farms, 2 battery energy storage systems, 4 industrial consumers, one grid interconnection and electricity market access. Every 15 minutes it reads the situation, decides what "good" means right now, computes a robust dispatch across forecast scenarios, validates it through a deterministic safety layer, routes high-impact actions to a human when required, executes, recovers from refused commands inside the same interval, and explains itself. Three baseline operators run on identical conditions so the improvement is measured rather than asserted.

Built for the ET AI Hackathon, Agentic Edition (Problem 4, Utilities). The structural architecture and 9-blocker justification are in [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md); the rubric mapping is in [docs/RUBRIC_MAPPING.md](docs/RUBRIC_MAPPING.md); a Word version of the backend architecture is in [docs/Backend_Architecture.docx](docs/Backend_Architecture.docx).

Contents

1. What the system does
2. Requirements
3. Run it in five minutes
4. Enable a model provider (Claude, Gemini or Grok)
5. Run with Docker
6. Using the dashboard
7. Demo script for judges
8. Evaluation
9. Tests
10. Architecture in one page
11. Component guide: what each part does
12. API reference
13. Configuration reference
14. Repository layout
15. Troubleshooting

1. What the system does

The portfolio is a digital twin with real physics: battery state of charge, efficiency and degradation, curtailment, demand respon