A first-of-its-kind reinforcement learning system built to design national energy strategies - starting with Ireland's path to net-zero. Fully interpretable, fully deterministic, and built to explain its reasoning.
GridAI is a custom reinforcement learning system developed entirely from scratch in Rust. It simulates over 250,000 full-scale national energy transitions to discover the most efficient, cost-effective, and publicly acceptable routes to a net-zero grid.
This isn't a black-box AI. GridAI is fully interpretable and built to explain its reasoning. It doesn't just give you the "what" - it tells you why.
After simulating over 2.5 million years of policy outcomes in less than an hour, GridAI delivered:
Achieves net-negative emissions while maintaining 100% power reliability across the national grid.
Total cost of €76B vs €120B in legacy reports - saving tens of billions through optimised planning.
Detailed breakdown including 41,000 acres of solar, 23,000 wind turbines, 18TWh battery storage, and 42TWh wave energy.
Achieves model-aligned public support score over 90% through multi-objective optimisation.
13,000+ lines of Rust code. Custom-built policy gradient reinforcement learning. Real-world population, generation, and cost data. Multi-objective optimisation across cost, emissions, reliability, and public opinion. Runs on consumer hardware - no cloud, no black boxes.
GridAI placed in the top 5 out of 170+ entries in EirGrid's CleanerGrid competition and was presented to national stakeholders including EirGrid's Head of Innovation. It has since gained traction with researchers, journalists, and educators.