Coalition Formation with Limited Information Sharing for Local Energy Management
Researchers solve privacy-compute tradeoff for smart grids using novel coalition formation method.
A team from Imperial College London led by Luke Rickard, Paola Falugi, and Eric C. Kerrigan has developed a breakthrough algorithm for coordinating energy exchange in distributed systems with prosumers. Their paper, "Coalition Formation with Limited Information Sharing for Local Energy Management," addresses a critical challenge in smart grids: how to form cooperative groups (coalitions) among energy producers and consumers without requiring them to share sensitive operational data. Traditional bottom-up coalition formation methods typically demand full information exchange, creating privacy concerns and imposing significant computational burdens that scale poorly with system size.
The researchers' key innovation is a limited information algorithm that constructs an upper bound on the value of potential coalitions. This approach eliminates the need to solve optimization problems for every possible merge, dramatically reducing computational complexity. The method is embedded within a model predictive control scheme and optimized using the Alternating Direction Method of Multipliers (ADMM), which further limits information sharing within formed coalitions. Crucially, the team mathematically proves their approach guarantees costs no greater than completely decentralized operation, ensuring no participant is worse off.
When evaluated on real-world energy data, the algorithm demonstrated significantly improved economic performance compared to decentralized control approaches, while maintaining computational costs substantially lower than full-information methods. This represents a practical solution to the privacy-compute tradeoff that has hindered widespread adoption of cooperative energy management systems. The work, submitted to CDC 2026, provides a scalable framework that could accelerate the transition to more efficient, resilient smart grids where households and businesses can safely collaborate to optimize energy usage and reduce costs.
- Algorithm reduces information sharing to aggregate data only, solving privacy concerns in energy coordination
- Method guarantees costs no greater than decentralized operation through mathematical proof of performance bounds
- Real-world testing shows improved economic performance with substantially lower computational costs than full-information approaches
Why It Matters
Enables practical, privacy-preserving smart grids where households can collaborate on energy management without exposing sensitive data.