Evolving robust policies for community energy system management

Abstract

Community energy systems (CESs) are shared energy systems in which multiple communities generate and consume energy from renewable resources. At regular time intervals, each participating community decides whether to self-supply, store, trade, or sell their energy to others in the scheme or back to the grid according to a predefined policy which all participants abide by. The objective of the policy is to maximise average satisfaction across the entire CES while minimising the number of unsatisfied participants. We propose a multi-class, multi-tree genetic programming approach to evolve a set of specialist policies that are applicable to specific conditions, relating to abundance of energy, asymmetry of generation, and system volatility. Results show that the evolved policies significantly outperform a default handcrafted policy. Additionally, we evolve a generalist policy and compare its performance to specialist ones, finding that the best generalist policy can equal the performance of specialists in many scenarios. We claim that our approach can be generalised to any multi-agent system solving a common-pool resource allocation problem that requires the design of a suitable operating policy.