Meta-RL with Shared Representations Enables Fast Adaptation in Energy Systems

TL;DR

This paper presents a novel Meta-Reinforcement Learning framework with shared representations and parameter sharing, enabling rapid adaptation and better generalization in complex, real-world energy management tasks.

Contribution

It introduces a bi-level optimization Meta-RL approach with shared feature extractors and parameter sharing, improving sample efficiency and transferability across tasks.

Findings

Enhanced task adaptation in energy systems.

Outperforms conventional RL and existing Meta-RL methods.

Effective knowledge transfer with shared representations.

Abstract

Meta-Reinforcement Learning addresses the critical limitations of conventional Reinforcement Learning in multi-task and non-stationary environments by enabling fast policy adaptation and improved generalization. We introduce a novel Meta-RL framework that integrates a bi-level optimization scheme with a hybrid actor-critic architecture specially designed to enhance sample efficiency and inter-task adaptability. To improve knowledge transfer, we meta-learn a shared state feature extractor jointly optimized across actor and critic networks, providing efficient representation learning and limiting overfitting to individual tasks or dominant profiles. Additionally, we propose a parameter-sharing mechanism between the outer- and inner-loop actor networks, to reduce redundant learning and accelerate adaptation during task revisitation. The approach is validated on a real-world Building Energy Management Systems dataset covering nearly a decade of temporal and structural variability, for which we propose a task preparation method to promote generalization. Experiments demonstrate effective task adaptation and better performance compared to conventional RL and Meta-RL methods.