Reinforcement learning (RL) is a branch of machine learning (ML) that focuses on optimal decision making using a trial-and-error approach— analogous to a child exploring and learning in their surrounding world. It is an exciting time for RL within the energy sector, but it is still early days and much of the content exists in highly technical academic literature or is seldom shared by industry.

This report bridges the gap between industry and academia by presenting the most salient advantages and challenges of RL, providing a framework to guide the reader on how to scope an RL approach to a given problem, and highlighting the landscape of near-term and future applications through a series of industry and academic use cases.

The report examines several use cases of RL in industry and explores the lessons we can learn from its application. Examples such as Google’s DeepMind, Carbon Re, and E.ON, are placed under the microscope as we explore how RL is applied, challenges and opportunities of RL’s applications, and alternative approaches to its deployment.

To maximise the potential for RL in the energy sector, we propose two calls to action:

Industry: As much as possible, share challenges and successes of real-world implementations of RL to inform future applications. Consider sharing datasets that can be used to develop RL solutions.

Academia: Conduct fair comparisons between RL and alternatives to better understand where RL might excel in the future and understand how they can potentially complement each other.