Scalable demand response of decentralised assets: A perspective on reinforcement learning as a control paradigm
B3 L5 R5209
In the last decade reinforcement learning has demonstrated tremendous progress in terms of being a model-free control paradigm for decision making in complex systems with uncertainty and partial observability, thus making it a candidate technology for demand response with a promise of true scalability.
Overview
Abstract
In the last decade reinforcement learning has demonstrated tremendous progress in terms of being a model-free control paradigm for decision making in complex systems with uncertainty and partial observability, thus making it a candidate technology for demand response with a promise of true scalability. Getting reinforcement learning to work in an industrial reality, however, is a path requiring exploration, exploitation, and many epochs. In this presentation, prof. dr. Claessens will present some typical decision making problems in the context of virtual power plants and demand response and how reinforcement learning concepts can be and are applied, from an industrial and academic perspective.
Brief Biography
Prof. dr. ir. Bert Claessens has been active in the field of demand response for nearly 14 years, as a practitioner (currently in the role of CPO at Beebop) and as an (applied) academic (part time professor at Ghent University).
Working for companies such as REstore, Centrica and now Beebop he has:
- Pioneered and developed In-front-of-the-meter battery projects optimized with a multi-market trading stack in Europe, currently over 100 MW worth of battery projects are optimized using the concepts developed by his teams.
- Pioneered and incubated residential demand response in different European countries, bringing new and innovative concepts to the energy system, currently over 10k residential assets are optimized with these concepts bringing true value to prosumers.