Optimal prosumer-based district heating and cooling using reinforcement learning agents. 01/11/2021 - 31/10/2023

Abstract

District Heating and Cooling (DHC) is a promising technology to shift to a sustainable energy supply, offering flexibility to the electric grid. Thermal storage can provide the necessary flexibility to balance production an demand for both electrical and thermal renewable energy sources (RES). Especially, the integration of decentralised thermal prosumers (e.g. boosters, thermal solar panel) in DHC have great potentials to improve the overall efficiency. Therefore, future DHC will need advanced control strategies facilitating the operation of prosumers-based DHC and providing flexibility to RES-dominated electric grids. Hereby, two main questions arise: (i) how should the temperature be controlled to improve the energetic, ecologic and economic performance of a DHC? And (ii) how to take into account the requirements of every direct stakeholder in the DHC? By simulating the DHC's behaviour, considering hydronics and prosumer behaviour, I will research the potential of a data-based control strategy, including multi-agent reinforcement learning (MARL). Every agent (per consumer, heat storage, etc.) pursues the local as well as the global objectives. The RL-agents are capable of self-learning a control strategy based on feedback (rewards). Besides valorisation throughout implementing such controls, the feedback and/or reward itself can be subject of follow-up research with respect to policy support.

Researcher(s)

Research team(s)

Project type(s)

  • Research Project