o:30448 Application of Reinforcement Learning for Control of Heat Pump Systems en Abstract— With the proliferation of heat pump systems for both heating and cooling applications for a wide range of space volumes, from isolated rooms to whole houses and buildings, their efficient operation is paramount to facilitate the transition to a more efficient building stock and reduction of greenhouse gas emissions. Also, phasing out polluting nonrenewable fossil fuel-based heating systems in favor of heat pumps contributes notably to the electrification of the thermal domain and allows for a more notable share to be facilitated by clean and renewable generation in the future. Therefore, on top of modeling approaches for these types of systems, adequate control algorithms need to be developed and deployed to ensure the proper utilization of flexibility that these devices offer. This paper presents a set of techniques based on reinforcement learning for heat pump control of room temperature based on varying source and user loop flow rates as control inputs and discusses the implications of a selection of different control strategies on the observed indoor temperature variables. 1552102 http://www.eventiotic.com/eventiotic/library/paper/688 2023-07-27T12:42:45.431Z 44 no 46 Dea Pujić Institut Mihajlo Pupin 0000-0002-3934-6346 Marko Jelić Institut Mihajlo Pupin person 0000-0002-0220-1688 Marko Batić Institut Mihajlo Pupin person 0000-0002-8443-3932 Nikola Tomašević Institut Mihajlo Pupin person 0000-0002-6620-479X application/pdf 613910 https://phaidrabg.bg.ac.rs/o:30448 no yes 1 70 92000004 11A20 12th International Conference on Information Society and Technology, ICIST 2022 Proceedings 1 4 2022