Pregled bibliografske jedinice broj: 1118044
Controlling the power distribution network with a deep reinforcement learning-based agent
Controlling the power distribution network with a deep reinforcement learning-based agent // KOI 2020 Book of Abstracts
Šibenik, Hrvatska, 2020. str. 52-53 (predavanje, nije recenziran, sažetak, znanstveni)
CROSBI ID: 1118044 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Controlling the power distribution network with a deep reinforcement learning-based agent
Autori
Brčić, Mario ; Jurić, Mislav ; Damjanović, Ivana ; Sudić, Ivan
Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni
Izvornik
KOI 2020 Book of Abstracts
/ - , 2020, 52-53
Skup
18th International Conference on Operational Research (KOI 2020)
Mjesto i datum
Šibenik, Hrvatska, 23.09.2020. - 25.09.2020
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Nije recenziran
Ključne riječi
deep reinforcement learning ; imitation learning ; distribution network
Sažetak
Controlling power distribution networks is becoming increasingly complex due to globalization, growing network size, the addition of new sources of uncertainty, rising consumption, and concurrent limitations on capacity expansion. Similar considerations can be applied to power distribution networks, supply chains, and information networks. This paper specifies and implements a deep learning-based agent to operate a simple power distribution network in a secure manner that prevents power outages. The agent is first trained by imitation learning from a heuristic agent and it is then further improved by reinforcement learning using proximal policy optimization.
Izvorni jezik
Engleski
Znanstvena područja
Matematika, Računarstvo, Interdisciplinarne tehničke znanosti
POVEZANOST RADA
Ustanove:
Fakultet elektrotehnike i računarstva, Zagreb
