Reinforcement learning for reactive power control

Vlachogiannis, JG; Hatziargyriou, ND

dc.contributor.author	Vlachogiannis, JG	en
dc.contributor.author	Hatziargyriou, ND	en
dc.date.accessioned	2014-03-01T01:21:18Z
dc.date.available	2014-03-01T01:21:18Z
dc.date.issued	2004	en
dc.identifier.issn	0885-8950	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/16189
dc.subject	constrained load flow	en
dc.subject	Q-learning algorithm	en
dc.subject	reinforcement learning	en
dc.subject.classification	Engineering, Electrical & Electronic	en
dc.subject.other	Approximation theory	en
dc.subject.other	Busbars	en
dc.subject.other	Closed loop control systems	en
dc.subject.other	Control system analysis	en
dc.subject.other	Convergence of numerical methods	en
dc.subject.other	Dynamic programming	en
dc.subject.other	Iterative methods	en
dc.subject.other	Learning algorithms	en
dc.subject.other	Learning systems	en
dc.subject.other	Nonlinear control systems	en
dc.subject.other	Power control	en
dc.subject.other	Reactive power	en
dc.subject.other	Constrained load flow	en
dc.subject.other	Multistage decision problem	en
dc.subject.other	Offline control settings	en
dc.subject.other	Q learning algorithm	en
dc.subject.other	Reinforcement learning	en
dc.subject.other	Electric load flow	en
dc.title	Reinforcement learning for reactive power control	en
heal.type	journalArticle	en
heal.identifier.primary	10.1109/TPWRS.2004.831259	en
heal.identifier.secondary	http://dx.doi.org/10.1109/TPWRS.2004.831259	en
heal.language	English	en
heal.publicationDate	2004	en
heal.abstract	This paper presents a Reinforcement Learning (RL) method for network constrained setting of control variables. The RL method formulates the constrained load flow problem as a multistage decision problem. More specifically, the model-free learning algorithm (Q-learning) learns by experience how to adjust a closed-loop control rule mapping states (load flow solutions) to control actions (offline control settings) by means of reward values. Rewards are chosen to express how well control actions cause satisfaction of operating constraints. The Q-learning algorithm is applied to the IEEE 14 busbar and to the IEEE 136 busbar system for constrained reactive power control. The results are compared with those given by the probabilistic constrained load flow based on sensitivity analysis demonstrating the advantages and flexibility of the Q-learning algorithm. Computing times with another heuristic method is also compared. © 2004 IEEE.	en
heal.publisher	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC	en
heal.journalName	IEEE Transactions on Power Systems	en
dc.identifier.doi	10.1109/TPWRS.2004.831259	en
dc.identifier.isi	ISI:000222975800009	en
dc.identifier.volume	19	en
dc.identifier.issue	3	en
dc.identifier.spage	1317	en
dc.identifier.epage	1325	en