Deep reinforcement learning based dynamic pricing for demand response considering market and supply constraints

Fraija, A., Henao, N., Agbossou, K., Kelouwani, S., Fournier, M. et Nagarsheth, S. H. (2024). Deep reinforcement learning based dynamic pricing for demand response considering market and supply constraints. Smart Energy, 14 . Article 100139. ISSN 2666-9552 DOI 10.1016/j.segy.2024.100139

Prévisualisation

PDF Disponible sous licence Creative Commons Attribution Non-commercial No Derivatives. Télécharger (2MB) | Prévisualisation

Résumé

Abstract

This paper presents a Reinforcement Learning (RL) approach to a price-based Demand Response (DR) program. The proposed framework manages a dynamic pricing scheme considering constraints from the supply and market side. Under these constraints, a DR Aggregator (DRA) is designed that takes advantage of a price generator function to establish a desirable power capacity through a coordination loop. Subsequently, a multi-agent system is suggested to exploit the flexibility potential of the residential sector to modify consumption patterns utilizing the relevant price policy. Specifically, electrical space heaters as flexible loads are employed to cope with the created policy by reducing energy costs while maintaining customers' comfort preferences. In addition, the developed mechanism is capable of dealing with deviations from the optimal consumption plan determined by residential agents at the beginning of the day. The DRA applies an RL method to handle such occurrences while maximizing its profits by adjusting the parameters of the price generator function at each iteration. A comparative study is also carried out for the proposed price-based DR and the RL-based DRA. The results demonstrate the efficiency of the suggested DR program to offer a power capacity that can maximize the profit of the aggregator and meet the needs of residential agents while preserving the constraints of the system.

Type de document:	Article
Mots-clés libres:	Demand response Demand response aggregator Dynamic pricing Market constraints Capacity limitation Reinforcement learning
Date de dépôt:	22 janv. 2026 16:02
Dernière modification:	22 janv. 2026 16:02
Version du document déposé:	Version officielle de l'éditeur
URI:	https://depot-e.uqtr.ca/id/eprint/12552

Actions (administrateurs uniquement)

Éditer la notice