Title: Deep-Q-network-based energy management of multi-resources in limited power micro-grid

Authors: Nabil Jalil Aklo; Mofeed Turky Rashid

Addresses: Electrical Engineering Department, University of Basrah, Basrah, Iraq; Electrical and Electronic Engineering Department, University of Thi-Qar, Thi Qar, Iraq ' Electrical Engineering Department, University of Basrah, Basrah, Iraq

Abstract: To overcome the shortage of power supply to the rural area, a hybrid connected mode micro-grid (MG) is proposed. It is suggested to include a diesel generator (DG) and renewable energy resources (RER) with a limited power of utility grid. To ensure the availability of fuel supply, the take-or-pay method is employed. In this paper, a smart energy management system (EMS) has been proposed to control the operation of hybrid MG, in addition to ensuring complete fuel disbursement under the scheduling of fuel supply. To facilitate the construction of EMS, a free model-based reinforcement learning (RL) algorithm has been employed for this purpose, in which the design of this algorithm depends on deep Q-network (DQN). The simulation of the algorithm has been achieved by MATLAB to validate the proposed system; the results showed a good performance of the technique compared with the performance achieved by improved particle swarm optimisation (IPSO) algorithm.

Keywords: energy management system; renewable energy resources; reinforcement learning; deep Q-learning; take-or-pay.

DOI: 10.1504/IJPT.2023.129667

International Journal of Powertrains, 2023 Vol.12 No.1, pp.25 - 53

Received: 17 Dec 2021
Accepted: 26 Mar 2022

Published online: 20 Mar 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article