An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids

Mohammed Alsaba; Mohammad Abido; Md Shafiullah

doi:10.1109/SmartGridComm65349.2025.11204593

An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids

Mohammed Alsaba^*
, Mohammad Abido
, Md Shafiullah

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Training deep reinforcement learning (DRL) agents for fault detection in microgrids requires significant computational resources and time, particularly for dense neural networks. This paper proposes an efficient framework to accelerate DRL model training using clustering and feature extraction techniques. K-medoids clustering with dynamic time warping (DTW) reduces the dataset size while retaining critical data representations. Feature extraction combines ResNet-18 and gated recurrent units (GRU) to capture key features. The combination of noise immunity and data drift detection improves fault detection reliability. evaluated using a temporal convolution attention-based neural network (TCAN) with advantage actor-critic (A2C) and proximal policy optimization (PPO) models. The framework achieves 99.2% fault detection accuracy (a 16.6% improvement over overcurrent relays) and reduces training time by 40.17%.

Original language	English
Title of host publication	2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9798331520847
DOIs	https://doi.org/10.1109/SmartGridComm65349.2025.11204593
State	Published - 2025
Externally published	Yes
Event	2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - North York, Canada Duration: 29 Sep 2025 → 2 Oct 2025

Publication series

Name	2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings

Conference

Conference	2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025
Country/Territory	Canada
City	North York
Period	29/09/25 → 2/10/25

Bibliographical note

Publisher Copyright:
© 2025 IEEE.

Keywords

AC microgrid
deep reinforcement learning
machine learning
power system protection

ASJC Scopus subject areas

Artificial Intelligence
Computer Networks and Communications
Computer Science Applications
Energy Engineering and Power Technology
Safety, Risk, Reliability and Quality
Control and Optimization

Access to Document

10.1109/SmartGridComm65349.2025.11204593

Cite this

Alsaba, M., Abido, M., & Shafiullah, M. (2025). An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids. In 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings (2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SmartGridComm65349.2025.11204593

Alsaba, Mohammed ; Abido, Mohammad ; Shafiullah, Md. / An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids. 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2025. (2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings).

@inproceedings{6a11cbff4714480db7da12ce7b08ccc3,

title = "An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids",

abstract = "Training deep reinforcement learning (DRL) agents for fault detection in microgrids requires significant computational resources and time, particularly for dense neural networks. This paper proposes an efficient framework to accelerate DRL model training using clustering and feature extraction techniques. K-medoids clustering with dynamic time warping (DTW) reduces the dataset size while retaining critical data representations. Feature extraction combines ResNet-18 and gated recurrent units (GRU) to capture key features. The combination of noise immunity and data drift detection improves fault detection reliability. evaluated using a temporal convolution attention-based neural network (TCAN) with advantage actor-critic (A2C) and proximal policy optimization (PPO) models. The framework achieves 99.2\% fault detection accuracy (a 16.6\% improvement over overcurrent relays) and reduces training time by 40.17\%.",

keywords = "AC microgrid, deep reinforcement learning, machine learning, power system protection",

author = "Mohammed Alsaba and Mohammad Abido and Md Shafiullah",

note = "Publisher Copyright: {\textcopyright} 2025 IEEE.; 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 ; Conference date: 29-09-2025 Through 02-10-2025",

year = "2025",

doi = "10.1109/SmartGridComm65349.2025.11204593",

language = "English",

series = "2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings",

address = "United States",

}

Alsaba, M, Abido, M & Shafiullah, M 2025, An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids. in 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings. 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings, Institute of Electrical and Electronics Engineers Inc., 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025, North York, Canada, 29/09/25. https://doi.org/10.1109/SmartGridComm65349.2025.11204593

An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids. / Alsaba, Mohammed; Abido, Mohammad ; Shafiullah, Md.
2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2025. (2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids

AU - Alsaba, Mohammed

AU - Abido, Mohammad

AU - Shafiullah, Md

PY - 2025

Y1 - 2025

N2 - Training deep reinforcement learning (DRL) agents for fault detection in microgrids requires significant computational resources and time, particularly for dense neural networks. This paper proposes an efficient framework to accelerate DRL model training using clustering and feature extraction techniques. K-medoids clustering with dynamic time warping (DTW) reduces the dataset size while retaining critical data representations. Feature extraction combines ResNet-18 and gated recurrent units (GRU) to capture key features. The combination of noise immunity and data drift detection improves fault detection reliability. evaluated using a temporal convolution attention-based neural network (TCAN) with advantage actor-critic (A2C) and proximal policy optimization (PPO) models. The framework achieves 99.2% fault detection accuracy (a 16.6% improvement over overcurrent relays) and reduces training time by 40.17%.

AB - Training deep reinforcement learning (DRL) agents for fault detection in microgrids requires significant computational resources and time, particularly for dense neural networks. This paper proposes an efficient framework to accelerate DRL model training using clustering and feature extraction techniques. K-medoids clustering with dynamic time warping (DTW) reduces the dataset size while retaining critical data representations. Feature extraction combines ResNet-18 and gated recurrent units (GRU) to capture key features. The combination of noise immunity and data drift detection improves fault detection reliability. evaluated using a temporal convolution attention-based neural network (TCAN) with advantage actor-critic (A2C) and proximal policy optimization (PPO) models. The framework achieves 99.2% fault detection accuracy (a 16.6% improvement over overcurrent relays) and reduces training time by 40.17%.

KW - AC microgrid

KW - deep reinforcement learning

KW - machine learning

KW - power system protection

UR - https://www.scopus.com/pages/publications/105022135228

U2 - 10.1109/SmartGridComm65349.2025.11204593

DO - 10.1109/SmartGridComm65349.2025.11204593

M3 - Conference contribution

AN - SCOPUS:105022135228

T3 - 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings

BT - 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025

Y2 - 29 September 2025 through 2 October 2025

ER -

Alsaba M, Abido M , Shafiullah M. An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids. In 2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2025. (2025 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2025 - Proceedings). doi: 10.1109/SmartGridComm65349.2025.11204593

An Efficient Framework for Training Deep Reinforcement Learning (DRL) Agents for Fault Detection in Microgrids

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this