Multi-Robot Cooperative Inspection Planning of Substation Based on Genetic Algorithm and Deep Reinforcement Learning
IEEE Access · 2025 · 13 · 148949-148960
DOI: 10.1109/access.2025.3598345 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2025. Verfasst von Jie Liu und 4 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: China Southern Power Grid Science and Technology Project Research on Key Technologies for Visual Detection of Live State of Substation Equipment.
Abstract (Originalsprache)
The safe and reliable operation of power systems relies on the intelligent operation and maintenance (O&M) of substations. However, multi-robot collaborative inspection, an essential component of intelligent O&M, faces challenges such as uneven task distribution, inefficient path planning, and underutilized resources. This paper proposes a multi-robot collaborative inspection planning framework that fuses an improved genetic algorithm with deep reinforcement learning. The goal is to achieve a global optimization of substation inspection resources and high-efficiency coordination among the robot systems. First, a three-dimensional intelligent inspection information-sharing system based on the Internet of Things and edge computing is constructed, enabling data interoperability and state perception among heterogeneous inspection devices. Next, a mathematical model for task allocation is established under multiple constraints, such as energy limitations, time windows, and equipment specialization, with a weighted single-objective optimization function integrating energy consumption, time, and reliability. An improved genetic algorithm featuring an adaptive crossover operator and an elite retention strategy is designed to solve the global optimal task allocation scheme. Finally, based on the allocation results, a two-layer deep Q network collaborative path planning method is proposed. Through the design of a reward-penalty mechanism and prioritized experience replay, dynamic obstacle avoidance and cooperative navigation among the robots are achieved. Validation in an actual 500kV intelligent substation environment shows that, compared with traditional single-robot inspection and simple partition-based inspection, the proposed method improves the total inspection time, energy consumption, and equipment utilization by 32.7%, 28.4%, and 41.6%, respectively. The research provides both a theoretical foundation and technical support for the intelligent O&M of power systems and is of great significance for promoting intelligent collaboration among substation inspection robot teams.
Förderung
- China Southern Power Grid Science and Technology Project Research on Key Technologies for Visual Detection of Live State of Substation Equipment
Zitieren
APA
Liu, J., Wang, J., Peng, J., Hong, Y., & Shao, T. (2025). Multi-Robot Cooperative Inspection Planning of Substation Based on Genetic Algorithm and Deep Reinforcement Learning. IEEE Access, 13, 148949-148960. https://doi.org/10.1109/access.2025.3598345
Vancouver
Liu J, Wang J, Peng J, Hong Y, Shao T. Multi-Robot Cooperative Inspection Planning of Substation Based on Genetic Algorithm and Deep Reinforcement Learning. IEEE Access. 2025;13:148949-148960. doi:10.1109/access.2025.3598345
BibTeX
@article{Liu2025,
author = {Liu, Jie and Wang, Jingsheng and Peng, Jiajun and Hong, Yongxue and Shao, Tianci},
title = {Multi-Robot Cooperative Inspection Planning of Substation Based on Genetic Algorithm and Deep Reinforcement Learning},
journal = {IEEE Access},
year = {2025},
volume = {13},
pages = {148949--148960},
doi = {10.1109/access.2025.3598345}
}
EUVIMED