Work

Intelligent Fault Diagnosis for Power Plant Arrester based on Reinforcement Learning and Condition Monitoring Data

Year: 2025
Type: article
Abstract: This paper proposes an intelligent fault diagnosis method for High-Voltage Arresters utilizing Live Online Monitoring data, addressing the complexity and latency of traditional diagnosis. The core of the method is a reinforcement learning model: it employs a Deep Q-Network (DQN) for state feature extraction and integrates the SARSA algorithm for real-time identification of arrester fault states. Furthermore, Monte Carlo Tree Search (MCTS) is introduced to enhance the attribution analysis of fault causes, thereby improving the depth and reliability of the diagnosis. Experimental results demonstrate that this approach achieves high-accuracy diagnosis (96.5%) across various fault types and exhibits fast response capability (0.45 s), providing an effective path for the intelligent analysis of abnormal data, such as Leakage Current.
Language: English
Open Access status: bronze
OpenAlex ID: W7125431908
FWCI:
Cites: 21
Cited by:
Related to: 10