Ai-based centralized real-time monitoring and diagnostic framework for lightning risk assessment in large-scale wind power plants

Abstract

Relevance: Lightning-induced failures significantly reduce operational reliability and safety in large-scale wind power plants worldwide.

Objective: This study develops a centralized AI-based monitoring and diagnostic framework for real-time lightning current analysis, multi-source data fusion, anomaly detection, and short-term forecasting in wind turbine systems operating under complex electromagnetic and meteorological conditions.

Methods: Multi-source sensor fusion, Bayesian inference, XGBoost prediction, adaptive normalization, and real-time transient signal classification were implemented.

Results: The proposed framework reliably distinguished lightning discharges from electromagnetic noise, achieving stable real-time diagnostics and predictive warning performance under experimental operating conditions.