针对风速波动的风力发电系统变流器开路故障鲁棒诊断方法

This paper proposes a robust fault diagnosis framework for open-circuit faults (OCF) in wind power converters, focusing on addressing the impact of two major challenges on fault diagnosis: wind speed fluctuations (induced by tower shadow effects, wind shear, and complex environmental conditions) and noise interference. The key contributions of this paper include: 1) A real-time current amplitude normalization (RCAN) mechanism using Park transformation for dynamic signal correction, 2) A wind-adaptive sampling (WAS) strategy for synchronized data acquisition, and 3) A novel hybrid deep learning architecture combining improved Gated Recurrent Neural Network (IGRNN) for temporal analysis and Markov Transfer Field Residual Network (MTF-ResNet) for spatial feature extraction. The method processes RCAN normalized three-phase current through WAS sampling, followed by parallel temporal-spatial feature extraction via IGRNN and MTF-ResNet branches. An attention mechanism integrates these features for final fault classification. Experimental results demonstrate significant performance improvements, achieving over 97% accuracy at 10dB signal-to-noise ratio (SNR) under fluctuating wind conditions, outperforming conventional approaches in reliability and robustness.