Effective Fault Detection and Diagnosis for Power Converters in Wind Turbine Systems Using KPCA-Based BiLSTM

The current work presents an effective fault detection and diagnosis (FDD) technique in wind energy converter (WEC) systems. proposed FDD framework merges the benefits of kernel principal component analysis (KPCA) model bidirectional long short-term memory (BiLSTM) classifier. In developed approach, KPCA is applied to extract select most features, while BiLSTM utilized for classification purposes. KPCA-based approach involves two main steps: feature extraction selection, classification. order efficient features final are fed distinguish between different working modes. Different simulation scenarios considered this study show robustness performance when compared conventional methods. To evaluate effectiveness we utilize data obtained from a healthy WTC, which then injected with several scenarios: simple generator-side, grid-side, multiple mixed on both sides. analyzed terms accuracy, recall, precision, computation time. Furthermore, efficiency shown by accuracy parameter. experimental results classical techniques (an 97.30%).