Research on Fault Diagnosis Model for Marine Transformer Digital Twin System Based on WOA-BiLSTM-TransFormer

As a core component of marine power systems, marine transformers are critical to ensuring safe ship navigation. Given the harsh and complex offshore environment that significantly increases transformer failure rates, developing a high-precision diagnostic model with interactive visualization capabilities is essential for enhancing maritime safety and energy efficiency. Therefore, this paper proposes a WOA-BiLSTM-Transformer-based fault diagnosis approach for addressing the limitations of low diagnostic accuracy and inadequate feature extraction in existing methods. Specifically, the Whale Optimization Algorithm (WOA) is used of optimizing the parameter settings for the Bidirectional Long Short-Term Memory (BiLSTM) network, and a Transformer decoder module with multi-head self-attention mechanisms is further used for extracting features to monitor winding faults. Performance analysis based on the fault data collected from CSB-5 marine transformers demonstrates that this method meets the requirements of digital twin diagnostic systems in terms of both accuracy and speed of diagnosis. The proposed model is further integrated into the Unity platform to construct a real-time condition monitoring and diagnostic visualization digital twin system for marine transformers, which can enhance ship safety and stability, improve operational and maintenance efficiency, and reduce repair costs.