Multi-Modal Integrated Monitoring Algorithm for Converter Valves Based on Variational Mode Decomposition and Bidirectional Long Short-Term Memory Network

This paper presents a novel multi-modal integrated monitoring algorithm for converter valves, employing Variational Mode Decomposition (VMD) and Bidirectional Long Short-Term Memory (BiLSTM) networks to enhance the accuracy of fault detection in high-voltage direct current (HVDC) systems. By leveraging multi-source data, including electrical and vibration signals, we effectively decompose the signals into intrinsic mode functions, facilitating the extraction of critical features. The BiLSTM model demonstrates superior performance compared to traditional methods,achieving lower MSE and MAE metrics. Our findings indicate that the proposed algorithm significantly improves the monitoring and diagnosis of converter valve states, contributing to the overall reliability of electrical networks.