Development of LSTM-AutoEncoder-Based Health Monitoring Technology for HRS Compressors

Purpose: This study focuses on monitoring the functional safety and fault health of hydrogen fueling station (HRS) components, including coolers, compressors, and dispensers. Using the HRS input power quality monitoring (PQM) device, this study aims to predict anomalies in the operations of these systems.BR/Method: A fault detection and diagnosis model based on long short-term memory (LSTM)-autoencoder was developed. Power data for model training was collected through the GEMS monitoring system at Jingok Station. Among the condition-based maintenance (CBM) functional safety monitoring conditions, error measurement parameters such as sound, vibration, and current were analyzed to detect and diagnose errors. The LSTM-autoencoder processes this data to predict anomaly occurrences.BR/Results: By integrating the PQM device with the HRS power input, the CBM+ of hydrogen compressors was enhanced. Data collected over three months from the developed PQM measurement device installed in the Gwangju HRS input power source revealed that the LSTM-autoencoder model could predict HRS failures approximately seven days in advance.BR/Conclusion: The LSTM-autoencoder model, trained on power data from the GEMS monitoring system at Jinkok Station in Gwangju, demonstrated accurate failure detection and early warning BR/capabilities. It provides alerts 1–7 days before potential failures, enhancing the reliability and safety of HRS operations.