Long Short-term Memory Time Series Modelling of Pressure Valves for Hydrogen-powered Vehicles and Infrastructure

Long-term reliability and accuracy of pressure valves are critical for hydrogen infrastructure and applications, particularly in hydrogen-powered vehicles exposed to extreme weather conditions like cold winters and hot summers. This study evaluates such valves using the Endurance Test specified in European Commission Regulation (EU) No 406/2010, fulfilling Regulation (EC) No 79/2009 requirements for hydrogen vehicle type approval. A long short-term memory (LSTM) network accelerates valve development and validation by simulating endurance tests. The LSTM model, with three inputs and one output, predicts valve outlet pressure responses using experimental data collected at 25 ◦C, 85 ◦C, and − 40 ◦C, simulating a 20-year lifecycle of 75,000 cycles. At 25 ◦C, the model achieves optimal performance with 40,000 training cycles and an R2 of 0.969, with R2 values exceeding 0.960 across all temperatures. This efficient, robust approach accelerates testing, enabling realtime diagnostics and advancing hydrogen technologies for a sustainable future.