Application of Hydrogen Leakage Consequence Prediction at Hydrogen Refueling Stations Based on Deep Learning

Hydrogen is a colorless and odorless gas. Hydrogen burns with a light blue flame that is barely visible during the day. Due to monitoring devices being sparse at hydrogen refueling­­­­ stations, hydrogen accidents are more challenging to detect in time. Emergency responders may struggle to observe the development of accidents and respond promptly, missing the critical period for accident response. Therefore, this paper first screened the maximum credible accident scenarios for hydrogen refueling stations and developed a 3D model of a hydrogen refueling station based on 3ds Max and Unity3D. Using virtual reality technology and the C# language, we have created a rapid construction system for the 3D visualization of maximum credible accident scenarios. The accident consequence scenario simulation system can adopt sparse monitoring data to reconstruct the concentration field, temperature field, thermal radiation field, and pressure field in the event of a hydrogen leakage and explosion accident scenarios. By transforming sparse point data into continuous field data, we can achieve dynamic visualization of accident scenarios. Emergency commanders can effectively monitor the development of the accident and accurately determine the possible consequences of the accident scenarios. The visualization of consequences can also be combined with VR development to achieve interaction between people and accident scenarios, providing the basis for disaster rehearsal, emergency simulation exercises, emergency decision-making, digital plan exercises, and 3D simulation skills training.