Green Hydrogen – Safety Assurance Framework across Project Lifecycle

This paper proposes an innovative approach to improving hydrogen safety knowledge accessibility by integrating GPT-4 with public hydrogen safety and incident databases. The current state of safety databases is plagued by poor formatting, duplicate entries, and low searchability, which hinder effective dissemination of safety information. Leveraging the capabilities of GPT-4, we trained the generative AI on public and private safety databases and incident reports, significantly reducing the time required to identify similar past incidents and retrieve crucial information.

In testing with LIFTE H2's own digital platform APM, we observed the benefits of integrating safety knowledge into daily operational workflows, as real-time telemetry and events aid in troubleshooting and generating comprehensive safety reports. The Hydrogen Academy, with its conversational-style database search interface, showed initial success in helping trainees absorb vital information and deduce root causes of safety issues. Furthermore, generative AI-assisted drafting of executive summaries for safety reports saves considerable time and effort for reporters, while presenting digestible summaries for a wider audience.

The paper explores the potential of GPT-4 integration in hydrogen safety training and operations, presenting a user-friendly, conversation-style interface that not only makes safety knowledge more interactive and discoverable but also promotes efficiency and effectiveness in safety management and training.