This project explores the potential use of Artificial Intelligence to inform and improve National Transmission System (NTS) network planning decisions, namely, to optimise the configuration of network assets to ensure sufficient pressure on the NTS. It aims to develop and deploy an AI-based software tool that could allow Gas Network Analysts at NESO to rapidly assess ‘Static’ day-to-day NTS scenarios, by bulk-solving more standard cases, allowing the prioritisation of less standard instances, and helping with the rapid assessment of different network configurations for more complex scenarios. If successful the developed software tools will also enhance finding start points for solving scenarios, check set limit violations alongside provide recommendations for removing pipes from the NTS.
Benefits
If successful the developed software tool will enhance decision-making around the NTS alongside saving Gas Network Analysts time to focus on understanding other components of their analyses, improving their productivity. It will significantly reduce time spent on analysis capability, allowing analysis to focus on other areas of their respective projects. Solving base scenarios and allowing the analyst to have a start point reduces the frequency of repetitive tasks. The recommended pipeline recommender would allow the analyst to determine which pipelines to isolate based on previous data.
Learnings
Outcomes
The FastPress Alpha+ project successfully demonstrated the feasibility of automating key elements of Gas Network Planning analysis, meeting or exceeding all core PEA objectives. The project advanced the Base Scenario Solver (BSS) and Limit Violation Checker (LVC) to near‑production level, developed the Capability Limit Finder (CLF) and Resilience Testing (RT) to proof‑of‑concept, and delivered a Local Component Demonstrator (LCD) enabling analysts to interact directly with solver and validation features. These outcomes provided strong evidence that automated scenario solving can materially improve analysis speed, consistency, and scalability across RIIO zones and demand days.
The project also removed the key technical risk for cloud deployment by successfully running SIMONE in an Azure Windows Container, establishing a viable architecture for future scaling and multi‑analyst use. Extensive user‑experience work and operational‑assumption validation ensured that the algorithmic outputs remained aligned with real analyst workflows and system rules. Collectively, these achievements demonstrate clear potential for automation to enhance NESO’s capability to deliver robust, repeatable network assessments and to reduce manual workload.
Lessons Learnt
- Early and continuous user engagement is essential. Iterative collaboration with analysts surfaced hidden requirements, corrected assumptions, and ensured the tool aligned with real workflows.
- Solver optimisation requires more time than expected. Achieving reliable performance demanded significant tuning of hyperparameters, solving strategies, and compressor configurations.
- Cloud deployment with legacy systems brings complexity. Running SIMONE in Azure Windows Containers required deeper engineering and risk‑mitigation work than originally planned.
- Clear definition of operational assumptions is critical. Validating pressure cover logic, supply/demand thresholds, and limit definitions early prevented mis‑alignment later.
- User‑centred design is a core, not auxiliary, workstream. Multiple UX iterations were required to deliver an intuitive interface which could be well-used by an analysts.
- Hands-on demonstration of tools increases the usability. Having analysts validate the Local Component Demonstrator enabled practical feedback to help improve the tool.
