Project Summary

The project's scope is tightly aligned with SIF's data and digitalisation challenge criteria:

• Automated pressure management software, and use of near real time data and machine-learning techniques, will contribute to better coordination, planning and network optimisation.
• Increased injection of biomethane (and, in the future, hydrogen) into networks will enable progress towards net zero and enable strategic outcomes from other challenges e.g., decarbonisation of heat.

The project directly addresses two points in the challenge scope:

• Point 7: this project will use novel sensor technology to improve visibility of network infrastructure condition, and make data-driven decisions about that infrastructure
• Point 9: this project will use data, combined with machine-learning (ML) and artificial intelligence (AI) techniques, to improve the forecasting abilities of both demand on networks, and required maintenance and interventions.

The principal innovation underscoring the project is use of data-driven techniques, based on ML & AI technology, acting in combination with remote pressure control and network extremity monitoring equipment deployed to networks in a distributed digitalised architecture. The innovation evolved from Utonomy's current solution, which uses manually-generated schedules, to offer a novel method for UK gas distribution network management and digitalisation.
From Alpha to Beta, the project evolved by taking a tiered approach to eight opportunity areas reflecting Discovery phase feasibility study outputs. Opportunities were progressed through targeted user research and collection of operational data, and, for highest ranked opportunities, Alpha phase activities extended to bench-testing proof-of-concept models designed to de-risk development and implementation of solutions at Beta phase. In addition, Alpha activities included a further evaluation of opportunities' benefits and definition of field trial stages.
The project's perception of the problem evolved from Alpha to Beta via a series of user interviews with representatives of key SGN stakeholders: Network Planning, Maintenance, Operational Technology, Policy, and Innovation. Key learnings were;

• Opportunities to reduce methane leakage and create holistic management dashboards were the most mature.
• Prediction of escapes was important to stakeholders but a definitive link between pressure and escapes could not be established from the data available in Alpha.
• Factors that lead to low pressure incidents showed the most potential from the second tier of opportunities.
• That increasing biomethane injection had a clear need within SGN.

In Beta, core users of innovations will come from: Maintenance teams, who are looking for faster resolution of network problems and fewer truck rolls to manually adjust governors; Network Planning teams who are looking for lower methane emissions; and Biomethane teams who are looking for greater injection rates.
Utonomy will be the main project partner at Beta, continuing its successful history of collaboration with SGN. The Utonomy engineering team has capabilities in electronics design for hazardous areas, data science and machine learning, industrial IoT and digital communications technologies, cyber security, and cloud-hosted software applications. Utonomy has collaborated successfully with SGN and Wales & West Utilities on the development and trial of its remote pressure control and management solution. Utonomy has developed and carried out initial field trials of Intelligent Control software via an Innovate UK funded project completed in March 2022 and is in the process of trialling a medium pressure variant of its pressure control equipment with Northern Gas Networks and Wales & West Utilities.
Utonomy will use Faculty Science Limited as lead subcontractor. Faculty has the unique capability to deliver state-of-the-art AI solutions from teams formed from over 200 professionals comprising both technical and commercial experts. In delivering AI solutions, in-house developed AI Engines allow specialised techniques to be applied to customer problems and to optimise performance.
As a world-class engineering consultancy with a range of multi-discipline technical specialists, DNV will leverage further expertise from their pool of Gas Industry experts to provide third-party verification and specialist support services. Acting as independent assurers, using their industry-leading recommended practices, they will ensure that any new solutions developed will provide real benefits in their application.
Other network partners will bring oversight and stakeholder governance to the project by contributing to regular steering discussions and stage gates, reviewing deliverables, and providing feedback on proposed solutions. Solutions could be trialled on other network partner's sites, and additional test data could also be provided. Having all four UK gas distribution networks in the project will also ensure that solutions are rolled-out as quickly as possible to benefit UK gas consumers.

Solutions will be primarily used by two sets of users: Network and Maintenance teams who have responsibility for managing pressure and carrying out maintenance, will use the solutions to adjust governor pressures remotely and automatically to minimise leakage and optimise biomethane feed-in; they will also use solutions to diagnose, and ultimately resolve, network faults; and network planning teams will use the solutions to track KPIs such as leakage reduction or biomethane injection, and take strategic network decisions based on the analysis provided by the solutions.

Innovation Justification

For leakage reduction, the most relevant current technology is profiling; these systems take the preceding days' average network pressure response and apply this historic view to the day ahead; this causes problems during sudden cold snaps as pressures are too low resulting in multiple alarms. This project's solution is innovative because it uses machine-learning technology to predict changes to network demand response ahead of time using external predictive factors including the next day's weather forecast. It also optimises governor set-point pressures individually and considers multiple different network extremities, which are features not available in the current technology.
For anomaly detection, current practise is manual and reactionary, for example escapes are mostly detected by gas being smelt by members of the public, low pressure incidents by public reporting boiler or other domestic faults, and governor faults after downstream knock-on effects are investigated by maintenance teams. This project's solution is innovative because maintenance teams will be alerted autonomously to anomalies by machine-learning algorithms that are constantly monitoring network data for small signs of change: the potential is that causes of faults could be further investigated and mitigated, or even rectified, before consumers become aware.
For biomethane injection, the current state of the art is for manual seasonal adjustments to be made at medium pressure supply governors to facilitate contractual minimum injection rates. Adjustments require site visits, which are costly and time-consuming, and the timing of changes is sensitive to unseasonal weather in shoulder months. An alternative solution involving compression is likely to be extremely expensive compared to this project's solution. A further alternative solution using remote pressure control technology from a governor equipment manufacturer has been trialled within a UK distribution network, but not been used commercially; this solution was shown to offer relatively simple pressure control but has limited ability to work for varying demands and lacks intelligence to set optimum pressures, as well as only integrating with pilot valves from the same supplier. In contrast, this project's solution would retro-fit to existing pilot valve equipment and offer complex network pressure control accounting for multiple entry points using machine-learning algorithms.
Across all three solution areas, the project goes beyond incremental innovation because the use of machine-learning algorithms, in conjunction with low-cost, scalable computing power that enables the algorithms to be 'always on', will represent a step-change in the data monitoring and optimisation that is possible when operating distribution networks.
For leakage reduction, current and estimated IRLs are 3 and 7, and for CRLs are 5 and 8 respectively.
For anomaly detection, current and estimated IRLs are 2 and 7, and for CRLs are 3 and 8 respectively.
For biomethane injection, current and estimated IRLs are 2 and 7, and for CRLs are 3 and 7 respectively.
The scale of the proposed project is appropriate as it balances the needs to deliver value to UK consumers with the need to safely and adequately develop and prove the solutions to the network stakeholder community. Two of the project's solutions are intended to deliver innovative pressure control applications; field trial demonstrations of these solutions will rightly require a level of scrutiny on safety and security of supply, which will be managed via the networks' existing policies and management procedures for introducing new technologies and products on live networks. Costs and time durations for these elements of field trials largely reflect the number of sites the trials will take place on and the sequential nature of the solutions' proof points to demonstrate safety and security of supply before efficacy. For the anomaly detection solution, demonstrable outcomes will depend on deploying the field trial over a sufficiently large proportion of the network that detectable anomalous events actually take place during the timescales.
If the project were funded under business-as-usual activities or within the price control, it would take significantly longer, and the solutions would likely arrive too late to enable the transition to net zero.
The project will not undermine the development of competitive markets because the solutions will rely on raw data (e.g., pressure measurements) gathered from the networks on which they will be installed and operate; however, ownership of all such raw network data will be retained by networks. This data could be made available by networks to potential new market entrants, which could be further used with knowledge made available by the project in how to operate networks with the solutions installed, to develop competitive solutions.
Counterfactual solutions investigated included pressure response models that did not consider seasonal effects such as temperature. From the Alpha bench-testing, these were proven to be outperformed by models that did include such features.

Bench-testing also indicated the need for solutions to vary pressure constantly during the day (to reflect intra-day demand changes), thereby discounting a counterfactual solution that used static or rarely changing intra-day pressures.

Benefits

The expected impacts and benefits from the project are set out below. Benefits will be delivered progressively from 2026 onwards as the solution is rolled out as business as usual.
1. Cost reductions in operating the networks:
a) Reduction in gas escapes:
The LP (low pressure) pressure control solution being developed will automatically adjust pressures in line with demand. This will enable networks to operate at lower average pressures, especially in the winter. Without this solution, networks are either fixed or seasonally adjusted which means that they are set at conservatively high pressures. The reduction in pressure especially during the winter leads to a reduction in gas escapes reducing the cost of repairs from 2024 to 2035 by £42m.
b) Elimination of seasonal adjustment of governors:
Most of the governors in the network today are manually adjusted. Many of these governors are adjusted seasonally i.e. 4 times per year. This is labour intensive, tying up valuable engineering resources for several weeks each year. Sometimes winter settings may need to be reinstated because of a late cold snap. The average cost of a seasonal adjustment is £280/governor/annum. If the governors were controlled automatically, costs would be reduced by £14m from 2024 to 2035.
c) Greater cost efficiency of mains replacement:
Mains replacement can either be carried out using an open cut method or an insertion method where a pipe of smaller diameter is inserted inside the mains to be replaced. The open cut method costs approximately £280/m and the insertion method costs approximately £140/m. Sometimes, mains replacement projects are unable to use the insertion method as the smaller diameter pipe would create too great a head-loss between the governors and consumers. Using pressure management will enable some of the open cut projects to be switched to insertion creating savings of £140/m. This can be done firstly by using live pressure data from the pressure management system rather than using network models which are sometimes inaccurate. Secondly, because pressure management enables the pressure to be increased just to meet the high peak demand and then brought back down again. There are projects where it is unacceptable to maintain a high pressure in the network for long periods but where it would be acceptable to increase the pressure for a short period to cover these high peak demands. A 2%switch from open cut to insertion would save £59m from 2024 to 2032.