Whole system asset modelling

Challenge

Unaccounted-for-Gas (UAG) arises predominantly from metering bias rather than physical gas loss and represents a significant operational and financial burden on National Gas Transmission (NGT). These errors when discovered can often lead to reconciliation processes that result in billing adjustments which can sometimes scale to millions of pounds. National Gas Transmissions are explicitly mandated to investigate and minimise UAG through metering accuracy and system transparency.

The challenge for NGT is the vast, distributed nature of their system with over 500 above-ground installations including terminals, compressor stations and multi-junctions, together with tens of thousands of additional meters. The question becomes where to look?

Despite several historical attempts by NGT to reduce UAG including meter validation programmes, baseline tolerance models, anomaly detection and predictive algorithms, UAG still represents approximately 1.67 GWh in 2024/2025. NGT turned to CFMS to help solve this persistent challenge.

Solution

Working in partnership with NGT subject matter experts, we proposed that using historic data alone would be insufficient to understand the UAG problem and that some amount of predictive modelling would be required to understand sources of uncertainty and errors in the system. Therefore, using 18-months of historic SCADA data, a forward-looking predictive mathematical model was developed at a high enough fidelity to rapidly predict the expected sensor readings. These were then compared to measured values and discrepancies identified.

The model was design to take in the historic telemetry data, such as pressure and volumetric flow rates, and apply time varying boundary conditions at NTS entry and exit sites. The model then simulates the expected network dynamics from any given sections including the pipe lengths, diameters and assets to simulate the gas dynamics in an ideal condition. This was developed and implemented using PETSc, enabling arbitrary scaling meaning the solution can be readily deployed across the entire NTS estate.

Result

To enable the NGT asset engineers to fully visualise the results, the model output was further integrated in a network performance digital twin. This allows visualisation of the identified assets including static, dynamic or live data to be shown in comparison to the model’s simulated data. During validation, the predictive model was shown to have an accuracy rate in excess of 90% against test cases.

Further development of the model has been identified which could extend the support for NGT in the earlier identification of deteriorating assets which are approaching their tolerance levels. This would allow National Gas asset engineers to complete more quantitative targeted investigations and interventions to further reduce operation costs, UAG and billing errors in the future.