Currently there are a significant number of assets which are not able to participate in Frequency Response services due to challenges in clearly demonstrating frequency response delivery separate from the delivery of other services. This means there is less competition in the markets and as a result the cost paid for response services is higher than it could be if these assets were able to participate.
This project will investigate analysis techniques and develop an algorithm to validate Response delivery from a large number of these assets which are unable to use conventional metering solutions.
This should enable service providers to participate in Dynamic Response markets with assets using forms of data processing to separate out dynamic response service delivery from other energy recorded by the meter.
Benefits
- Significant increase in assets able to participate in services (initially estimated at 443MW, rising to 672-1126MW in 2025), greatly increasing competition and liquidity ultimately resulting in lower Response procurement costs (currently approximately £20m/month)
- Reduced barriers to entry supporting overall frequency market ambitions
- Provide global industry leading solution to use data processing innovation which could later be adapted to unlock capacity in other markets
- Reduced requirement and dependency on constrained new connections with benefits to ESO and wider industry
Learnings
Outcomes
This project delivered a synthetic data generation engine tool (through the use of Python coding) to flag susceptible unit behaviour to NESO. Specific features of this include:
- A data collation module for fetching and processing necessary data
- Algorithms for running gaming checks
- A metric that aggregates scores across all checks into a single value, indicating the likelihood of gaming
- Demonstrations with visualisations of checks and scores
Lessons Learnt
Future projects could test the solutions in practice and train the models in NESO’s Dynamic Response markets using historical data instead of synthetic data. This would help to inform any necessary revisions, such as weighting some checks more heavily than others and identifying thresholds where gaming is more likely, reflecting where events are confirmed or discarded. This approach could provide NESO with greater confidence in the robustness of the checks and ultimately reduce barriers for participation from units with variable baselines in its dynamic response services.
Additionally, other projects could explore applying similar techniques to other ancillary services. While different data flows and IT integration would need to be considered, the fundamental principles of gaming detection and the solutions reviewed in this project could be applicable to other services.
Future projects could also evaluate the applicability of the solution developed in this project for different use cases. The primary use case considered in this project was that of a battery providing an ancillary service while at the same time serving a variable load. Comparing the effectiveness of the solution across different use cases could help determine whether refinements are required including whether different checks should be carried out or weighted differently, for various scenarios.
