The project aims to explore, identify factors, analyse and manage the future fluctuations through disaggregated information of interconnector (IC) flows caused by operation and market dynamics, which lead to imbalanced volumes and unnecessary operational costs. In addition, this also impact the overall system security in terms of procuring additional responses and reserve holdings in managing the IC movements. The project seeks to identify the factors influencing future IC flow changes, improve forecast movement accuracy, and develop a robust operational strategy to mitigate these risks. The goal is to enhance the efficiency and security of the NESO operations by anticipating and optimising future IC movements.
Benefits
This project is expected to benefit the Electricity National Control Centre (ENCC) by improving confidence in operational dispatch decisions through more accurate interconnector (IC) movement analysis. It quantifies IC movements against key factors, helping mitigate risks such as ramping errors. Projected IC movements could also enable quicker responses to imbalance volumes, improving operational reliability, efficiency, and safety.
While not directly measured in this project, longer-term outcomes may include more accurate quantification of reserve and response holdings that incorporate IC movements. Moderated IC movements could give UK energy consumers better access to affordable electricity from international sources, potentially reducing costs for end users and lowering the need for ancillary market services.
Learnings
Outcomes
Phase 1 has successfully delivered a final report that includes a list of factors ordered by relevance and contributing to the volume and direction of the flows for each interconnector.
The data source of each factor is specified along with their contribution utilising the correlation method and SHAP values.
The characteristics of the unexpected interconnector moves have been identified and thorough analysis has been conducted on the historical data between January 2023 and November 2025 to conclude the frequency of their occurrence and the degree of impact on the flows.
This list of factors and the results of the analysis are being utilised during Phase 2 for the development of the Machine Learning (ML) models to forecast the interconnectors flows.
Lessons Learnt
Testing multiple model architectures demonstrated that different interconnectors may require different forecasting approaches, highlighting the importance of comparing model performance before selecting a final solution. Specifically, three distinct model types are currently being tested (Gradient Boosted Trees, NBEATS and TFT)
Additionally, the Final Physical Notification (FPN) of IC flows can be predicted either directly or indirectly, by calculating the delta between forecasted FPN and Day Ahead (DA) nominations.
