Learnings

Outcomes

The outputs produced in this study are as follows:

• Work package 1 report: A review of the current modelling capabilities of UK network operators for flexible gas generation.
• Work package 2 report: Key drivers of flexible gas generation operation and data availability for forecasting flexible generation.
• A model and accompanying documentation / user guide (work package 3 outputs). Also a slide set with the model calibration results.
• Work package 4 report: Recommendations & highlighting the value of data sharing 

The model developed forecasts of operation of flexible gas generators which will enable GDNs to more accurately forecast the operation of flexible generators on their networks, supporting renewable energy to help the UK meet Net Zero. 

As mentioned above, once more data becomes available on real life operation of flexible gas generation sites (either from new sites connecting to the gas system or from gathering more data from existing sites) this can lead to improvements in the accuracy of the model (via recalibrations) that could enable the model to become part of BAU network operators. 

This study and the model itself could be used to explore further how flexible gas generation could be used to support generation challenges in the future / deployment of more renewables in the system. The model is built around several key drivers (‘levers’) that influence the need for flexibility in the power system. Exploring how these levers will change in the future could be carried out to inform the need for and potential role for flexible generators in the future to support renewable generation deployment. 

Lessons Learnt

The model developed in this project forecasts the operation of flexible gas generation sites with a good level of accuracy, which could add more granularity to forecasting approaches currently used by GDNs.  We would encourage GDNs and DNOs to now use and test this model to get a better understanding of how it works and to explore if and how this can be integrated with current modelling approaches (with a view to making this new model part of BAU processes).  

This project highlighted several learnings where actions can be taken that would improve the model and its accuracy in the future / streamline processes or that would enable more value to be captured from this project.  These are:

•  As more flexible generation sites connect to distribution networks (which should be added to the model) and as better quality data is gathered on the existing sites (via deploying more data loggers), recalibrations of the model should be carried out. This will further increase the accuracy of the model, potentially enabling the model to become part of BAU (integrated into existing models or modelling processes).
• The model could also be useful for flexible gas generators (to support their forecasting) and the logger data gathered by GDNs could be useful for flexible gas generators (where metering issues may be identified). 
• Gathering more accurate data for flexible generation sites will support engagement with flexible gas generators and between GDNs and DNOs for more data sharing and collaboration.
• Further standardisation of data between GDNs, and across GDNs and DNOs, can facilitate more collaboration and will lead to more accurate and easier to use datasets for network companies.
• Changes in network codes and regulations will make it easier for network operators (in particular, DNOs and GDNs) to share useful data with each other that can support improvements in forecasting and network operation.
• NG ESO and NG GT could learn from this project / use this model to support and inform their understanding of and thinking around improving forecasting of embedded gas generation.
• This project has also highlighted the importance of improving transparency for data availability on sites that don’t fall under categories with lots of publicly available data (e.g. non-BMU sites). NG ESO could support improving transparency of information for these sites.
• The reports and material produced in this study could be distributed more widely to support building a common understanding of, and alignment on terminology for, the different elements of the operation of gas and electricity networks. For example, the WP2 report could be used as a guide to support understanding and knowledge building in electricity / flexibility markets.
• The WP4 report contains links to lots of relevant electricity network data sets that are publicly available. This could be used as a good reference for the network industry.
• The WP4 report also contains the approaches and assumptions that Delta-EE has used to handle data gaps. These can be used as guideline to define a common approach in handling data gaps for this model or for other forecasting / modelling work.

The only limitation experienced in this project was limitations in accessing some flexible generation site-specific data & information. In particular, key data & information that were not accessible (and where assumptions were made, or other datasets were brought together to plug gaps) are:

1. Knowledge on whether a power generation site connected to the gas network is a flexible gas generation site or something else (e.g. a CHP system).
2. Gas Distribution Use of System rates have not always been available. Some DNOs don’t publish this directly but information can still be inferred from other data sets.
3. Knowledge on whether sites are Balancing Mechanism Units (BMU) or non-BMU and if they are active in providing Short Term Operating Reserve (STOR) services. There is limited visibility, currently, from National Grid on non-BMU STOR sites.

Likelihood that the method will be deployed on a large scale in the future:

There is a high likelihood that this model will be regularly used by all GDNs. Engagement with all UK GDNs has been high on this project. The format of the outputs have been informed by what GDNs need and so this can work alongside existing modelling work undertaken by the GDNs.  It is also very likely that the model developed will be used by NG ESO and SPEN to inform their model development.

The model developed is very effective and robustly forecasts (within a reasonable error margin) the operation of around two thirds of the sites where historical operation data was provided by the GDNs. The model has been built in a way that users can add more sites to it in the future and it can be expanded to include any changes in the operational drivers of sites