Learnings

Outcomes

The outcomes of the project are as follows; 

• The project has delivered the trial area selection report that outlines the trial areas that have been selected and their key characteristics. 
• A combined planning process has been developed and documented. As part of the process to produce this, working documents have been produced outlining the data model, the approach to disaggregation, the options for sensitivity analysis and how energy efficiency impact can be modelled.
• The local energy plans have been created with the input of the local authorities which has involved a great deal of data preparation, disaggregation and the creation of dummy substations to support modelling.
• The NIFT tool has been adapted to allow for modelling of energy efficiency and to upgrade the analysis engine from WinDebut to Connect LV before being used to analyse the trial areas and generate results in a suitable format for the CBA tool.
• The HV NAT tool has been specified, developed, tested and used to analyse the networks in the trial areas. It has generated results in a suitable format for the CBA tool.
• The process to analyse the gas networks, determine the appropriate costs to be used and create suitable output files has been trialled.
• The project has highlighted some key areas where data quality is insufficient and would be problematic if the process were to be adopted at scale.
• The WS CBA tool has been configured with appropriate costs and metrics for benefits.   The configuration allows for the analysis of multiple use cases and for output files from multiple network analysis tools. The CBA tool has also been enhanced by the inclusion of standard charts.
• The learning from EPIC to date has been fed into the workshops held by Energy Systems Catapult in February 2022 in relation to Local Area Energy Planning.
• The analysis has been carried out to understand the impacts of different approaches or values for each use case.  Each Use Case has been analysed with the results written in a report.
• The combined learning for the project has been captured in the Evaluation and Learning report which includes recommendations for future work and for using the work and tools from EPIC. Learning has been shared via a webinar with interested parties.

Lessons Learnt

WP1 Trial Area Selection and Initial Data Gathering         

• The  boundary of the Strategic Planning Area (SPA) i.e. the area to be analysed with the EPIC Process, is critical and will be influenced by a variety of factors including location of significant new developments and the boundaries of the Electricity Supply Areas (ESA), which is the area supplied by a primary substation, and the boundary of the Gas Supply Area (GSA) the smallest area of the gas network that can be modelled in isolation.  In common with other projects, gathering this data took longer than planned and needed multiple iterations between the project partners, networks and local authority stakeholders. Significant efficiencies could be obtained if the main data elements were already prepared..  Standard data formats are under investigation as part of the Open Networks project, Work Stream WS1b P4.  

LV Tool Specification  

• Agreeing the format of the input and output files from the different modelling stages was beneficial reducing risks at the analysis stage.  It was possible to keep a degree in flexibility about some of the processes at the specification stage whilst still agreeing data formats.
• An existing tool was used which reduced costs overall, and means that the same tool can be used for other network areas relatively easily

WP2 Energy Plan development         

• Active local authority engagement is critical to the success of gathering the data required to generate accurate local energy plans. Continued, regular engagement is crucial and building sufficient time into the project plan to allow local authority stakeholders to refer to published (or draft) policies between the two workshops could be advantageous for future users of the EPIC process to ensure that the local energy requirements plans are as accurate as possible.
• To develop a fully integrated plan it is necessary to have local authority plan data and defined energy policies. This could come from a LAEP type process that would proceed EPIC.
• Use cases provided a useful starting point for network analysis and options appraisal. The number of use cases and sensitivities needs to be balanced against the increased network planning resource that is required.
• Existing stakeholder engagement with stakeholders, should be extended to capture potential future use cases that may require modelling.

WP3 – Investment and Options appraisal tool development and testing          

• The use of the WS CBA tool saved time and gave confidence that best practice was being applied. For future CBA studies, use of pre-exiting tools should be considered before any tool development.
• For user-defined financial metrics (e.g. Weighted Average Cost of Capital, capitalisation rate etc.), it’s important to ensure the most up-to-date and accurate values are used as these will change with time.
• To ensure compatibility of network analysis tools with the Whole Systems CBA tool, it would be useful to pre-define a live “EPIC CBA inputs” workbook where the outputs of the network analysis tools can be stored for effective data integration with the CBA tool. This would minimise the data manipulation required by the ‘EPIC energy planner’ and would be the most efficient way to collect and input the required data into the CBA tool. 
• To support potential future use of the CBA tool by a wide range of local authorities that may have differing views on which metrics to include in the tool, stakeholder engagement should take place to determine if standardised sets can be used.
• Future users of the EPIC process may want to align an approach to reference and locate network demand in the gas and electricity network analysis models. Although a postcode approach was used in project EPIC, a database based on Unique Property Reference Numbers (UPRNs) or a combination of gas and electricity meter numbers could ensure more effective, common language that is relevant and meaningful for both the gas and electricity networks.
• For technologies that impact both the gas and electricity networks, it is essential that the same forecasting methodology is used for these technologies by both networks and early agreement on an appropriate forecasting approach will be useful.

WP4 HV NAT development       

• It had been assumed that the “gaps” between the bottom up and top down analysis, due to distribution substations that could not be modelled within the NIFT and HV connected customers would be simple to fill but this process ended up being very time consuming.  This highlights a general point of requiring good quality data to support automated analysis processes. 
• Similarly the SINCAL network model that was used did not support network analysis one feeder at a time as parts of the model HV feeder attribution were missing. This meant that the network model did not include details of the HV feeder that each cable, transformer or item of switchgear formed part of and that models for each HV feeder from the source circuit breaker to the normal open points could not easily be generated. Modelling each HV feeder separately is expected to have benefits in terms of speed of modelling.
• In order to improve processing times, hourly rather than half hourly analysis was used and comparison of the results showed this improved running times without detrimentally affecting the results. 

WP 5 Gas Network Analysis tool development       

• The Gas network analysis tool is still under development and now has a focus on understanding the gas network impacts of a switch to hydrogen rather than whole system integration with electricity systems.
• Aligning the areas used for DFES disaggregation and gas networks was hampered by comparing postcodes with lat/long systems and issues with historic postcode changes.
• The maximum reduction in peak gas demand across all scenarios and SPAs was 13% but information wasn’t available for network analysis to determine whether this was because of local growth from new developments being outweighed by reductions in load via energy efficiency and / or switching to heat pumps. As a result it wasn‘t possible to identify the reinforcement that would be needed for new developments or any decommissioning if whole areas were moved to other technologies. In order to provide an opportunity to follow a process for gas network analysis and costing, work was done to generate dummy reinforcement based on arbitrarily modelled demand increases, or a change in the properties of the gas being transported, even though this wouldn’t influence the CBA or the integrated investment plans for this project.
• There was a lack of data on heat pumps and the evolution of boilers and other assets to use hydrogen over longer timescales which limited the hydrogen modelling that could be carried out. There is a need for profiles for hydrogen variants and a longer term view of prices and carbon intensity of gas vs. electricity.

WP 5 HV  Network Analysis       

• The SINCAL model contains cables with no thermal rating information as this has been sourced from the Geographic Information System (GIS) data.  Using a value of 99A allows us to prevent the tool over-reporting the required investment upgrades.
• The SINCAL model generates dummy transformers of 100kVA capacity at the locations of HV connected customers. These would have triggered  investment upgrades on non-existent transformers. Similarly they introduce an impedance which is not correct for network modelling. 
• The lack of HV feeder attribution in the underlying network model has resulted in the HV NAT needing to model an entire primary at a time rather than modelling each HV feeder separately.  It is possible that this is slowing down the overall processing time for the tool but it can’t be confirmed without having a comparable network model and changing how the HV NAT operates. This should be investigated as we are likely to make use of more automated network analysis in the future.
• The HV NAT running time was very slow, partially due to the number of nodes being processed in SINCAL. There were amendments that were made to speed up the process without compromising the results. One was to carry out analysis for 120 half-hour timesteps rather than 240 half-hour timesteps in the time series reflecting the representative days i.e. hourly rather than half-hourly analysis. Similarly, calculating Capacity Health Index (CHI) in the same power flow analysis, in which Network Investment (NI) and Flexibility Service (FS) calculations were carried out, saved time. 
• The HV connected sites had no transformer rating data with all of them reading zero.  This is correct as unless we have details of customer equipment the site will not contain National Grid owned transformers. However this resulted in issues with the disaggregation approach which was based on transformer ratings. Therefore, transformers for HV connected sites was assumed to be 2 MVA so that they get disaggregated load in the top down approach.
• The LV DFES data has got profile class (PC) information only for non-hybrid heat pumps i.e. a distribution substation had heat pumps allocated for PC1 and PC2 separately. This profile class split information is used by EA Technology. As PC information is not needed in HV NAT this PC split was seen by HV NAT as duplication of heat pump volume allocation and only PC2 volume was getting picked up in the analysis thereby underestimating the demand due to Heat Pumps (HPs).   
• Upgrading of 6.6 kV cables to 11 kV cables was intended to be captured in HV NAT; however, it has been decided not to consider this upgrade programmatically but to consider it as a one off. Hence it is not considered in HV NAT.
• The number of representative days in this kind of long term analysis can be reduced from five to three. The “Int_Warm” and “Summer MinGeneration” representative day recorded the least level of investment. Dropping these representative days would lead to lesser computational effort as the number of Half Hourly (HH) time steps reduces by a one fourth of the processing time.

WP5 LV network analysis          

• The time taken to prepare, complete and analyse the results is much more dependent on the number of use cases/scenarios modelled, rather than the total number of substations.  Where possible the number of scenarios should be minimised in order to reduce the costs involved.
• The availability of accurate, high quality network data for the area to be studied is key.  In this project timescales did not allow for an existing model to be updated, resulting in older, less accurate data being used.  As digitalisation of network data increases the availability of accurate models of the network should improve, and this should be a pre-requisite for future modelling.

WP 5 Assessment / Development of the CBA tool  

• To ensure compatibility with the Whole Systems CBA tool, it would be useful to pre-define a live “EPIC CBA inputs” workbook where the outputs of the three network analysis tools can be stored for effective data integration with the CBA tool. This would minimise the data manipulation required by the ‘EPIC energy planner’ and would be the most efficient way to collect and input the required data into the CBA tool.
• Future users of the EPIC process may want to align an approach to reference and locate network demand in the gas and electricity network analysis models. Although a postcode approach was used in project EPIC, a database based on UPRN (Unique Property Reference Numbers) or a combination of gas and electricity meter numbers could ensure more effective, common language that is relevant and meaningful for both the gas and electricity networks.
• For technologies that impact both the gas and electricity networks, it is essential that the same forecasting methodology is used for these technologies by both networks and early agreement on an appropriate forecasting approach will be useful.
• While it was intended to reflect the network benefit from reinforcement work that created spare capacity, it was very hard to specify a metric for this that could be applied consistently across the LV and HV networks and that did not have a value so large as to overwhelm the other benefits and costs in the network analysis. This was overcome by relating the metric to the change in capacity rather than reflecting the entire network capacity.

WP6 Use of the CBA tool to assess the use cases  

• Despite the time taken to set-up and gain familiarity, the Open Network Whole System CBA tool proved itself very useful and could be more widely adopted.
• There is a general need to improve the quality of (low voltage) network data and the assumptions underpinning LV network planning.

• Care needs to be taken when applying “whole system” cost benefits to understand the relationship between different cost/benefit drivers, some of which may counteract each other.

• Improved modelling of flexible Time of Use (ToU) tariffs is needed to better reflect how they would act to reduce peak demand.

• The energy efficiency and ‘Fit for the Future’ results suggest that further work should be completed to articulate the benefits of either approach.  

• Flexibility services may become cost effective for managing HV and LV constraints when there is a larger pool of LV connected flexibility service providers, therefore features to support future flexibility services should be built into domestic EV chargers and batteries.

• Given the similar costs of both EV charging scenarios, a policy that initially emphasises installing on-street charging points then moves to installing rapid charging hubs at a later stage is likely to be cost effective.

• While hybrid heat pumps can reduce network costs, the exclusion from incentive schemes may result in this opportunity being difficult to realise.