Learnings

Outcomes

The main outcomes and outputs achieved are discussed below.  
Project trial: The project successfully recruited 60 participants, designed a trial specification which included the trial offer and recruitment approach, and gathered energy consumption data 
Consumer research:  
Consumer research was conducted to understand consumer energy usage and attitudes towards flexibility. Four consumer/network operator flexibility value propositions were assessed during interviews and surveys with the project participants. 
It was concluded that there is no one tariff fits all. Future energy tariffs, potentially paired with LCTs, could encourage off-peak energy use, but diverse tariff frameworks may be necessary. Flexibility with heating control is welcomed if it ensures thermal comfort without increasing bills. 
Flexibility in extreme weather scenarios can be limited due to the thermal comfort required by the consumers. 
People need education on heat pump operation, which operates more effectively in a constant temperature setting throughout the day compared to gas boilers, which can heat a home quickly with short-term operation during the day. 
ADMD model and analysis:  
The project developed an algorithm to simulate energy demand in 1-in-20 year winter scenario. It also developed a Matlab based tool to calculate ADMD figures for houses with heat pumps. Two different sets of ADMDs were developed: one for 1-in-20 winter extreme weather scenario using the trial dataset, and another for the winter of 2021-22 using the EoH dataset. The latter is considered a robust dataset and is proposed to be used by the business and the industry. The ADMD figures for that can be seen below: 
1-2-bedroom houses with heat pumps: 2.1kW 
3-bedroom houses with heat pumps: 2.6kW 
Larger than 3-bedroom houses with heat pumps: 3.4kW 
Heat pumps will typically contribute between 0.6 kW and 1.2kW to diversified load in winter, and less than 0.2 kW in summer. 
While the addition of a heat pump increases a household’s peak load significantly, the ADMD does not increase by the full heat pump peak load, as the heat pump and appliance peaks do not coincide. 
ADMD associated with flats and terraced houses is around 15% lower than the average value across similarly sized homes. 
Network simulations:  
The project identified and modelled in RSCAD (RTDS) three network topologies (rural, urban and suburban) that were used to assess the impact of heat pump uptake to electricity networks. 
Most distribution networks will be able to cope with low levels of heat pump uptake, but higher levels could cause network stress and damage.  
The daily load profile for homes is changing with the use of heat pumps, resulting in a flatter profile with a morning peak and more considerate evening peak. Consequently, the transformers and cables will be used differently, potentially resulting in the need for usage of continuous rating instead of cyclical ratings when sizing them. 
The heat pump uptake in the UK Power Networks is quite low, thus a top-down approach in different voltage levels cannot be performed with accuracy at this stage. 
Under severe weather conditions, heat pumps with a small rating compared to the heating demand of the house will run at full capacity for periods of several hours to reach the desired house temperature. Installing larger heat pumps would reduce the period of continuous running but increase the instantaneous load. 
Flexibility baseline: Heat pump profiles, to be utilised during the procurement of flexibility from heat pump aggregators to support billing activities, were developed. 
UK Power Networks’ network analysis:  
The project conducted a top-down regression analysis using actual UK Power Networks’ network data from high voltage feeders to examine the impact of heat pump technologies on the network. 
Observations revealed the heat pump profile changes the profile observed in feeders with heat pumps installed. However, no feeder in UK Power Networks had a high heat pump uptake, with the highest penetration being 8% and most feeders having less than 5% penetration. Despite observing a shift in load profiles due to heat pump penetration, the analysis was limited by the low numbers of heat pumps and the granularity of monitoring data. 
 

Lessons Learnt

Some lessons learned for future projects are described below. 
 
Recruitment activities  
It was particularly difficult to recruit participants that were not already part of a community of households that participate in trials, such as the Living Lab used in the project. In cases where a large dataset is required, it would be recommended to identify these datasets ahead of the project to mitigate risk to project outcomes. 
It was challenging to identify and recruit small houses (1 bedroom) and large houses (5+bedroom) with heat pumps and smart meters around GB. Medium sized houses (2-bedroom, 3 bedroom & 4 bedroom) were easier to recruit. This discrepancy can be attributed to the lower availability of small and large houses in comparison to medium-sized ones.  
Access to data outside of the project trial 
The project sought additional half hourly meter data from houses with heat pumps that were not part of the project to reinforce the calculations for the ADMD figures for heat. It was challenging to get access to data from innovation projects ran by other DNOs as they cited concerns due to GDPR related issues raised by the energy suppliers involved  
Dataset size 
Overall, while a small dataset can provide some interesting high-level insights, it is not sufficient to generate analysis outputs with a high level of confidence that can be directly used by the business to change standards or processes. This is why, the project team decided to seek additional datasets. The identification of the EoH dataset, which comprised approximately 700 heat pumps’ metered data over a year, was valuable for the analysis and provided the opportunity of more robust outcomes.  
Consumer research activities 
Although the project obtained some interesting learnings from the consumer research conducted, further testing would be required to understand customer behaviours to the flexibility propositions discussed and how much energy they would shift in respond to them. Some existing projects within UK Power Networks and the rest of the industry are already testing these concepts, and Neighbourhood Green has shared its insights with those projects. 
Lab testing activities 
Hardware of the PNDC Lab was replaced during the project impacting the project timelines for delivery. However, all activities were completed by the end of the project. When real time lab testing activities are included the scope of works, the project should make sure that all hardware will be available when required and any replacement and/or maintenance will not affect any project milestones.  
Some future work that could be undertaken is discussed below: 
Investigate heat pump demand under severe winter conditions using a larger dataset. This could involve analysing smart meter data during prolonged spells of unreasonably cold temperatures. An alternative would be to use a large dataset of a moderate winter with an improved algorithm that can simulate a 1-in-20 year winter. The algorithm developed as part of the project assumed a linear relationship between the heat pump consumption and weather which is the respective relationship between gas boiler consumption and weather. More work to establish the exact relationship is required. 
Understand the heat pump behaviour across houses with different energy efficiency levels by performing an analysis with large dataset that also includes information about the houses’ Energy Performance Certificates (EPC). The EoH dataset did not have sufficient EPC data to incorporate into the overall analysis.  
Perform an analysis with a larger dataset of houses with both heat pump and other LCT datasets to better understand the overall demand profile within homes. The trial should consider when different types of loads are used within the home and how they impact the demand profile of the home, e.g. time of day EV charged and time of day when heat pump is running. 
Combine a bottom-up approach for ADMD calculation using domestic smart meter data with a top-down approach from feeders with high heat pump uptake. This would provide improved insights into ADMD values and the respective impact of a heat pump uptake on the network at different voltage levels. This analysis can be performed once feeders with high heat pump uptake are identified, along with access to the domestic smart meters connected to these feeders.