The UK public is increasingly supportive of measures that support the transition to a net-zero carbon economy, including the rapid adoption of alternatives to fossil fuel-powered vehicles. A recent YouGov poll found that 56 per cent of respondents would consider buying an EV if it were affordable. However, 81 per cent of those polled raised concerns about the range of the vehicles and the lack of existing charging infrastructure.
Behind the scenes however, the rollout of EVs and associated infrastructure will pose a challenge to the UK’s six Distribution Network Operators (DNOs). This is because, a failure to prepare for heightened electrical demand could leave DNOs unable to meet expectations and regulatory targets.
With the adoption of EVs depending on a wide set of variables and a broad swathe of stakeholders, leveraging anonymised data, such as mobility patterns, is proving valuable to the ones who plan and support the localised rollout of necessary EV infrastructure. Mobility data is harnessed from mobile network operators and when combined with a variety of alternative datasets, either those in the public domain, or through private partnerships, it can assist with predictive infrastructure modelling.
Model behaviour
This is where scenario modelling and simulation exercises can shape the preparedness and responsiveness of DNOs and the wider utilities sector. The interconnectedness of the industry means that future planning cannot be siloed, the energy sector must work ever more closely with local authorities, EV suppliers and consumers to ready itself for fluctuations in energy supply and demand in particular geographic areas.
Mobility data serves a critical purpose in that it allows for the planning and understanding of future scenarios as it provides an opportunity to understand aggregated patterns of behaviour and means that forecasters can provide estimates and predictions of where (geographically) and when (time of day) there are likely to be peaks and troughs in electricity usage.
Such trends can be analysed through comparative measurements that allow modellers to observe mobility patterns – and the associated impact on the electrical grid – over a given period of time. Once observed and validated, these factors can be parameterised and leveraged to quantitatively model future scenarios. For example, mobility data could indicate key commuter routes in towns and cities and how these evolve over time with the introduction of EV charging hubs, and it can help to understand which areas are likely to experience a quicker EV uptake and will therefore see the requirements for enabling infrastructure increase.
Mobility mapping and infrastructure installations
Predictive modelling will play a key role in the phased rollout of EV enabling infrastructure over the next 15 years. Local authorities, private forecourt operators, and housing developers will need to understand where to install charging points, while DNOs can better understand when and where to reinforce their networks in order to increase capacity and meet consumer demand.
Through mobility data it is possible to analyse patterns of travel and by monitoring specific measures we can look to build geographical profiles to identify areas where the adoption of EVs is likely to be high and so will require a greater number of charging hubs and terminals. In these areas, we will build out ‘occupancy profiles’ which detail whether an area is largely residential, commercial or mixed-use, while also accounting for other factors such as the number of incoming and outgoing trips to the area in relation to existing charging infrastructure and the level of average dwell times. This provides an indication of how frequently an area is visited and how long travellers typically stay for.
Once the picture of a town, village or city has been constructed in this way, we can then run simulation exercises to inform future plans for the deployment of charging infrastructure. This predictive modelling allows local authorities and private forecourt operators to deploy charging infrastructure in areas of high demand where they will be of most need to drivers. This would allay fears from buyers of EVs around a lack of infrastructure and reduce the so-called ‘range anxiety’ experienced by many today. This strategy could also complement the objectives of private businesses by affording them a means of attracting further custom and maximising their returns.
Catering to demographics
An understanding of the demographics in a given area coupled with the mobility patterns of those people will provide purveyors of EV charging infrastructure with a clearer insight into local needs and demand. For example, in the aforementioned YouGov poll, 73 per cent of those polled aged between 18 and 24 indicated that they would purchase an EV.
With this in mind, areas where this demographic makes up a high proportion of the population – such as university towns and cities – could require EV infrastructure to be deployed at a greater rate than a coastal town populated by retirees. Leveraging anonymised demographic and mobility data can assist with predictive patterning that will play a fundamental role in the rollout of EV technology.
DNOs would benefit from such predictive analytics as areas of high risk would be identified. This means that enabling interventions can be made in locations where the available network capacity is more likely to be constrained. This would prevent an overburdening of the network and ensure that supply and demand are balanced.
DNO preparedness
Predictive analytics is already shaping the preparedness of one electricity DNO in the UK as it looks to improve the performance and resilience of its Low Voltage (LV) electrical network.
Due to an historical lack of available data for electrical load at secondary (LV) level, network operators currently have to take into account a degree of uncertainty when planning specific network interventions or designing new low voltage connections. This often leads to increased costs for onsite inspections / manual monitoring and in some instances, a less-than-optimal utilisation of the installed capacity.
By leveraging a number of data sets including mobility data, network data, available load data and other relevant data sets, a machine learning capability has been developed that is able to estimate maximum load profiles for every LV substation on the DNO’s network. The model is able to generate these estimates without relying on consumer smart meter data.
Proving its capabilities, the model is capable of providing full visibility of the maximum demand profiles for LV distribution networks, an estimated improvement of up to 20 per cent in released network capacity and it has the potential to reduce the need for onsite inspections – reducing the DNOs related costs.
This is not to say that leveraging data is not without its challenges. Currently, data sources employed are fragmented as there is no one singular source of data that can be called upon. However, by leveraging data where it is available and building this out with other available data sets, a degree of confidence can be garnered through machine learning which help to create high confidence around what the data indicates.
Where next?
The successful rollout of EV infrastructure depends on a significant number of variables – some of which are within our power to predict and prepare for. We have 15 years to begin taking action to support the transition to a “green” economy – predictive analytics undoubtedly has a role to play in the UK’s net-zero story. Our collective next step is clear: embrace the art of the prediction through quality data and machine learning.