What will electric vehicles mean for transport and energy systems? Introducing BatSim
Understanding interactions between transport and energy by modelling the behaviour of electric vehicles
Decarbonising the transport sector is a key priority for countries across the world. In addition to reducing car trips through mode shift to walking, cycling, and public transport, encouraging remaining car users to adopt electric vehicles (EVs) is a major part of this strategy.
With ambitious policy targets such as ending the sale of petrol and diesel vehicles in the EU and UK by 2035, the infrastructural implications of widespread EV adoption will become much more important to consider. A handful of chargers in a supermarket car park will probably not be sufficient, especially where EVs are bought by those without their own garages who will be less likely to charge at home.
The delivery of this infrastructure will generate many questions for decision makers across the globe. Where should we locate en route charging facilities? How many chargers should we install in each location? How will levels of access to charging at home, on-street, and at destinations such as workplaces and supermarkets influence the need for en route facilities?
There will also be implications for the electrical grid. In areas of high demand, will EV demand overwhelm existing capacity? Is it realistic to upgrade grid capacity enough to cope with scenarios where EV mode share is very high?
Introducing BatSim
In the City Modelling Lab we have developed BatSim, a tool that can help us answer some of these questions. It models EV battery consumption and charging demand by postprocessing the outputs from our agent-based transport simulations.
We have programmed agents with charging logic from literature on how real-life agents (people!) make decisions around charging. Of course, there are issues with this and as with many things that require modelling, there is a great deal of uncertainty about how current behaviour will map to future behaviour.
We’ve started with some simple logic — each agent’s battery level is tracked as they drive and recharge. Agents aim to recharge in a way that minimises their cost in both time and money. Home charging (or depot charging for freight vehicles), where available, is thus the most attractive option as agents are generally based there overnight.
If an agent is unable to charge sufficiently while at home or during other activities, the drop in their battery level will trigger a ‘desire to charge’ while driving. Agents will, however, try to avoid en route charging due to its significant time penalty compared to refuelling for petrol or diesel.
The diagram below demonstrates how three people would choose to charge. Alice only decides to charge at home, although she also has access to charging at work or the supermarket. Bryan cannot charge at home, so instead chooses to charge at work. Carl, meanwhile, cannot charge at home, work, or the supermarket, and thus has to charge en route both on the way to work and on his way home.
BatSim outputs EV charging demand datasets that are highly detailed spatially and temporally, which can be analysed in a multitude of different ways. Thanks to clever programming, BatSim runs multiple scenarios extremely rapidly, allowing us to test and analyse many combinations of EV-related scenarios.
Here is a selection of the key insights that we can and have drawn from BatSim:
Where will people be able to charge?
For people with a home charger, one of the appeals of EVs is they can charge while parked at home overnight, eliminating the need to find somewhere to refuel. While many early adopters of EVs are more affluent and have the space to install a home charger, the ability to charge at home cannot be assumed if EVs are adopted more widely.
People living in flats or terraces are less likely to be able to charge at home, and BatSim can account for this when determining a given agent’s access to home charging. A previous post details how we have combined OpenStreetMap and census data to more accurately capture house types for agents based on where they live. This in turn helps us to determine through a probabilistic model whether a given agent will be able to charge at home.
BatSim can likewise take account of probabilities that an agent may be able to charge on-street or at a destination. These probabilities can vary with the type of activity (such as work, leisure, shopping etc.) or with the location (such as urban or rural areas). Combined with access to home charging, this enables us to build a picture of how each agent may be able to charge over the course of the day and indicates when and where an agent may need to charge en route.
How much electrical demand will EVs generate system-wide?
One of the beauties of an Agent-Based Model (once the model is sufficiently validated of course!) is that granular outputs can be aggregated based on many different attributes and a range of scales. This means BatSim can estimate the annual electrical demand from EVs split by charging type (e.g. home, en route), battery type (fully electric or plug-in hybrid), levels of income, and geographic area among others.
Importantly, we can visualise how both overall demand and demand per characteristic will vary depending on the uptake of EVs and access to different types of charging. This helps inform discussions on the scale of EV demand relative to overall grid capacity and indicates how particular measures — increasing access to destination charging for instance — will influence different types of demand at the system level.
How will EV charging demand vary across a day?
BatSim’s outputs allow us to analyse EV charging both geospatially and by time. This means we can analyse how different types of charging demand vary during the course of a day, which is a key consideration for electrical grids. For instance, we can see how en route demand may peak during typical commuting peak hours while destination charging peaks as people arrive at workplaces in the morning. We can also see the dominance of home charging, which spikes as people return home in the early evening.
We can see how this pattern varies geographically too. In rural areas - where most people have access to home charging - the proportion of en route charging is lower. Patterns can be substantially different in urban areas, where destination and on-street charging are much more significant, and in areas crossed by major road routes, which may have a much higher propotion of en route demand.
This analysis allows us to identify potential challenges — large spikes in demand as people plug-in at home after returning from work, for instance — that help decision makers to focus on which policy responses will have the biggest impact. This helps inform responses to the challenges EVs pose to the electrical grid, which we have touched on in a previous post.
What scale of en route charging infrastructure could be required at different locations?
As noted above, if an agent’s battery drops below a certain level while driving, BatSim will generate a ‘desire to charge’ event. At this point in the real world, a driver will search for a charge point, and we replicate this in postprocessing by allocating each ‘desire to charge’ to the nearest en route charging facility. We are working to enhance this more simple approach by adding charging events that are activities in the daily plans of our agents. This will allow us to assess how charging will influence wider travel behaviour in a more sophisticated way.
As each charging event has a spatial and temporal component, we can analyse the scale of demand at each charging facility across the day and by factors such as vehicle type. We can evaluate, for instance, how a fleet with a higher proportion of plug-in hybrids relative to fully electric EVs would affect charging demand (as plug-in hybrids would be more likely to refuel rather than charge en route) and see how peak demand compares to average levels of demand at different sites. This gives insight into the extent of infrastructure that may be required at particular facilities, which can be assessed against local capacity on the electrical grid.
We are currently applying this in work with Transport Infrastructure Ireland as they aim to meet the requirements of the EU’s Alternative Fuels Infrastructure Regulation (AFIR) for en route charging. Our analysis is helping build a picture on whether AFIR requirements alone will be sufficient to handle demand, or if larger scale infrastructure will be required in certain locations.
What else could we do?
There is plenty of potential for us to expand BatSim’s scenario testing to include variations in the transport network and different travel patterns. We could test the potential for enhanced public transport to encourage agents to switch from driving, which may reduce demand for EV charging infrastructure. Or we could consider patterns of demand on a weekend or during a holiday period, where demand in some regional areas may be much higher than on a typical weekday — large differences could have drastic consequences for local electricity networks. These will help enhance our understanding of how EVs can contribute to decarbonisation.
Conclusion
The push for decarbonisation globally means that radical changes are in store for the transport sector. With time at a premium, it is essential that decision makers have the insights to make informed decisions on which investments will most effectively bring us to a low-carbon transport future.
The capabilities of BatSim are a great step for us in analysing the impact of a wide range of EV scenarios. The tool brings us to a better understanding of interactions between different systems as we begin to model transport’s impact on the electrical grid. We already have ideas on how we could further enhance BatSim’s capabilities and look forward to bringing further updates in the coming months.
