The opportunities and challenges of range prediction for electric vehicles

MOBILErange generates a model for the energy consumption of the respective vehicle and, as a central service, calculates the range prediction required for the respective subsystem, e.g. for the charging management system, the depot management system or the ITCS. © INIT

For the seamless operation of electric vehicles, it is important to know the battery's state of charge at all times and whether there is enough charge to reach the end of the block. This is particularly relevant in order to give dispatchers enough time to take appropriate action if there is insufficient range, but also for the management of charging processes and for block assignments. With MOBILErange, INIT offers a solution for range predictions which reliably calculates and provides ranges for other systems. For this purpose, MOBILErange generates a model of the vehicle’s energy consumption and calculates the range forecast required, e.g. for the charge management system, the depot management system or the ITCS.

The range prediction is calculated using historical vehicle and operating data from the MOBILEefficiency evaluation system as well as current data such as the weather. MOBILErange can be used to avoid vehicle breakdowns and make planning and vehicle dispatching more efficient. The system is already in use and helping customers to optimize the cost of deploying their e-bus fleet.

Advantages of range prediction

Accurate range prediction is even more important in the day-to-day operation of electric buses as battery capacity does not decrease on a linear basis. Ideally, the prediction should accurately represent the actual energy demand over the course of the operating day. Accordingly, the predicted consumption should reflect the actual consumption as accurately as possible. Accurate predictions benefit many different areas for the company – firstly, deployment planning: Planning reserves can be reduced and consequently, electric vehicles can be deployed more efficiently. In other words, when fewer reserves are needed, a higher proportion of energy can be used productively. Transport companies with large fleets benefit even more because by using their electric vehicles more efficiently, they can reduce the number of required vehicles and ultimately reduce their costs. Good range prediction also brings many advantages for the control center. Dispatchers have to intervene less and, if the range does run low, it gives them more time to take appropriate action. At the end of the block, an accurate range prediction provides a reliable database for the charge management system, which can plan the required amount of charge for the next day based on the remaining range. For the next operating day, the depot management system can also use range prediction information to assign blocks.

Information about the range is important for the operation of the e-fleet, for example, for the management of charging processes and for block assignments. © INIT/Ulrike Kabel

Parameters influencing the prediction

A valid prediction that approximates actual consumption is the result of complex computing processes because energy consumption depends on a large number of influencing factors. These include the temperature during the operating day, because additional energy consumers such as heating and cooling systems consume a considerable amount of energy. Vehicle type, passenger volumes, topology, etc. also play a role. Most of these factors are known in advance for each block, so they can be taken into account for the prediction. Temperature information can be retrieved via a weather service. Operational information such as block, line and trip are of course also known in advance. Many of these factors depend heavily on the transport company’s general conditions: An electric vehicle used in a mountainous area consumes more energy than a similar vehicle on flat terrain. Based on recorded consumption data on the individual route sections, MOBILErange predicts the expected consumption for the blocks, trips and route sections, taking into account the current conditions.

Another factor that affects energy consumption is the volume of passengers. In the future, an interface could be used to predict occupancy levels from MOBILEguide, INIT's system for occupancy information, to further refine the range prediction.

Necessary planning reserves

Unforeseen events such as accidents or other factors which are not known in advance make it essential to plan energy reserves. These must be calculated in quantities as small or as large as necessary in order to run the vehicles as efficiently as possible. If a vehicle consumes more energy than predicted, it can run out of charge and may not reach the next charging station. Consequently, if energy consumption is predicted to be higher, this poses fewer problems than if it is predicted to be lower.

When determining optimal planning reserves, it is important to take into account the respective planning period because different systems used to manage electric vehicles all operate with different planning periods. For example, the planning system plans for the next few weeks, whereas the charge and depot management system plan for the next one or two operating days, and the ITCS usually only considers the current operating day. The longer the planning period, the larger the planning reserve should be, as there are likely to be greater uncertainties such as the weather. MOBILErange can adjust this planning reserve individually for each system.

In general, the more electric buses of a certain vehicle type are in operation on the same line or block, the faster the required amount of data can be obtained in order to refine the prediction so that only minimal planning reserves are required.

Basis for the prediction: vehicle data via the FMS interface

The prediction is based on the recorded data provided by the vehicle's on-board computer via the FMS interface. This is a communication protocol developed by a consortium of vehicle manufacturers for the uniform, manufacturer-independent transfer of vehicle data to the on-board computer. However, there are some challenges for the processing of this data, because the FMS interface was originally developed for diesel vehicles and does not yet contain sufficient e-mobility parameters. Among other things, this means that vehicle manufacturers do not all supply the same data. In addition, some vehicle manufacturers provide consumption data in a very low resolution (for example, in 0.5% or 1.5 kWh increments). For this reason, consumption cannot be explicitly defined on short sections of the route. If, for example, the distance between stops is very short, then there is no change in the state of charge on the sections in between, and a change only occurs after the third or fourth stop. The necessary processing of the data or allocation of the consumption to the short route sections is carried out in MOBILEefficiency.

The data supplied via the FMS interface is not always plausible. Implausible jumps in the consumption data (e.g. 85% -> 5% -> 84% on a route section of 300 meters) are filtered out and not taken into account for the prediction because they would greatly distort the result. Another challenge comes from the fact that consumption of additional energy consumers such as air conditioning, heating or on-board electronics is often not explicitly specified.

The VDV (Association of German Transport Companies) has already recognized the need to improve the quality of vehicle data. The VDV publication 238 is currently being worked on, which will, among other things, define a standardized data set for electric buses. MOBILErange is already making an important contribution to the efficient use of e-buses. With an improvement in the data basis is improved, it will be possible to further refine the range predictions.