Climate change and environmental impact are hot topics in Germany. In 2019, EU parliament issued a directive that stated by 2025, 45% of public buses should be low- or zero-emission. This ratio was expected to grow further over the next five to ten years.
In light of this, SimPlan, a simulation and optimization service provider based in Germany, ran a project with the city of Hanau and Frankfurt University of Applied Science. The goal was to model and develop a digital twin of municipal electric vehicle (EV) fleet operations.
Electric vehicle modeling problem
For the project, SimPlan considered two vehicle types: public buses (HSB - Hanauer Straßenbahn) and garbage collection trucks (HIS - Hanau Infrastruktur Service). For the vehicle types, they needed to assess three areas:
- The fleet mix of electric vehicles powered by lead-acid batteries (BEV) and EVs powered by hydrogen (FCEV).
- The optimal charging and refueling infrastructure configuration and how much power it would require.
- The operational consequences of transport schedules and routes.
Additionally, SimPlan encountered two challenges that they also needed to consider for electric vehicle modeling.
Firstly, the electric vehicle fleet has travel range limitations compared to conventional diesel-fueled vehicles. A conventional bus has a range of 500 km, while a BEV and FCEV will only be able to cover 300-350 km.
The problem is that this EV travel range is less than the public transport’s current tour distance. Consequently, vehicle schedules and routes would need to be adjusted for the new EV-based fleet mix.
Secondly, the charging process is non-linear. How quickly an electric vehicle charges depends on its state of charge, the amount of power available at the charging stations, and the number of vehicles charging at the same time.
Solution
To build an EV operations simulation model, SimPlan first needed to gather data on the vehicles (their type, travel range, consumption), schedules, locations of bus stops and garbage containers, costs, and so on. The data was imported from Excel into AnyLogic.
In AnyLogic, they used an agent-based modeling approach to build two electric vehicle operations models. In the models, agents represented vehicles, facilities, transport management, and charging management. SimPlan described the charging process with statecharts and used Java code to develop heuristics for location mapping, vehicle scheduling, the arc-routing problem, and tour assignment.
During modeling, SimPlan discovered four additional routing restrictions that had to be considered to accurately capture the systems’ behavior:
- The direction of a street (one- or two-way).
- The side from which a truck should collect garbage (left or right side of a street).
- Avoiding U-turns. Large vehicles should avoid them in order not to block traffic.
- Hardly accessible garbage containers. SimPlan developed algorithms to determine the closest accessible road to such a container so that a truck can collect garbage.
For the electric vehicle operations models, SimPlan also created a custom user interface so that users can easily configure them and run scenarios.
Modeling electric vehicle operations (bus)
One of the models that SimPlan built for this project was an electric bus operations model. On the settings screen a user can choose vehicle types (diesel, BEV, FCEV), select the number of each type in a fleet, and set the number of charging stations in the depot, and their maximum power. The user can also select an option to consider the temperature effect on vehicle energy consumption.
On the map view of the model, the user can see different types of buses moving on the streets of Hanau, their states of charge, energy types, stops, and the charging station (the depot).
Modeling electric vehicle operations (truck)
The second model SimPlan built was a garbage collection simulation. Because the settings are similar to the previous model, we will only highlight the differences in the map view.
On the model screen, a user can see not only garbage trucks moving on the streets of Hanau, their states of charge, energy types, and the charging station (the depot), but also containers for specific garbage types. The user can also choose the preferred weekdays and the number of a week and observe changes in trucks’ schedules. If a truck reaches its maximum load, it drives to the disposal facility, unloads the picked-up garbage, and resumes the tour.
Statistics
While a model is running, AnyLogic’s visualization capabilities provide dynamic insight and track KPIs – showing the state of charge, tour duration, tour distance, and assignments per vehicle.
After running the models, they exported the data to an external database, evaluated the results, and compared scenarios.
Electric vehicle modeling results
In this project, SimPlan specialists built models of municipal electric vehicle fleet operations. In the models, they considered two vehicle types: buses and garbage trucks (both conventional, battery- and fuel-cell-powered) and modeled them as agents. To accurately describe the systems’ behavior, they used AnyLogic statecharts and Java for heuristics development.
The company used simulation modeling because it can accurately forecast large-scale EV fleet tasks and assess their daily operation.
After running simulation scenarios and processing the results, SimPlan analyzed a report on a number of required vehicles of each type. It showed that the cold season required more zero- and low-emission vehicles compared to warmer months, and more zero- and low-emission vehicles than conventional vehicles in general.
Additionally, they gained insight into potential power needs for the charging infrastructure. With these statistics, they could analyze what would happen to the networks’ performance if they had a limited number of charging stations or limits on power in general. Furthermore, they could evaluate how that would impact vehicle schedules.
In the future, SimPlan plans to investigate opportunity charging instead of overnight charging and whether on-demand bus tours (without a certain schedule) can fulfill real-time customer requirements. For day-to-day operations, the company plans to develop the simulation model into a digital twin.
The project and its findings were presented by Dr. Nadia Galaske of SimPlan at the AnyLogic Conference 2021 – Agent-based model to design and support e-mobility transformation for municipal vehicle fleets.
