Simulating individual charging behaviour in MATSim

Sebastian Hörl

13 June 2025

MATSim User Meeting 2025

Electric vehicles become more and more prevalent
Various studies on infrastructure placement exist
But behavior plays a crucial role
- Charging at home, at work, using public chargers?
- Only last component is covered in most studies
Goal: Let agents intelligently choose chargers in MATSim so that they optimize their daily score, like in a standard simulation
Additional features: fallback behavior if a station is taken, ...

Introduction

Electric vehicles (ev) contrib simulates decharging and charging of vehicles, provides the data and code infrastructure for charging stations, ...
EV charging in DRT
- Send vehicle to charging stations when idle
- Decide when to stop charging
UrbanEv package by TU Berlin
- Insert charging activities based on plan structure
- Charging inserted along the way
- Based on SoC thresholds
Last package is used by for optimizing charging service provision (Montreal)

Existing work

Concept

Each standard plan of MATSim obtains a new attribute that is a container for charging plans
Always one charging plan is selected while the others are there for memory
A charging plan indicates when an agent wants to charge during the day. The relevant information is saved in a ChargingActivity
- During an activity (start activity index, end activity index, charger)
- Along a leg (leg index, start/end time, charger)

Concept

While running the simulation, we obtain charging plan score:
- Large penalty for reaching SoC = 0
- Penalty for falling below a configurable SoC
- Penalty for performing a detour to get to a charger
- ...
A new replanning strategy for startegic charging is introduced:
- In X% of the cases, a new charging plan is proposed
- In (1 - X)%, an existing charging plan is selected from the charging plan memory, according to the charging score
Fully integrated with other replanning strategies: A plan that is duplicated for mode choice / time mutation / ... keeps its attributes and thus the charging plan memory

Components

Two new major components:

Within-day electric vehicle charging (WEVC)

Computational infrastructure for implementing charging activities during a day
Provides interfaces that return a list of planned charging activities at the beginning of the day and implements them at time step 0
Provides interfaces for online decision-making if desired

Strategic electric vehicle charging (SEVC)

A standard implementation of the interfaces of WEVC
Reads the planned charging activities from the plan attribute and lets WEVC implement them
Provides standard routines for online decision-making
Provides comprehensive scoring with charger types, tariffs, subscriptions, ...

Within-day electric vehicle charging

Handles the insertion of new plug and unplug activities:
- Useful for analysis
- Act as triggers in the modeling chain ("when a plug activity starts, do this ...")
Charging slots describe when the respective activities should be inserted (around a chain of activities, or along a leg)

Within-day electric vehicle charging

Handles the insertion of new plug and unplug activities:
- Useful for analysis
- Act as triggers in the modeling chain ("when a plug activity starts, do this ...")
Charging slots describe when the respective activities should be inserted (around a chain of activities, or along a leg)
Charging alternatives describe what the agent does when a planned charger is not available anymore

Within-day electric vehicle charging

En-route alternatives

When an agent goes on a leg that leads to a planned plug activity, we can call the alternative provides
Allows the agent to check whether the plug is currently taken; allows us to implement short-term reservation systems

Proactive charging

Send an agent to a charger at any time while a leg is ongoing
Allows to implement spontaneous charging decisions
With computational impact!

Strategic electric vehicle charging

Streamlined interfaces that are implementing the WEVC functionality
Replanning strategy for MATSim

Selection

Innovation

Choose among the existing charging plans based on the charging score

Find all possible activity-based charging slots
Find all possible leg-based charging slots
Randomly activate/deactivate each activity-based slot
Remove all incompatible leg-based slots
Randomly activate/deactivate each leg-based slot
Randomly select a charger for each retained slot

Charging infrastructure

Define different classes of chargers that ease the charger innovation process

Person-based chargers (e.g. household chargers)

Added to the choice set by person identifier
Per-charger attribute sevc:facilities (comma-separated)

Facility-based chargers (e.g. workplace chargers)

Added to the choice set by the facility identifier of the first activity along a charging slot
Per-charger attribute sevc:persons (comma-separated)

Public chargers

Searched within a radius of (configurable) meters
Per-charger attribute sevc:public = true

Charging infrastructure

Subscriptions

Person attribute sevc:subscriptions contains the subscriptions that an agent possesses
Charger attributes sevc:subscriptions indicates which subscriptions are necessary to use a charger
Only chargers without subscriptions or compatible ones are selected

Charging scoring

Various dimensions that are already included

Charging tariffs

Different configurable cost structures

Global for all chargers

Individually per charger

Using charger attributes

Tariff-based

All tariffs are configured by name in configuration
Chargers can have sevc:tariffs charger attribute
Tariffs can be accessible conditional on the possession of subscriptions
Agents choose the cheapest tariff that is available to them

Usage

Activate the relevant modules in the run script
Otherwise, everything encoded in
- Configuration
- Population (subscriptions, desired SoCs, ...)
  - Activate functionality by setting wevc:active = true !!!
- Chargers (types, ...)
Considerable work of setting everything up coherently and in a realistic way.
Plan is to provide a bootstrap script that sets up a basic configuration from scratch.
Next step: set up a fully parameterized realistic scenario!

Usage

Full MATSim loop

Use SEVC together with other replanning strategies
ChargingSlot activation works such that a slot is only implemented if it is still compatible with the current standard plan configuration

Standalone charging simulation

Disable all other replanning strategies
Agents will try to find the optimal charging plan for a fixed standard plan from a baseline simulation
Allows to adapt the infrastructure according to the driving needs of the population
Allows to explore the optimal charging decisions given current mobility patters

Output

Detailed trace for all
- charging processes (from first attempt to finishing or giving up)
- charging attempts
Detailed tracking of all scoring events
- When, where, who, how much (cost, travel time, ...), and how much score