The growing importance of climate issues is supported by alarming data. In recent decades, the number of climate disasters has doubled, generating global economic costs exceeding $2.5 trillion between 1990 and 2020. Faced with this reality, the European Union has set ambitious objectives, aiming to reduce its greenhouse gas emissions by 55% by 2030 compared to 1990, as part of a global strategy to combat climate change.
In order to encourage this dynamic, initiatives are deployed in various sectors, notably that of transport, responsible for 25% of greenhouse gas (GHG) emissions within the European Union. The road transport sector, alone responsible for nearly three quarters of these emissions, constitutes a priority where actions must be taken to mitigate its environmental impact.
Numerous measures have already been put in place to help decarbonize this sector such as the implementation of Low Emission Zones prohibiting access to the most polluting vehicles, the ban on the sale of new thermal vehicles from 2035 or even the financial aid supporting the purchase of an electric vehicle or the installation of an electric charging station. The implementation of these measures requires transport companies to review their fleet in order to respect the environmental constraints imposed at governmental or European level.
The Fleet Decarbonation tool helps support companies in their fleet decarbonation process by providing them with a clear and actionable decarbonation strategy. Based on the company’s fleet data (number of vehicles, types of vehicles, types of fuel, etc.), the tool indicates the replacements to be made to guarantee the GHG emissions reduction objectives set by the company to the desired horizon while ensuring its budgetary constraints.
Clear and interactive visuals as well as a decarbonation plan report are provided by the tool to enable an in-depth understanding of the decarbonation strategy and highlight key business indicators such as financial indicators (opex, capex), environmental indicators (GHG emissions, energy intensity) and fleet indicators (number of zero-emission vehicles, etc.).
In the rest of the article, one will discuss in more depth the functioning of the Fleet Decarbonation tool, whether from a use point of view or construction of the tool model. One will first present the Fleet Decarbonation tool, then look at the methodology used to build the model behind the tool, and finally conclude with the results of the tool and discuss the areas for improvement and the next steps in this project.
The Fleet Decarbonation tool
Input data
The tool uses a large amount of data to recommend the decarbonation plan for companies’ vehicle fleets. Some data is integrated into the tool while others must be entered by the user.
The data to be completed by the company can be grouped into 3 main categories:
- Fleet data: vehicle types, fuel types, years of vehicle purchase, annual distances traveled, etc.
- Infrastructure data: type of charging (electric/hydrogen), number of charging stations, years of purchase of the charging stations, etc.
- Various constraints: opex & capex budget, the objective of reducing CO2 emissions, regional policies, political constraints, etc.
Global data such as emissions factors, fuel and vehicle prices, mileage consumption, etc. All this data is available by vehicle type and fuel type for each year.
This information is crucial to enable the most accurate modeling possible of the emissions and costs associated with the entire fleet and thus estimate the most appropriate decarbonation plan for each company.
Simulation results
The Fleet Decarbonation tool provides information regarding the state of the fleet before and after the decarbonation process.
a. Initial fleet tab
Different KPIs are used to describe the initial fleet such as:
- The composition of the fleet including the number of thermal vehicles, hybrid vehicles and zero-emission vehicles.
- Current fleet emissions broken down into scope 1 emissions (emissions emitted directly by vehicles during their use) and scope 3 emissions (emissions emitted indirectly by vehicles, linked to fuel production and vehicle manufacturing ).
- The energy intensity of the fleet indicating the average mileage consumption.
Several graphs are made available to the user to have a visual representation of the distribution of vehicle types according to fuels, emissions and vehicle age.
All of these initial fleet analyzes can also be filtered by the department the vehicles belong to.
b. Results tab
The “Results” tab includes the fleet vehicle replacement strategy in order to achieve the user’s emissions reduction objectives while respecting their constraints (budgetary, political, etc.).
The KPIs included on this tab are mainly of two types :
- Financial: capex, opex and total cost of ownership (TCO) linked to the decarbonation process
- Environmental: energy intensity as well as CO2 emissions saved thanks to the energy transition strategy.
Six visuals are available to understand the replacements carried out each year and see how the composition of the fleet is changing. The CO2 emissions reduction trajectory is also plotted between the initial and final years of the decarbonation plan as well as the evolution of the share of scope 1 and scope 3 emissions. The latest visuals show waterfall graphs to easily identify additional capex and opex costs linked to decarbonation processes.
The Methodology
The model behind the Fleet Decarbonation tool is a combination of a prospective model and a genetic algorithm. The genetic algorithm indicates how a certain type of vehicle running on a certain type of fuel should be replaced in a certain year. The prospective model applies this replacement strategy provided by the genetic algorithm when it considers that vehicles need to be replaced and evaluates the costs and emissions linked to the transformation of the fleet.
Both models are briefly described in the sections that follow.
The prospective model
The prospective model performs the same steps for each year between the initial and final year of the decarbonation plan.
- Replacement year calculation: The prospective model identifies vehicles that need to be replaced for the given year by taking into account the age and total number of kilometers traveled by the vehicle.
- Fleet update: The template replaces the vehicles that need to be replaced for the year in question. A replaced vehicle will always be of the same type (private vehicle, light commercial vehicle, heavy goods vehicle) but may have a different fuel from the vehicle it replaces. The choice of the new fuel of the replaced vehicle is indicated by the replacement strategy provided by the genetic algorithm in matrix form, called replacement rate matrix. This matrix indicates for each fuel to be replaced the share to be replaced with new fuels, for each year and each type of vehicle.
- Updating infrastructure: The prospective model checks whether it is necessary to purchase new infrastructure (charging stations) in view of the change in fleet composition.
- Calculation of emissions and costs: The prospective model calculates GHG emissions and costs associated with the vehicle fleet.
- Update of the mileage of fleet vehicles: each vehicle sees its total number of kilometers increased by the average annual number of kilometers.
The genetic algorithm
The genetic algorithm is the origin of the replacement rate matrix. It is an iterative algorithm inspired by the reproduction process in biology. The objective of this operations research algorithm is to generate a replacement rate matrix that satisfies the company’s budgetary and environmental constraints and that minimizes the TCO of the fleet decarbonation process at the final year.
The steps of the genetic algorithm are as follows:
- Initialisation: N replacement rate matrices are generated randomly.
- Evaluation: These N matrices are used in the prospective model to evaluate the emissions and costs brought by the vehicle replacements proposed by each replacement rate matrix.
- Selection: among these N matrices, only K matrices are kept, those which give the best TCO at the final year (lowest TCO). The N and K parameters were set by testing several pairs of values for different fleets and different objectives, finding a compromise between the best performance and calculation times.
- Crossing: the K matrices retained (parent matrices) are randomly crossed with each other to form new matrices (child matrices) and fall back on N matrices in total. Crossovers consist of the random selection of the replacement rates of the parent matrices.
- Mutation: From the N replacement rate matrices kept and crossed, we introduce mutations into certain child matrices so as to introduce diversity. A mutation corresponds to changing certain replacement rates randomly while ensuring that they respect the constraints entered by the user (for example, if there is a ban on the sale of diesel from 2030, the replacement rate for diesel from this year will be 0%).
- Iterations: steps 2 to 5 are repeated i times so as to be able to explore a maximum of solutions given their large number and to move towards a set of increasingly optimal matrices with each iteration.
- Decision: Among these N matrices, the best matrix is kept, that is to say, the one which gives the lowest TCO and which verifies the constraints. In the case where the matrices cannot respect all the constraints, the matrix selected is the one which reduces emissions the most.
Improvements & Next Steps
The solution currently being developed meets the need for a decision-making tool in the energy transition of corporate fleets. It is aimed at companies that have a fleet of vehicles and want to reshape it to reduce their carbon footprint to meet government or European environmental expectations, while ensuring that they stay within their budget.
The Fleet Decarbonation tool provides a clear and customized decarbonation plan that enables companies to anticipate the vehicle changes they will need to make in the near future and measure the environmental and financial impact of these changes. Customers have access to KPIs and visuals describing their current and future fleet via a user-friendly and intuitive dashboard.
The first case study was carried out for a Canadian municipality with a fleet of 750 vehicles. In less than 4 hours, the Fleet Decarbonation tool found a vehicle replacement solution that reduced CO2 emissions by 33% and mileage consumption by 20%, while keeping within the municipality’s budget.
Given the success of this first version of the Fleet Decarbonation tool, further improvements to the tool will be carried out over the coming months in order to reduce calculation time by rethinking the algorithm used, to produce new impactful visuals and to integrate new functionalities such as the possibility of changing a vehicle for another type of vehicle or of reducing/increasing the size of the fleet for example.
Victoire VINCENT, Nahel ZIDI, Yanis SELLAÏ and Pierre MORDANT
