Shell.

ai Hackathon for Sustainable

and Affordable Energy
Fleet Transition Challenge: Problem Statement

Introduction
Welcome to the fifth edition of the Shell.ai Hackathon for Sustainable and Affordable Energy.
Shell.ai Hackathon brings together brilliant minds passionate about digital solutions and AI, to
tackle real energy challenges and help build a lower-carbon world where everyone can access and
afford energy.

In the previous four editions, we addressed some of the digital challenges around the energy
transition: windfarm layout optimization (2020), irradiance forecasting for solar power generation
(2021), optimal placement of electric vehicle (EV) charging stations (2022), and supply chain
optimization for biorefineries (2023). This year, we focus on fleet decarbonization - a transition
problem in the mobility sector.
Challenge
Professional, delivery and operational fleets are a significant contributor to global greenhouse
emissions. Fleet owners aspire to achieve net-zero emissions promptly; however, the transition
presents a complex dilemma. Balancing the urgency of achieving net-zero emissions with business
sustainability and customer satisfaction requires a decision-making framework that considers
factors such as timing, location, and approach. In this hackathon, you will have a chance to
develop mathematical models to optimize fleet decarbonization strategies, to help fleet owners
make informed decisions that align with their energy transition objectives and business outcomes.
By harnessing the power of data and mathematical models, you will navigate the complexities of
demand forecasts, dissect emission profiles, and find ways to meet ambitious emission targets. The
end game is to develop ingenious solutions that strike a balance between operational effectiveness
and environmental impact.

Problem Statement
Road transport is the backbone of supply chain, playing a pivotal role in moving goods and
bolstering the economy. This mode of transport’s advantages are flexibility, door-to-door service, and
connectivity between cities, towns, and villages. While it comes with convenience and advantage,
professional, delivery and operational fleets are a significant contributor to global greenhouse
emissions. Fleet owners aspire to achieve net-zero emissions promptly; however, the transition
presents a complex dilemma. Balancing the urgency of achieving net-zero emissions with business
sustainability and customer satisfaction requires a decision-making framework that considers factors
such as timing, location, and approach.

In this hackathon, you will have a chance to develop mathematical models to optimize fleet
decarbonization strategies, to help fleet owners make informed decisions that align with their energy
transition objectives and business outcomes.

By harnessing the power of data and mathematical models, you will navigate the complexities of
demand forecasts, dissect emission profiles, and find ways to meet ambitious emission targets. The
end game is to develop ingenious solutions that strike a balance between operational effectiveness
and environmental impact. We will provide various yearly ‘demand’ data from a fleet operator that
must be met. The demand data is further divided into various size and distance buckets which indicate
what vehicle sizes should be used and how much distance per day they can cover. These are some
additional constraints imposed on meeting the customer demand. We provide various vehicles from
the following 3 drivetrains: Diesel, LNG, and BEV (Battery Electric Vehicle). For each of these
vehicles, the cost, operational yearly range, distance bucket they can cover, and the vehicle ID
(unique identifier which helps you reference them in your solution) is provided. We include the
information on the fuel consumption by every model of the vehicle and the corresponding fuel types.
Furthermore, we also include the cost for every fuel type along with the amount of carbon emissions
by each of them, for every single year. Finally, you are also provided with the total carbon emission
limits that must not be exceeded every year. All the data provided spans the years 2023 to 2038
(both years inclusive), for a total of 16 years.

Your solution should provide an optimal fleet composition over the years, which meets all supply-
chain demand and constraints while abiding by the carbon emission limits for every year and has the
lowest overall cost possible. The data provided to you and the solution expected from you has been
further explained in the next section.

Data
We are providing the following datasets:

1. Demand.csv: This file gives you the total yearly distance demand (in kms) that needs to be
satisfied with vehicles of size Sx (size bucket) which can travel at least a minimum of Dx
(distance bucket) per day. For example, row 1 indicates that there is a yearly demand of
869181 km for the S1 sized vehicles which can travel at least a minimum of D1 distance bucket
per day.
2. Vehicles.csv: This file gives you the vehicle ID (model), type of vehicle (drivetrain), size bucket,
year in which you can purchase it, purchase cost, yearly range (in kms), and the daily
maximum distance bucket it can travel.

ID Vehicle Size Year Cost ($) Yearly Distance

range (km)
BEV_S1_2023 BEV S1 2023 187000 102000 D1
: : : : : : :
Diesel_S2_2024 Diesel S2 2024 107120 106000 D4
: : : : : : :
LNG_S4_2038 LNG S4 2038 301395 118000 D4

3. Vehicles_fuels.csv: This file gives you the fuel consumption (unit of fuel consumed/km) for
every vehicle ID using a certain type of fuel.
Year Size Distance Demand

2023 S1 D1 869181

2023 S1 D2 2597094

: : : :

2038 S4 D4 306910
 ID Fuel Consumption (unit_fuel/km)
BEV_S1_2023 Electricity 0.893043
: : :
Diesel_S1_2023 B20 0.223016
: : :
LNG_S4_2038 BioLNG 0.154568

4. Fuels.csv: There are 5 fuel types, and for each, this table provides the carbon emission per
unit fuel and the (median) cost per unit fuel across all the years. It also includes the uncertainty
in the fuel cost.

Fuel Year Emissions Cost Cost Uncertainty

(CO2/unit_fuel) ($/unit_fuel) (±%)
B20 2023 3.04858 1.220845 0
: : : : :
Electricity 2024 0.0 0.184113 2
: : : : :
LNG 2038 2.486188 0.926499 30

5. Carbon_emissions.csv: Provides the total carbon emissions limits that should not be violated
for every year. It is a decreasing profile over the years.

Year Total Carbon emission limit

2023 11677957
2024 10510161
: :
2038 2404387

Notations:
Ctotal ← Total cost of fleet ownership and operations across all the years.
𝑦𝑟
𝐶𝑏𝑢𝑦 ← Total cost of buying vehicles in year 𝑦𝑟.
𝑦𝑟
𝐶𝑖𝑛𝑠 ← Total insurance cost incurred on the vehicles in the fleet for the year 𝑦𝑟.
 𝑦𝑟
𝐶𝑚𝑛𝑡 ← Total maintenance cost incurred on the vehicles in the fleet for year 𝑦𝑟.
𝑦𝑟
𝐶𝑓𝑢𝑒𝑙 ← Total fuel cost incurred on the operating fleet in the year 𝑦𝑟.
𝑦𝑟
𝐶𝑠𝑒𝑙𝑙 ← Amount received by selling some vehicles in the fleet in the year 𝑦𝑟.

𝑉𝑦𝑟 ← Set of all vehicles purchased in the year 𝑦𝑟.

𝐶𝑣𝑦𝑟 ← Purchase cost of a single vehicle with ID 𝑣𝑦𝑟 .

𝑁𝑣𝑦𝑟 ← Number of vehicles of ID 𝑣𝑦𝑟 that have been purchased.

𝐹𝑦𝑟 ← Fleet of vehicles in the year 𝑦𝑟.

𝐶𝑣𝑦𝑟𝑝 ← Cost of vehicle in fleet purchased in the year 𝑦𝑟𝑝. 𝑦𝑟𝑝 is the year of purchase and 𝑦𝑟
is the year of operation such that 𝑦𝑟𝑝 ≤ 𝑦𝑟.

𝑁𝑣𝑦𝑟𝑝 ← Number of vehicles currently in the fleet in the year 𝑦𝑟𝑝.

𝑣𝑦𝑟𝑝
𝐼(𝑦𝑟−𝑦𝑟𝑝) ← Insurance cost in the year 𝑦𝑟 for vehicle 𝑣𝑦𝑟𝑝 purchased in the year 𝑦𝑟𝑝.
𝑣
𝑦𝑟𝑝
𝑀(𝑦𝑟−𝑦𝑟𝑝) ← Maintenance cost in the year 𝑦𝑟 for vehicle 𝑣𝑦𝑟𝑝 purchased in the year 𝑦𝑟𝑝.
𝑣𝑦𝑟𝑝
𝐷(𝑦𝑟−𝑦𝑟𝑝) ← Depreciation cost in the year 𝑦𝑟 for vehicle 𝑣𝑦𝑟𝑝 purchased in the year 𝑦𝑟𝑝.

𝑈𝑦𝑟 ← Vehicles being used (driven) in the year 𝑦𝑟. This is a subset of 𝐹𝑦𝑟 .

𝐹𝑣 ← All fuel types applicable for vehicle 𝑣.

𝐷𝑠𝑣𝑓 ← Distance travelled by vehicle v using fuel f.

𝑁𝑣𝑓 ← Number of vehicles of type 𝑣 driving fuel type 𝑓.

𝑓
𝑚𝑣 ← Fuel Consumption of vehicle type 𝑣 driving with fuel type 𝑓.
𝑦𝑟
𝐶𝑢𝑓,𝑓 ← Cost of unit fuel of type 𝑓 in the year 𝑦𝑟.
𝑠𝑒𝑙𝑙
𝑁𝑦𝑟,𝑣 𝑦𝑟𝑝
← Number of vehicles 𝑣𝑦𝑟𝑝 to be sold in the year 𝑦𝑟.
𝑦𝑟
𝐶𝑎𝑟𝑏𝑜𝑛𝑡𝑜𝑡 ← Total carbon emission in the year 𝑦𝑟.

𝐶𝐸 𝑓 ← Carbon emission for the fuel type 𝑓.

Objective:
2038
𝑦𝑟 𝑦𝑟 𝑦𝑟 𝑦𝑟 𝑦𝑟
𝐶𝑡𝑜𝑡𝑎𝑙 = ∑ 𝐶𝑏𝑢𝑦 + 𝐶𝑖𝑛𝑠 + 𝐶𝑚𝑛𝑡 + 𝐶𝑓𝑢𝑒𝑙 − 𝐶𝑠𝑒𝑙𝑙
𝑦𝑟=2023
 𝑦𝑟
𝐶𝑏𝑢𝑦 = ∑ 𝐶𝑣𝑦𝑟 ∗ 𝑁𝑣𝑦𝑟
𝑣𝑦𝑟 ϵ𝑉𝑦𝑟

𝑦𝑟 𝑦𝑟𝑝 𝑣
𝐶𝑖𝑛𝑠 = ∑ 𝐶𝑣𝑦𝑟𝑝 ∗ 𝐼(𝑦𝑟−𝑦𝑟𝑝) ∗ 𝑁𝑣𝑦𝑟𝑝
𝑣𝑦𝑟𝑝 ϵ𝐹𝑦𝑟

𝑦𝑟 𝑦𝑟𝑝 𝑣
𝐶𝑚𝑛𝑡 = ∑ 𝐶𝑣𝑦𝑟𝑝 ∗ 𝑀(𝑦𝑟−𝑦𝑟𝑝) ∗ 𝑁𝑣𝑦𝑟𝑝
𝑣𝑦𝑟𝑝 ϵ𝐹𝑦𝑟

𝑦𝑟 𝑓 𝑓 𝑓 𝑦𝑟
𝐶𝑓𝑢𝑒𝑙 = ∑𝑣ϵ𝑈𝑦𝑟 ∑𝑓ϵ𝐹𝑣 𝐷𝑠𝑣 ∗ 𝑁𝑣 ∗ 𝑚𝑣 ∗ 𝐶𝑢𝑓,𝑓

𝑦𝑟 𝑣
Your
𝑦𝑟𝑝 𝑠𝑒𝑙𝑙
𝐶𝑠𝑒𝑙𝑙 = ∑ 𝐶𝑣𝑦𝑟𝑝 ∗ 𝐷(𝑦𝑟−𝑦𝑟𝑝) ∗ 𝑁𝑦𝑟,𝑣 𝑦𝑟𝑝
𝑣𝑦𝑟𝑝 ϵ𝐹𝑦𝑟

Constraints:
1. Vehicle of size Sx can only cater to the demand of size bucket Sx.
2. Vehicle belonging to distance bucket Dx can satisfy all demands for distance bucket D1 to
Dx. For example, vehicle belonging to distance bucket D4 can satisfy demand of D1, D2,
D3, D4buckets; similarly, D3 can satisfy D1, D2, D3 but NOT D4.
3. Total carbon emitted by fleet operations each year should be within the respective year’s
carbon emissions limits provided in carbon_emissions.csv. Total carbon emissions for a year
is calculated using:

𝑦𝑟
𝐶𝑎𝑟𝑏𝑜𝑛𝑡𝑜𝑡 = ∑ ∑ 𝐷𝑠𝑣𝑓 ∗ 𝑁𝑣𝑓 ∗ 𝑚𝑣𝑓 ∗ 𝐶𝐸 𝑓
𝑣ϵ𝑈𝑦𝑟 𝑓ϵ𝐹𝑣

4. Total yearly demand for each year must be satisfied for each distance and size buckets.
5. Vehicle model of year 20xx can only be bought in the year 20xx. For example,
Diesel_S1_2026 can only be bought in 2026 and not in any subsequent or previous years.
6. Every vehicle has a 10-year life and must be sold by the end of 10th year. For example, a
vehicle bought in 2025 must be sold by the end of 2034.
7. You cannot buy/sell a vehicle mid-year. All buy operations happen at the beginning of the
year and all sell operations happen at the end of the year.
8. Every year at most 20% of the vehicles in the existing fleet can be sold.

Evaluation:
You are expected to provide the solution in a .csv format file. The column names that should exist in
the .csv along with “valid” entries are provided in the table below.
 Column Name Valid entries
Year 2023, 2024, ...., 2038
ID Should be among list of IDs provided in
vehicles.csv
Num_Vehicles >=1
Type Should be among “Buy”, “Use”, “Sell”.
Fuel Should be among “Electricity”, “B20”,
“LNG”, “BioLNG”, “HVO”.
Distance_bucket Should be among D1, D2, D3, D4.
Distance_per_vehicle(km) Should > 0 and <= Yearly range of that
model.

Note: Distance bucket of the solution file corresponds to the distance bucket in the demand.csv file.
Note that this is not the distance bucket of the vehicle itself (we get that from vehicles.csv anyways
using the provided ID).

Additional info:
1. Vehicle resale value, insurance cost and maintenance costs as a percentage of its purchase
cost is given below for each year after the purchase of vehicle.

% of Purchase Cost

End of Year Resale Value (%) Insurance Cost (%) Maintenance Cost (%)

1 90% 5% 1%

2 80% 6% 3%
3 70% 7% 5%

4 60% 8% 7%

5 50% 9% 9%

6 40% 10% 11%

7 30% 11% 13%

8 30% 12% 15%

9 30% 13% 17%

10 30% 14% 19%
To illustrate the calculations, let us take an example of purchase cost = $100 for a vehicle bought on
Jan 1st, 2025. Using the percentages in the above table, the values can be calculated as follows:

Year
End Resale
of Resale Insurance Maintenance Insurance & Maintenance
of Value
Oper Date Cost (IC) ($) Cost (MC) ($) Period
Year (RV) ($)
ation
31st Dec,
2025 1 90 5 1 1st Jan – 31st Dec, 2025
2025
31st Dec,
2026 2 80 6 3 1st Jan – 31st Dec, 2026
2026
31st Dec,
2027 3 70 7 5 1st Jan – 31st Dec, 2027
2027
31st Dec,
2028 4 60 8 7 1st Jan – 31st Dec, 2028
2028
31st Dec,
2029 5 50 9 9 1st Jan – 31st Dec, 2029
2029
31st Dec,
2030 6 40 10 11 1st Jan – 31st Dec, 2030
2030
31st Dec,
2031 7 30 11 13 1st Jan – 31st Dec, 2031
2031
31st Dec,
2032 8 30 12 15 1st Jan – 31st Dec, 2032
2032
31st Dec,
2033 9 30 13 17 1st Jan – 31st Dec, 2033
2033
31st Dec,
2034 10 30 14 19 1st Jan – 31st Dec, 2034
2034

2. Distance bucket mappings.

Distance bucket Name of category

Up to 300 km daily D1
Up to 400 km daily D2
Up to 500 km daily D3
Up to 600 km daily D4
 3. Vehicle size bucket mappings

Vehicle size Name of category

17 tons S1
44 tons S2
50 tons S3
64 tons S4

4. Please note that insurance and maintenance costs should be calculated for all vehicles in the
fleet (irrespective of whether they are used in a particular year), whereas fuel costs are only
for those vehicles that will be used (driven) in that year.

Scoring:
If your solution satisfies all constraints, we will first calculate your solution’s total cost of fleet
ownership and operations using the overall cost function. Your cost (the lower the better) will then
be converted to a score (the higher the better) between 30 to 100 using the following transformation
function for the leaderboard ranking:
70×𝑐𝑜𝑠𝑡
𝐿𝑒𝑎𝑑𝑒𝑟𝑏𝑜𝑎𝑟𝑑 𝑆𝑐𝑜𝑟𝑒 = max [30, (100 − 𝑅𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝐶𝑜𝑠𝑡 )]

Scores between 0 to 26 are reserved for the error codes detailed below.

As you can see from the fuels.csv, there is a median fuel cost along with an uncertainty band that
has been provided. The final cost (which will also comprise of fuel costs), will be a statistical
aggregate over 1000 randomly drawn samples from the provided fuel distribution. There are two
leaderboards: Public and Private. The public leaderboard is shown to you during the entire duration
of the competition. Here, you will be evaluated until the year 2028. The private leaderboard will not
be shown to you and will only be revealed at the end of the competition. This is evaluated for the full
course of 16 years from 2023 to 2038. The reference cost for private leaderboard is 172,000,000,
and for public leaderboard it is 65,000,000.

Error codes:
0 → Any unaccounted error code.
1 → Year entry must be of type integer.
2 → ID entry must be a string.
3 → Num_Vehicles entry must be of type integer.
4 → Type entry must be a string.
5 → Fuel entry must be a string.
6 → Distance_bucket entry must be a string.
7 → Distance_driven_per_vehicle(km) must be of type float.
8 → Year entry must be >=2023 and <=2038.
9 → ID must be from the given list of vehicle IDs.
10 → Num_Vehicles must be > 0.
11 → Type must take values among “Buy”, “Sell”, “Use”.
12 → Fuel must be among “Electricity”, “LNG”, “BioLNG”, “HVO”, “B20”.
13 → Distance bucket must be among “D1”, “D2”, “D3”, “D4”.
14 → Distance driven must be >= 0 and <= Yearly range for that model.
15 → Constraint 2 violations.
16 → Constraint 5 violation. Vehicle bought in year YYYY should have YYYY in its ID.
17 → Constraint 6 violation. Vehicle purchased in YYYY can only be used for 10 years.
18 → Demand.csv has the demand pertaining to 16 combinations (S1_D1, S1_D2, …, S4_D4) for
each year. Your solution.csv file should also have all these combinations present.
19 → Constraint 4 violation. For example, in demand.csv for year 2023 S1_D1 demand is 869181
km. In the solutions you provide, sum of all distance travelled in the year 2023 by vehicles with S1
size which satisfy D1 distance demand must be >= 869181 km.
20 → You should use the right type of fuel for a given vehicle.
21 → Constraint 8 violation. You must sell only a max of 20% of the fleet every year and nothing
more.
22 → You can only "Use" vehicle IDs that you have in fleet.
23 → You can only "Use" as many vehicles that you have in fleet for each ID.
24 → You can only "Sell" vehicle IDs that you have in fleet.
25 → You can only "Sell" as many vehicles that you have in fleet for each ID.
26 → Constraint 3 violation. Total Carbon emission by fleet operations <= Carbon budget for that
year.