Embed Size (px)
Citation preview
E-Car-Sharing: Modeling, Planning, IncentivesWocomoco 2017, Berlin
Bin Hu
AIT Austrian Institute of Technology
Center for Mobility Systems
Dynamic Transportation Systems
2
Methods and strategies for
optimized planning and
operating e-car sharing systems
Strategic level
recharging stations
fleet configuration
Tactical level
incentives
Operational level
relocation and recharging
Topic
© Drivenow
© car2go
© Flinkster
© multicity
3
Input
Business area + road network
Recharging stations can be built on certain street
junctions
Investment budget (for cars + stations)
Demand prediction for user trips
Goal
Maximize the expected profit by
Balancing the number of cars and stations
Planning the locations for recharging stations
Choosing the right incentives
Viennese Use Case
4
Modeling User Choices
5
Modeling User Choices
6
Modeling User Choices
Battery SOC too low
7
Modeling User Choices
Nearest car
8
Modeling User Choices
Fully charged car
(high incentive)
9
Modeling User Choices
“Stranded” car
(incentive)
10
Modeling User Choices
70%
25%
0%
5%
11
Modeling User Choices
Desired destination
Charging station
(high incentive if
SOC under 20%)
High future demand
(high incentive)
70%
25%
0%
5%
12
Questionnaire
Income
Typical car-sharing usage behavior
Goodies/incentives that are appealing
Computation of “shadow wages” – value of time
Results
~39% of the users are not likely to accept goodies
at all for taking detours (shadow wage of 66 EUR/h)
~61% of the users will potentially accept goodies
for taking detours (shadow wage of 28 EUR/h)
Amount of incentive that is reasonable, e.g.:
10 min for charging a car
max 25% reduction for taking a stranded car
Incentives and User Acceptance
Always pick up
nearest car
Depends on incentive
and detour distance
Online questionnaire by
13
Scenario settings
Investment budget of 500k – 1M Euro
Different business area sizes
Vehicle fleet: Smart ED with average 50% SOC
Users are willing to walk 3 min to a car / to their destination
Demand model based on Taxi data
Solution algorithm
Variable Neighborhood Search
Solution evaluation with Monte Carlo Sampling
Simulation for one week of operation
Maximize the expected profit
Computational Experiments
14
Results
Recharging stations location Cars after one week, with incentives
SOC: full – empty
Cars after one week, without incentives
SOC: full – empty
15
Average usage of one day
Number of trips: 7
Total travel time: 53 min
Total charging time: 9 min
Total waiting time: 23 h
Average SOC
Begin of scenario: 50%
End of scenario: 45%
End of scenario (without incentives): 1%
Car Statistics
3%
1%
96%
Total travel time
Total charging time
Total waiting time
16
Accepted Trips Statistics
24%
61%
1%14%
76%
Pickup locations
Unfulfilled requests
Nearest car
Fully charged car from a station
Car with incentive
23%
56%
2%19%
77%
Return locations
Unfulfilled requests
Desired destination
Charging station
Other incentivized location
17
Profit, Number of Trips, and Paid Incentives
0.00
1000.00
2000.00
3000.00
4000.00
5000.00
6000.00
7000.00
8000.00
9000.00
10000.00
0.00
200000.00
400000.00
600000.00
800000.00
1000000.00
1200000.00
1400000.00
Area size 1 Area size 2 Area size 3 Area size 4
Profit Paid incentives Number of accepted requests
Method for modeling, optimizing and simulating free-floating e-car sharing systems
Each area size requires a minimum budget to make it work
Network effect → necessary density of cars
Charging stations → necessary amount of chargers
Applying the right incentives is essential on operational level
Less staff personnel required for recharging
Less staff personnel required for rebalancing
Conclusions
