View Conference Presentation - USAEE
The Electric Vehicle Charging Station Location Problem: A Parking-Based Assignment Method T. Donna Chen Dr. Kara M. Kockelman Moby Khan Department of Civil, Architectural, & Environmental Engineering University of Texas at Austin
Overview • • • •
Motivation Existing research Puget Sound region data Study methodology
Photo by Daniel Reese for KUT News
▫ Forecasting zone-level parking demand ▫ Forecasting trip-level parking demand ▫ Locating optimal charging infrastructure (Seattle application)
• Conclusions
Background • Trend: Increased number of EVs in the market ▫ Higher fuel economy requirements ▫ Higher fuel costs
• Caveat: Range anxiety for EV owners/potential buyers affects ▫ ▫ ▫ ▫
EV adoption rates Electrified mile shares Petroleum demand Power consumption across times of day
• Potential solution: Public charging station provision
Background • Charging station installation costs ▫ $3,000 to $15,000 per station using existing infrastructure ▫ $40,000+ per station when requiring local infrastructure upgrade
• Energy providers, cities, & MPOs need a methodology to optimally locate public charging stations that… ▫ Serve charging demand, ▫ Minimize access costs for EV drivers, ▫ Meet constrained budgets.
Existing Research • Wang et al.’s (2010) numerical method for Chengdu, China used distribution of gas-station demands as proxy for charging demands. • Sweda & Klabjan’s (2011) Chicago case study uses an agent-based model to identify residential patterns of EV ownership & driving activities to identify strategic station locations. • Frade et al.’s (2011) coverage model for Lisbon seeks locations that maximize match for charging demands. • Our work most closely tracks that of Hanabusa & Horiguchi (2011): minimizing driver walk costs while ensuring coverage (via buffer distances between stations, in Seattle).
Our Approach • Behavioral models calibrated to predict when & where EVs are likely parked. ▫ Zone-level parking demand based on land use attributes of destination zones. ▫ Trip-level parking demand based on individual trip characteristics.
• Optimization routine (MIP) identifies charging station locations in order to… ▫ Minimize station access penalties for EV drivers, while… ▫ Satisfying budget constraints, & ▫ Ensuring minimum station spacing requirements.
The Data • Puget Sound Regional Council’s (PSRC) 2006 household travel survey data. ▫ 4 Washington Counties (King, Kitsap, Pierce & Snohomish) ▫ 4,741 households ▫ 10,510 individuals ▫ 3,700 traffic analysis zones (TAZs) ▫ 1,177,140 parcels ▫ 87,600 total person-trips ▫ 48,789 trips by light-duty vehicles
Seattle’s 3,700 Traffic Analysis Zones
Summary Statistics of PSRC Data Mean
St Dev.
Min
Max
Age (years)
41.9
21.8
0
99
Male Indicator
0.47
0.50
0
1
Driver’s License Indicator
0.78
0.42
0
1
0.21
0.4
0
1
Household Size
2.22
1.21
1
8
#Workers in Household
1.13
0.85
0
5
#Vehicles in Household
1.89
1.07
0
10
$71,400
$42,300
$5,000
$175,000
Person Records (N=10,510)
Student Indicator Household Records (N=4,741)
Annual Household Income
Determining Parking Location & Duration • Starting with 48,789 trips (55.7% of all persontrips). ▫ Only trips ending away from one’s home parcel are counted. ▫ Only trips with adequate parking duration (at a single parcel) are counted (at least 15 minutes), to help ensure adequate time for charging.
• 30,085 candidate parking periods emerged.
Forecasting Zone-Level Parking Demand Summary Statistics of PSRC Zone Attributes Variable (n = 3962) Parking duration (mins/mile2) Population density (persons/mile2) Employment density (jobs/mile2) Student density (students/mile2) Housing density (units/mile2) Average price of daily paid parking ($) Average price of hourly paid parking ($) 3-way intersections (1/2 mile radius) 4-way intersections (1/2 mile radius) Express bus stops (1/4 mile radius) Bus stops (1/4 mile radius)
Mean 2.41E+04 7.99E+03 1.26E+04 1.64E+03 3.43E+03 $0.145 $0.066 45.8 36.8 2.25 6.21
Std. Dev. 1.18E+05 1.88E+04 8.27E+04 2.48E+04 8.20E+04 $1.027 $0.465 22.52 45.91 6.369 9.016
Min 0 0 0 0 0 0 0 0 0 0 0
Max 2.43E+06 2.91E+05 2.07E+06 1.02E+06 1.27E+05 $21.3 $11.0 119.3 251.8 55.6 69.6
OLS Results for Total Demand/SqMile Y= Total Parking Demand in Zone (minutes/mile2) Variable Constant Density Population density (residents/mile2) Employment density (jobs/mile2) Student density (students/mile2) Parking Prices (within ¼ mile) Average price of daily parking ($) Transit Access & Network Connectivity #3-way intersections (within ½ mile) #4-way intersections (within ½ mile) #Express bus stops (within ¼ mile) #Bus stops (within ¼ mile) Number of Observations Adjusted R-squared
Parameter Estimate 3268
Standardized Coef.
-0.294 0.583 0.226
-0.047 0.408 0.047
-3.50 27.0 4.11
2.22
0.193
11.0
-0.030 0.062 0.083 0.124 3,692 TAZs 0.521
-2.41 2.94 3.29 4.17
-158.0 160.8 1537 1624
t-stat 1.06
Forecasting Trip-Level Parking Demand Summary Statistics of PSRC Trip Attributes Trip Attribute Activity: Work Activity: School (K-12) Activity: College Activity: Eating out Activity: Personal business Activity: Everyday shopping Activity: Major shopping Activity: Religious/community Activity: Social Activity: Recreation-participate Activity: Recreation-watch Activity: Accompany someone else Activity: Pick up/drop off Activity: Turn around Vehicle: Car Vehicle: SUV Vehicle: Van Vehicle: Truck
Avg Parking Duration (min. per trip) 379.7 338.8 222.5 46.1 46.8 27.7 47.6 116.8 127.6 103.5 107.4 58.8 15.5 53.0 147.2 133.2 103.3 173.7
Trip Attribute Parking duration (min/trip) Trip distance (miles) Passengers (excluding driver)
Mean
Std. Dev.
Min
Max
142.0
199.5
15.0
2120
6.71
7.14
0.23 0
67.6
0.421
0.811
0
6
OLS Results for Parking Durations/Trip Y = Parking Duration of Trip (min/trip) Variable Constant Activity: Work (base case) Activity: School (K-12) Activity: College Activity: Eating out Activity: Personal business Activity: Everyday shopping Activity: Major shopping Activity: Religious/community Activity: Social Activity: Recreation-participate Activity: Recreation-watch Activity: Accompany someone else Activity: Pick up/drop off Activity: Turn around Number of Observations Adjusted R-squared
Parameter Estimate Standard. Coef. 372.2 -21.28 -0.009 -157.6 -0.069 -306.6 -0.396 -313.2 -0.603 -324.5 -0.609 -308.0 -0.196 -246.3 -0.170 -241.9 -0.238 -259.7 -0.302 -254.2 -0.161 -298.9 -0.141 -344.1 -0.587 -303.8 -0.102 30,085 Parking durations 0.590
t-stat 125.5 -2.37 -18.4 -95.0 -135.6 -133.9 -51.5 -44.6 -60.7 -75.6 -41.8 -37.2 -127.3 -27.5
Anticipating Best Sites for Charging Stations Mixed Integer Programming (MIP) in GAMS: Minimizes total access cost as function of walk distances (cij) weighted by parking demand duration (yij).
Parking demand (di) in zone i is met by station in zone j. Ensures all demand is met. Total # of stations won’t exceed budgeted #of stations (L). Ensures minimum spacing (spread) between stations. Ensures met parking demands are non-negative. Indicator (xj) is 1 when station is assigned to zone j; 0 otherwise. Indicator (δij) is 1 if minimum spacing (r) is met; 0 otherwise.
Application: Optimal Station Locations in Seattle • 218 TAZs (i & j =218) • Number of charging stations limited to L=20 • Minimum station spacing r=1 mile
Application (2) Selected TAZs 13 19 30 31 37 39 53 56 65 70 90 93 101 152 153 162 174 180 199 209
Own-Zone Demand (mins) 2232 69 6449 2311 552 771 1481 749 118 8959 5063 678 2445 376 0 341 2466 421 603 593
Demand Rank (out of 218) 20 167 2 19 93 74 40 76 152 1 7 77 18 109 207 113 17 106 85 87
Comparing MIP Results to Simple Assignment Optimal Solution
Top 20 Zones by Demand
127,510 mile-minutes
132,762 mile-minutes
Average parking access distance
0.69 miles
0.72 miles
Maximum parking access distance
1.53 miles
3.80 miles
94.5%
79.6%
Total Cost (z)
% of drivers accessing parking within 1 mile
Conclusions Parking demand at the zone level… ▫ Rises significantly with job & student densities. ▫ Is higher in more connected networks & transitserved zones.
Parking duration at the trip level… ▫ Is influenced most by trip purpose, with work & school trips having longest durations. ▫ Also rises with trip distance & vehicle type (cars park longer than trucks, SUVs, & vans).
Conclusions (2) This study provides a basic framework for MPOs, cities, & energy providers to… ▫ Anticipate parking demand, & ▫ Efficiently locate EV charging infrastructure in new settings subject to a variety of constraints.
Potential extensions include… ▫ Determining optimal capacity at each charging station ▫ Behavioral models to anticipate type of vehicles parked in zones, & ▫ Sensitivity tests to determine fluctuations in charging station zone selection due to changes in inputs.
References Frade, I., Ribeiro, A., Goncalves, G. & Antunes, A.P. (2011) Optimal Location of Charging Stations for Electric Vehicles in a Neighborhood in Lisbon, Portugal. In Transportation Research Record: Journal of the Transportation Research Board, No. 2252: 91-98. Hanabusa, H. & Horiguchi, R. (2011) Lecture Notes in Computer Science in Knowledge-based & Intelligent Information & Engineering Systems, Volume 6883/2011, pp. 596-605. Sweda, T. & Klabjan, D. (2011) An Agent-Based Decision Support System for Electric Vehicle Charging Infrastructure Deployment. 7th IEEE Vehicle Power & Propulsion Conference, Chicago, Illinois. Wang, H., Huang, Q., Zhang, C. & Xia, A. (2010) A Novel Approach for the Layout of Electric Vehicle Charging Station. Apperceiving Computing & Intelligence Analysis Conference.
Thank you for your time! Questions?
Overview • • • •
Motivation Existing research Puget Sound region data Study methodology
Photo by Daniel Reese for KUT News
▫ Forecasting zone-level parking demand ▫ Forecasting trip-level parking demand ▫ Locating optimal charging infrastructure (Seattle application)
• Conclusions
Background • Trend: Increased number of EVs in the market ▫ Higher fuel economy requirements ▫ Higher fuel costs
• Caveat: Range anxiety for EV owners/potential buyers affects ▫ ▫ ▫ ▫
EV adoption rates Electrified mile shares Petroleum demand Power consumption across times of day
• Potential solution: Public charging station provision
Background • Charging station installation costs ▫ $3,000 to $15,000 per station using existing infrastructure ▫ $40,000+ per station when requiring local infrastructure upgrade
• Energy providers, cities, & MPOs need a methodology to optimally locate public charging stations that… ▫ Serve charging demand, ▫ Minimize access costs for EV drivers, ▫ Meet constrained budgets.
Existing Research • Wang et al.’s (2010) numerical method for Chengdu, China used distribution of gas-station demands as proxy for charging demands. • Sweda & Klabjan’s (2011) Chicago case study uses an agent-based model to identify residential patterns of EV ownership & driving activities to identify strategic station locations. • Frade et al.’s (2011) coverage model for Lisbon seeks locations that maximize match for charging demands. • Our work most closely tracks that of Hanabusa & Horiguchi (2011): minimizing driver walk costs while ensuring coverage (via buffer distances between stations, in Seattle).
Our Approach • Behavioral models calibrated to predict when & where EVs are likely parked. ▫ Zone-level parking demand based on land use attributes of destination zones. ▫ Trip-level parking demand based on individual trip characteristics.
• Optimization routine (MIP) identifies charging station locations in order to… ▫ Minimize station access penalties for EV drivers, while… ▫ Satisfying budget constraints, & ▫ Ensuring minimum station spacing requirements.
The Data • Puget Sound Regional Council’s (PSRC) 2006 household travel survey data. ▫ 4 Washington Counties (King, Kitsap, Pierce & Snohomish) ▫ 4,741 households ▫ 10,510 individuals ▫ 3,700 traffic analysis zones (TAZs) ▫ 1,177,140 parcels ▫ 87,600 total person-trips ▫ 48,789 trips by light-duty vehicles
Seattle’s 3,700 Traffic Analysis Zones
Summary Statistics of PSRC Data Mean
St Dev.
Min
Max
Age (years)
41.9
21.8
0
99
Male Indicator
0.47
0.50
0
1
Driver’s License Indicator
0.78
0.42
0
1
0.21
0.4
0
1
Household Size
2.22
1.21
1
8
#Workers in Household
1.13
0.85
0
5
#Vehicles in Household
1.89
1.07
0
10
$71,400
$42,300
$5,000
$175,000
Person Records (N=10,510)
Student Indicator Household Records (N=4,741)
Annual Household Income
Determining Parking Location & Duration • Starting with 48,789 trips (55.7% of all persontrips). ▫ Only trips ending away from one’s home parcel are counted. ▫ Only trips with adequate parking duration (at a single parcel) are counted (at least 15 minutes), to help ensure adequate time for charging.
• 30,085 candidate parking periods emerged.
Forecasting Zone-Level Parking Demand Summary Statistics of PSRC Zone Attributes Variable (n = 3962) Parking duration (mins/mile2) Population density (persons/mile2) Employment density (jobs/mile2) Student density (students/mile2) Housing density (units/mile2) Average price of daily paid parking ($) Average price of hourly paid parking ($) 3-way intersections (1/2 mile radius) 4-way intersections (1/2 mile radius) Express bus stops (1/4 mile radius) Bus stops (1/4 mile radius)
Mean 2.41E+04 7.99E+03 1.26E+04 1.64E+03 3.43E+03 $0.145 $0.066 45.8 36.8 2.25 6.21
Std. Dev. 1.18E+05 1.88E+04 8.27E+04 2.48E+04 8.20E+04 $1.027 $0.465 22.52 45.91 6.369 9.016
Min 0 0 0 0 0 0 0 0 0 0 0
Max 2.43E+06 2.91E+05 2.07E+06 1.02E+06 1.27E+05 $21.3 $11.0 119.3 251.8 55.6 69.6
OLS Results for Total Demand/SqMile Y= Total Parking Demand in Zone (minutes/mile2) Variable Constant Density Population density (residents/mile2) Employment density (jobs/mile2) Student density (students/mile2) Parking Prices (within ¼ mile) Average price of daily parking ($) Transit Access & Network Connectivity #3-way intersections (within ½ mile) #4-way intersections (within ½ mile) #Express bus stops (within ¼ mile) #Bus stops (within ¼ mile) Number of Observations Adjusted R-squared
Parameter Estimate 3268
Standardized Coef.
-0.294 0.583 0.226
-0.047 0.408 0.047
-3.50 27.0 4.11
2.22
0.193
11.0
-0.030 0.062 0.083 0.124 3,692 TAZs 0.521
-2.41 2.94 3.29 4.17
-158.0 160.8 1537 1624
t-stat 1.06
Forecasting Trip-Level Parking Demand Summary Statistics of PSRC Trip Attributes Trip Attribute Activity: Work Activity: School (K-12) Activity: College Activity: Eating out Activity: Personal business Activity: Everyday shopping Activity: Major shopping Activity: Religious/community Activity: Social Activity: Recreation-participate Activity: Recreation-watch Activity: Accompany someone else Activity: Pick up/drop off Activity: Turn around Vehicle: Car Vehicle: SUV Vehicle: Van Vehicle: Truck
Avg Parking Duration (min. per trip) 379.7 338.8 222.5 46.1 46.8 27.7 47.6 116.8 127.6 103.5 107.4 58.8 15.5 53.0 147.2 133.2 103.3 173.7
Trip Attribute Parking duration (min/trip) Trip distance (miles) Passengers (excluding driver)
Mean
Std. Dev.
Min
Max
142.0
199.5
15.0
2120
6.71
7.14
0.23 0
67.6
0.421
0.811
0
6
OLS Results for Parking Durations/Trip Y = Parking Duration of Trip (min/trip) Variable Constant Activity: Work (base case) Activity: School (K-12) Activity: College Activity: Eating out Activity: Personal business Activity: Everyday shopping Activity: Major shopping Activity: Religious/community Activity: Social Activity: Recreation-participate Activity: Recreation-watch Activity: Accompany someone else Activity: Pick up/drop off Activity: Turn around Number of Observations Adjusted R-squared
Parameter Estimate Standard. Coef. 372.2 -21.28 -0.009 -157.6 -0.069 -306.6 -0.396 -313.2 -0.603 -324.5 -0.609 -308.0 -0.196 -246.3 -0.170 -241.9 -0.238 -259.7 -0.302 -254.2 -0.161 -298.9 -0.141 -344.1 -0.587 -303.8 -0.102 30,085 Parking durations 0.590
t-stat 125.5 -2.37 -18.4 -95.0 -135.6 -133.9 -51.5 -44.6 -60.7 -75.6 -41.8 -37.2 -127.3 -27.5
Anticipating Best Sites for Charging Stations Mixed Integer Programming (MIP) in GAMS: Minimizes total access cost as function of walk distances (cij) weighted by parking demand duration (yij).
Parking demand (di) in zone i is met by station in zone j. Ensures all demand is met. Total # of stations won’t exceed budgeted #of stations (L). Ensures minimum spacing (spread) between stations. Ensures met parking demands are non-negative. Indicator (xj) is 1 when station is assigned to zone j; 0 otherwise. Indicator (δij) is 1 if minimum spacing (r) is met; 0 otherwise.
Application: Optimal Station Locations in Seattle • 218 TAZs (i & j =218) • Number of charging stations limited to L=20 • Minimum station spacing r=1 mile
Application (2) Selected TAZs 13 19 30 31 37 39 53 56 65 70 90 93 101 152 153 162 174 180 199 209
Own-Zone Demand (mins) 2232 69 6449 2311 552 771 1481 749 118 8959 5063 678 2445 376 0 341 2466 421 603 593
Demand Rank (out of 218) 20 167 2 19 93 74 40 76 152 1 7 77 18 109 207 113 17 106 85 87
Comparing MIP Results to Simple Assignment Optimal Solution
Top 20 Zones by Demand
127,510 mile-minutes
132,762 mile-minutes
Average parking access distance
0.69 miles
0.72 miles
Maximum parking access distance
1.53 miles
3.80 miles
94.5%
79.6%
Total Cost (z)
% of drivers accessing parking within 1 mile
Conclusions Parking demand at the zone level… ▫ Rises significantly with job & student densities. ▫ Is higher in more connected networks & transitserved zones.
Parking duration at the trip level… ▫ Is influenced most by trip purpose, with work & school trips having longest durations. ▫ Also rises with trip distance & vehicle type (cars park longer than trucks, SUVs, & vans).
Conclusions (2) This study provides a basic framework for MPOs, cities, & energy providers to… ▫ Anticipate parking demand, & ▫ Efficiently locate EV charging infrastructure in new settings subject to a variety of constraints.
Potential extensions include… ▫ Determining optimal capacity at each charging station ▫ Behavioral models to anticipate type of vehicles parked in zones, & ▫ Sensitivity tests to determine fluctuations in charging station zone selection due to changes in inputs.
References Frade, I., Ribeiro, A., Goncalves, G. & Antunes, A.P. (2011) Optimal Location of Charging Stations for Electric Vehicles in a Neighborhood in Lisbon, Portugal. In Transportation Research Record: Journal of the Transportation Research Board, No. 2252: 91-98. Hanabusa, H. & Horiguchi, R. (2011) Lecture Notes in Computer Science in Knowledge-based & Intelligent Information & Engineering Systems, Volume 6883/2011, pp. 596-605. Sweda, T. & Klabjan, D. (2011) An Agent-Based Decision Support System for Electric Vehicle Charging Infrastructure Deployment. 7th IEEE Vehicle Power & Propulsion Conference, Chicago, Illinois. Wang, H., Huang, Q., Zhang, C. & Xia, A. (2010) A Novel Approach for the Layout of Electric Vehicle Charging Station. Apperceiving Computing & Intelligence Analysis Conference.
Thank you for your time! Questions?
