Redistributing fleets based on new control totals

When the user has data on the expected joint distribution of the fleet across multiple categories, including Vehicle class, power train, ful and its vintage, as well as the distribution of the existing fleet, it is possible to use the Vehicle Redistribution procedure to run an Iterative-Proportional fitting over the original distribution to match targets, including the total fleet for each census tract in the modeled area.

Imports

from os.path import join
import os
from pathlib import Path
from tempfile import gettempdir

import pandas as pd

from polaris.prepare.vehicle_distribution_updater.vehicle_distribution_updater import RedistributeVehicles

Input data

Let’s take a look at the input data

pth = Path(os.getcwd()).resolve(True).parent / "data" / "vehicle_redistribution"

• Vehicle file

Note that this file is TAB separated

pd.read_csv(pth / "vehicle_distribution_chicago.txt", delimiter="\t").head()

	TRACT	POLARIS_ID	VEHICLE_CLASS	FUEL	POWERTRAIN	VINTAGE	PROPORTION
0	17007010100	1000	CAR_COMPACT	Elec	BEV	0	0.003972
1	17007010100	1009	CAR_COMPACT	Diesel	Conventional	0	0.000000
2	17007010100	1010	CAR_COMPACT	Diesel	Conventional	1	0.000000
3	17007010100	1011	CAR_COMPACT	Diesel	Conventional	2	0.000000
4	17007010100	1012	CAR_COMPACT	Gas	Conventional	0	0.006155

• Target file

pd.read_csv(pth / "target_2040_low.csv").head()

	VEHICLE_CLASS	POWERTRAIN	FUEL	VINTAGE	PROPORTION
0	CAR_COMPACT	Conventional	Gas	0	0.01755
1	CAR_COMPACT	Conventional	Gas	1	0.01750
2	CAR_COMPACT	Conventional	Gas	2	0.01830
3	CAR_COMPACT	Conventional	Diesel	0	0.00000
4	CAR_COMPACT	Conventional	Diesel	1	0.00000

• Vehicle codes

pd.read_csv(pth / "polaris_vehicle_codes.csv").head()

	POLARIS_ID	VEHICLE_CLASS	POWERTRAIN	FUEL	AUTOMATION	CONNECTIVITY	VINTAGE
0	0	DEFAULT	BEV	Elec	NaN	No	0
1	1	DEFAULT	BEV	Elec	NaN	No	1
2	2	DEFAULT	BEV	Elec	NaN	No	2
3	3	DEFAULT	BEV_short	Elec	NaN	No	0
4	4	DEFAULT	BEV_short	Elec	NaN	No	1

• Zone weights

pd.read_csv(pth / "veh_by_zone_chicago.csv").head()

	TRACT	veh_count
0	17007010100	1353
1	17007010200	991
2	17007010300	1015
3	17007010400	1303
4	17007010500	2183

All our files need to be inside the same folder (pth, below)

rv = RedistributeVehicles(
    model_dir=pth,
    veh_file="vehicle_distribution_chicago.txt",
    target_file="target_2040_low.csv",
    veh_codes_file="polaris_vehicle_codes.csv",
    zone_weights="veh_by_zone_chicago.csv",
    fleet_mode=False,
)

Processing and saving results

The convergence threshold can be as low as required, as it is nearly guaranteed to converge

rv.process(conv_threshold=0.001, max_iterations=50)

# Two convergence scatter plots are also saved with the outputs for a quick sanity check
rv.save_results(join(gettempdir(), "veh_distr.csv"))

• 
• 

Iteration 1: 0.0s , error 0.01668
Iteration 2: 0.0s , error 0.00136
Iteration 3: 0.0s , error 0.00011
Total processing time: 0.1s

Looking at the outputs

import pandas as pd

df = pd.read_csv(join(gettempdir(), "veh_distr.csv"))
df.head()

	TRACT	POLARIS_ID	VEHICLE_CLASS	FUEL	POWERTRAIN	VINTAGE	PROPORTION
0	17007010100	1034	CAR_COMPACT	Gas	ISG	1	0.013509
1	17007010100	1035	CAR_COMPACT	Gas	ISG	2	0.014122
2	17007010100	1042	CAR_COMPACT	Gas	PHEV	0	0.000844
3	17007010100	1043	CAR_COMPACT	Gas	PHEV	1	0.000844
4	17007010100	1044	CAR_COMPACT	Gas	PHEV	2	0.000885