Westport adds natural gas F-150 to lineup
Irving Oil and TransCanada to form JV to develop new Saint John Marine Terminal for Energy East pipeline

“Big data” analysis of Beijing taxi fleet suggests maximum electrification subsidy benefit from targeting medium-range plug-in hybrids

A pair of researchers at the University of Michigan have used “big data” mining techniques to evaluate the impact of adopting plug-in electric vehicles (PEVs) in the Beijing taxi fleet on life cycle greenhouse gas emissions based on the characterized individual travel patterns.

Although the results are based on a specific public fleet, the study demonstrates the benefit of using large-scale individual-based trajectory data (big data) to better understand environmental implications of fleet electrification and inform better decision making, Hua Cai and Ming Xu suggested in a paper published in the ACS journal Environmental Science & Technology. This research represents the first of a series of studies exploring the role of big data in environmental systems analysis for the emerging PHEV/BEV systems.

While PHEV/BEV technology has developed rapidly in recent years, there exist great uncertainties in terms of market acceptance. Previous studies evaluated factors that impact PHEV/BEV adoption, including infrastructure support, economies of scale, word of mouth effects (influence from other people’s perception of electric vehicles), age of current vehicle, consumer income, and travel patterns.

In particular, consumer travel patterns (i.e., travel behavior) have increasingly received significant attention, because they directly determine whether PHEV/BEV is acceptable to consumers and how it is utilized for daily travels. However, previous research has predominately used aggregated travel pattern data, such as the often cited National Household Travel Survey (NHTS), which assumes that everyone follows the same travel pattern as the aggregated average and neglects the heterogeneity of individual users and their specific travel patterns. Recent attempts to differentiate the impacts of individual travel patterns on PHEV/BEV market acceptance have also been constrained by the size of travel pattern samples (usually in the dozens or hundreds) due to the difficulty in collecting travel behavior data from the private fleet.

Fortunately, the rapid development of information and communications technology has increasingly made a massive amount of travel behavior data available at a much larger scale. The availability of these “big data” (commonly referring to large-scale data sets) on individual travel patterns, especially for public fleets, represents untapped opportunities to better understand how individual travel behavior affects the PHEV/ BEV market acceptance and the associated environmental impacts.

—Cai and Xu

In their study, Cai and Xu used a large-scale data set containing real-time trajectories of 10,375 taxis in Beijing (approximately 15% of the taxi fleet) retrieved by GPS systems for one week to explore the impacts of individual travel patterns to PHEV acceptance and associated implications on GHG emissions.

Each data point includes a unique taxi ID, the time (to the seconds) of the recording, and the position (longitude and latitude) of the taxi at the specific time. Depending on the GPS device setup in each vehicle, the frequency of recording ranges from five seconds to ten minutes—but stays consistent for the same vehicle.

To clean up the data, they applied a filter to eliminate 1) empty data points; 2) duplicate data points; 3) taxis with less than seven data points; and 4) unreasonable off-the-chart positions.

Because taxis do not have uniformly regular parking times—and some drivers pair up to drive the same taxi in rotation to minimize costs—“daily driving distance” is not a good metric to characterize taxi trips, the team noted. Taxis may have significantly different starting and ending times of each “day” and the length of a “day” may also be different from taxi to taxi (e.g., one-driver taxi versus two-driver taxi).

To address this issue, Cai and Xu introduced the concept of “driving segments”: the total distance driven between two major resting periods when the vehicle is parked with a predetermined length threshold. One segment may contain several separate trips, which is similar to “trip chains” used in previous studies. Resting periods between segments represent charging opportunities.

For the study, they ranged the predetermined resting threshold from 30 min to eight hours to test the impact of charging opportunities on PHEVs adoption. In the journal paper, they focused on two extreme cases: the “home-charging only” scenario and the “ubiquitous charging” scenario. The home-charging only scenario represents a relatively conservative case that vehicles can only be charged at home thus requiring a longer resting period eight hours in this study). The ubiquitous charging scenario represents an optimistic case that public charging stations are ubiquitously available, allowing drivers to charge their vehicles as long as they have more than half an hour of resting time.

Among the findings of the study were:

  • The largest gasoline displacement (1.1 million gallons per year) can be achieved by adopting PHEVs with modest electric range (approximately 80 miles) with current battery cost, limited public charging infrastructure, and no government subsidy.

  • While the battery range is one of the major concerns from consumers’ perception, the study shows that a larger battery actually decreases the VMT electrification rate when the unit battery cost is higher than $200/kWh. Only when the unit battery cost is lower than $200/kWh, extended electric drive range can increase the adoption and thus electrification rate.

  • Government subsidies can be more effective to increase the VMT electrification rate and gasoline displacement if targeted to PHEVs with modest electric ranges (80 to 120 miles).

  • Beijing belongs to the North China Grid with a GHG emission factor of 236.7 g CO2-eq/km traveled. GHG emissions of conventional gasoline vehicles are approximately 224.4 g CO2-eq/km. While taxi fleet electrification can increase greenhouse gas emissions by up to 115 kiloton CO2-eq per year with the current grid in Beijing, emission reduction of up to 36.5 kiloton CO2-eq per year can be achieved if the fuel cycle emission factor of electricity can be reduced to 168.7 g/km.

  • At current battery cost (approximately $500/kWh), a larger battery does not necessarily imply higher rate of adoption, utilization, and electrification of PHEVs due to the heterogeneous individual travel patterns.

  • Subsidy can effectively increase the VMT electrification rate by filling the gap between fuel cost savings and the premium cost of PHEVs. Results show that focusing on PHEVs with modest electric ranges (80 to 120 miles) can most efficiently boost VMT electrification with a fixed amount of budget.

Our study demonstrates how individual travel patterns, charging opportunities, and battery size influence life cycle GHG emissions due to PHEV adoption and utilization at the individual vehicle level. It also sheds lights on the utilization of large-scale vehicle trajectory data for enhancing assessment of environmental impacts of PHEV/BEVs.

—Cai and Xu


  • Hua Cai and Ming Xu (2013) Greenhouse Gas Implications of Fleet Electrification Based on Big Data-Informed Individual Travel Patterns. Environmental Science & Technology doi: 10.1021/es401008f



I would not call 80 miles EV range "modest" for a PHEV.
It is more like a Nissan Leaf's battery range.

I have no idea where they get the figures for the GHG for the north China grid - it seems very high, but I suppose most of the power comes from coal.

The fact that taxis tend to be driven all day, and to spend a lot of time waiting at many different locations, and crawling along in queues, suggests that you might be as well off with HEVs as PHEVs for less cost.

I suppose electrifying taxis for non-ghg pollution makes sense, PHEVs will have much lower soot, NOx, HCs etc when running in electric mode, which can only be a good thing in a polluted city like Beijing.

The question of how to keep PHEVs topped up by taximen remains a tricky problem.

(Maybe you use a "moving head queue system where the taxis drive into a queue and plug in. The notion of the head of the queue moves towards the taxis, rather than the other way round. If you could police it, it would work.
[ It would also work for ICE taxis, as they could switch their engines off till they picked up a fare. ]

In a big taxi rank, you could have a man indicate the head of the queue, in a smaller one, it might not be worth bothering.


Wireless charging in taxi ranks is a no-brainer.


@Dave, how accurately do you have to position the car ?

You would need 2 queues - one for the EVs getting charged, one for the other cars (applies to both charge/queue systems).

Starts getting complex if you mix charging and queuing - maybe better keep them separate.


It's not particularly difficult to arrange mobile charging cords along a curb.  All you need is a cable going around a couple of sheaves at each end to pull the cords along with the vehicles and recirculate them to the beginning.  Electric can be supplied through sliding brushes inside the support rail.  Mechanical support is just rollers.

If you really wanted to get fancy, you could use a car-wash pusher system to move the cars (current J1772 spec is to shut down the drive systems when they're plugged in) and build auto-eject and cord retract in for when the car reaches the end of the queue.


The receiver in the car can communicate with the pad in the road, so that it is easy to position over the pad.


Another option would be a two stage taxi rank. The first stage would be a holding area where the taxi could charge: upon entering they get a queue number. When their number is called they enter the second stage which operates like a normal taxi rank. Can use normal charges and the customer experience is the same - the complexity is in the queue ticketing system.

I don't understand the paper. The largest gasoline displacement is to have 100% BEV in an unconstrained problem. So what are they constraining - as the answer is dependent upon that. Previous analyses with the average trip and the constraint of cost came up with the answer of HEV.


This is not a surprise. BEVs, PHEVs and even HEVs are more efficient than ICEVs and produce less GHG and other pollution.

The comments to this entry are closed.