A large-scale empirical study on impacting factors of taxi charging station utilization
Introduction
With innovations in battery technology and increasing concerns about environmental problems, electric vehicles (EVs) have attracted increasing attention in recent years (Morrissey et al., 2016a). Many cities are actively promoting the electrification of their public transport systems (Tu et al., 2016), for instance, electric taxis (e-taxis) and buses. Many charging stations (CSs) need to be built to satisfy the increasing charging demand. Especially for e-taxis, they represent a large percentage of urban charging demand (Yang et al., 2018), and their daily operation exhibits complex spatial–temporal dynamics (Wong et al., 2014, Qian and Ukkusuri, 2015). Furthermore, charging stations occupy a large number of parking facilities, which is a limited resource in many cities. In this context, it is essential to understand what factors affect charging facility utilization in more significant detail. This can support decision-making concerning the placement of charging stations and potentially achieve better utilization across urban areas.
There are two major gaps in the literature. First, there are few studies that systematically and empirically explore the impact factors of charging station utilization. In fact, good knowledge of what affects charging station utilization is a prerequisite for a successful planning scheme. Second, previous studies generally predefine a relationship between CS utilization and impacting factors (Wagner et al., 2014, Wolbertus et al., 2018, Olk et al., 2019, Mortimer et al., 2022). However, such a linear relationship may only be effective within a certain range of impacting factors (Tao et al., 2020).
The objectives of this study are to explore the impacting factors of charging station utilization and investigate the potential non-linear relationship between them. To this end, in this study, we use a Random Forest Regression (RFR) model to explore the factors influencing charging station utilization using data sets collected from Shenzhen, a city with the largest e-taxi fleet in the world. Furthermore, a SHapley Additive exPlanations (SHAP) value method (Lundberg and Lee, 2017) is applied in the RFR to explain the non-linear relationships between charging station utilization and impacting factors (e.g., built environment and taxi demand).
The study contributes in three aspects to the literature and practice. From the empirical perspective, it enriches the understanding of the relationship between collective factors and charging station utilization. By examining the non-linear relationship and threshold effects, the study provides a further understanding of how the factors effectively affect charging station utilization. From the methodology perspective, a random forest model, explained by the SHAP value method, is applied to capture complicated relationships between charging station utilization and explanatory variables. The methodology is proved to be more effective in revealing non-linear relationships than conventional linear regression models. From the perspective of planning practice, the study quantifies the relative importance of factors and gives guidance on prioritizing factors with tense land resources and limited planning alternatives. The study also offers implications for policy interventions to stimulate higher and more balanced e-taxi utilization at charging stations.
The remainder of the article is organized as follows. In the next section, we review the literature on the utilization of charging stations and the impacting factors analysis and identify research gaps. Section 3 illustrates the study area and data preparation and introduces the dependent and explanatory variables used in the model. Section 4 explains the methodology of the study. The results and discussions are provided in Section 5. In Section 6, we summarize the key results and discuss the associated policy implications.
Section snippets
Related work on utilization of charging stations
A voluminous body of literature has been published regarding the utilization of charging stations. These studies have paid particular attention to the following aspects: (1) Maximizing charging station utilization. This aspect can be divided into 3 approaches, namely optimizing the location and size of charging stations (Yang et al., 2017, Asna et al., 2021), optimizing real-time charging price (Wolbertus and Gerzon, 2018, Li et al., 2020) and coordinating charging demand (Zeng et al., 2019,
Study area
In this work, we select Shenzhen as the study area. As an economically dynamic city of more than 13 million residents, Shenzhen is one of the pioneers in promoting electric vehicles worldwide. The taxi electrification process in Shenzhen dates to 2010. At the end of 2018, there were more than 20 000 e-taxis (which represents 99.06% of the total number of taxis), making Shenzhen the city with the largest e-taxi fleet in the world. Many charging stations have been constructed to meet the rapidly
Linear regression model
In this study, we first use a linear regression (LR) model to interpret the relationship between utilization and explanatory variables. The linear regression takes the form: where is the charging station utilization, is the coefficient vector of the explanatory variables, is the vector of the standardized explanatory variables, is an intercept term. The result of collinearity statistics shows strong collinearity between pick-up and drop-off
Feature importance analysis
This section uses the SHAP method to explore the relationship between utilization rate and selected features. The association is illustrated in the SHAP summary graph as Fig. 7, which combines the feature importance with their effects. The summary plot sorts the features on the -axis according to the feature importance and the value of the feature on the -axis according to the Shapley value. The -axis units are the same units as the model output. The points in the summary plot represent
Conclusions
Study efforts. Understanding the factors affecting charging station usage is the basis for a successful planning scheme. However, previous studies have neglected comprehensive investigation of these factors. In this work, we systematically and empirically explored the factors affecting the utilization of charging stations. To this end, a random forest regression model is constructed based on extracting variables from large-scale detailed data sets. The Shapley value method is then used to
Acknowledgments
This research work was supported by the China Scholarship Council , and we express our gratitude to Wu Li, who helped us finish this work. The contents of this paper reflect the views of the authors responsible for the facts and the accuracy of the data presented herein. This paper does not constitute a standard, specification, or regulation.
References (44)
OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks
Comput. Environ. Urban Syst.
(2017)- et al.
Mixed land use and walkability: Variations in land use measures and relationships with BMI, overweight, and obesity
Health Place
(2009) - et al.
Examining non-linear built environment effects on elderly’s walking: A random forest approach
Transp. Res. D
(2020) - et al.
Data-driven smart charging for heterogeneous electric vehicle fleets
Energy AI
(2020) - et al.
Modeling determinants of ridesourcing usage: A census tract-level analysis of Chicago
Transp. Res. C
(2020) - et al.
Modeling the preference of electric shared mobility drivers in choosing charging stations
Transp. Res. D
(2022) - et al.
Stochastic dynamic itinerary interception refueling location problem with queue delay for electric taxi charging stations
Transp. Res. C
(2014) - et al.
Increasing effectiveness of the transportation network by using the automation of a Voronoi diagram
Transp. Res. Procedia
(2018) - et al.
Exploring electric vehicle charging patterns: Mixed usage of charging infrastructure
Transp. Res. D
(2020) - et al.
Understanding charging dynamics of fully-electrified taxi services using large-scale trajectory data
Transp. Res. C
(2022)
Investigating urban metro stations as cognitive places in cities using points of interest
Cities
Failure mode and effects analysis of RC members based on machine-learning-based SHapley Additive exPlanations (SHAP) approach
Eng. Struct.
Future standard and fast charging infrastructure planning: An analysis of electric vehicle charging behaviour
Energy Policy
Spatial variation of the urban taxi ridership using GPS data
Appl. Geogr.
Exploring the non-linear associations between spatial attributes and walking distance to transit
J. Transp. Geogr.
Optimizing the locations of electric taxi charging stations: A spatial–temporal demand coverage approach
Transp. Res. C
Fully charged: An empirical study into the factors that influence connection times at EV-charging stations
Energy Policy
A cell-based logit-opportunity taxi customer-search model
Transp. Res. C
To walk or not to walk? Examining non-linear effects of streetscape greenery on walking propensity of older adults
J. Transp. Geogr.
A data-driven optimization-based approach for siting and sizing of electric taxi charging stations
Transp. Res. C
Analysis of an optimal planning model for electric vehicle fast-charging stations in Al Ain City, United Arab Emirates
IEEE Access
Random forests
Mach. Learn.
Cited by (6)
Understanding the daily operations of electric taxis from macro-patterns to micro-behaviors
2024, Transportation Research Part D: Transport and EnvironmentAtmospheric water demand constrains net ecosystem production in subtropical mangrove forests
2024, Journal of HydrologyElectrification pathways for public transport systems
2024, Transportation Research Part D: Transport and EnvironmentBook-ahead ride-hailing trip and its determinants: Findings from large-scale trip records in China
2023, Transportation Research Part A: Policy and Practice