The Usage of Big Data in Electric Vehicle Charging: A Comprehensive Review

Abstract

With major effects on power grids and people’s lifestyles, the quick uptake of electric vehicles (EVs) poses serious problems for the robustness of charging infrastructure. By enabling spatiotemporally optimal charging strategies that optimize grid operations, big data technologies provide game-changing solutions. In order to solve the following issues, this paper summarizes state-of-the-art applications of EV charging big data, which are derived from vehicles, charging stations, and power grids: (1) optimized control of grid operation; (2) charging infrastructure layout; (3) battery development; and (4) safety of charging equipment. Future research opportunities include: (1) deep integration of intelligent transportation and smart grids; (2) renewable energy and intelligent energy management optimization; (3) synergizing smart homes with EVs; and (4) AI for energy demand forecasting and automated management. This study establishes big data as a pivotal tool for low-carbon EV transition, providing actionable frameworks for researchers and policymakers to harmonize electrified transport with energy sustainability goals.

1. Introduction

The intensification of the global warming crisis has made carbon neutrality a central goal, with the transport sector as the main source of emissions, accounting for more than 21.2% of energy-related carbon emissions [1]. Electric vehicles (EVs), capable of reducing lifecycle emissions by 23 tonnes per vehicle and cumulatively 1.8 billion tonnes by 2035 [2], face adoption barriers: 59% of consumers cite inadequate charging infrastructure planning [3], compounded by long charging times [4,5], unsatisfactory battery performance [6], and inefficient power resource scheduling [7]. These challenges stem from the multi-dimensional complexity of the EV charging ecosystem—defined in this paper as an integrated system centered on EV charging demands, connecting four key entities (drivers, EVs, charging infrastructure, power grids) and relying on multi-source data integration. Such complexity arises from its integration of multi-source data: driver behavior [8,9,10], vehicle battery status [11,12], charging infrastructure operations [13], and grid load dynamics [14,15]. However, the rational use of big data in ecosystems offers a beacon of hope, enabling predictive modeling [16,17], optimization of operations [18,19], and management of equipment [20]. This potential of big data to revolutionize the EV charging ecosystem is a promising development that should inspire optimism about the future of EVs.

The core of integrating big data technology into charging ecosystems lies in analyzing its role between drivers, EVs, infrastructure, and the power grid. From the perspective of drivers, high-quality charging services can attract their participation [21,22,23]. Big data technology can analyze the historical data of drivers to obtain their activity preferences, thereby providing personalized charging services [23]. Analyzing household travel data [23], travel and parking, charging information [24], and charging session data [25] provides a basis for the rational layout and configuration of charging facilities, thereby promoting the realization of high-quality charging services.

The purpose of high-quality charging services is to enhance the user experience. Improving user experience can also start with understanding the remaining charging time of EVs [26]. Shi et al. [26] used a residual charging time algorithm to analyze EVs’ historical and real-time charging data. The proposed method reduced the estimation error of remaining charging time by 73.6–84.4 compared to traditional methods, achieving errors of 2.0165 min and 2.735 min in constant current and constant voltage charging stages, respectively. Big data technology can also analyze current, voltage, and charging duration data to detect the battery’s health status [27]. Correct battery remaining time estimation and safety guarantee can help drivers save time.

In addition to relying on proper control of the remaining charging time for EVs, a stable and efficient charging infrastructure can also improve users’ charging experience. The deployment location, charging technology type, charging convenience, charging waiting time, charger power, and capacity of charging infrastructure affect whether drivers can complete the charging process smoothly and quickly, thereby affecting user experience and market acceptance of EVs [28,29]. Big data technology can measure charging demand indicators, thereby revealing the spatial relationship between charging demand and infrastructure and assisting in infrastructure layout planning [30]; it can also filter useful information from massive operational data to identify data features that can characterize the health status of charging devices, construct fault diagnosis models to detect the operational status of charging devices [20], and help improve user experience.

The power grid is responsible for providing stable power support for charging infrastructure. With the increase in EVs, the power grid faces tremendous pressure, especially during peak hours. From the power grid perspective, EV charging demand growth represents a significant threat to grid stability [31,32,33]. Because the charging demand for EVs fluctuates wildly, the uncertainty of its charging and discharging mode will affect the supply and demand balance of the power grid. Big data technology optimizes charging scheduling by analyzing historical and real-time data of EVs in the power environment, effectively reducing peak consumption of the power grid by 6–7% [34] and providing load regulation solutions. In addition, big data technology can optimize the access and distribution of renewable energy [35,36]; through data analysis and prediction, we can ensure the stable input of renewable energy, improve the intelligent management level of the power grid, and establish a more efficient, green and environmentally friendly energy system.

In order to cope with the challenges brought by the increasing demand for EV charging, the power grid is gradually transforming towards intelligence. Traditional charging infrastructure makes it difficult to cope with the dynamic changes in traffic flow, charging demand, and grid load [37], resulting in insufficient accuracy in grid supply and demand regulation and low efficiency in traffic flow management. If big data technology is applied correctly, data processing and prediction models can be used to estimate traffic flow [38], charging demand [39], and electricity consumption [40], to help intelligentize the power grid.

The green and low-carbon development of smart grids has increased the proportion of renewable energy in the grid, posing new challenges to load management [35]. As one of the driving forces behind low-carbon power grids, EVs’ charging and discharging behavior also affects the supply-demand balance of the power grid [41]. Big data technology can help coordinate the power grid and renewable energy scheduling, promoting joint management between the power grid, EVs, and renewable energy [42,43].

Smart homes can help alleviate the pressure on the power grid. The collaborative management between EVs and smart homes can improve the flexibility of the power grid. Vehicle-to-home (V2H) technology has gradually become a research hotspot. As Higashitani et al. [44] pointed out, V2H technology and EV batteries can serve as temporary energy storage devices to meet the demand of insufficient grid load. However, not everyone brings electric cars home to charge. Song et al. [45] identified EV owners who charge at home based on smart meter data and big data technology. The above achievements prove that integrating EVs with smart homes is a promising future research direction.

As EVs’ impacts on the electricity grid deepen, big data technology is no longer limited to grid scheduling optimization. However, it is also integral to intelligent charging management [46]. With analysis based on user data, big data can give drivers personalized and intelligent charging plans [47] to optimize charging efficiency and user experience [48].

To illustrate the importance of systematically analyzing big data within the EV charging ecosystem, we first review the latest research progress related to charging big data in drivers, EVs, charging infrastructure, and the power grid. We found that although big data has a wide range of application scenarios, most of the relevant review literature is still limited to the classification of big data-related technologies themselves or fragmented analysis of the ecosystem. These studies fail to characterize the EV charging ecosystem systematically and exhibit research gaps in the deep integration of multi-source data and its diverse application scenarios. This article summarizes the core focus of relevant reviews and their discussions on the EV charging ecosystem in

Table 1. Summary of review studies.

Authors	Timeline	Number of Papers	Core Focus	EV Charging Ecosystem
Pevec et al. [49]	2011–2018	96	Exploring research on EVs from the perspectives of social economy (market acceptance, sales forecasting) and social technology (batteries, charging facilities, Vehicle-to-Grid (V2G))	Fragmentation analysis of various parts of the EV charging ecosystem
Shahriar et al. [50]	2010–2020	96	Machine Learning (ML) methods in EV charging behavior (mainly applied in analysis and prediction)	Fragmented discussion from a technical perspective
Tappeta et al. [51]	2011, 2013–2022	130	The protocols, standards, emerging communication and computing technologies emerging in EVs	Fragmented application of computational technology in EV charging systems
Ali et al. [52]	2018–2022	86	Application of digital twin technology in EVs and autonomous vehicles	Focusing on technical analysis of EVs, charging infrastructure, and power grids
Mololoth et al. [53]	2011–2014, 2016–2022	123	Research on the application of blockchain technology and ML in smart grids	Discussing EVs and the power grid
Ali et al. [54]	2015–2024	120	Review the impact, application, and models of ML/Deep Learning (DL) in the field of mobile electrification	Isolation analysis of various parts of the EV charging ecosystem from ML/DL perspective
Cavus et al. [55]	2010–2016, 2018–2024	123	Exploring the application of Artificial Intelligence (AI) technology in battery management Systems and system control technologies for EVs	Focusing on the optimization and technological application of EVs

(in cases where the literature search timeline is unclear, the primary timeline of references is adopted; when the number of retrieved documents is unspecified, the number of references serves as a proxy).

From a systemic perspective, EVs are an opportunity to achieve zero-carbon transportation and the key to achieving low-carbon power grids, households, and society. As a bridge between transportation and the power grid, EVs promote the zero-carbon transformation of transportation via big data. By leveraging their role as a distributed energy source with dual capabilities of energy consumption and storage, they facilitate the integration of distributed renewables and advance grid decarbonization. Based on dual capabilities, the deep integration of smart homes and EVs in household electricity presents excellent low-carbon household application prospects. At the level of a low-carbon society, big data-driven charging management adopts a user-centric approach, fully taps into user data, meets personalized needs, and supports the development of a low-carbon society.

Therefore, compared with the existing research, the main innovations of this paper are as follows: (1) from the perspective of four key stakeholders of the EV charging ecosystem (drivers, EVs, charging infrastructure, and power grid), it explores the generation of big data of EV charging; (2) it explains the key issues that big data can address in EV charging; (3) based on these issues, it explores data collection and data processing in EV charging applications; and (4) it analyzes future research opportunities for EV charging big data. The research structure is shown in Figure 1.

2. Review Methodology

This section will elaborate on the complete process of literature search and screening, aiming to clarify the scope of literature sources, screening logic, and the research basis of this review, and improve the transparency, reproducibility, and scientific rigor of conclusions.

2.1. Databases Searched

As a review article, data collection of the research field literature is crucial as it defines the domain of knowledge from which conclusions are drawn from the study. Given the interdisciplinary nature of this topic, the Web of Science database was selected as our database, as it enables access to significant research findings across various academic fields [56].

2.2. Exclusion Criteria

In order to avoid missing literature, we used the keyword “big data ev charg*” in the fielded search interface of the Web of Science Core Collection dataset and selected the search scope as “All fields” (Searches all of the searchable fields using one query) for search. Meanwhile, the publication time range of the literature is set between 30 June 2000 and 30 June 2025. To select literature that is highly relevant and of reliable quality to the research topic, we excluded conference papers. Due to the relatively strict peer review process of journal articles, the quality of journal articles is usually higher compared to other document types such as conference papers. We obtained 307 articles in our first search.

As the main purpose of this study is to review the literature on EV charging in the context of big data, it is necessary to filter out all papers outside the scope of the research. Therefore, we further screened literature related to “big data” and “EV charging” based on titles, keywords, abstracts, methods, etc. In the end, a total of 142 papers were retained. For these 142 eligible papers, we systematically extracted, analyzed, and summarized them based on six dimensions, data collection, data analysis, modeling methods, research advantages, limitations, and future prospects, forming the core literature foundation of this review.

3. Big Data Generated by EV Charging

3.1. The Chain of EV Charging

The chain of EV charging involves several elements. The user drives the EV and determines the behavior of charging [57]. The EV is the object of charging, and the capacity of its battery affects the range [58]. Charging infrastructures including charging stations and private charging posts are supported by the grid and decide charging locations. The power grid provides energy and is influenced by the charging load [59]. The whole charging chain is as follows: the user drives the EV to the charging station, the charging pile or charging device transforms the grid’s power into Direct Current, and the electricity is then transmitted to the EV’s battery. The electrical energy will be stored in the EV’s battery to power the car after it has been received [60]. The EV charging process not only consists of the behavior of energy transmission [61] but is also the process of data generation and interaction [62]. The chain of the EV charging process is shown in Figure 2.

3.2. Big Data Generated by EV Drivers

The user’s decision directly influences how the EV is used and charged and indirectly influences the big data generated by EV charging. EV trajectory data includes miles traveled, travel time, location information and speed [63,64]. These data are closely related to the user’s travel behavior. The user’s activity range, which includes residential, commercial, industrial, public service, road facility and green areas, affects dynamic charging demand and the distribution of charging locations [65]. Drivers’ charging behavior can be influenced by a variety of factors, including charging preferences and habits [66] and charging methods such as fast charging and slow charging [46]. Additionally, charging expenses were taken into account, along with the quantity of coupons, coupon ID, and discount amount [67]. The analysis of these data will help to gain insights into drivers’ charging behaviors and preferences to better meet the needs of EV customers.

3.3. Big Data Generated by EVs

Big data of EVs is crucial for EV charging. The configuration data for various EV types varies. For example, Shi et al. [68] used configuration big data of pure EVs and plug-in hybrid vehicles to explore issues related to EV charging capacity; differences in battery State of Charge (SOC), before and after charging; charging duration; and driving range. They concluded that the actual charging capacity of most EVs is lower than the rated capacity and that the charging interval mileage is positively correlated with battery capacity. In addition, more detailed data of EVs, including engine status, electric motor data, electric motor drive systems, vehicle modes, vehicle manufacturing time, and 100 percent charge range, are collected and analyzed [69].

The battery plays the most important part when charging an EV. EV charging is significantly impacted by a number of factors, including the battery management system (BMS), battery pack voltage, battery current, SOC minimum and maximum battery voltage, and minimum and maximum temperature data [70]. Analyzing historical data from battery swapping stations to predict the number of swapping events at each station during specific future time periods is also an important area of research [71]. Big data from EVs will help to diagnose charging status and improve battery performance.

3.4. Big Data Generated by Charging Infrastructures

EVs need charging infrastructure for charging, and charging infrastructure is built and developed to meet the charging needs of EVs. The interdependent relationship between EVs and charging infrastructure generates a large amount of data. First of all, the location information of charging infrastructure is often of concern. This includes charging station and equipment characteristic information such as station ID, charger ID, location, charger type, timestamp, current status (whether it is charging an EV), charging facility ID number, and facility type [18,46,72]. Second, new data emerges along with the start of charging for EVs. This includes charging data characteristics during the charging process, such as charging duration [72], charging frequency, charging energy [18,73], average transacted power [18], and charging charges [46]. In addition, information related to charging start time, charging duration, charging start SOC, charging end SOC, and charging consumption will be tracked [74]. In addition to the data on the charging process, the charging infrastructure can also record the information of the charging station users, including the user’s perception of the charging station and the environment, the user’s information, and the user’s charging card number [46]. The organization and classification of these data provide strong support for an in-depth understanding of EV charging system operation and user behavior.

3.5. Big Data Generated by Power Grid

Compared with the traditional grid, more and more smart devices are installed in the smart grid to make it more reliable and efficient. On the one hand, these devices generate a large amount of data that is used for monitoring, recording, and communication [75]. For example, the generated residential total load data will be utilized in EV charging [76]. On the other hand, the total EV charging load also affects the grid by reducing grid quality and system stability [77]. Therefore, during EV charging, there is a need to document the load impacts of charging on the grid, including grid loads during charging, grid stability, and operational data.

3.6. Big Data Generated by Others

In addition to the types mentioned above, there are some other types of big data that can influence EV charging behavior, although they are not generated by EV charging. Weather big data (maximum temperature, average temperature, average humidity, and average wind speed) can significantly affect EV bus traffic [78]. The congestion index varies with the hourly traffic flow, and a specific congestion index can be used to determine the charging demand at different times of the day [79]. Data on electricity generation from solar power plants can help people utilize electricity to charge EVs, increasing energy efficiency, reducing operator costs, and mitigating the negative effects of grid overloads [80]. The use of big data tools to obtain information on parking and EV spaces, such as parking lots, on-street parking spaces, and plazas, can optimize current layouts and provide new layout options for further research and applications [81]. Moreover, wireless charging for EVs is an emerging technology that is gaining increasing attention. Due to its principle of magnetic field coupling, there are common concerns about potential risks such as magnetic radiation and magnetic-thermal coupling [82]. With the help of big data, safety analysis and assessment models can be trained to generate real-time safety management decisions [83]. Due to space limitations, this paper will not discuss these aspects in detail.

4. Key Issues in EV Charging That Big Data Can Address

EV charging faces a number of problems that hinder its enhancement and expansion. Big data, as an emerging technology, provides huge amounts of information, which provides new ideas to solve these problems.

4.1. Optimized Control of Grid Operation

As the energy demand for EVs continues to increase, there is growing regulatory uncertainty in the grid, placing significant deployment pressure on the grid [84]. EV charging has a large impact on the operation of the power system through loads and harmonics, as well as losses in the grid. The uncertainty and randomness of user vehicles and charging behaviors increase the difficulty of controlling the power grid and are prone to EV charging peaks, which lead to peak power consumption and increased load on the grid [31]. Moreover, EV battery charging is a nonlinear load, and the electronic power equipment will generate harmonics, which will interfere with the power equipment connected to the power grid. This undoubtedly presents new challenges to the hardware equipment of the grid, the planning of the grid system, and its operation [85]. Big data on EV charging can be applied to address the above issues. For example, using charging big data to calculate peak or off-peak contribution factors to set a flat rate can shift electricity consumption from peak to off-peak hours without affecting the rest of consumers’ daily lives [34]. The above program enhances the stability of the grid by shifting peaks and valleys to address grid congestion.

In addition, V2G technology, in which the onboard battery of an EV is used as a distributed energy storage unit to realize the two-way transfer of energy and information between the vehicle and the grid, is also inseparable from big data. In the context of V2G, rapid response to electricity demand is required [86]. Therefore, Sun et al. [87] used trajectory data to estimate the charging and discharging needs of leased EVs, in order to guide the strategic and policy formulation of V2G technology charging and discharging events in the power grid.

4.2. Charging Infrastructure Layout

Charging stations are the core of the charging infrastructure and an important infrastructure for the charging behavior of EVs. Long charging times and a limited supply of chargers hinder the rapid popularization of EVs, and one of the solutions to this problem is to expand the charging infrastructure. In recent years, charging station layout optimization has been one of the hotspots of research in the field of EV charging, and charging station layout optimization has a close relationship with big data on EV charging. Tu et al. [88] developed a location model for electric cab charging station location optimization. The study extracted public cab demand with spatiotemporal attributes from massive cab GPS data and constructed a location model under spatial and temporal constraints such as Electric taxi (ET) range, charging time, and charging station capacity by modeling the periodic interactions between cab demand, ETs, and charging stations with a spatiotemporal path tool. Jia et al. [89] collected and processed large-scale cellular signal data to generate travel demand beyond commuting trips to capture the real travel patterns of EVs. Charging demand locations are grouped based on travel patterns, and then charging station sites are identified through optimization. In addition, GPS trajectory datasets with seasonal characteristics (travel, parking, and charging information [24], The charging event dataset [90], Mobile Signal Data [90], Urban Statistical Information and Geographic Data [90], Road density data [91], and Population density data [91] of charging stations have all been applied to optimize charging stations.

4.3. Battery Development

The battery system is an important, expensive, and partially interchangeable component of EVs [69,92], which directly affects the performance and range of EVs. Studying and developing battery voltage and capacity is one of the key directions of the technologies facing EV charging.

Battery voltage refers to the voltage of the battery cells. In order to achieve the required output voltage and range of an EV, hundreds of battery cells are connected in parallel and series to form a battery pack [93]. Therefore, increasing the voltage of a single battery will improve the voltage level of the entire battery pack, thereby enhancing the performance and efficiency of the vehicle. However, the pursuit of high voltage poses a great challenge to battery safety [94]. The detection of battery safety can be associated with big data technology. Zhou et al. [95] designed an online State of Health (SOH) estimation method for in-service EV batteries based on the charging data collected under the normal operating conditions of EVs, with the error of online SOH estimation controlled within 8%. Sun et al. [94] employed two parameters, namely correlation and variability, to capture abnormal voltage fluctuations, and applied an improved K-means method to identify faulty battery cells. This method reduced the computation time by approximately three times compared with existing methods, enabling accurate detection of different types of faults such as progressive faults and sudden faults. Its reliability and robustness were also verified using real-world EV fault data and thermal runaway accident data.

Battery capacity is typically defined as the total available capacity of a battery discharged from the upper cut-off voltage to the lower cut-off voltage under specified conditions set by the manufacturer [96]. Battery capacity is a key performance parameter reflecting battery aging, and capacity estimation is a prerequisite for determining the SOC and state of energy (SOE) of a battery [97,98,99]. Scholars use big data to predict battery capacity. Reference [12] combined particle swarm optimization with convolutional neural networks and long short-term memory networks to extract various health indicators from the real EV operation and environmental temperature data. The error in predicting the battery health status in real EV scenarios was about 3%. Reference [11] used big data collected from 707 EVs, including detailed records of driving, charging, and resting periods, to predict battery capacity using Gaussian process regression with an error of 2.07%. If a deep fusion transfer learning network is used, the average error is as low as 1.42%.

4.4. Dynamic Pricing of Charging Network

The dynamic pricing in the EV charging network is to adjust the charging price in real time according to factors such as charging demand, traffic flow, and charging station location [100,101], thereby unleashing the flexibility of EV customers [102]. In practice, big data can help charging stations to make reasonable dynamic pricing. For example, Lu et al. [103] analyzed EV behavior data to implement dynamic pricing, thereby estimating the utility and demand information of EVs. The results showed that the revenue loss of EV charging stations did not exceed 20%, even in the case of the worst prediction outcomes. Yang [104] designed and verified the Price-Responsive Pre-charging Adaptive Control (PRECC) scheme focusing on dynamic pricing: PRECC factors were extracted from historical EV data via data mining, and simulations were conducted by combining base load data and dynamic electricity price data. The results showed that by accurately responding to dynamic pricing signals, PRECC achieved an average profitability of 5.87%, outperforming the online (dynamic) scheduling by B&C algorithms (linear programming-based dynamic scheduling, 1.33%) and the fixed SOC threshold scheme (4.01%). Moreover, by means of dynamic pricing, charging service providers can reshape space-time charging demand through pricing signals, reduce the impact on the power grid [105] and contribute to stable energy distribution [106].

4.5. Safety of Charging Equipment

The security of EV charging is limited not only to the safety of the battery but also to the security and reliability of the charging equipment. A large number of smart charging stations have been deployed in the past few years, and most of them are online and connected, thus presenting potential threat risks. While there is relevant work on smart vehicle security, little work has been performed on charging device security [107]. Big data analysis can be used to monitor the operational status of charging equipment. By collecting real-time data from the equipment, potential problems can be detected, and preventive maintenance can be carried out in time to improve the safety and reliability of the charging equipment. For example, detecting anomalies in smart car charging and station power supply systems in real-life scenarios. Gao et al. [20] used big data technology to extract useful information from massive operational data, filter out data features that characterize the health of charging equipment, and construct a fault diagnosis model. The model has been tested in practical engineering, and has certain progressiveness and high classification accuracy, which can be used for fault diagnosis of charging equipment. Xu et al. [108] utilized a front-end and back-end feature information fusion module, combined with deep learning techniques, to diagnose charging pile faults by mining and fusing spatial and temporal features. In subsequent experimental tests, the fault diagnosis accuracy of this method reached 96.36%, far higher than that of the comparison model.

5. Data Collection and Data Processing in EV Charging Applications

5.1. EV Charging Big Data Collection

In the field of EV charging, there are more methods of data acquisition. As can be seen in Figure 3, open data sources are one of the main methods of acquiring big data for EV charging. Open data sources refer to the use of open data interfaces (API, Application Programming Interface) to obtain data from public databases, government agencies or other organizations. The National New Energy Vehicle Monitoring and Management Platform (NNEVMP), as a governmental organization, provides a large amount of EV big data that can be used for scientific research. The raw data on the NNEVMP used by Li et al. [70] were collected by the on-board terminals of electric buses and then uploaded to the data platform. The dataset includes 38 entries of big data such as sampling time, BMS number, battery pack voltage, battery current, SOC, minimum battery voltage, maximum battery voltage, minimum temperature, maximum temperature, and so on. Li et al. [109] utilized data collected from five EVs of the same brand on the NNEVMP including voltage, current, SOC, and temperature values. Huo et al. [12] also used battery data, motor data, electronic control unit data, and alarm data from the National New Energy Vehicle Big Data Alliance dataset for estimating battery health status

Data can also be obtained through database queries. Database queries are usually performed for data stored in specific databases, and access to such databases usually requires authorization. Andrenacci et al. [48] extracted the required data via the Octo Telematics database containing information such as, for example, vehicle ID; instantaneous GPS longitude, latitude, and speed; timestamp; and distance from the previous trajectory. Kim et al. [110] accessed the Korea Environment Corporation (KEC) database via the KEC API to determine whether any EVs were occupying any EV chargers for charging in the target area at the time of performing the data query. Sensor collection is also a major method; Chen et al. [111] used a real charging dataset collected from 20 EVs over a period of 29 months, including battery pack current and voltage SOC. The maximum and minimum battery voltages within the battery pack, as well as the maximum and minimum battery temperatures, predict the capacity of the battery. In addition to the specific data collection methods described in detail above, web crawlers [112], The questionnaire survey [113] and other methods are used in the EV charging data collection set. The specific collection methods and data categories can be found in Figure 3.

However, current EV charging data collection has obvious gaps: key data like long-term charging demand [110], real-time high-dimensional data [39,50], seasonal trajectories and renewable energy data [24,114] are insufficient or missing. Influencing factor data—holiday effect, charging tech, pricing policies [115], population density [116] and traffic flow—are also underincorporated. These gaps limit coverage of transport, weather, etc., leading to one-sided analyses and impractical strategies.

Figure 3. EV charging big data collection [11,12,16,24,30,39,65,68,69,70,72,73,77,80,88,89,90,94,109,110,111,112,113,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133].

Figure 3. EV charging big data collection [11,12,16,24,30,39,65,68,69,70,72,73,77,80,88,89,90,94,109,110,111,112,113,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133].

5.2. EV Charging Big Data Analysis

With the growth of big data and the rapid development of information technology, big data analysis tools and methods are emerging. Algorithms have become effective tools for processing and analyzing massive data, and they occupy a major position in EV charging big data. Table 2 shows several big data analysis algorithms commonly used in EV charging, such as K-means clustering, random forest, and genetic algorithm.

Clustering is an essential tool in data mining research and applications. It is the subject of active research in many fields of study, such as computer science, data science, statistics, pattern recognition, artificial intelligence, and machine learning [134]. One of the most popular clustering methods is the k-means algorithm, which identifies clusters by minimizing the clustering error. In EV charging k-means algorithm mainly analyzes charging data, driver data. Random forest is a statistical or machine learning algorithm used for prediction [135]. In EV charging Random Forest algorithm is analyzing charging behavior characteristics, fall and spring weekday data. Genetic Algorithm (GA) is a representative example of a group of methods known as evolutionary algorithms. Genetic algorithms are iterative algorithms based on the concept of generation, but they are also highly parallel in themselves, as they model the evolution of a series of solutions [136]. Genetic algorithms are used in EV charging to solve the spatio-temporal demand coverage localization model and the dual objective deployment model for electric truck battery exchange stations. In addition to this, the KNN algorithm, logistic regression algorithm and support vector machine algorithm are used to analyze which variables affect the maximum range of a vehicle after fully charging. An a priori algorithm can be used to analyze and study the effect of weather and holidays on the user’s charging. The greedy heuristic algorithm analyzes the optimal location of charging posts. A more detailed representation is in Table 2.

Table 2. Big Data Analytics Algorithms in EV Charging.

Big Data Analytics Algorithm	Problems Solved	References
DBSCAN and K-means clustering algorithm	– To classify EV drivers. – Analysis of EV charging demand location.	[46,89]
K-means and direct and hierarchical and DPC clustering algorithm	– EV load clustering.	[137]
K-means clustering algorithm	– To identify the user’s charging mode. – Zoning of charging stations. – To classify drivers. – To analyze drivers’ preferences. – To explore the inherent laws of charging behavior. – To analyze charging habits. – To analyze destination GPS coordinates. – To cluster similar dates	[9,66,69,74,117,118,138,139]
PCA algorithm	– To analyze EV charging station data. – To reduce the dimensionality of data while preserving key information.	[75,140]
Random forest algorithm	– To analyze the charging behavior characteristics of EV. – To analyze household electricity consumption data on weekdays in spring and autumn to identify households that charge EVs.	[119,141]
GBDT algorithm	– To analyze the factors that influence user frequency and vehicle preferences in shared EVs.	[142]
Genetic algorithm	– Solving the space-time demand coverage positioning model. – Solving the dual-objective deployment model of electric truck battery swapping station.	[10,88]
KNN algorithm and logistic regression and SVM	– To analyze the variables that affect the maximum cruising range of the car after it is fully charged.	[120]
Greedy heuristic algorithm	– To analyze the best position of the charging pile.	[143]
Two-stage guided constrained differential evolution algorithm	– Combining electricity load data, optimizing multi-objective queuing strategies, and finding win-win strategies between the power grid and users.	[84]
Dynamic Time Warping algorithm	– To obtain the most similar mode in the driving cycle.	[121]
Ridge Regression	– To predict the driving range of EVs	[144]
Gradient Boosting algorithm and XGBoost algorithm	– To analyze the charging demand factors affecting station utilization.	[145]
XGBoost algorithm	– To analyze historical photovoltaic power and weather data for short-term photovoltaic power prediction.	[146]

Table 2. Big Data Analytics Algorithms in EV Charging.

Big Data Analytics Algorithm	Problems Solved	References
DBSCAN and K-means clustering algorithm	– To classify EV drivers. – Analysis of EV charging demand location.	[46,89]
K-means and direct and hierarchical and DPC clustering algorithm	– EV load clustering.	[137]
K-means clustering algorithm	– To identify the user’s charging mode. – Zoning of charging stations. – To classify drivers. – To analyze drivers’ preferences. – To explore the inherent laws of charging behavior. – To analyze charging habits. – To analyze destination GPS coordinates. – To cluster similar dates	[9,66,69,74,117,118,138,139]
PCA algorithm	– To analyze EV charging station data. – To reduce the dimensionality of data while preserving key information.	[75,140]
Random forest algorithm	– To analyze the charging behavior characteristics of EV. – To analyze household electricity consumption data on weekdays in spring and autumn to identify households that charge EVs.	[119,141]
GBDT algorithm	– To analyze the factors that influence user frequency and vehicle preferences in shared EVs.	[142]
Genetic algorithm	– Solving the space-time demand coverage positioning model. – Solving the dual-objective deployment model of electric truck battery swapping station.	[10,88]
KNN algorithm and logistic regression and SVM	– To analyze the variables that affect the maximum cruising range of the car after it is fully charged.	[120]
Greedy heuristic algorithm	– To analyze the best position of the charging pile.	[143]
Two-stage guided constrained differential evolution algorithm	– Combining electricity load data, optimizing multi-objective queuing strategies, and finding win-win strategies between the power grid and users.	[84]
Dynamic Time Warping algorithm	– To obtain the most similar mode in the driving cycle.	[121]
Ridge Regression	– To predict the driving range of EVs	[144]
Gradient Boosting algorithm and XGBoost algorithm	– To analyze the charging demand factors affecting station utilization.	[145]
XGBoost algorithm	– To analyze historical photovoltaic power and weather data for short-term photovoltaic power prediction.	[146]

Besides algorithms, there are other ways to analyze big data. Han et al. [147] used Monte Carlo simulation to generate charging and V2G scenarios and analyze the impact of V2G on peak hours. Hu et al. [46] used an extended RFM model, i.e., RFMLT model, a combination of a two-stage clustering method and entropy weighting method for clustering and analyzing EV drivers. Cao et al. [148] used decentralized computing to process driving big data of EVs. Shi et al. [68] used time series data analysis to study EV data.

Although many methods exist for analyzing EV charging big data, existing models have limitations: insufficient scenario adaptability, reliance on short/long-term ML to improve prediction accuracy [50,107], and unproven geographical portability (e.g., Long Short-Term Memory (LSTM); [149]). Data-wise, multi-source data real-time processing faces technical barriers [150], and balancing privacy and algorithm performance remains unsolved.

5.3. Applications for EV Charging Big Data

Applications of big data for EV charging include forecasting, detection, and charging station optimization layout. Forecasting is the process of analyzing historical data as well as real-time data to help predict charging demand and future trends. Detection refers to the use of historical data to detect the safety status of battery packs and identify malfunctions and to detect anomalies in the power supply system of smart car charging stations. EV charging big data is also often used to solve the problem of optimizing the layout of charging stations. By analyzing the literature, we found that machine learning (ML) techniques dominate the application techniques of EV charging big data. So, we categorize big data applications in EV charging into machine learning techniques and other application methods.

ML is used in data applications to extract useful information from large amounts of data, to make predictions and decisions, and to help automate data processing [151]. There are various machine learning techniques, and the machine learning used for EV charging big data is shown in

Table 3. Application of ML in EV charging.

Category	Application in EV Charging	Machine Learning Method	References
Power Grid	– Day-ahead forecast of green energy power generation.	Hybrid Neural Network	[80]
	– Family load forecasting.	LSTM and Bayesian Neural Networks	[122]
Battery	– Predicting battery swapping demand for EVs.	LSTM, GRU, Bi-LSTM and Bi-GRU	[71]
	– Predicting the normal IRs of temperature, mileage and charging state input.	Hybrid Neural Network	[153]
	– Predicting the SOC of the real driving cycle.	LSTM	[121]
	– Predicting battery charging capacity.	XGBoost regression model	[154]
	– Detecting the abnormal points of the voltage.	K-means algorithm	[94]
Charging Infrastructure	– Industrial Control System network traffic anomaly detection.	Multiple Head Attention model	[107]
	– EV charging equipment monitoring and fault diagnosis system.	Deep Belief Network	[20]
	– Determining the optimal charging scheduling strategy for low-power EV	Deep Reinforcement Learning	[47]
	– EV capacity prediction under privacy protection.	Convolutional Neural Network (CNN) and Feedforward Neural Network (FNN)	[111]
	– Abnormal energy consumption vehicle detection.	XGBoost regression	[123]
	– Estimating the SOH of lithium batteries.	Particle Swarm Optimization and CNN and LSTM	[12]
	– Estimating the SOH of the battery.	Gaussian process Regression	[11]
	– SOC prediction of EVs.	Transformer	[155]
	– Estimating SOC and SOH.	Transformer	[156]
EV	– Predicting the number of EVs in charging stations in the next few hours.	LSTM	[149]
	– Online health state estimation method for in-use EVs	Iterative extended Gaussian process regression-Kalman filter	[95]
	– Estimation of electric bus energy consumption.	CatBoost Decision Tree Model	[70]
	– Obtaining optimal parameters of the boost and low power charging stages.	Reinforcement Learning	[73]
Driver	– Forecasting EV charging demand.	Heterogeneous spatio-temporal graph convolutional network	[39]
	– Predicting the charging level, EV location and parking lot connection time when EV enters the parking lot.	LSTM	[114]
	– Forecasting EV charging demand.	Decision Tree	[78]
	– Explaining the influencing factors of charging demand in different regions of South Korea	Binary Logic Model	[110]
	– Predicting travel time, queuing time of charging stations, and charging consumption at different times of the day.	Gradient Hoist Model	[124]
	– Revealing the cycle travel mode of online ride-hailing drivers.	K-means clustering	[125]
	– Assigning BEV drivers to stations by convenient standards.	Fuzzy model	[126]
	– Establishing the charging choice behavior model of EV.	Hybrid Choice model	[127]
	– To capture user behavior in car sharing systems.	Gradient Boosting Decision Tree	[142]

. It can be seen that prediction with ML (such as LSTM, a variant of recurrent neural networks [152]) occupies an absolutely dominant position. The predictions include grid generation, battery exchange demand, battery capacity, battery internal resistance, remaining battery power, charging station capacity, number of EVs, charging behavior, and charging demand. In addition, machine learning is used for fault detection and user classification.

In addition to machine learning, there are many methods from other fields that have been applied to the field of big data for EV charging. Methods from the field of operations research are also commonly used to analyze big data on EV charging. He et al. [157] used a two-tier planning framework, collaborative capacity optimization (CCO) method, large M method to linearize the nonlinear problem, and analytic target cascading (ATC) technique to optimize charging station siting. Discrete mathematical optimization is also used to find the best solution for electrification of urban bus fleets [158]. Kernel density functions and p-median models in statistics are used for the deployment of charging facilities [24,30]. In terms of evaluation, Liang et al. [143] used the entropy weight-TOPSIS method to form an index system for evaluating the location of EV charging piles, proposed an ensemble coverage model based on the evaluation, and used a greedy heuristic algorithm to find the optimal location of charging piles. Yang et al. [18] used the sequential construction research method, which is commonly used in social science and education research fields, to analyze the regional public EV charging facilities. Zhao et al. [128] employed a decoupling framework in the field of software engineering in order to investigate the partial relationship between EV performance degradation and each individual variable (e.g., temperature and total driving distance). Strielkowski et al. [159] employed an energy internet in the field of internet to design a market design for a high-renewable power system. Hong et al. [80] also used 6G Internet of Things (IoT) in the Internet domain to optimize charging configurations at multiple charging station locations for fleet operators.

Big data is widely used in EV charging but has clear limitations. On one hand, while it helps control charging to ease grid transformation/operation costs [119], long-term adaptation still needs major grid reforms (reallocating generation sites, integrating renewables, etc. [48]). On the other hand, though big data supports V2G (e.g., demand forecasting) with economic potential [160], it is limited by weak infrastructure and low driver participation (battery degradation concerns [161]), requiring big data-driven battery monitoring to balance benefits and risks.

6. Future Research Opportunities

The aforementioned study indicates that there are four primary areas for future research opportunities: (1) Transportation system, (2) Energy system, (3) Home energy system and (4) EV charging system. Detailed research directions are shown in Figure 4.

6.1. Opportunity 1: Deep Integration of Intelligent Transportation and Smart Grid

The deep integration of intelligent transportation and smart grid aims to achieve high coordination between transportation systems and energy networks. Intelligent transportation systems provide services in navigation, railways, waterways, and air transport systems, resulting in a large amount of information [162], while smart grids also need to provide real-time information when achieving a near-immediate balance between grid supply and demand [163], both of which generate large amounts of data. By fully leveraging the big data of the traffic and power grid and using EVs as a bridge between them, we could expect to achieve the coordinated optimization of the road traffic network and the power grid, with zero-carbon traffic as the guide.

Context 1: Synergistic optimization of transportation systems and power grids

The large-scale application of EVs makes the power grid and the transportation network deeply coupled, showing an interdependent state [164]. In real life, using the electricity price signal of a smart grid to guide EVs to charge in low peak hours can alleviate the power grid’s load pressure and reduce drivers’ charging costs [165]. At the same time, combined with traffic flow data, the layout of charging piles is optimized to reduce vehicle waiting time [166]. This coordination addresses resource waste from independent operations by aligning transportation demand with grid load distributions. The traffic flow of the transportation system, the grid load of the power grid, and the charging period of the EV big data are analyzed to achieve the purpose of forecasting the load [167], optimizing the charging schedule, and improving the efficiency of resource utilization [124], so as to help the coordinated optimization of the transportation system and the power grid.

Context 2: Zero-carbon transportation and energy transition

Transportation electrification is a key strategy to decarbonize the transportation system [168]. As a representative of electrified transportation, many countries actively promote EVs. For example, China plans to build a large-scale car network interactive system and intelligent charging network by 2030 [169]. Similarly, Germany plans to have 7 to 10 million EVs on the road by 2030 and to increase investment in charging infrastructure [170]. These policies are moving toward reducing or eliminating carbon emissions from transportation and achieving zero-carbon transportation. Achieving the goal of zero-carbon transportation requires an assessment of transportation carbon emissions. Big data on transportation [171], electricity [172], and user travel behaviors [173] can help evaluate and predict transportation carbon emissions. Then, providing a basis for achieving zero-carbon transportation. More importantly, integrating traffic flow and power grid load big data enables dynamic matching of EV charging demand with the grid’s low-carbon supply periods, reducing high-carbon energy use [174] and supporting zero-carbon transportation networks in line with energy transition.

6.2. Opportunity 2: Renewable Energy and Intelligent Energy Management Optimization

The intermittency and randomness of renewable energy [175] increase the difficulty of energy management [176]. However, the intelligent energy management system can maximize renewable energy consumption through demand response services, interact well with EVs [177], and help energy management. Renewable energy sources such as wind and light energy [178] and EVs are distributed energy sources [179]. To make the two collaboratively optimize energy management, it is necessary to fully integrate distributed energy sources.

Context 3: Integration and optimization of distributed energy

Driven by global net-zero commitments [180] and rising fossil fuel costs, centralized thermal power generation is facing increasingly strict environmental regulations and resource constraints, while distributed energy resources (DERs) are gradually becoming an important supplement and development direction for the energy system due to their flexibility [181] and low-carbon advantages. DERs are resources connected to the distribution network that provide services to nearby or on-site energy loads. These resources include wind energy, solar photovoltaic systems, batteries, natural gas fuel cells, EVs, demand response, and energy efficiency measures [179]. DERs are becoming essential to power grids, and while their integration with the power system offers benefits, it also presents challenges [182]. EVs as mobile DERs should be integrated. Yi et al. [183] proposed a hierarchical collaborative optimization planning framework that includes energy storage and various flexible resources on the demand side, such as hydrogen energy and V2G technologies. The case study showed that V2G reduced the average load variance of microgrids by 56.22%. Similarly, Huang et al. [184], in optimizing the design of the campus microgrid, formulated a charging and discharging policy for participating in power grid dispatching according to the residence time of EVs in the parking lot, narrowing the gap between peak and off-peak hours. The above research results show that EVs’ full bidirectional charging function can reduce the peak-valley difference.

Context 4: EVs and energy storage optimization in urban energy systems

EVs are gradually regarded as flexible energy storage resources in urban energy systems [185], and their role in the distribution network is increasingly prominent [186]. EVs help to improve the stability of the power grid and promote the flexible scheduling of the power system. For example, Borozan et al. [160] showed that among the three modes of Grid-to-Vehicle (G2V), V2G, and Vehicle-to-Building (V2B), V2G has the most significant economic potential in network planning, and V2B has important application value at the transmission level. It is not difficult to see that EVs can enhance the stability of the power grid [187], support building-level energy management [188], and further improve the operating efficiency of the overall energy system.

6.3. Opportunity 3: Synergizing Smart Homes with EVs

Integrating smart homes and EVs provides important opportunities for flexible planning and resource sharing, enhancing energy management capabilities [189]. With the help of V2H technology, various problems related to power consumption and consumer energy demand in smart homes can be solved [190]. At the same time, artificial intelligence (AI) can use energy consumption data in smart homes to make energy management decisions [191] and combine ML to predict the charging demand of EVs [192]. Therefore, combining AI and V2H can optimize EVs’ charging and discharging timing and strengthen energy management.

Context 5: V2H Technology

V2H technology, as a subset of V2X (Vehicle-to-Everything) frameworks, enables bidirectional interaction between EVs and home energy systems. León et al. [193] p proposed and evaluated a strategy that combines time of use pricing (TOU) with V2H operation to promote the popularization of EVs. Similarly, Wu et al. [194] found that V2H technology can reduce household energy costs by at least 22.92%. The above research demonstrates the potential of V2H in reducing household energy expenditure and reveals that EVs have transformed from simple transportation to mobile energy for households. Tostado-Véliz et al. [195] combined V2H technology with the flexible requirements of smart homes in their residential electrification system design, which can reduce total costs by up to 41.5%.

Context 6: AI for energy demand forecasting and automated management

Combining AI, particularly ML, improves the accuracy of predicting power demand at charging stations and helps determine optimal charging station characteristics [192]. If smart home systems apply AI, they could predict household energy demand, allowing residents to adjust their consumption plans for optimal efficiency [196]. Adelman and Uçkun [197] found that the dynamic pricing of smart homes dramatically reduces users’ electricity costs. The EV, with the function of a mobile energy storage unit combined with the vehicle-home-network data interaction platform driven by artificial intelligence, can realize home energy management and further reduce user costs.

6.4. Opportunity 4: Data-Enabled and User-Driven Smart EV Charging Management

This section discusses the necessity of a user-oriented EV charging strategy. Based on user data analysis, the user’s charging habits can be used to construct a personalized charging strategy to improve user satisfaction. Further, the user satisfaction measurement index is not limited to the charging process but extended to the charging management network, including charging infrastructure location and charging infrastructure.

Context 7: Personalized user needs and data-driven decision-making

A data-driven strategy is an important method to meet the personalized needs of EV drivers, and it can improve the efficiency and adaptability of charging systems [198]. Yang et al. [9] found differences in charging modes among different groups of EV users (such as commuters and non-commuters) by analyzing their charging locations. Similarly, Ahmadian et al. [16] used deep neural networks to predict EV drivers’ dwell/charge time and energy consumption. The above research results provide theoretical support for the design of personalized charging strategies. Personalized charging strategies can recommend the best charging time and location and combine time of use dynamic electricity pricing strategies to optimize charging efficiency and cost, achieving the goal of efficient charging management based on user behavior.

Context 8: User-driven intelligent management of charging networks

User-driven charging network management models can better address user charging demands and improve user satisfaction [199], thus facilitating the widespread popularization of EVs [200]. More and more studies have regarded user satisfaction as an important indicator for measuring the quality of charging management systems. For example, Tang et al. [201] proposed a multi-objective EV charging station location method based on a travel chain that can improve location selection efficiency, reducing user search time by 47.91% and charging waiting time by 65.83%, thereby improving user satisfaction. At the same time, Menos-Aikateriniadis et al. [202] pointed out in their case study that when the charging management strategy follows the principle of “user orientation”, the charging strategy based on Deep QNetwork can reduce daily electricity costs by up to 49.83% and charge up to 86.22% within the same period as the user’s historical charging mode. These studies indicate that user-centered charging management can optimize charging resources and improve users’ charging experience and satisfaction.

7. Conclusions

EVs are a crucial part of the solution to meet global carbon reduction targets, and big data on EV charging offered fresh perspectives on a number of issues. This study focuses on reviewing the applications of big data in EV charging, aiming to provide a reference for the future development of big data in the field of energy science. Based on a literature review and other methods, we reviewed the following:

(1) By combining EV charging chains, we analyzed four main kinds of big data generated by the EV charging process: (i) big data generated by drivers; (ii) big data generated by EVs; (iii) big data generated by charging infrastructures and (iv) big data generated by the power grid.

(2) We reviewed key issues in EV charging that big data can address, including: (i) optimized control of grid operation; (ii) charging infrastructure layout; (iii) battery development; (iv) dynamic pricing of the charging network and (v) safety of charging equipment.

(3) We reviewed the collection and processing of big data applications, especially analytics algorithms in EV charging.

(4) We discussed future opportunities of EV charging big data, including the opportunities in four systems: (i) the transportation system; (ii) energy system; (iii) home energy system and (iv) EV charging system.

This study systematically sorts out the sources, core applications, and future directions of charging big data by leveraging the EV charging chain. It helps researchers clearly grasp the overall context of this field while also providing references for subsequent studies focusing on specific issues (such as charging infrastructure optimization and power grid coordination) to facilitate more targeted advancement of research in this domain. The limitation of the proposed research is that it did not take into account the development of EV charging big data over time and did not assess the probability of each future opportunity.