The Use of Business Intelligence and Analytics in Electric Vehicle Technology: A Comprehensive Survey

Abstract

The emerging urbanization and the extensive increase of the transportation sector are responsible for the significant increase in carbon dioxide emissions. Therefore, replacing traditional cars with Electric Vehicles (EVs) is a promising solution, offering a clearer alternative. EVs are becoming more and more well-known and are being quickly used worldwide. However, the exponential rise in EV sales has also raised a number of issues, which are becoming important and demanding. These challenges include the need of driving security, the battery degradation, the inadequate infrastructure for charging EVs, and the uneven energy distribution. In order for EVs to reach their full potential, intelligent systems and innovative technologies need to be introduced in the field of EVs. This is where business intelligence (BI) can be employed, along with artificial intelligence (AI), data analytics, and machine learning. In this paper, we provide a comprehensive survey on the use of BI strategies in the EV transportation sector. We first introduce the EVs and charging station technologies. Then, research works on the application of BI and data analysis techniques in EV technology are reviewed to further understand the challenges and open issues for the research and industry community. Moreover, related works on accident analysis, battery health prediction, charging station analysis, intelligent infrastructure, locating charging stations analysis, and autonomous driving are investigated. This survey systematically reviews 75 peer-reviewed studies published between 2020 and 2025. Finally, we discuss the fundamental limitations and the future open challenges in the aforementioned topics.

1. Introduction

In the past decades, Information and Communication Technology (ICT) has taken a tremendous leap forward and we expect technology to only accelerate in the coming years. More specifically, the rapid development of the Internet of Things (IoT), artificial intelligence (AI), and business intelligence (BI) have enabled innovative solutions that can be employed in every aspect of everyday life. Technology affects people’s everyday lives, routines and practices including their workday, entertainment, leisure, health, and self-care.

At the same time, it is observed that the world is becoming urbanized. Reports show that more than 4 billion people now live in urban areas that lead to increasingly dense cities. In recent years, there has been a mass immigration of people that move from rural areas to urban regions, since cities promise greater number of jobs and prosperity than rural areas. However, urbanization creates a majority of problems for the people living in the cities, which are transformed into unsustainable and costly environments. Recent studies show that urban settings can also lead to significant difficulties in urban mobility that challenge safety and security, excess energy waste, and environmental degradation. Therefore, policymakers, the government, and the research community need to work together to tackle these challenges through efficient city planning and management challenges.

Nowadays, cities provide a wide range of opportunities to deal with the emerging challenges so as to create urban spaces and environments with enhanced livability for the people. Towards this direction, the concept of smart cities arises as a way to address urbanization, environmental concerns, and economic growth. The primary goal of a smart city is to create an urban environment that yields a high quality of life to its residents while generating overall economic growth. In order to achieve these goals, smart cities employ a wide range of technologies including business intelligence and analytics, cloud computing, machine learning, machine-to-machine communication, big data, IoT, AI, and Wireless Sensor Networks (WSNs). In fact, these emerging technologies are essential for smart cities because they form a quintessential and operational feasible environment.

Among the key advancements of smart cities is Electric Vehicle (EV) technology. Smart cities aim to improve urban living through sustainable and environmentally friendly solutions, whereas the transportation sector accounts for almost 24% of global $C{O}_2$ emissions, with road transport being the largest contributor. Thus, the replacement of conventional cars with EVs can significantly reduce carbon emissions and improve air quality. The EVs are powered by electricity and can even use renewable energy sources including solar and wind energy. Therefore, EVs are by design pollution-free and environmentally efficient. In order to understand how EVs can improve the transportation sector in smart cities, several issues must be investigated.

EV technology is not new. The first EVs were invented in the early 1830s but their batteries were improved upon in the 1880s. The golden age of EVs was in the early 1990s, but soon, EVs began to lose their share on the vehicle market. More specifically, their use was limited and few EVs were produced and actually used. For many years, little advancements were achieved in EV technology. Nevertheless, the finite nature of vehicle fuels and the rising environmental concern created the perfect conditions to exploit the capabilities of EV technology. In the beginning of 2000s, EV technology enjoyed ascending improvement, leading to an exciting increase in the EV market. Furthermore, EV technology received significant attention among the research community. At present, electric car markets are seeing exponential growth as sales exceeded 10 million in 2022. The share of EVs in total sales has more than tripled in three years. Looking ahead, it is expected that the EVs will demonstrate a steady annual growth rate (Compound Annual Growth Rate (CAGR) 2024–2028) of $9.82$%, and the unit sales of EVs market are anticipated to reach $17.07$ million vehicles by the year of 2028.

The swift advancement in electric car technology has made room for a more eco-friendly and sustainable mode of transportation. But in order to fully realize the potential of EVs, intelligent systems that can improve user experience, maximize performance, and solve a variety of issues related to electric mobility are required. This is where AI and BI come into play. BI and analytics enhance performance and develop innovative technologies in the areas of safety and security, charging station infrastructure and management and autonomous driving.

1.1. Contribution

The main contributions of this survey are multifold and aim to provide a consolidated, in-depth understanding of how BI and analytics can enhance the functionality, reliability, and intelligence of EV technologies.

1.

This work presents a comprehensive and unified review of BI- and AI-driven solutions across the EV ecosystem, including accident analysis and prediction, battery health estimation, charging station utilization modeling, intelligent charging infrastructure planning, and autonomous driving systems. These domains are traditionally studied sperately. Thus, we provide a holistic perspective that highlights their interdependencies for future EV innovation.

2.

We propose a structured taxonomy that systematically classifies existing research according to analytical goals, computational methods, and application scenarios. This taxonomy not only supports clearer comparison of methodologies but also reveals emerging patterns, dominant techniques, and underexplored areas across EV-related BI applications.

3.

In addition, this survey highlights the importance of stakeholder cooperation within the EV ecosystem and discusses how BI- and AI-driven mechanisms, such as dynamic pricing, decentralized data-sharing, and predictive decision support systems can motivate EV owners and charging operators to participate in coordinated strategies that optimize system performance.

4.

We offer an in-depth discussion of the fundamental limits that constrain the effectiveness of BI-driven EV solutions, such as fragmented and heterogeneous data sources, lack of standardized reporting frameworks, limited access to high-quality real-world datasets, reliance on centralized cloud architectures, model interpretability challenges, and cybersecurity vulnerabilities across interconnected EV infrastructures.

5.

We outline major future research challenges and opportunities, emphasizing the need for decentralized and privacy-preserving data management, physics-informed and explainable AI models, grid-aware charging coordination, scalable optimization for large EV fleets, and secure, robust autonomous driving intelligence.

6.

Finally, this survey goes beyond summarizing existing works by providing critical insights, cross-domain analysis, and forward-looking recommendations that collectively establish a roadmap for the development of next-generation BI-enabled EV technologies. By bridging the gap between EV engineering and data-driven intelligence, this study positions itself as a valuable reference for researchers, industry practitioners, and policymakers aiming to advance the digital transformation of electric mobility.

1.2. Paper Organization

The remainder of the paper is organized as follows. In Section 2, we introduce a comprehensive description of the electric vehicles market, along with a classification of the EVs’ types. Additionally, we study the charging methods that enhance and motivate the use of EVs. Section 3 provides a comprehensive review of BI- and AI-driven applications in the EV ecosystem. This section is structured into five key domains: (i) accident analysis and prediction, (ii) battery health estimation, (iii) charging station utilization analysis, (iv) intelligent charging infrastructures, and (v) autonomous driving, introducing the problem context and the state-of-the-art methodologies. In Section 4, the willingness of diverse stakeholders to cooperate and adopt optimal strategies is discussed. Section 5 offers an in-depth discussion of the fundamental limits that constrain current BI-enabled EV technologies, followed by an exploration of future research challenges and opportunities across data management, analytics models, grid integration, and autonomous mobility. Section 6 concludes the paper by summarizing the main insights and highlighting the importance of BI and analytics in enabling scalable, reliable, and intelligent EV systems.

2. Electric Vehicles: Preliminaries

This section provides a technical overview of electric vehicle architectures and charging technologies, not merely as background information, but as a foundation for understanding the data characteristics and analytical challenges addressed in subsequent sections. The diversity in EV configurations and charging modalities directly affects sensor deployment, data volume and velocity, operational constraints, as well as the design of BI and analytics solutions for performance optimization, safety, and grid interaction.

2.1. Electric Vehicles Market

The market of EVs is growing rapidly in recent years, and their use is expected to decarbonize road transport. Additionally, their use poses great challenges for the power grid network due to charging and discharging actions that great significant energy transfers. It is evident that the use of a single vehicle does not affect the energy market to a large extent. The energy market is impacted when the number of EVs becomes large. The sizeable number of EVs’ owners using the power grid, especially during high load periods, require their cooperation in order to achieve remarkable energy reductions. The EVs’ owners do not have common interests in the majority of times and sometimes their objectives are contradictory, which challenge the collaboration.

It is crucial to discuss two questions in order to provide solutions for the increasing adoption of EVs in the market share, as they make a meaningful difference in the energy and cost efficiency of the smart cities. First, which are the crucial sectors of the EVs market that the research communities need to focus to provide sustainable and cost effective solutions? Second, how will the different EVs’ owners and all the involved market counterparts be motivated to enter in a cooperative environment and decide the optimal solutions? In the interest of answering the first question, it is necessary to briefly highlight the main challenges associated with the rapid growth in EV adoption. These challenges include the increased pressure on power grids during peak charging periods, the need for scalable charging infrastructure, the variability of user driving and charging behaviors, as well as the technical limitations related to battery performance and degradation. Understanding these issues sets the foundation for identifying where BI and analytics can provide meaningful, data-driven solutions. The answer to the second question is more demanding and thus will be discussed in the paper in great detail. The second question is more complex, as it requires understanding how to motivate diverse EV owners, grid operators, mobility providers, and charging-station operators to cooperate within a unified ecosystem. This challenge calls for advanced BI and analytics frameworks capable of aligning stakeholder objectives, forecasting energy and mobility demands, and supporting optimal decision-making. These aspects will be examined in detail throughout the remainder of the paper.

2.2. Electric Vehicles Classification

EVs are generally experiencing an increase in popularity in recent years. Therefore, EV technology gains great attention in the research and industry community. In this section, the different types of EVs are discussed, including battery EVs, plug-in hybrid EVs, hybrid EVs, fuel cell EVs, and solar EVs (shown in Figure 1) [1,2,3,4,5,6,7].

2.2.1. Battery Electric Vehicles

Battery EVs (BEVs) are also known as all-electric vehicles, or pure vehicles, since they have only one rechargeable energy source [8,9]. Thus, the battery pack (usually lithium batteries) is connected to one or more electric motors, as shown in Figure 1a. The battery pack is charged periodically through an EV charger, and the charging time ranges between 6 and 8 h for a full charging period. The typical range of a fully charged BEV is around 100 to 150 km; however, the research community is working towards the extension of the battery life with a range up to 350 km [10]. Recent advances in battery technology have significantly extended the driving range of battery electric vehicles, with many contemporary models achieving typical ranges of 300–500 km or more under standard driving conditions, as reported in recent studies and industry reports [11,12]. The BEVs are popular in the market share on account of the multitude of advantages they bring together. More especially, the BEVs produce nearly zero emissions when driving, are quiet and cost-efficient, and have low maintenance costs. On the other hand, their limited range, especially during cold seasons, poor charging infrastructure, and battery life concerns are some of the disadvantages of BEV technology.

2.2.2. Plug-In Hybrid Electric Vehicles

The second category of EVs are the plug-in hybrid EVs (PHEVs), which employ both batteries to power an electric motor as well as incorporate an engine that is powered by conventional fuel (such as petrol or diesel), as depicted in Figure 1b. This is the reason why these EVs are also called hybrid. The battery can be charged through external plug-in sources. Usually, the PHEVs use the battery until it is nearly depleted before automatically switching to the fuel engine [13,14,15]. The advantages of the PHEVs include the zero emissions when driving on batteries. However, they are expensive, complex to maintain, and there are major concerns with regards to their battery pack. Their use is increasing since the automobile manufactures are working towards the improvement of PHEVs.

2.2.3. Hybrid Electric Vehicles

Hybrid EVs (HEVs) are also called hybrid or parallel hybrid, since they have both electric motor and an engine, as observed in Figure 1c [13,16,17,18]. The engine receives energy from fuel, and the motor receives electricity from batteries. This operation refers to the parallel hybrid series. The hybrid series uses the motor powered by the battery to only drive the wheels, and these EVs are often called extended-range EVs. The HEVs differ from the PHEVs, since the HEVs cannot be plugged into an EV charger and the battery is charged through the engine and the brakes system [19,20]. The HEVs’ operation allows their motor to be smaller, while also reducing energy losses. On the contrary, maintenance and repair of HEVs are complex and expensive, while the complexity in their manufacture escalates due to the parallel nature of the engine and the motor.

2.2.4. Fuel Cell Electric Vehicles

Unlike BEVs, PHEVs, and HEVs, fuel cell EVs (FCEVs) have a hydrogen tank to store the energy in the form of hydrogen. Energy that is fuel-produced is converted directly into electric energy [21,22]. The FCEVs additionally have a Power Control Unit (PCU) that decides when to use stored electric energy from the battery pack or draw directly from the fuel cell stack. Figure 1d depicts the infrastructure of a typical FCEV [23]. The main advantages of FCEVs are their extended range of up to 300 km and reduced energy losses due to their battery technology. In the future, FCEVs are expected to grow in popularity, since great improvements in their operation and energy system infrastructure are needed [24].

2.2.5. Solar Electric Vehicles

Last but not least, solar EVs (SEVs) include photovoltaic cells to convert solar energy into electric and a battery pack similar to the ones used in BEVs (shown in Figure 1e) [25,26,27,28]. Among the advantages of the SEVs is their extended range and zero costs since solar energy is used when the sun is available. When there is lack of sunlight, the SEVs operates as a simple EV. Automobile manufacturers and research community deal with the open challenges of SEVs.

BEVs, PHEVs, and HEVs differ significantly in terms of powertrain architectures, energy management strategies, and control complexity. For instance, BEVs rely entirely on high-capacity lithium-ion battery packs and electric drivetrains, requiring continuous monitoring of state-of-charge, state-of-health, and thermal dynamics, while PHEVs and HEVs involve coordinated control between internal combustion engines and electric motors. These architectural differences lead to heterogeneous operational data profiles, which in turn demand tailored analytics pipelines for fault diagnosis, energy optimization, and predictive maintenance.

2.3. Types of Electric Vehicle Charging Systems

Having described the different types of EVs, it is important to investigate how the EVs can be charged, namely through conductive and inductive charging systems.

Conductive charging technologies use direct contact of the EVs to the charging inlet. AC and DC chargers are employed within conductive charging technologies. AC charging, which is further categorized into Level-1, Level-2, and Level-3, is carried out using AC chargers. Level-1 is also known as home charging and is used by BEVs, PHEVs, and SEVs. Slow charging is achieved through Level-1 charging technology. Level-2 charging is usually used in private and public locations, and it is commonly known as fast AC charging. Level-2 is used in PHEVs and SEVs. DC charging or Level-3 charging is carried out using a DC charger, and they are faster chargers with maximum power capacity usually found in private or public locations. Level-3 is used in BEVs, PHEVs, and SEVs. Inductive charging technologies use electromagnetic field to transfer energy between two objects, i.e., charger and EV. The charging of the vehicles is performed in a wireless mode, and electromagnetic waves are employed for coupling. The EVs’ inductive charging is, however, slower and more expensive. Nowadays, conductive charging is widely used, but inductive charging technologies draw the attention of the research community, and their extended use is expected as their challenges and open issues are well investigated.

EV charging technologies, including Level-1, Level-2, and DC fast charging, exhibit distinct electrical characteristics in terms of voltage levels, current profiles, and power ratings, which significantly impact battery degradation rates, grid load patterns, and charging behavior modeling. For example, DC fast charging introduces high C-rates that accelerate thermal stress and aging, generating high-frequency data streams that can be exploited by BI and machine learning models for battery health prediction and charging optimization. Wireless and bidirectional charging further introduce complex control and communication requirements, enabling advanced applications such as Vehicle-to-Grid (V2G) services and demand response.

2.4. Charging and Discharging Methods

Charging EVs can be unidirectional and bidirectional. Firstly, unidirectional charging refers to the case when energy is transmitted solely from the energy grid to the EV. In unidirectional charging, electricity is converted to energy that can be stored in the battery pack of the EV so as to be used in powering the vehicle. This method of charging is the most popular nowadays. Nonetheless, a lot of effort is given towards a new charging technology, referred to as bidirectional charging. When employing bidirectional charging, an EV not only can receive electricity from the power grid, but additionally, the EV can send energy to different recipients, i.e., the power grid, home or business appliances, or even to other EVs. The four different types of bidirectional charging are the following: Vehicle-to-Grid (V2G), Vehicle-to-Home (V2H), Vehicle-to-Load (V2L), and Vehicle-to-Vehicle (V2V). V2G is the most common application of bidirectional charging, which allows EVs to send energy directly back to the power grid. V2H charging sends energy to home. V2L can use the EV’s battery pack to power appliances on the go. Lastly, in V2V, energy from one vehicle to another is exchanged. The research community is working towards bidirectional charging, which clearly has major advantages. First, bidirectional charging allows EVs to store backup energy that may be used in home or business appliances. When energy is transferred from one vehicle to another, EVs’ owners can even gain money through the exchange. Second, the EV is transformed into a portable energy source, and since energy storage is challenging, EVs appear as a promising solution. Third, bidirectional charging guides EVs’ owners towards more flexible charging practices. For example, EVs can choose to charge their vehicles during off-peak hours or when renewable energy resources are available.

3. Electric Vehicles Data Analysis with Business Intelligence

Electric vehicles generate massive volumes of heterogeneous data from multiple sources, including battery management systems, power electronics, onboard sensors, GPS modules, charging stations, and grid interfaces. These data streams vary in sampling rates, formats, and reliability, requiring robust BI architectures capable of handling high volume, velocity, and variety. A typical BI pipeline for EV systems consists of data acquisition at the edge, preprocessing and cleaning, feature extraction, integration into data warehouses or data lakes, as well as analytical processing through online analytical processing (OLAP), machine learning models, and visualization dashboards to support operational and strategic decision-making.

The application of AI and BI techniques in transportation technology improves performance, fosters creativity, enables intelligent charging of EVs, and builds intelligent environments. As required by today’s and tomorrow’s societies, BI aids in the control of energy consumption, safety, security, and the creation of an eco-friendly, pollution-free environment. BI and EVs are the solutions that society needs to see intelligent breakthroughs in user-friendly EVs. Predictive models based on machine learning and deep learning are employed to forecast battery degradation, energy consumption, traffic risks, and charging demand, while prescriptive analytics can recommend optimal charging schedules, routing strategies, and maintenance actions. These advanced analytics capabilities transform BI platforms into intelligent decision-support systems that can autonomously adapt EV operations to dynamic environments.

The EV industry is rapidly evolving especially due to the automation technologies and the electrification. The EVs owners’ expectations are constantly increasing, e.g., they want safer vehicles with more connectivity, and governments are constantly coming up with new laws to keep up with the worldwide movement toward a more environmentally friendly society. Furthermore, applications in EVs, electrical/electronics, embedded systems, and sensing technologies supporting autonomous vehicles are just a few areas where innovative technologies are applied. These technologies are present everywhere, from traffic lights in urban regions to the EVs’ systems. One of the main forces behind the current transformation in the automotive sector is the electrification of vehicles. AI and BI can be leveraged in different aspects of EVs, of which few are outlined and investigated in the next subsections (depicted in Figure 2):

1.

EV Accident Analysis: The engineering and scientific community investigate thousands of accidents involving EVs and various types of transportation vehicles in general. Given the large amount of data, intelligent systems can be proposed in order to decrease the number of transportation accidents. The exploitation of big data enables us to overcome the limitations and challenges of traditional transportation systems. Therefore, machine learning methods along with BI analysis can be developed to enhance safety in the roads [29].

2.

EV Battery Health Prediction: For the EVs to be reliable and safe, the examination of battery health is essential. For batteries to be trustworthy, their condition must be accurately estimated and predicted accurately. In recent years, there has been a huge discussion over EV technology and transportation of big data. In the realm of battery condition estimation, this tendency has prompted the use of data-driven techniques. Thus, BI and AI methodologies are appealing solutions that promise substantial contributions to both guaranteeing the vehicle’s safe operation and comprehending the internal state of the battery in real time [30].

3.

EV Charging Station Analysis: Another crucial objective in EV systems is how to determine the utilization of EV charging stations. With the exploitation of the data, we are able to determine which stations are used the most and which stations are used the least. These data are important for decision-making in various problems concerning the charging stations, such as the installation of new charging stations. When EV charging stations are not positioned and sized properly, increased energy losses may appear, which can negatively impact the growth of EVs, the design of the city traffic network, and the convenience of the EVs’ owners [31].

4.

Intelligent Charging Station Infrastructures: Since the power grid must keep up with the rising demands of EVs, voltage fluctuations and higher power consumption are just a few of the serious problems that the adoption of EVs has brought. Wind and solar power are not predictable because of weather fluctuations and changes. Thus, it is now crucial to balance supply and demand with temporary energy storage. EVs are capable of acting as energy storage systems due to their large battery capacities. Additionally, in order to transmit energy between the grid and EVs in an efficient manner, bidirectional information exchange needs to be achieved. As a result, it is obvious that smart charging infrastructure is required in order to address these issues and maintain the power grid [32].

5.

Locating EV Charging stations: As the number of EVs is increasing, it is important for the EVs’ owners to have an easy access to the charging stations. BI and AI strategies can provide effective solutions to locating a charging station that is deployed in an appropriate location to balance demand and supply. Prediction models can be used to improve user experience in the transportation system. Additionally, sharing charging solutions could be introduced in order to maximize the utility of the charging stations. Locating the available charging stations is very important in conducting the appropriate charging techniques [33].

6.

Autonomous Driving: With the use of advanced technology, including BI and AI, autonomous driving allows EVs to navigate on their own without the need of human interaction. In addition, it promises to facilitate transporting people with mobility impairments and reduce accidents while offering users safe and convenient services. However, autonomous driving still has to overcome a number of formidable obstacles, including privacy and data breaches, inefficiencies and poor fault tolerance in centralized management, and the difficulty of guaranteeing data accuracy. Towards this direction, blockchain technology, data analysis, sensors and actuators, complex algorithms and machine learning systems could guarantee data security, integrity, and accuracy [34].

To systematically investigate the role of BI and analytics across the EV ecosystem, in our work, we review relevant studies across multiple application domains. The study selection process followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines to ensure methodological transparency and reproducibility. A structured search strategy was applied across major scientific databases, and the identification, screening, eligibility, and inclusion stages are summarized in the PRISMA flow diagram (Figure 3), using combinations of keywords related to electric vehicles, business intelligence, data analytics, machine learning, charging infrastructure, battery health, accident analysis, and autonomous driving. Only peer-reviewed journal articles and conference papers published in English were considered. The inclusion criteria were as follows: (i) studies focusing on electric vehicles or EV-related infrastructures, (ii) application of BI, data analytics, machine learning, or AI techniques, and (iii) clear methodological descriptions and technical contributions. The exclusion criteria included the following: (i) studies not related to EV systems, (ii) works lacking data-driven or analytical components, (iii) papers focused solely on policy, market analysis, or mechanical design without analytics, and (iv) studies with insufficient technical detail. The diagram illustrates the number of records retrieved, duplicates removed, reports excluded during title and abstract screening, full-text articles assessed for eligibility, and studies ultimately included in the qualitative synthesis. By explicitly documenting each stage of the review process, the PRISMA flowchart provides a clear and systematic overview of how relevant literature was selected and ensures compliance with established standards for systematic and scoping reviews. More specifically, during the identification phase, a total of 480 records were retrieved through structured searches across major scientific databases, including IEEE Xplore, Scopus, Web of Science, and ScienceDirect. After removing 120 duplicate records, 360 unique studies remained for further evaluation. In the screening phase, titles and abstracts of these records were reviewed, leading to the exclusion of 250 articles that did not directly address business intelligence or analytics applications in electric vehicle systems. The remaining 110 articles underwent full-text eligibility assessment, during which 35 studies were excluded due to insufficient methodological detail, lack of relevance to EV-related analytics, or focus on non-data-driven approaches. Finally, 75 studies met all inclusion criteria and were selected for qualitative synthesis as well as detailed analysis in this survey. This systematic selection process ensures transparency, reproducibility, and compliance with the PRISMA 2020 guidelines.

Due to the heterogeneity of datasets, evaluation metrics, and experimental settings across EV application domains, this survey focuses on qualitative synthesis and comparative analysis rather than a quantitative meta-analysis. This survey emphasizes data-driven BI and analytics methodologies across EV subsystems, rather than in-depth physical modeling of individual components, to provide a holistic system-level perspective on intelligent EV ecosystems.

3.1. Electric Vehicles Accident Analysis

The rapid adoption of EVs worldwide is inevitably accompanied by growing concerns regarding accident risks. While traditional accident analysis has focused primarily on internal combustion engine vehicles, EVs present unique safety challenges due to their distinct design and operating characteristics, e.g., the weight distribution of batteries, the risk of post-crash fires, the quiet operation of EVs that may endanger pedestrians. Therefore, systematic accident analysis is critical, not only to improve the safety of EVs themselves but also to inform infrastructure design, regulatory frameworks, and insurance practices. A number of research works focused on accident analysis and prediction in order to offer safer driving. In this paper, we provide a systematic review of the state-of-the-art works that study vehicle accidents using artificial intelligence and machine learning techniques.

Predicting accidents involving EVs is a growing research focus. Machine learning and deep learning are used to identify risk factors, forecast the likelihood of crashes and assess the severity of an injury. The idea is to move to proactive prevention by forecasting potential accidents based on historical and real-time data. The models analyze factors such as vehicle speed, driver behavior, weather, road topology, and energy consumption patterns—even battery temperature, charging state, and regenerative braking data—to improve prediction accuracy. These predictive models support accident prevention, early warning systems, and safer vehicle and infrastructure design.

Recent literature reviews have systematically examined the application of machine learning and deep learning techniques for accident prediction in connected and autonomous vehicles, providing comprehensive overviews of current advancements and challenges [35,36,37,38,39]. Systematic reviews also address the broader impact of artificial intelligence on safety-critical systems and autonomous vehicle safety, emphasizing the need for robust, interpretable, and standardized approaches [40,41]. In [42], the authors examine how machine learning techniques can accurately predict car accidents, helping to identify key factors and propose effective prevention strategies for safer and community-friendly cities. In [43], machine learning methods for vehicles, pedestrians, and traffic detection show promising outcomes, but require standardization and experimentation for optimal results. The research works that focus on AI-based object detection and traffic prediction in autonomous vehicles are investigated thoroughly in [44]. Another study [45] categorizes factors affecting road accident risk, presents prediction algorithms, and outlines methods to explain risk assessment based on driving behavior. Deep learning approaches using camera vision to develop cost-efficient and more secure autonomous vehicle systems as well as the corresponding mechanisms are studied in [46].

Table 1. Taxonomy of techniques in EV accident analysis.

Category	Methods	Related Works
Traditional Machine Learning	Decision Trees, Random Forest, Support Vector Machines, Logistic Regression, Naïve Bayes, k-Nearest Neighbor (k-NN)	[35,42,45,47,48,49,50]
Deep Learning	Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs, LSTM, BiLSTM), Deep Belief Networks, Multilayer Perceptron (MLP), Graph Neural Networks (GNNs)	[42,44,45,48,50,51,52,53,54,55]
Hybrid/Integrated Models	CNN, LSTM, BiLSTM, Feature Fusion (e.g., combining feature engineering and deep learning)	[45,50,56]
Clustering and Unsupervised	K-means, DBSCAN, Unsupervised Feature Extraction	[48,49]
Vision-Based/Perception	Object Detection, Semantic Segmentation, Scene Analysis, Traffic Sign/Light Recognition	[42,44,46]
Application Domain	Perception (detection, recognition), Decision Making, Path/Motion Planning, Risk Assessment, Severity Prediction	[40,42,44,50]

summarizes existing research works on accident analysis for EVs. The studies are categorized studies according to their methodological approach and primary research focus. The table highlights how related works employ statistical modeling, machine learning techniques, or hybrid analytical frameworks to examine accident causes, severity patterns, and safety-related EV characteristics. By structuring the literature into clear categories, the table illustrates the diversity of analytical strategies applied in this domain. It also emphasizes the increasing importance of data-driven approaches and BI tools for improving EV safety and understanding accident dynamics. Overall, Table 1 provides a consolidated review of the state-of-the-art and establishes a foundation for identifying gaps that motivate future research.

The studies summarized in Table 1 indicate a clear shift from traditional statistical accident analysis toward data-driven models based on machine learning and deep learning, leveraging heterogeneous data sources such as vehicle sensors, traffic information, and environmental conditions. While recent approaches demonstrate improved prediction accuracy and real-time capability, most works rely on limited datasets collected under specific scenarios, which raises concerns about generalization across different regions, driving behaviors, and road conditions. Moreover, few studies explicitly address model interpretability and robustness, which are critical for safety-critical EV applications. These observations suggest that future research should focus on large-scale multi-source data integration, explainable accident prediction models, and edge-enabled analytics frameworks capable of supporting low-latency safety decision-making in real-world deployments.

3.2. Electric Vehicles Battery Health Prediction

Battery health prediction is a critical component of EV safety, reliability, and lifecycle management. As lithium-ion batteries age, their capacity, internal resistance, and thermal stability gradually degrade, affecting driving range and overall performance. Accurately estimating the state of health and forecasting future degradation patterns enables manufacturers, fleet operators, and EV owners to schedule timely maintenance, reduce unexpected failures, and optimize charging strategies. Traditional model-based approaches rely on electrochemical models and empirical degradation profiles. However, these methods are often limited by their dependence on predefined assumptions and difficulty in capturing nonlinear aging factors such as temperature fluctuations, variable load cycles, and driver-specific behaviors. With the rapid growth of EV deployments and the increasing availability of operational data, data-driven battery health prediction has become a promising and scalable solution for real-world applications.

In recent years, AI and machine learning techniques have shown exceptional potential in addressing the complex nature of EV battery degradation. Machine learning models, such as Support Vector Regression (SVR), Random Forests, Gradient Boosting, and Neural Networks, have been widely adopted for state-of-health estimation due to their ability to learn patterns from large datasets without requiring explicit physical modeling. Deep learning methods, including Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNNs), are particularly effective in capturing the temporal dynamics of battery aging using voltage, current, temperature, and charging/discharging profiles. Moreover, these models can be integrated into BI platforms to generate real-time dashboards for fleet monitoring, predictive maintenance scheduling, and early fault detection. Despite the research progress, challenges remain, including limited access to high-quality labeled battery datasets, the need for model interpretability, and variability in battery types. Ongoing research aims to combine physics with data-driven learning to enhance robustness and generalization while maintaining transparency and safety in EV battery health prediction.

Researchers investigated the battery health prediction problem thoroughly in the literature. A number of advanced AI techniques that can accurately predict lithium-ion batteries’ life by identifying key indicators for performance degradation are described in [57]. Machine learning and more specifically, deep learning approaches are proposed to estimate EVs’ battery health and charging levels in [58,59,60,61]. The authors in [62] review state-of-health estimation methods for lithium-ion batteries in HEVs, highlighting the need for tailored techniques and use realistic scenarios to outline the performance and the accuracy of the algorithms. Sui et al. [63] examine how Support Vector Machine and Neural Network algorithms estimate the state of health of batteries. Last but not least, data-driven methods are offering great accuracy and adaptability for estimating EVs batteries’ health in [64,65,66,67]. Recent studies have also highlighted the importance of accurate state-of-health estimation for enabling second-life applications of EV batteries in grid-scale energy storage, where technical, economic, and regulatory factors must be jointly considered, as illustrated in the SWOT-based analysis reported in [68]. Battery health analytics are investigaed beyond first-life usage by integrating degradation tracking and lifecycle KPI monitoring into adaptive management frameworks for second-life EV batteries in [69].

Table 2. Taxonomy of techniques in EV battery health prediction.

Category	Methods	Related Works
Direct measurement	Capacity tests, resistance measurement	[33,70,71,72]
Model-based	Electrochemical, circuit models, adaptive filtering	[33,70,71,72,73,74,75,76]
Data-driven	Machine Learning/Deep Learning (SVM, RF, LSTM, CNN, Digital Twin)	[33,64,65,70,71,73,75,77,78,79,80,81]
Hybrid	Physics-informed ML, co-estimation	[70,71,73,75,76,78,82,83,84]

summarizes representative studies on EV battery health prediction by comparing their data sources, analytical methods, and application objectives. Beyond highlighting dominant research directions such as state-of-health estimation and remaining useful life prediction, the table also reveals common methodological limitations, including reliance on laboratory datasets, limited generalization across battery chemistries, and high computational complexity. This comparison provides insight into the critical role of BI and analytics frameworks in enabling scalable, real-time battery monitoring and predictive maintenance in practical EV deployments.

From the comparative analysis in Table 2, several important insights emerge. First, most existing approaches rely heavily on supervised learning models trained on historical cycling data, which limits their adaptability to real-world driving conditions and evolving battery usage patterns. Second, while deep learning methods achieve high prediction accuracy, their lack of interpretability and high computational cost pose challenges for onboard deployment. Third, there is a clear trend toward integrating multi-source data, including temperature, voltage, current, and driving behavior, indicating that future BI frameworks must support multimodal data fusion and real-time analytics. These observations suggest that scalable, explainable, and edge-enabled BI architectures will be essential for next-generation EV battery intelligence.

3.3. Electric Vehicles Charging Station Analysis

As EVs become more widespread, analyzing the utilization and performance of CSs has become a fundamental component of intelligent transportation systems. Efficient charging station planning requires understanding when, where, and how frequently charging stations are used, as well as identifying patterns in charging duration, user behavior, and energy demand. Traditional analytical approaches often rely on statistical methods to evaluate temporal and spatial utilization trends. However, these methods may fall short in capturing the complex interactions between charging demand, urban traffic flow, and heterogeneous user profiles. High utilization rates may signal insufficient infrastructure or inadequate charging power, whereas low utilization may indicate suboptimal station placement or misalignment with user mobility patterns. Therefore, accurate and data-driven charging station analysis is essential for improving user satisfaction, reducing waiting times, optimizing charging infrastructure investment, and ensuring the reliable integration of EVs into the power grid.

Recent advances in BI, AI, and machine learning have significantly enhanced the precision and depth of EV charging station analysis. Machine learning techniques, such as clustering, time-series forecasting, neural networks, and reinforcement learning, are widely applied to classify station usage patterns, predict future charging demand, and generate insights for infrastructure expansion. For instance, clustering methods can categorize stations based on their usage intensity or user demographics, while forecasting models can anticipate peak loads to support smart grid management. Spatial analytics integrated with Geographic Information Systems (GIS) enable the identification of high-demand zones and charging bottlenecks, facilitating optimal placement of new charging stations. Additionally, BI dashboards allow city planners, power utilities, and mobility providers to visualize real-time utilization data and perform scenario analysis. Despite these advances, challenges persist in terms of data heterogeneity, limited interoperability between charging networks, and the need for scalable predictive models capable of adapting to rapidly changing EV adoption patterns. Continued research in BI-driven charging infrastructure analysis will be crucial for developing efficient, resilient, and user-centric EV charging ecosystems.

AI models have been thoroughly investigated for dynamic pricing, for identifying charging patterns, for charging scheduling, and for traffic routing [85,86]. The authors in [87,88] provide comprehensive state-of-the-art reviews considering machine learning techniques and reinforcement learning for solving the problem of optimal charging scheduling. Refs. [89,90] review various methodologies for identifying the suitable location for EVs’ charging in urban environments. Business models focusing on smart CSs planning and EVs charging operation are examined in [91].

Table 3. Taxonomy of techniques in charging station analysis and utilization.

Category	Methods	Related Works
Statistical and Time Series	ARIMA, SARIMA, Seasonal Decomposition, Regression	[92,93,94,95]
Machine Learning	Random Forest, XGBoost, CatBoost, LightGBM, SVM, k-NN	[95,96,97,98,99]
Deep Learning	RNN, LSTM, Bi-LSTM, GRU, CNN, Transformer, Hybrid Models	[94,100,101,102,103,104]
Reinforcement Learning	Q-Learning, Multi-Agent RL	[102,105]
Clustering and Pattern Mining	K-means, Matrix Profiles, Symbolic Aggregate Approximation	[55,106]

reviews research works that investigate the utilization, performance, and operational behavior of EV charging stations. These studies apply BI- and AI-based analytics to evaluate charging patterns, station occupancy rates, user behavior, and energy demand fluctuations. The categorization in the table makes clear how different techniques, ranging from statistical analysis and clustering to advanced prediction models, facilitate accurate assessment of charging infrastructure performance. Through this structured overview, the table shows how intelligent analysis can support decision-making for improving charging efficiency, reducing waiting times, and planning capacity expansions. Table 3 provides a comprehensive snapshot of analytical methods that underpin modern charging infrastructure management.

From Table 3, it can be observed that existing research on charging station utilization increasingly employs spatiotemporal data analytics and demand forecasting models to characterize user behavior and charging patterns. Although many studies report promising results in predicting station occupancy and waiting times, most approaches assume static user preferences and simplified mobility patterns, which may not fully capture the dynamic nature of urban transportation systems. In addition, the limited consideration of real-time grid constraints and pricing mechanisms restricts their practical applicability. These findings highlight the need for adaptive, context-aware BI frameworks that jointly model mobility, user behavior, and grid conditions to enable more accurate and operationally relevant charging utilization analytics.

3.4. Intelligent Charging Stations Infrastructure

The increasing penetration of EVs places significant stress on existing power grids, leading to challenges such as voltage instability, peak load amplification, and fluctuating power quality. These issues become more pronounced when considering the growing reliance on renewable energy sources, such as wind and solar, whose output is inherently variable and weather-dependent. Intelligent charging station infrastructures aim to mitigate these challenges by optimizing the interaction between EVs, charging stations, and the grid. Key strategies include load balancing, demand-side management, and dynamic pricing schemes that shift charging activities to off-peak hours. Furthermore, Vehicle-to-Grid (V2G) and Vehicle-to-Home (V2H) technologies enable EVs to serve as distributed energy storage units, allowing them to discharge stored energy back into the grid or local loads during periods of high demand. This bidirectional flow of electricity transforms EVs from passive energy consumers into active participants in grid stability, supporting energy resilience and improving the integration of intermittent renewables.

Advancements in BI, AI, and communication technologies are crucial to realizing intelligent charging infrastructures. Machine learning models can forecast charging demand, renewable energy availability, and grid conditions, enabling real-time optimization of charging schedules and power distribution. Smart charging algorithms, such as reinforcement learning, model predictive control, and multi-agent systems, coordinate charging events across networks of EVs to minimize grid impact and maximize economic efficiency. At the same time, bidirectional communication via IoT sensors, edge computing nodes, and advanced metering infrastructure ensures seamless information exchange between EVs, charging stations, and utility operators. BI dashboards provide actionable insights for monitoring energy flows, predicting congestion, and supporting grid operators in decision-making processes. Despite significant progress, challenges remain in terms of ensuring interoperability among different charging networks, safeguarding data privacy, and scaling intelligent control systems to accommodate future EV adoption. Continued research into smart charging infrastructure is essential for enabling a stable, flexible, and sustainable energy ecosystem.

Comprehensive reviews of energy management techniques for EVs and their integration in current and future smart grids are investigated in [107,108,109,110]. The authors in [111] examine the integration of smart metering in and the integration of AI and machine learning algorithms are investigated in [112]. The studies [113,114,115] explore various smart charging technologies and their impact on the power grid. The use of renewable resources is examined in [116].

Table 4. Taxonomy of techniques in intelligent charging stations infrastructure.

Category	Methods	Related Works
Load Balancing and Demand Response	Machine Learning (LSTM, DNN, RL, federated learning), dynamic pricing, demand prediction	[117,118,119,120,121,122]
Smart Scheduling and Optimization	Metaheuristics (ACO), multi-agent systems, adaptive priority, dynamic reservation	[119,122,123]
Communication and IoT Integration	IoT sensors, edge computing, 5G, Zigbee, LoRa, OCPP, cloud/edge platforms	[124,125,126,127]
Dynamic Pricing and Incentive Schemes	RL-based pricing, user digital twins, behavioral economics	[120,122,124,126,128]
Security and Privacy	Blockchain, federated learning, cybersecurity protocols	[13,112,121,125]

categorizes recent contributions that focus on developing intelligent charging infrastructures capable of supporting the growing energy demands introduced by EVs. The selected works explore smart charging strategies, bidirectional energy exchange (V2G, V2H), demand–response mechanisms, and grid stability optimization. By organizing studies according to their technological solutions—such as optimization algorithms, machine learning forecasting tools, or IoT-enabled control systems—the table highlights how intelligent infrastructures enhance grid resilience, balance supply and demand, and integrate renewable energy sources. This overview demonstrates the central role of BI-driven insights and AI-enhanced control in enabling next-generation, flexible charging systems.

The works summarized in Table 4 reveal a strong research focus on optimizing charging schedules and coordinating EV charging with renewable energy integration and grid stability objectives. Optimization-based and learning-based methods dominate this domain, reflecting the increasing complexity of charging infrastructures. However, many existing solutions are validated primarily through simulations and small-scale case studies, with limited real-world deployment evidence. Furthermore, interoperability across heterogeneous charging networks and the scalability of centralized optimization frameworks remain open challenges. These observations suggest that future intelligent charging infrastructures should incorporate decentralized control, edge intelligence, and standardized data exchange mechanisms to achieve scalable and resilient grid-aware charging operations.

3.5. Locating Electric Vehicles Charging Stations

The rapid growth of EV adoption has intensified the need for strategically placed charging stations to ensure reliable and convenient access for EV owners. Poorly located charging infrastructure can result in long waiting times, uneven utilization of charging stations, increased range anxiety, and overall dissatisfaction with the transportation system. Consequently, optimal placement of charging stations is a critical factor in fostering widespread EV adoption and supporting sustainable urban mobility. Traditional location planning often relies on static demographic and traffic data; however, such approaches are insufficient in dynamically evolving urban environments where mobility patterns frequently change. BI tools enhance this process by integrating diverse datasets, such as travel demand, population density, land-use patterns, grid constraints, and socio-economic data, to identify areas where charging demand is likely to grow. Predictive analytics further enables the anticipation of future hotspots by analyzing temporal patterns, seasonal variations, and the projected increase in EV penetration, thereby ensuring that infrastructure investments remain both adaptive and cost-effective.

AI and machine learning techniques further advance charging station location planning by enabling data-driven optimization and simulation-based decision-making. Methods such as clustering algorithms, multi-criteria decision analysis, and reinforcement learning can identify optimal locations that balance accessibility, grid stability, and economic efficiency. GIS combined with machine learning models allow cities to visualize spatial relationships between mobility flows, existing charging infrastructure, and potential candidate sites. Moreover, emerging concepts such as shared charging solutions, where charging infrastructure is jointly used by private EV owners, and public users can be optimized using BI to maximize station utilization and reduce infrastructure redundancy. Real-time data from IoT-enabled charging stations can also support dynamic recommendations, guiding drivers to the nearest available station and reducing congestion. Despite substantial progress, challenges remain in integrating heterogeneous datasets, accommodating uncertainties in future EV adoption, and ensuring equitable access across urban and rural areas. Continued advancements in BI- and AI-driven location planning will be essential for creating efficient, resilient, and user-centered EV charging ecosystems.

A comprehensive review of exact, heuristic and metaheuristic optimization solutions are investigated for the complex multi-criteria placement problem [129,130,131,132]. GIS-based methods and multi-criteria decision analysis are increasingly used to integrate spatial, demographic, and technical data for site selection, especially in urban contexts [89,133,134].

Table 5. Taxonomy of techniques in charging stations infrastructure placement.

Category	Methods	Related Works
Mathematical and Optimization Modeling	Linear programming, Genetic Algorithms, Particle Swarm Optimization, Peer-to-peer negotiation, Hybrid metaheuristics	[135,136,137,138,139,140]
Machine Learning and Data-Driven Approaches	Use of machine learning (e.g., Random Forest, Linear Regression), Clustering, Predictive analytics	[141,142,143]
Simulation and Agent-Based Models	Agent-based demand simulation, Discrete choice models	[137,144]
Swarm Intelligence and Evolutionary Algorithms	Artificial Bee Colony (ABC), Genetic Algorithms, Simulated Annealing, Social Network Optimization	[138,145,146]
Game Theory and User Behavior Modeling	Game-theoretical frameworks, Discrete choice models	[136,137,147]

compiles research works dedicated to the optimal placement of EV charging stations, a key challenge for ensuring accessibility and balancing charging demand with available capacity. The table outlines different methodological approaches—including GIS, heuristic optimization, clustering techniques, and prediction-based frameworks—that determine ideal charging-station locations. By classifying studies based on their chosen analytical tools and optimization objectives, Table 5 illustrates the major trends in charging station placement research and reveals how BI and AI strategies can significantly improve infrastructure deployment. This categorization highlights the importance of integrating spatial, demographic, transportation, and energy-grid data into the decision-making process.

As shown in Table 5, research on charging station location planning largely adopts spatial optimization, clustering, and multi-criteria decision-making approaches to balance cost, accessibility, and demand coverage. While these methods effectively identify candidate locations under static assumptions, many studies overlook the temporal evolution of EV adoption, mobility patterns, and charging demand. Additionally, limited attention is given to uncertainty modeling and real-time adaptability in planning frameworks. This suggests that future BI-driven location models should integrate long-term demand forecasting, dynamic urban mobility data, and scenario-based analytics to support more flexible and future-proof infrastructure deployment strategies.

3.6. Autonomous Driving

Autonomous driving represents a transformative advancement in the EV ecosystem, enabled by the integration of BI, AI, and sophisticated sensing technologies. By leveraging data-driven perception, decision-making, and control systems, autonomous EVs are capable of navigating complex environments with minimal or no human intervention. These vehicles rely on a diverse set of sensors -including LiDAR, radar, cameras, ultrasonic sensors, and GPS- to capture real-time information about their surroundings. AI algorithms, such as deep neural networks and reinforcement learning models, analyze this data to detect obstacles, recognize traffic patterns, predict human behavior, and determine optimal driving actions. BI tools support this ecosystem by aggregating and visualizing large-scale mobility data, enabling stakeholders to evaluate system performance, optimize traffic flows, and design safer transportation networks. Ultimately, autonomous EVs hold the potential to reduce accident rates, improve mobility for elderly and disabled individuals, and enhance overall transportation efficiency through coordinated and intelligent routing.

Despite these promising advancements, fully autonomous driving faces significant technical, ethical, and operational challenges. One of the primary obstacles is ensuring the accuracy, integrity, and security of the massive volumes of data required for autonomous decision-making. Centralized data management systems may suffer from latency, single points of failure, and vulnerabilities to cyberattacks or data breaches. Additionally, variations in sensor reliability, unpredictable environmental conditions, and the inherent uncertainty of machine learning models introduce risks that must be carefully managed in safety-critical applications. To address these issues, emerging technologies such as blockchain are being explored to create decentralized and tamper-resistant data-sharing frameworks, ensuring transparency and trustworthiness across autonomous driving systems. While blockchain can enhance trust and data integrity, its latency and energy overheads limit its applicability in real-time control loops, necessitating lightweight, permissioned, and off-chain designs that complement rather than replace time-critical autonomous driving functions. Advanced sensor fusion techniques, edge computing, and explainable AI also play critical roles in enhancing reliability, reducing computational delays, and improving fault tolerance. Continued research into robust algorithms, secure communication protocols, and regulatory frameworks will be essential for enabling autonomous EVs to operate safely, efficiently, and ethically within future intelligent transportation environments.

A number of literature review papers examine the integration of AI in EVs technology for safety, localization, mapping, and quick decision-making [40,148,149,150,151]. Innovative methods concering autonomous vehcicle technology are the main focus in [152] and in [153]. Deep learning approaches improve autonomous driving through the use of sensors and cameras [46,154,155].

Table 6. Taxonomy of machine learning techniques in autonomous driving.

Category	Methods	Related Works
Deep learning	Deep learning, Reinforcement learning, Convolutional and recurrent neural networks	[156,157,158,159,160,161,162,163]
Decision-making	Path planning, optimization, Metaheuristic optimization, Hybrid solutions	[164,165]
Sensors	Deep learning	[148,166]
Hybrid	Deep and reinforcement learning	[167,168,169]

summarizes representative studies related to autonomous driving technologies within the EV ecosystem, emphasizing analytical, algorithmic, and data-centric methods. The selected works apply machine learning, computer vision, sensor fusion, and intelligent decision-making frameworks to enable perception, prediction, planning, and control in autonomous vehicles. The categorization clarifies which studies focus on object detection, trajectory prediction, system reliability, or cybersecurity, thereby outlining the breadth of challenges addressed by current research. By providing a structured overview, Table 6 underscores the pivotal role of BI, AI, and advanced analytics in achieving safe, efficient, and reliable autonomous driving systems.

Table 6 highlights the growing adoption of deep learning, sensor fusion, and real-time data analytics in autonomous EV systems to support perception, decision-making, and control tasks. Despite significant progress, most approaches require large labeled datasets and intensive computational resources, which constrain their scalability and real-time deployment in resource-limited environments. Moreover, issues related to data security, privacy preservation, and system robustness are often treated as secondary concerns. These insights indicate that future autonomous EV research should prioritize lightweight and explainable learning models, edge–cloud collaborative analytics, and secure data management frameworks to ensure reliable and trustworthy autonomous driving services.

4. Cooperation and Incentive Mechanisms in EV Ecosystems

The successful deployment of intelligent EV ecosystems relies not only on technological advancements but also on the ability to motivate diverse stakeholders, including EV owners, charging station operators, grid utilities, and manufacturers, to participate in cooperative decision-making processes. These stakeholders often have conflicting objectives, such as minimizing charging costs, maximizing station utilization, maintaining grid stability, reducing operational risks, or enhancing mobility services. BI and analytics play a pivotal role in aligning these objectives by providing transparent insights, quantifying benefits, and enabling data-driven coordination strategies.

4.1. Economic Incentive

A primary mechanism to foster cooperation is the implementation of dynamic and fair economic incentive schemes. BI-driven demand forecasting and user clustering allow charging operators and utilities to design pricing models, such as time-of-use tariffs, real-time congestion pricing, rebates for off-peak charging, and compensation for V2G participation that motivate users to contribute to grid-friendly behaviors. These incentives transform what would otherwise be voluntary cooperation into economically beneficial choices for EV owners. However, it should be noted that user responses to dynamic pricing schemes are influenced by heterogeneous preferences, risk attitudes, and contextual factors, and may deviate from rational optimization assumptions, highlighting the need for behavior-aware and human-in-the-loop analytics models. Similarly, charging station operators can optimize revenue and operational cost by leveraging analytics dashboards that reveal patterns in user behavior, peak load periods, and energy procurement costs, thus aligning their incentives with system-level optimization goals.

4.2. Information Transparency and Decision Support

Cooperation increases when stakeholders have access to reliable, real-time information that reduces uncertainty. BI platforms enable transparent visualization of charging demand, station availability, grid stress levels, renewable energy forecasts, and mobility patterns. This transparency allows both EV users and infrastructure operators to make informed decisions that collectively minimize congestion, improve service quality, and support grid stability. Decision-support systems powered by predictive analytics can recommend optimal charging times, node balancing strategies, or fleet-charging schedules, ensuring that local decision-making aligns with broader system requirements.

4.3. Data-Sharing Frameworks and Digital Platforms

Effective cooperation requires robust data sharing mechanisms that facilitate coordinated actions while protecting privacy. Emerging architectures, such as cloud–edge hybrid systems, federated learning, and secure IoT communication frameworks, enable the aggregation of insights from distributed stakeholders without exposing sensitive data. Charging operators can share utilization statistics, grid operators can broadcast network status indicators, and EVs can contribute predictive mobility information, all without compromising individual privacy. BI platforms serve as integration layers that unify these distributed data streams, enabling multi-stakeholder coordination in real time.

4.4. Trust, Transparency, and Blockchain-Based Coordination

Trust is a critical prerequisite for cooperation, especially when multiple commercial entities are involved. Blockchain offers decentralized, tamper-resistant mechanisms for recording energy transactions, verifying V2G contributions, establishing smart contracts, and managing peer-to-peer charging and energy trading. Smart contracts can automatically enforce cooperation rules, such as compensation for grid services or penalties for overloading nodes, ensuring transparency and eliminating the need for a central authority. This strengthens stakeholder confidence and enables more flexible market interactions.

4.5. Multi-Objective Optimization and Fairness Considerations

EV ecosystems inherently involve multi-objective optimization problems, balancing user convenience, grid reliability, economic efficiency, and environmental sustainability. BI-enabled optimization algorithms can incorporate fairness constraints to ensure that benefits are equitably distributed among stakeholders—for example, guaranteeing minimum charging access for users in underserved regions or preventing price discrimination during high-demand periods. By reflecting diverse stakeholder preferences and constraints in the analytical models, cooperation becomes sustainable and socially acceptable.

4.6. Long-Term Strategic Coordination

Beyond real-time operations, long-term planning plays a critical role in motivating cooperation. BI tools support scenario analysis, forecasting of EV adoption trends, infrastructure expansion planning, and investment decision-making for charging networks and grid upgrades. When stakeholders can clearly see long-term benefits—such as reduced operational risks, improved profitability, or enhanced service quality—they are more willing to engage in coordinated strategies. For policymakers, analytics can reveal where regulatory interventions, subsidies, or market frameworks may be needed to reinforce cooperative behavior across the EV ecosystem.

In summary, cooperation among EV stakeholders is achievable when incentive mechanisms, transparent information flow, secure data-sharing frameworks, and fairness-aware optimization methods work together to align individual decision-making with system-level performance goals. BI and analytics serve as the foundational technologies that make such alignment possible, enabling EV ecosystems that are efficient, scalable, reliable, and equitable.

5. Discussion

5.1. Fundamental Limits

5.1.1. Fragmented and Heterogeneous Data Ecosystems

A major limitation in BI-enabled EV applications lies in the fragmented structure of the underlying data sources. Accident datasets, battery degradation logs, charging events, autonomous driving sensor data, user mobility patterns, and grid operational metrics are often collected independently, using different sampling rates, proprietary formats, and inconsistent quality standards. This fragmentation hinders the integration of data into unified analytical frameworks and limits the scalability of cross-domain predictive models. Moreover, manufacturers, charging operators, and grid utilities impose strict restrictions on data sharing due to privacy, competition, and regulatory concerns, resulting in data silos that are difficult to harmonize. Consequently, many analytics models are trained on narrow datasets that fail to capture the diversity of real-world EV behavior. It should be noted that the competitive and proprietary nature of the automotive industry limits large-scale data sharing, and thus future BI frameworks must prioritize privacy-preserving, incentive-compatible, and interoperable solutions rather than assuming fully open data ecosystems.

5.1.2. Limited Availability of High-Quality Training Data

Although machine learning and deep learning techniques have shown promise across various EV applications, their performance is fundamentally constrained by the availability of large, high-quality labeled datasets. In areas such as accident prediction or autonomous driving, data scarcity arises from both the rarity of safety-critical events and the high cost of annotating sensor-rich recordings. Similarly, battery health prediction models often rely on laboratory datasets that do not accurately represent real-world conditions involving driver variability, extreme weather, or irregular charging practices. This mismatch between controlled and operational environments leads to reduced model generalization and weak robustness under dynamic conditions.

5.1.3. Over-Reliance on Centralized Architectures

Many BI and AI systems for EVs rely on centralized cloud platforms for data storage, processing, and model inference. While cloud computing provides scalable computation, it also introduces latency, bandwidth constraints, and single points of failure. For time-sensitive applications—such as fault detection in battery systems, real-time accident prediction, or autonomous vehicle decision-making—reliance on cloud-dependent pipelines can compromise safety and operational reliability. Centralized systems also expand the attack surface for cyber threats, making EV infrastructures vulnerable to data breaches, denial-of-service attacks, and model manipulation.

5.1.4. Lack of Explainability and Transparency in AI Models

A persistent challenge across EV analytics is the reliance on black-box machine learning models whose internal decision-making processes are not easily interpretable. In domains like safety assessment, grid–EV coordination, or autonomous navigation, the inability to justify model outputs poses serious concerns for accountability, regulatory approval, and public acceptance. Most deep learning models used today for trajectory prediction, anomaly detection, or charging optimization lack mechanisms to provide explainable reasoning. This gap limits trust and restricts deployment in mission-critical or legally governed environments.

5.1.5. Cybersecurity and Adversarial Vulnerabilities

As EVs, charging infrastructures, and grid operators become increasingly interconnected, cybersecurity risks become a fundamental limitation that cannot be overlooked. Malicious actors may exploit vulnerabilities in communication channels, tamper with sensor signals, or launch adversarial attacks against AI models, leading to incorrect predictions or dangerous control actions. The absence of robust authentication, secure communication protocols, and adversarially resilient AI methods poses significant threats to the safety and stability of EV ecosystems. The risk escalates further in autonomous driving, where manipulated perception data could directly trigger hazardous decisions.

5.2. Future Challenges

5.2.1. Decentralized and Privacy-Preserving Data Management

Future EV ecosystems will involve large volumes of heterogeneous data generated by vehicles, charging stations, road infrastructures, and grid components, making centralized data management increasingly vulnerable to scalability, privacy, and single-point-of-failure issues. A promising technical path lies in blockchain-enabled data platforms, where distributed ledgers can ensure data integrity, traceability, and trust among multiple stakeholders, such as EV owners, utilities, and service providers. In parallel, federated learning frameworks can enable collaborative model training across distributed EV nodes and edge devices without sharing raw data, thus preserving user privacy while benefiting from large-scale data intelligence. Typical application scenarios include collaborative battery health modeling across fleets, secure sharing of charging demand profiles among operators, and decentralized accident data reporting for real-time safety analytics.

5.2.2. Unified, Multimodal BI Pipelines for EV Ecosystems

Future research must focus on developing unified BI architectures capable of integrating heterogeneous datasets—ranging from telematics and sensor logs to energy forecasts and grid analytics—into coherent decision-support systems. Such pipelines must support multimodal data fusion (text, images, signals, geospatial data) and continuous model updating. Interoperability between automotive manufacturers, charging operators, and grid utilities must be standardized to enable large-scale data harmonization. This integration is essential for holistic EV intelligence across safety, energy, charging, and autonomous functions.

5.2.3. Next-Generation Battery Intelligence and Predictive Maintenance

Battery health prediction will require more advanced models that combine data-driven analytics with physics-informed machine learning and electrochemical knowledge. Future research should address the long-term prediction of battery aging under irregular operating conditions, rapid charging behaviors, and extreme climates. Real-time battery diagnostics integrated into BI dashboards must support predictive maintenance, fault prevention, and second-life assessment of battery packs. The integration of digital twins (e.g., virtual replicas of battery systems) represents a promising direction for improving accuracy and safety.

5.2.4. Grid-Aware Charging Optimization and Renewable Integration

As EV penetration increases, uncontrolled charging may lead to peak load amplification, voltage deviations, and transformer overloading in distribution networks. Future research should focus on grid-aware charging optimization frameworks that tightly couple EV charging decisions with real-time grid states. Technical paths include the use of deep reinforcement learning and model predictive control to dynamically schedule charging and discharging based on electricity prices, grid congestion levels, renewable generation forecasts, and user mobility patterns. Typical application scenarios involve coordinated charging of EV fleets in residential areas, workplace parking lots, and fast-charging corridors, where predictive analytics can minimize grid stress while ensuring user satisfaction. Moreover, integration with Vehicle-to-Grid (V2G) schemes enables EVs to act as distributed energy storage units that actively support grid stability and renewable energy smoothing.

5.2.5. Advanced Autonomous Driving Intelligence and Safety Certification

The next phase of autonomous driving research must strengthen reliability, explainability, and safety certification. Integrating explainable AI, uncertainty estimation, risk-aware prediction, and formal verification techniques is essential for guaranteeing safe decision-making under unpredictable real-world conditions. Future systems must handle adversarial attacks, sensor degradation, extreme weather, and rare safety-critical scenarios using robust perception and planning algorithms. Edge–cloud collaborative architectures will also be required to reduce latency and improve fault tolerance in real-time autonomous operations.

5.2.6. Robust Cybersecurity Frameworks for EV Infrastructures

As EV systems become more intelligent and interconnected, cybersecurity will remain a significant challenge. Future research must design secure communication protocols (e.g., secure OCPP and ISO 15118 extensions), develop intrusion detection systems tailored to EV networks, and create adversarially robust machine learning models. Ensuring the integrity of sensor data, vehicle behavior logs, and energy transactions is critical for protecting users and maintaining public trust. Beyond identifying cybersecurity threats, future EV analytics frameworks should incorporate defense-in-depth strategies, including ML-based intrusion detection, secure V2X authentication protocols, end-to-end encryption, and continuous anomaly monitoring at both vehicle and infrastructure levels. Evaluating the trade-offs among security strength, latency, and computational overhead will be essential for practical deployment in large-scale EV networks.

6. Conclusions

The rapid evolution of electric vehicles and their integration into modern transportation and energy systems have generated a growing need for intelligent, data-driven solutions capable of addressing operational, environmental, and infrastructural challenges. This survey provided a comprehensive examination of how business intelligence (BI) and analytics can support and enhance a wide range of EV applications, including accident analysis, battery health prediction, charging station utilization, intelligent charging infrastructures, and autonomous driving. By synthesizing existing research and organizing it into a structured taxonomy, we highlighted the diversity of analytical techniques, data sources, and methodological approaches employed across the EV ecosystem. The detailed literature tables and in-depth analysis offered a clear view of current capabilities, technological trends, and the limitations that continue to shape the landscape of BI-enabled EV solutions. In total, this survey analyzed 75 recent studies across five EV analytics domains, including accident analysis (29 studies), battery health prediction (40 studies), charging station utilization (22 studies), intelligent charging infrastructures (22 studies), charging station placement (20 studies), and autonomous driving (24 studies).

Our discussion revealed several fundamental constraints that hinder large-scale deployment of intelligent EV systems, such as fragmented and heterogeneous data environments, limited availability of high-quality training datasets, challenges in explainability and transparency of machine learning models, and significant cybersecurity vulnerabilities arising from the increasing digitization of EV infrastructures. At the same time, our exploration of future challenges underscored promising research directions, including decentralized and privacy-preserving data management frameworks, unified multimodal analytics pipelines, physics-informed battery intelligence, grid-aware charging optimization, robust autonomous driving models, and secure communication protocols for next-generation EV networks.

Overall, this survey demonstrates that BI and advanced analytics constitute foundational pillars for the future of electric mobility. By enabling predictive insights, optimizing operational decisions, and supporting real-time system intelligence, BI-driven methods hold the potential to significantly enhance safety, efficiency, sustainability, and user experience in EV ecosystems. We hope that the insights, taxonomies, and research directions presented in this paper will serve as a valuable resource for researchers, practitioners, and policymakers seeking to develop scalable, secure, and intelligent solutions that can accelerate the global transition toward sustainable electric transportation.