Advances in Battery Technology for EV Applications PDF

Received 12 September 2023, accepted 16 September 2023, date of publication 22 September 2023, date of current version 3 October 2023. Digital Object Identifier 10.1109/ACCESS.2023.3318121 Advances in Batteries, Battery Modeling, Battery Management System, Battery Thermal Management, SOC, SOH, and...

Received 12 September 2023, accepted 16 September 2023, date of publication 22 September 2023, date of current version 3 October 2023. Digital Object Identifier 10.1109/ACCESS.2023.3318121 Advances in Batteries, Battery Modeling, Battery Management System, Battery Thermal Management, SOC, SOH, and Charge/Discharge Characteristics in EV Applications R. RANJITH KUMAR1,2 , C. BHARATIRAJA 2 , (Senior Member, IEEE), K. UDHAYAKUMAR1 , S. DEVAKIRUBAKARAN 2 , K. SATHIYA SEKAR3 , AND LUCIAN MIHET-POPA 4 , (Senior Member, IEEE) 1 Department of Electrical and Electronics Engineering, Anna University, Chennai 600025, India 2 Center for Electric Mobility, Department of Electrical and Electronics Engineering, SRM Institute of Science and Technology, Chennai 603203, India 3 Department of Electrical and Electronics Engineering, KSR Institute for Engineering and Technology, Tiruchengode, Tamil Nadu 637215, India 4 Faculty of Information Technology, Engineering and Economics, Østfold University College, 1757 Halden, Norway Corresponding authors: C. Bharatiraja ([email protected]) and Lucian Mihet-Popa ([email protected]) This work was supported by the Department of Science and Technology, Government of India, Promotion of University Research and Scientific Excellence (PURSE), under Award SR/PURSE/2021/65. ABSTRACT The second-generation hybrid and Electric Vehicles are currently leading the paradigm shift in the automobile industry, replacing conventional diesel and gasoline-powered vehicles. The Battery Manage- ment System is crucial in these electric vehicles and also essential for renewable energy storage systems. This review paper focuses on batteries and addresses concerns, difficulties, and solutions associated with them. It explores key technologies of Battery Management System, including battery modeling, state estimation, and battery charging. A thorough analysis of numerous battery models, including electric, thermal, and electro-thermal models, is provided in the article. Additionally, it surveys battery state estimations for a charge and health. Furthermore, the different battery charging approaches and optimization methods are discussed. The Battery Management System performs a wide range of tasks, including as monitoring voltage and current, estimating charge and discharge, equalizing and protecting the battery, managing temperature conditions, and managing battery data. It also looks at various cell balancing circuit types, current and voltage stressors, control reliability, power loss, efficiency, as well as their advantages and disadvantages. The paper also discusses research gaps in battery management systems. INDEX TERMS Electric vehicle, battery management, battery modelling, state of charge, state of health, cell balancing, battery thermal management system. I. INTRODUCTION greenhouse gas emissions (GHGE). Utilizing renewable The effects of fossil fuel depletion on the ecosystem have energy sources and electrifying the transportation sector, increased the urgency to transition to renewable energy as shown in Fig.1, can reduce the GHGE by up to 40%. sources and alternative transportation technologies. The Renewable energy, such as solar, wind, wave, and tidal power excessive extraction and utilization of fossil fuels result in provides a greener, more sustainable alternative to fossil fuels the generation of significant quantities of CO2 and other. However, the intermittent nature of these energy sources poses a challenge to maintaining a consistent and reliable The associate editor coordinating the review of this manuscript and power supply. To tackle this challenge, energy storage sys- approving it for publication was Branislav Hredzak. tems (ESSs) are utilized to store surplus energy generated This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. VOLUME 11, 2023 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ 105761  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 3. Schematic diagram of Li-ion cell. FIGURE 1. Vehicle CO2 emission levels. with the electrical grid. During periods of low demand or high renewable energy generation, EVs can supply stored electric- ity back to the grid, thereby assisting in balancing supply and demand and promoting grid stability. Fig. 4 demon- strates the worldwide battery industry’s explosive expansion, projecting a surpass of 2500 GWh within the next decade. Fig. 4 (b) showcases the increasing demand for bat- teries across different applications and regions, with electric mobility being a major driving force behind the growth of the modern battery industry. The popularity of electric and alternative fuel vehicles is accelerating the research and development of battery materials and automotive technology, FIGURE 2. A standard electric vehicle. which supports smart mobility. China has made plans to meet its peak emissions before 2030, in keeping with the global goal of achieving carbon neutrality. To make electric from renewable sources during peak production periods and vehicles comparable to fossil fuel vehicles, Li-ion Batteries release it to the grid during high demand or when renewable (LIBs) are expected to come to an energy density goal of energy generation is low. approximately 500 Wh kg−1 for EV applications. Numerous The ESSs play a crucial part in boosting the viability electric car models have made extensive use of both Li-ion and stabilizing the power grid of the widespread adoption batteries and nickel-metal hydride (Ni-MH) batteries. of renewable energy sources. EV (shown in Fig.2) and The popularity of Li-ion batteries stems from their improved hybrid electric vehicles (HEVs) have gained popularity as reliability, power density, energy density, and efficiency. potential replacements for automobiles powered by inter- Additionally, the decreasing manufacturing costs of Li-ion nal combustion engines, offering numerous benefits such as batteries have contributed significantly to their widespread reduced greenhouse gas emissions, decreased air pollution, commercialization, enabling their adoption across multiple and improved energy efficiency. industries. Efficient battery management is crucial to ensure EVs and HEVs are powered by batteries, which offer safe use, increase driving range, improve power management features include high energy density, low environmental techniques, lengthen battery life, and lower costs. Batter- impact, and durable performance. The wider adoption of EVs ies require specific attention in electric vehicle applications. depends on advancements in battery technology. Efforts are Overcharging, over discharging, or other improper activi- being made to enhance energy storage capacity, reduce charg- ties can pose serious safety threats to the batteries, hasten ing times, and lower costs. Currently, Lithium-ion (Li-ion) their ageing process, and potentially result in fire or explo- batteries are the most prevalent type used in EVs due to their sion accidents. Battery systems in electric vehicles not favorable characteristics, but researchers are also exploring only power the electric motor but also different electrical other battery chemistries as shown in Fig.3. components. These vehicles often operate under complex This concept allows EVs not only to consume energy but conditions characterized by frequent acceleration and decel- also to function as energy storage systems, actively engaging eration, and human charging behavior can be unpredictable. 105762 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 4. Global battery industry. (a) Growth. (b) Demands by applications. VOLUME 11, 2023 105763  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System Additionally, because the battery is an electrochemical sys- the battery electric current is going in and out. Communi- tem, state determination is quite challenging due to the battery cation interfaces facilitate the information transfer between high nonlinearity and time-varying characteristics. There- external devices and the BMS such as the vehicle’s control fore, creating precise and dependable BMS technologies is system or a battery management network. To protect the still a challenging effort to guarantee that batteries and the battery from potentially harmful circumstances, the BMS also associated energy systems operate in a secure manner and includes safety functions including over-current protection, function to the best of their abilities. This paper aims to over-voltage protection, and under-voltage protection. Fur- give detailed review is Focuses on a Battery management thermore, The BMS is in charge of managing the charging and system and key technologies for BMS in Section.II. The discharging procedures, ensuring they are carried out within typical batteries used in EV are reviewed in Section.III. safe and optimal parameters. Discussed Various types of Battery Modelling the typical In the market, there exist various types of integrated BMS batteries used in EV in Section.IV. Various SOC estimation chips that offer different functionalities. These chips are Techniques are discussed for Battery Cell and Battery Pack designed to perform specific tasks within the BMS architec- in Section V. Comprehensive review of Various Battery SOH ture. Some of the common functional components found in estimation in Section.VI. Several Important and Conventional BMS chips are a fuel gauge monitor, a cut-off field effect battery charging Strategies are covered, along with the related transistor, a cell voltage monitor, a state machine, temperature optimization techniques in Section VII. Focuses on various monitors, and a real-time clock. The organization and cell balancing topologies has been recommended in recent integration of these components can vary depending on the years in Section VIII. It provides and overview of the most specific BMS chip. BMS chips can range from simpler ana- recent advance in LIB thermal management for high charge/ log front ends with microcontrollers capable of monitoring discharge cycles in Section VIII. The problems with BMS are and balancing to fully integrated solutions that can oper- discussed. the viewpoint of BMS improvement is examined ate autonomously. The level of integration and complexity in Section IX. A summary of the viewpoints of the current depending on the applications needs the desired functionality study and the Suggested future research activity of BMS is of the BMS. In EVs, Different types of actuators, controllers, Provided in Section.X. Finally, the conclusion of the paper is and sensors can be included in BMS. These components summarized in Section XI. work together to ensure the safe and wide range of actuators, controllers, and sensors can be used with BMS. The BMS II. BATTERY MANAGEMENT SYSTEM also performs accurate monitoring of battery parameters, The state of charge (SOC), state of health (SOH), state providing valuable information for battery health assessment, of energy (SOE), state of power (SOP), and state of life state of charge estimation, and overall battery performance (SOL) are just a few examples of estimations covered by optimization. In terms of hardware architecture, there are battery management technologies (SOL). Among these, SOC three basic types of topologies that are frequently employed and SOH monitoring are particularly crucial as they serve in BMS: modular architectures, centralized systems, and dis- as the foundation for enhancing reliability and ensuring tributed systems. BMS can also be categorized according safety. A software and hardware device called a BMS is to the particular features they have. These ideas offer intended to control batteries and optimize their performance a comprehensive framework with fundamental functionality , as depicted in Fig. 6. for BMS design. Within the battery pack, various sensors are The BMS software serves as the central component of strategically placed to collect data at the monitoring layer the system, responsible for controlling hardware operations. All of the battery pack’s elements and the vehicle control and analyzing sensor data to make informed decisions. processor are connected to the BMS. Safety has always been a Online data processing plays a critical role in detecting most top priority for BMS. The suggested BMS designs, however, faults, and intelligent data analysis is necessary to provide use more sensors than the safety circuits now in use, allow- timely battery malfunction warnings. Data collection is of ing for improvements like accurate warnings and controls paramount importance to identify potential issues before they to prevent overcharging, over discharging, and overheating. manifest as faults. Hardware components within the BMS, A system of sensors is necessary to track and quantify bat- such as sensors, make it easier to measure battery voltage tery properties such cell voltage, current, and temperature. and current. The general block diagram of a BMS is illus- However, the practical viability of these measurements is trated in Fig. 7. A BMS comprises various functional units, hindered by space limitations and the cost of devices. As a including cell voltage balancing, temperature monitoring, result, accurate measurements of current, temperature, and current sensing, and communication interfaces. Cell voltage voltage are crucial to improving state tracking capabilities in balancing guarantees that each battery pack’s individual cells practical applications. Based on these data, SOC, SOH, State are are maintained at consistent voltage levels, maximizing estimations were been obtained. Also the surface tempera- the overall pack performance and extending its lifespan. ture is measured to attain the thermal characteristic and the Temperature monitoring is crucial for preventing overheating impact of temperature with the battery SOC and SOH were and managing thermal conditions within the battery. Current obtained. Along with this the battery joint state estimation sensing enables accurate measurement and monitoring of has been measured using the above two data. This joint state 105764 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 5. Organization of the review article. estimation is a important measure for the battery to effectively vehicles, renewable energy storage, and so on. These attained manage and operate the battery and increases the battery parameters has been used for defining the charging behavior, life span in different types of applications such as electric fault monitoring, fault/abnormal detection, predictive control VOLUME 11, 2023 105765  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System 1. Protection: This entails preventing the battery from being damaged by high temperatures, overcharging, overcurrent, and short circuits. 2. In the field of ‘‘high-voltage control and sensing,’’ tasks including measuring temperature, voltage, cur- rent, thermal management, contactor control, pre-charge functionality, and ground-fault detection are included. 3. Diagnostics: The SOL estimate, SOH estimation, and abuse detection functions of the BMS are used to assess the battery overall health and condition. 4. Performance Management: This encompasses tasks such as power-limit computation, cell balancing or equalization, and SOC estimation, which is crucial for optimizing battery performance. FIGURE 6. Overview of the BMS hardware and software components. 5. Interface: The BMS facilitates data recording, report- ing, communications, and range estimation, allowing for effective communication and integration with other and fault diagnosis. The various steps like, obtaining the vehicle systems. appropriate data, modeling, data collection, and data storage By fulfilling these functions, the BMS ensures the battery sys- as shown as a block diagram in Fig.8. tem’s effectiveness, dependability, and safety while providing One part of the system that controls the charge-discharge essential information for the management and utilization of cycle is the charge controller. A variable resistor could be vehicular energy needed to maintain cell balance or check internal resis- tance. Cell balancing management, which aims to balance III. BATTERY TYPES IN EV the battery pack’s cells and accurately gauge the battery Various types of batteries can be utilized as the power EV health, is one of the most crucial design factors. BMS applications as given in Fig. 9. The BMS consists of multiple subsystems must communicate internally because they are functional modules. In this study, popular battery types and independent modules. A Controller Area Network (CAN) key BMS technologies are analysed and condensed. Accord- bus is used as the main means of communication within the ing to their capacity for charging, batteries can be divided BMS for the transfer of data. By implementing intelligent into two general categories: primary batteries and secondary batteries with embedded microchips that can communicate batteries. Secondary batteries can be recharged following the with users and chargers, more information can be obtained. discharge process, however primary batteries can only be In order to increase connection between the battery and used once after being entirely depleted. Secondary batteries charger, radio and communication technologies are also being with a high cycle life, a low power density, a low energy rapidly included into charging systems. Because temperature loss, and sufficient safety levels are required for EV and variations can have an impact on cell imbalance, depend- HEV applications. Some commonly used battery types in ability, and performance, a thermal management module is EVs include Li-ion, lead acid, nickel-cadmium (NiCd), and required. Reduced temperature differences between cells are NiMH, among others and the evolution of the batteries with critical, ensuring they operate under appropriate temperature respect to its timeline is shown in Fig. 10. Key details for conditions to maintain optimal performance and longevity. these well-liked battery types are presented in Table 2. This Different sensors, actuators, controllers, and signal lines are clearly demonstrates that Li-ion batteries exhibit significant all included in BMS. Its main job is to make sure that the advantages over other types, in terms of their longer cycle battery stored energy is used safely and optimally while giv- life, which is essential for ensuring long service life in EVs ing the car’s energy management system reliable information (typically 6-10 years). Additionally, Li-ion batteries are about the battery condition. In the sample circuit depicted made of environmentally acceptable components, don’t emit in Fig.6 , Using the gating signal that is received from any hazardous gases, and provide a high level of safety. the control circuit as a starting point, the primary goal is As a result, Li-ion batteries are now the most widely used to measure current, voltage, and temperature. The control kind of EV power. Lithium-based batteries have the high- circuit utilizes advanced algorithms to estimate the SOC, est cell potential and the lowest reduction potential when SOH, SOP, and SOL of the batteries. These estimates are compared to other elements as given Table 3. Lithium is obtained from measurements of battery current, voltage, and one of the single-charged ions with one of the smallest temperature, which are converted from analog signals. The ionic radii, making it the third-lightest element in terms of resulting information is then sent to the vehicle controller, mass. These qualities allow Li-based batteries to attain high giving key deciding elements for the management and distri- power density, gravimetric capacity, and volumetric capacity bution of power in vehicles , ,. The functionality. The Li-ion battery exhibits an energy density range of of a BMS can be categorized as follows : 200-250 Wh/kg and boasts a high columbic efficiency of 105766 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 7. Battery management system functional block diagram. FIGURE 8. Key technologies of BMS. nearly 100%. It is also free from memory effect. Due to alternative electrochemical energy storage systems. One such its superior energy and power density compared to lead-acid technology is the lithium-sulfur (Li-S) battery, which offers and Ni-Cd batteries, lithium-ion batteries are now the pre- advantages for instance, increased energy density, enhanced ferred choice. It is widely utilised in many different products, security, a larger operational temperature range, and maybe such as electric automobiles, power equipment, and portable lower prices due to the abundance of sulfur. These factors gadgets , ,. Li-ion battery development is ongo- make Li-S batteries a promising option for EV applications ing with the goal of increasing their cycle life and safety. Energy density and specific energy of various batteries in both normal and abusive situations , and overall per- at cell level is shown in Fig. 11. However, widespread com- formance characteristics. In the pursuit of higher energy mercialization of lithium-sulfur technology has not yet been density for electric vehicles, researchers have explored achieved due to certain limitations. These excessive discharge VOLUME 11, 2023 105767  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 9. Classification of electrochemical energy storage sources. current, self-discharge, poor cycle life and capacity decline developed only two technologies in this field. These include brought on by cycling, low columbic efficiency, uncontrolled sodium/sulfur (Na/S) batteries and sodium/metal chloride dendrite development, and other factors. (Na/MCl2) batteries. These batteries must function at high temperatures between 270 and 350 ◦ C in order to achieve the A. BATTERY TECHNOLOGIES BEYOND LITHIUM necessary ionic conductivity. Extensive research has been done on battery technologies other than lithium as LIBs get close to their natural limits in 3) SODIUM/METAL CHLORIDE (NA/MCL2) BATTERY terms of specific energy and energy density. Three different Na/MCl2 batteries use transition metal chloride as the cath- battery types have developed as alternative technologies in ode material. In particular, Na/FeCl2 and Na/NiCl2 batteries recent decades: are made using iron chloride and nickel chloride, respectively. Among these, the Na/FeCl2 battery has undergone more 1) 3.1.1 METAL/AIR BATTERIES significant development compared to the Na/NiCl2 battery. Anodes made of metal and cathodes made of air are The Na/NiCl2 battery offers several advantages, including used in metal/air batteries. The energy capacities of these increased power density, a wider working temperature range, batteries are primarily determined by the anode capac- and less corrosion of metallic elements. ity and the handling process. Despite this limitation, they offer exceptionally high energy density and specific energy, 4) SODIUM/SULFUR BATTERY with maximum values of 400 and 600 Wh/L, respectively. The Na/S battery uses beta-alumina ceramic electrolyte, Zinc/air, aluminum/air, iron/air, magnesium/air, calcium/air, sodium anode, and sulphur cathode. However, the perfor- and lithium/air batteries are only a few examples of the mance of Na/S batteries tends to decline as the internal several kinds of metal/air batteries that are available. These resistance increases, which is further exacerbated by deeper batteries can be classified as primary (non-rechargeable), discharges. In recent research, there has been exploration into electrically rechargeable, or mechanically rechargeable. room-temperature Na/S batteries that demonstrate robust and Among them, mechanically rechargeable batteries provide consistent cycling performance ,. the convenience of refueling and recycling. IV. BATTERY MODELLING 2) SODIUM-BETA BATTERIES The core of BMS design is building an accurate battery High energy density is a well-known characteristic of model, which is essential for estimating the battery status. sodium-beta batteries, although researchers have successfully Battery models vary in terms of accuracy and complexity, 105768 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 1. Nomenclature. TABLE 1. (Continued.) Nomenclature. with three primary categories: battery electric models, battery thermal models, and battery coupled models, as illustrated in Fig. 12. A. BATTERY ELECTRIC MODEL The models that need batteries include electrochemical mod- els , , , , , equivalent circuit models , , , , , , , , , , , , , , , and data-driven models , , , , , , , , , , , , , , , , , , , , ,. 1) ELECTROCHEMICAL MODEL (EM) Electrochemical models describe battery behavior by uti- lizing partial differential equations that consider electrolyte concentration, electrode size, and electrochemical processes within the battery. While electrochemical models provide precise battery parameters, they require significant computa- tional power and time to solve multiple equations pertaining to the battery current, temperature, electrolyte concentration, solid concentration, open circuit potential, over potential, and electrolyte potential, and more. Implementing them in real-time applications is challenging. Researchers have proposed various approaches to address these challenges. Doyle et al. introduced a Pseudo-2-D (P2D) elec- trochemical model, however because there are so many nonlinear equations, it takes longer to simulate and It reduces the effectiveness of its computation for BMS applications. Domenico et al. developed a reduced-order electrochem- ical model by instead of taking into account its dispersion throughout the electrodes, averaging the solid electrolyte concentration, enabling real-time implementation on board buses. However, parameter identification remains a difficult task. Ahmed et al. , employed a SOC estimation and genetic algorithms are used to identify parameters, but VOLUME 11, 2023 105769  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 2. Key details of batteries used in EV. FIGURE 10. Milestones and foresight of battery Technologies. the model’s accuracy is compromised due to assumptions and capacitors. A high-value capacitor or a regulated made to reduce its order. Han et al. provided a rough voltage source serves as ECM representation of the battery model that keeps track of the diffusion process and how Open Circuit Voltage (OCV), a vital metric for state esti- electrolyte concentration is distributed inside the battery. mation approaches. The Rint model, Thevenin model, Zou et al. A reduced-order model based on singular PNGV model, and GNL-model are examples of analogous perturbation and averaging theory was presented for Li-ion circuit models that are frequently employed, as illustrated in battery SOC estimation and discharging capacity forecasting. Fig. 13. The Rint model, which represents the battery as a This model simplification approach is applicable to all battery voltage source with series resistance, is the simplest basic types. However, building a high-fidelity model that takes ECM. However, this simple model is not capable of into account age, capacity fading, and temperature increases accurately capturing the specific characteristics of batteries complexity while also improving accuracy. Table 4 com- used in EVs. To enhance the representation of battery dynam- pares various electrochemical battery model types in a brief ics, the Rint model is extended by incorporating a single manner. Resistance-Capacitance (RC) parallel network, resulting in the widely used Thevenin model. The dynamic behavior 2) EQUIVALENT CIRCUIT MODEL (ECM) of batteries is well captured by the Thevenin model. The The electrical activity of the battery is modeled by the ECM Partnership for a New Generation of Vehicles (PNGV) model, using electrical elements including voltage sources, resistors, or FreedomCar model a modified variation of the Thevenin 105770 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 3. Li-ion battery types. TABLE 4. Comparison of various electrochemical models of battery. suitable for low SOC areas, it may not accurately represent high SOC regions. Resistance-Capacitance (RC) network- based models are among the several ECM models explored in the literature that have found widespread use for online applications. These models include the one RC network ECM , , two RC network ECM , , , and three RC network ECM ,. Table 5 lists the model equations and parameters in accordance with circuit theory. The two RC network model is one of them. It is particularly notable for its high accuracy in predicting the relationship between input current and output voltage (I-V), as well as the charg- ing and discharging times of the battery. Since batteries are nonlinear systems, their dynamics vary under different operating conditions such as SOC, temperature, and charg- ing/discharging rates. Therefore, parameterizing the model FIGURE 11. Energy density and specific energy of various batteries at cell level. becomes an ‘‘identification problem’’ or ‘‘optimization prob- lem’’ to fit the model to measured data ,. The SOC, model, includes a fictive capacitor to account for variations temperature, and charge-discharge rate of the battery must all in OCV ,. The PNGV model consists of OCV, be taken into account when updating the model parameters polarization resistance, a capacitor, an imaginary capacitor, because the ECM circuit parts do not accurately reflect bat- and an ohmic resistance. While the PNGV model is teries physically. VOLUME 11, 2023 105771  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 5. Models of various RC network-based equivalent circuits. 3) DATA DRIVEN MODEL (DDM) In these cases, a data-driven control approach, facilitated by Data-driven models (DDMs) offer a more efficient alternative DDMs, can provide significant benefits by accurately captur- to ECM and EM models, with the ability to approximate ing the behavior of the system without relying on a known highly nonlinear battery characteristics. DDMs rely on data mathematical model as given in Table 6. Modeling batteries and computational intelligence to describe battery behav- accurately is challenging using traditional methods like ECM ior, without the need for prior understanding of the battery and EIM internal chemical processes and hazy environmental internal structure. Various types of DDMs, such as Artificial operating conditions. Black-box models, on the other hand, Neural Network (ANN) , Adaptive Neuro-Fuzzy Infer- offer benefits like parallel distributed processing, high com- ence Systems (ANFIS) , Deep Neural Network (DNN) putation rates, fault tolerance, and the capacity to adapt to , and Support Vector Machine (SVM) , , , deal with this complexity by utilizing the nonlinear connec- have been employed for battery modeling, as shown in tion of input data for training. Fuzzy systems’ subjectivity Fig. 20. DDMs have several advantages, particularly in sit- and flexibility are combined with neural networks’ capacity uations where : for learning in ANFIS. The inherent multiple-model 1. The controlled system has no known global mathematical structure of the T-S fuzzy model allows it to manage the model. nonlinear dynamics of batteries. Black-box models, on the 2. Unknown is the controlled system entire global mathemat- other hand, produce accurate results. Rule-based modeling, ical model. however, has accuracy that varies with the number of rules 3. Building a mathematical model to depict the controlled at the cost of increased computational complexity and lim- system with an undetermined structure while it is in oper- ited interpretability. SVM, on the other hand, uses a small ation is not practical. number of samples with the kernel trick to describe system 4. The Regulated System’s Mechanism model has too many dynamics. parameters, is overly complicated, or is difficult to study Although SVM has a simpler design than ANN, it requires and create using conventional methods. solving a Costly optimization in terms of computing problem 105772 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 6. Comparison of DDM of battery. VOLUME 11, 2023 105773  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 6. (Continued.) Comparison of DDM of battery. FIGURE 13. The following Li-ion battery models are: (a) Rint model (b) Thevenin model (c) PNGV model (d) GNL-model. procedures. In all the aforementioned methods, data prepro- cessing and noise removal are essential. The computational complexity associated with DDMs can pose obstacles in FIGURE 12. Types of battery modelling. economic applications. Data collection is vital for developing accurate DDMs, as these models require a large amount of training data. As batteries are increasingly deployed across to determine kernel parameters. The RBF kernel is commonly various applications, their degradation rates vary under differ- used due to its strong generalization capability. How- ent operating conditions, necessitating application-specific ever, SVM struggles with handling large amounts of data, data for accurate battery modeling. To mitigate the time making SOC estimation and It might be difficult to estimate and cost associated with data collection, researchers can uti- SOH for battery packs. SVM, ANN, DNN techniques use lize publicly available data instead of conducting extensive machine learning algorithms to forecast nonlinear parameters experiments. To increase accuracy while lowering complex- and estimate battery SOC based on statistical data. Among ity, care should be taken when choosing the right model, these, DNN outperforms ANN and SVM. DDM neces- model parameter type, and parameter identification proce- sitates intensive calculations for real-time understanding of dure. Table 7 provides a performance comparison of different battery properties by means of training and data-acquisition battery models, highlighting their strengths and weaknesses. 105774 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System Overall, both accuracy and simplicity are critical considera- tions when selecting a battery model for BMS design. B. BATTERY THERMAL MODEL Due to its large impact on battery performance and lifes- pan, thermal behavior, in particular temperature, is a vital component of EV BMS. To accurately represent the thermal behavior of batteries, a variety of models have been cre- ated, including heat transfer models, heat generation models, reduced-order thermal models, and data-driven models. The distribution of elements including activation, concentration, and ohmic losses, which vary within the battery, are taken into FIGURE 14. Schematic view of Li-ion battery pack. consideration by different methodologies used by heat pro- duction models to characterise heat generation in batteries. Abada et al. presented a thermal model for the thermal management system of a Li-ion battery pack, based on the electro-thermal model that determines heat generation and energy balance between heat generation and heat dissipa- calculates battery SOC. A three-dimensional temperature dis- tion. The thermal model can be represented by the following tribution model plus a two-dimensional potential distribution equation: model make up this model. By utilizing this coupled model, battery SOC and temperature distribution can be effectively d ∂T d d determined under both constant and dynamic currents. Qaccu = ρCp = Qgen − Qdis (1) dt ∂t dt dt In another study , a Batteries with three distinct In this equation ρ, C p , t, and T are the cell den- cathode materials were used to validate a simplified low- sity, heat capacity, time and cell temperature respectively. temperature electro-thermal model. This reduced model In addition, Qgen , Qaccu , and Qdis are the accumulated demonstrates sufficient accuracy and enables the develop- heat, generated heat, and dissipated heat, respectively. ment on Under low-temperature circumstances, quick heating Qgen encompasses the heat produced by chemical reac- and optimal charging techniques. Basu et al. used a tions that is both reversible and irreversible. Qdis includes linked three-dimensional electro-thermal model to investigate heat-transferring processes like conduction, convection, the impacts of various battery operations, such as coolant flow and radiation, The electrochemical-thermal model and the rate and discharge current, on battery temperature. Through electro-thermal model were developed on the basis of this the analysis of this coupled model, it was observed that thermal model as summarized in Table 8. ,. They contact resistance plays a vital role in determining battery take into account things like chemical processes, ion mobility temperature. in the solid electrolyte interphase (SEI), over potentiation at In another study , a Batteries with three distinct the reaction surface, Ohmic loss in electrodes, and entropy cathode materials were used to validate a simplified low- during charging and discharging when analyzing the ther- temperature electro-thermal model. This reduced model mal behaviour of batteries. Table 9 lists the symbols and demonstrates sufficient accuracy and enables the develop- characteristics related to the electrochemical-thermal model. ment on Under low-temperature circumstances, quick heating The electrochemical-thermal model provides a comprehen- and optimal charging techniques. Basu et al. used a sive understanding of battery operation by considering both linked three-dimensional electro-thermal model to investigate electrochemical and thermal aspects. However, one drawback the impacts of various battery operations, such as coolant flow of this model is its high computational burden, which arises rate and discharge current, on battery temperature. Through from the large number of equations required to accurately the analysis of this coupled model, it was observed that predict battery temperature. contact resistance plays a vital role in determining battery temperature. C. BATTERY COUPLED ELECTRO-THERMAL MODEL There have been several coupled electro-thermal models V. STATE OF CHARGE established to capture the strong coupling between battery Battery charging requires careful consideration and effec- electric and thermal behaviors. These models allow for tive measures to ensure a smooth and efficient process. The the simultaneous consideration of battery electric param- SOC is a crucial factor in battery operation, representing the eters (e.g., voltage, current, SOC) and thermal parame- level of charge relative to the battery capacity as shown in ters and behaviours Shown in Fig.14. (e.g., surface and Fig. 15. comparable to a fuel gauge in a gasoline-powered internal temperature). There have been several developed car, SOC indicates the remaining amount of energy in a bat- linked electro-thermal models have been proposed in the tery to power an EVs. Various critical performance aspects, literature to achieve this coupling , ,. For such as range and fuel economy, heavily depend on SOC. instance, Goutam et al. established a three-dimensional SOC is typically expressed as a percentage (0% = empty; VOLUME 11, 2023 105775  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 7. Comparison of various battery models. TABLE 8. Comparison of electrochemical-thermal model and the electro-thermal model. 100% = full) and is commonly used to define a battery greatly improve SOC prediction accuracy by merging data current status while it is in operation. from many sources. Here are a few fusion models and meth- (Q0 + Q) ods that are frequently employed for estimating battery SOC: SOC% = 100 × (2) Qmax Extended Kalman Filter (EKF): The EKF is a Kalman The SOC calculation can be performed using Equation (2), filter modification created to handle nonlinear systems. where Q0 (mAh) is the battery initial charge. Q (mAh) is By integrating voltage and current data with a bat- the quantity of electricity delivered by or supplied to the tery model that accounts for the nonlinear relationship battery. it is negative during the discharge and positive during between SOC and battery voltage, it is frequently used the charge. Qmax (mAh) is the maximum charge that can be to estimate battery SOC. stored in the battery. The determination of battery SOC is Particle Filter (PF): Both nonlinearities and uncertainties a fundamental aspect of BMS. Accurate and reliable SOC are handled by PFs. They operate by converting the SOC estimation is crucial for vehicle energy management and the into particles and changing their weights in accordance optimal design of control systems. To achieve real-time SOC with voltage and current readings. When dealing with estimation, numerous methods have been proposed. To pro- complicated battery behavior and shifting operating cir- vide a more detailed comparison of these methods, they can cumstances, PFs are very helpful. be categorized into four groups, as illustrated in Fig. 16. Recursive Least Squares (RLS) with Adaptive Gain: Battery SOC estimate is an essential component of battery RLS algorithms can use voltage and current observa- management systems, and fusion models and algorithms may tions to estimate battery characteristics and SOC in an 105776 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 9. Electrochemical thermal model parameters and symbols. FIGURE 16. Classification of the SOC estimation methods. Model-Based Adaptive Filters: These filters use adap- tive algorithms to combine a battery model with in- the-moment data, modifying the model’s parameters to reflect actual battery performance. The accuracy of the long-term SOC estimate is improved by this method. Neural Networks and Deep Learning: Current and volt- age data can be combined over time using recurrent neural networks (RNNs) and long short-term memory (LSTM) networks to calculate SOC. These networks are capable of complicated connection learning and battery condition adaptation. Multiple Model Estimation (MME): Each battery model or estimating technique that MME combines is suited FIGURE 15. Charging and discharging process of battery. for a particular set of operating conditions. Based on the present operational condition, the most suitable model is chosen. adaptive manner. RLS is capable of coping with fluc- Unscented Kalman Filter (UKF): The UKF is an EKF tuations in battery behavior by gradually modifying the substitute that employs a deterministic sampling method estimate gain. to capture the real statistical moments of the battery VOLUME 11, 2023 105777  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System model’s nonlinearities. Compared to the EKF, it can offer a more precise SOC estimation. Fuzzy Logic: Fuzzy logic enables nonlinearity and uncertainty to be included into SOC estimate. In order to provide an accurate SOC estimate, it can integrate volt- age, current, temperature, and other sensor information. Ensemble Methods: Ensemble approaches can deliver a more reliable and accurate SOC calculation by merging different estimating techniques, such as EKF, UKF, and neural networks, especially when dealing with shifting operating circumstances. Hybrid Approaches: In order to accurately estimate SOC, hybrid models integrate physics-based and data- driven methodologies, utilizing both battery models and FIGURE 17. LiFePo4 OCV-SOC relationship. real-time observations. efficiency are taken into consideration in the estimation per- The selection of a fusion model or algorithm is influenced by formed using the Ah technique. The equation to calculate a number of variables, including the precision of the avail- battery SOC using the Ampere-hour method is presented in able measurements, the complexity of the battery’s behavior, Eq. (3) available computing power, and the required level of accu- Rk racy. Battery SOC estimate may be made more accurate and K ηI (t)dt trustworthy by information fusion, which is crucial for the SOC(k) = SOC(K0 ) + 0 (3) Cn dependable and safe functioning of battery systems. where, η stands for the efficiency of battery charging or discharging, SOC (k0 ) is known initial SOC, I(t) is the cur- A. LOOKUP TABLE BASED METHOD rent value which is positive for charging and negative for The SOC of batteries has a direct correlation with their discharging, Cn stands for the battery nominal capacity. The extrinsic identifying characteristics, such as impedance and Ah method has several drawbacks that need to be addressed. OCV. The relationship between SOC and OCV has been Firstly, it has to be aware of the battery initial SOC, which plotted in Fig. 17. Therefore, by measuring these parameters may not always be readily available. Secondly, there are and utilizing a look-up table that establishes the relationships inherent measurement errors in the battery current because to between SOC and one or more parameters, to estimate the sporadic disruptions like noise and temperature drift, which SOC of batteries ,. For example, the SOC of the can affect the accuracy of the SOC estimation. Lastly, the battery can be determined by the knowledge of OCV. This value of Q, which represents the capacity of the battery, may approach is commonly used in battery management tech- need to be recalibrated due to variations in the battery age nologies for SOC estimation. However, obtaining precise and operating circumstances. When all of these conditions real-time measurements of OCV is challenging because it are present, the Ah method’s accuracy may suffer. Therefore, requires disconnecting the power source and allowing the it is more suitable to use the Ah method in conjunction with battery to rest for an extended period. Additionally, measure- other supporting techniques, such as model-based methods, ment relying on battery impedance on specific measurement to increase the SOC estimation’s precision and dependability. devices, making it impractical for use in operating EVs. Instead, impedance measurement is more suitable for labora- C. MODEL BASED ESTIMATION METHODS tory environments where accurate and controlled testing can The Model-Based Estimation methods for SOC can be be conducted. broadly classified into three types: Electrochemical method (EM), Equivalent Circuit Model (ECM), and Electrochemical B. AMPERE-HOUR INTEGRAL METHOD Impedance Model (EIM). These methods involve express- By directly measuring the battery voltage and current, the ing battery models as nonlinear state equations and utilizing SOC can be determined. One commonly used method is state estimation algorithms and adaptive filters to infer the the Ampere-hour (Ah) method, which estimates the battery internal state of the batteries. Various algorithms, such as state by integrating the charging and discharging currents. Kalman Filter (KF) , , , Extended Kalman Fil- This method is straightforward and computationally efficient ter (EKF) , , , , Unscented Kalman Filter. However, there are some challenges associated with (UKF) , , , , , Fading Kalman Filter the Ah method in Dynamic applications. Accurately measur- (FKF) , Cubature Kalman Filter , , , Parti- ing the initial SOC is challenging, because SOC estimation cle Filter , H∞ observer method , , Adaptive is constrained by things like the battery unknown begin- Extended Kalman Filter (AEKF) , , , and Adap- ning capacity, its self-discharge rate, and the reduction in tive Unscented Kalman Filter (AUKF) , , , battery capacity. Typically, Peukert’s impact and coulombic , are commonly employed in these methods. The general 105778 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System Various algorithms can be utilized for black-box modeling, such as fuzzy controllers , , support vector machines , , neural networks , , , , , , , , , , , and combinations thereof. These algorithms, however, are quite sensitive to the parameters, and incorrect parameter selection may lead to non-convergence if the training data does not adequately cover the operating conditions. ANNs , , , , , have gained popularity for validating complex nonlinear models due to their self-learning capabili- ties. Although ANNs heavily rely on training with collective FIGURE 18. A general Block Diagram of model based SOC estimation information, they offer computational efficiency at a lower method. cost. However, overlearning poses a challenge with ANN models. A summary of SOC estimation using different com- block diagram of the model based SOC estimation method is binations of methods is provided in Table 10 and Table 11. shown in Fig. 18. It is clear from the table that the Ah method is Simple and Kalman Filter is an optimal estimator widely used for lin- It is clear from the table that inexpensive, but unsuited to ear systems. KF for nonlinear systems necessitates intricate real-time applications. Adaptive filter and observer methods computations. Plett developed the EKF approach specifically offer high accuracy and are suitable for real-time appli- for nonlinear battery model SOC estimation. Although EKF cations, but they suffer from computational complexity, addresses nonlinearities, it suffers from linearization errors configuration effort, and implementation challenges. Data- and increased computational effort and the flowchart of EKF Driven Model (DDM) complexity, making them suitable for method is shown in Fig.19. UKF, on the other hand, can real-time applications with lower computational complexity. provide accurate results for highly nonlinear models by elim- The block diagram of the DDM is shown in Fig.20. However, inating linearization errors. However, it involves Cholesky successful implementation of DDM methods requires appro- factorizations and sigma point selection, which impact per- priate model selection, hyper parameter tuning algorithms, formance. EKF incorporates a fading concept to correct proper training algorithms, and extensive data collection and modeling errors but demands more computational power. normalization. Filter parameters like noise covariance matrices significantly influence estimation accuracy and convergence rate. KF algo- E. SOC ESTIMATION FOR BATTERY PACK rithms struggle with non-Gaussian noises. Accurate estimate is achieved by the development of AUKF algorithms, which The use of battery packs, consisting of multiple connected automatically update noise covariance matrices. However, cells, introduces challenges in accurately estimating the they come with increased computational time and complex- SOC due to variations in individual cell performance and ity. The H∞ observer method is another suitable approach non-uniform characteristics within the pack. While the but shares similar issues as KF-based methods, including capacity and SOC of a single cell can be measured through dependence on gain for accuracy and convergence rate. discharge testing, these measurements are not directly appli- KF algorithms possess self-correcting capabilities, making cable to battery packs. The complexity, time-varying, nonlin- them suitable for estimating the situation of quickly shifting ear, and non-uniform properties of battery packs make it chal- systems with accurate models. However, challenges persist, lenging to assess capacity and SOC accurately. By calculating such as handling initial SOC errors. Therefore, KF algo- the SOC of a battery pack, one can determine the internal rithms should be employed alongside other techniques to condition of a complicated hybrid-connected battery system. enhance estimation accuracy and reliability in practical Accurate SOC estimates for battery packs have been sought applications. after, and these efforts can be categorized into three types: D. DATA-DRIVEN BASED ESTIMATION METHODS 1) CELL CALCULATION BASED METHODS The black-box model method is an effective approach for The ‘‘Big cell’’ method determines the SOC by treating solving nonlinear problems in battery modeling and state the battery pack as a single cell and using the voltage and estimation, providing high prediction accuracy. The Data- current of the pack. However, this method overlooks the Driven Model, explained in detail in Section. IV, utilizes inconsistencies in cell performance, compromising the safety methods for modeling nonlinear statistical data that are prac- of the battery pack. The ‘‘Short board effect’’ method uses the tical for capturing complex relationships and patterns in the extreme cell (with the lowest or highest voltage) to estimate data. For instance, neural networks have been employed the SOC of the battery pack. While this method improves to develop SOC estimators, with inputs including current, safety, it reduces energy utilization within the desired oper- temperature, battery SOC, and voltage as the output layer. ating range of the battery pack (30% - 80% SOC). The This method has demonstrated high computational accuracy. ‘‘One by one’’ calculation approach determines the SOC VOLUME 11, 2023 105779  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 19. Flow diagram of EKF. FIGURE 20. Data-driven model ,. for each individual cell before calculating the battery pack’s pack. Due to the pack’s good consistency, the SOC of a total SOC. Although this method offers accurate estimations, single cell is then utilized to represent the SOC of the com- it incurs high computational costs and is unfit for real-time plete battery pack. A second-level screening process can be applications in electric cars. The flowchart of the cell filtering employed to select suitable cells for packaging the battery method is shown in Fig. 21. pack, ensuring better consistency among the cells. 2) SCREENING PROCESS BASED METHODS 3) BIAS CORRECTION METHODS These methods involve selecting battery cells with similar In this procedure, a notional model of the battery pack is characteristics (capacity, resistance, etc.) building a battery constructed before a bias-correction technique is used to 105780 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 10. SOC estimation using neural network. Cell SOC, discharge/charge rate, and the maximum achiev- able capacity differential between the cell and the average value of the battery pack are all functions of the uncertainty factor in the equation.   j j Ut = Uoc − UD1 −... −UDn − iL Ri +δ Crate , zj , 1Qj (4) where 1Qj between cell j and average value of battery pack, uncertainty δ is the function of cell, -zj ,maximum avail- j able capacity difference, discharge/charge rate -Crate , cell SOC. the j is denote the cell number in battery pack. This method reduces computational costs and improves real-time performance. It shows promise for SOC estimation in With their time-varying, nonlinear, and uneven features, battery packs. However, if the number of battery cells in an electric FIGURE 21. Cell filtering approach Procedure. vehicle is large, Costs associated with computing must be significantly reduced. In summary, accurately estimating the determine the discrepancies between the nominal model and SOC of battery packs is challenging due to variations in the actual battery cells. The revised model is used to do the cell performance and non-uniform characteristics within the SOC estimation, which determines the SOC of the battery pack. Different methods, such as cell calculation, screen- pack by calculating the SOC of each individual cell.. ing processes, and bias correction, have been proposed to VOLUME 11, 2023 105781  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System TABLE 11. SOC estimation at different combinations. address this issue, each with its own advantages and consider- below a certain level, the cycle life gradually goes below ations in terms of computational cost, accuracy, and real-time 15◦ C or exceeds 45◦ C. Further temperature increase leads to applicability. a sharp decrease in cycle life due to thermal runaway , ,. VI. STATE OF HEALTH SOH estimation does not have a fixed definition, and each It is crucial to distinguish between two ideas: battery health battery manufacturer establishes their own criteria. A num- state and remaining useful life prediction. The battery cycle ber of battery properties, including capacity and internal life is the maximum number of cycles a battery can with- resistance, can be used to compute SOH. However, rather stand given its kind, construction, and the manufacturer’s than being a precise measurement, it is an evaluation and recommended usage. The SOH compares the health and per- judgement. Some factors that determine how well Li-ion formance of a used battery to a brand new battery of the same batteries perform over time include the phase shift of the type. SOH is determined by calculating the ratio of electrode material, electrode dynamic performance, elec- the current actual capacity QC of the battery to its nominal trolyte breakdown state, and the creation of SEI films , capacity Qn , as shown in Equation 5. ,. Battery aging is characterized by irreversible changes in electrolyte characteristics, anode and cathode SOH = QC Qn  (5) properties, and alterations in battery component structures as SOH is a subjective metric that has been defined differently shown in Fig. 22. Aging can be categorized as cycle aging, by many studies by taking into account various quantitative due to periods of battery use and calendar aging that take battery performance metrics, including current, resistance, place while batteries are stored. Changes in capacity, internal voltage, self-discharge rate, temperature, stress, and strain. resistance, and power fade are indicators of aging and are Although SOH depends on these parameters, it is a compares closely related to the estimation of SOH. The choice a used battery performance and health to those of a brand-new of the most suitable parameter for SOH estimation depends battery of the same type. Temperature also plays a significant on the specific circumstances and the changes observed in role in battery performance as shown in Fig. 23, where the the external actions of the battery, such as a reduction in cycle life of a cell is optimal when The operating temperature rated capacity or an increase in temperature brought on by is kept between 15◦ C and 45◦ C. When the temperature falls internal modifications like corrosion the relationship between 105782 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 22. Lithium-ion battery aging. FIGURE 23. Li-ion battery lifecycle vs ◦ C diagram. FIGURE 24. Li-ion battery lifecycle vs temperature diagram. cycle life and cell operating temperature, highlighting the optimal range between 15◦ C and 45◦ C. Operating below 15◦ C or above 45◦ C gradually decreases cycle life, while can hasten deterioration and reduce battery life. The battery further temperature increases lead to a sharp decline due to degradation rate, nevertheless, varies depending on the rate thermal runaway as shown in Fig. 23. Fig. 24 illustrates the of charge or discharge, which is impacted by stress factors, Li-ion batteries typically function within a certain current and and it is not constant within the permissible range. The dis- voltage range. According to the battery nominal capacity, the charge rate is dynamic and directly influenced by operating x-axis shows the current (C-rate), while the y-axis represents conditions such as route slope, vehicle weight, speed, and the voltage (V). Positive current values indicate the discharge acceleration. A threshold is frequently established by EV process, while negative current values correspond to charging designers to restrict the maximum discharge current rate. The or regenerative processes. charge rate is fairly stable during the charging procedure. Critical thresholds, depicted as gray zones, are defined Though it can speed up battery charging, a higher pace may based on the specific Li-ion battery type. When the also shorten battery life. Therefore, designers strive to voltage rises over the maximum defined charging voltage or strike a balance between the charging rate and its impact on falls below the stated cut-off discharge voltage, these thresh- battery life, which influences the available charging rate in olds prohibit overcharging and over-discharging, respec- charging stations (e.g., level 1 and level 2). The BMS tively. Operating within the acceptable voltage range is cru- regulates the charging rate to ensure optimal charging. The cial for battery longevity, as any overcharge or Overcharging process of charging batteries is further temperature-sensitive. VOLUME 11, 2023 105783  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System the battery cell. Such insights enable a more accurate estima- tion of SOH and contribute to the overall understanding of battery performance and longevity. Various SOH estimation methods are given in Figure 26. A. ESTIMATION EXPERIMENTAL METHODS 1) MEASUREMENT OF BATTERY INTERNAL RESISTANCE A battery internal resistance, which controls the voltage drop during current flow, is a key factor in determining its SOH. This parameter is significantly affected by aging and degra- dation, with an increase in value indicating a decrease in battery SOH. Consequently, the internal resistance is fre- quently utilized as a robust indicator for estimating battery SOH. Several researchers have investigated techniques to measure this internal resistance, with the most prevalent method known as current pulse ,. This method applies Ohm’s Law by measuring the voltage drop across the battery for a specified current and then employs the following FIGURE 25. (a) Current-temperature SOA zone, (b) voltage temperature formula : SOA zone. OCV (SOC, T) − Vbat (SOC, T) Rb (SOC, T) = (6) Operate within the safe operating area (SOA) advised by the Ipulse manufacturer to ensure battery safety. where Rb represents the internal resistance of the battery, The SOA may need to be adjusted based on battery aging OCV the open circuit voltage, Vbat the voltage, and Ipulse the and environmental conditions, as battery function deterio- current applied. This technique is frequently used in labora- rates due to factors like resistance and capacity degradation, tories to accurately describe the internal resistance behaviour as shown in Fig. 25. of batteries under various operating situations. This method Furthermore, predicting the highest possible instantaneous is better suited for stationary and laboratory applications due power capacity becomes important as ESSs are utilized to to its time-consuming nature, which necessitates allowing meet higher power demands. As a battery state indicator, the battery to relax and attain equilibrium first, which takes State of Function (SOF) is utilized to determine the maxi- around an hour. mal instantaneous output capability and guarantee operating within the SOA. Causes, impacts, and outcomes of 2) BATTERY INTERNAL IMPEDANCE MEASUREMENT the drop in li-ion batteries with respect to health and battery life is given in Table 12 Various approaches are employed An indication of battery SOH is known to be a battery internal for battery state estimation, which can be categorized into impedance, which includes both internal resistance and reac- three main methods, as shown in Fig. 26. Internal resis- tance. It has been observed and substantiated that a battery tance, impedance, and capacity are the three basic parameters internal impedance tends to rise with time, making it a valu- used to estimate battery SOH. While internal resistance and able SOH indicator. The most commonly employed method impedance show the battery ability to deliver power, capacity for measuring impedance is Electrochemical Impedance displays how much energy it can hold. In contrast to EVs, Spectroscopy (EIS) ,. EIS is a non-destructive where battery energy is more important, power capability is of method that determines an electrical system’s impedance by higher significance in hybrid applications. Due to ageing fac- running a sinusoidal AC current through it and gauging the tors, these signs vary over the battery lifespan. By comparing voltage response. The impedance is measured across a range the actual indicator value (capacity, impedance, or resistance) of frequencies. One notable advantage of this method is its with its original value, SOH can be computed. Two different ability to accurately identify the aging phenomena occur- approaches, experimental and adaptive methods, can be used ring within the battery. In a specific study , the author to predict these changes. Experimental methods involve stor- employed EIS to investigate two key aging phenomena in ing the using battery cycling data history and newly acquired batteries: the movement of lithium ions through the Charge knowledge to determine SOH. By considering the impact of transport at the positive electrode and the SEI layer. key parameters on battery lifespan, it becomes possible to estimate the SOH of the battery. This estimation process 3) BATTERY ENERGY LEVEL necessitates a thorough understanding of the relationship The fundamental aspect of capacity shows the total energy between battery cell operation and degradation, which can be storage capacity of a battery. With aging, this capacity is obtained through physical analysis or by evaluating extensive known to decrease. Therefore, one of the most reliable tech- datasets that combine operation history and SOH testing of niques for calculating the battery SOH is by experimentally 105784 VOLUME 11, 2023 R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System FIGURE 26. Various SOH estimation methods. TABLE 12. Causes, impacts, and outcomes of the drop in Li-ion batteries. measuring the fading capacity over time. In a study conducted levels and temperatures. The utilization of a data map is by the author , multiple charging/discharging cycles necessary for accurate long-term predictions and because the were performed on a Li-ion battery until it reached its End calculation of a reliable IR value may require some time. of Life (EoL). The objective was to examine the relation- However, a drawback of this approach is that each map needs ship between the battery charging capacity and its voltage to be parameterized for individual cell references. In , at different levels of degradation (cycle numbers). Another Using the idea of a severity factor map, a strategy based on study focused on estimating battery capacity through a weighted ampere-hour throughput model of the battery is experimental testing conducted under varying temperature introduced. Within this framework, the investigation focuses conditions, ranging from 25◦ C to 40◦ C, with the battery on two primary factors that contribute to battery life reduc- subjected to 800 cycles. The offline data obtained from The tion: DOD and temperature. development of an online SOH estimate method was then done through experimentation. The battery is tested up to 2) COULOMB COUNTING its EoL using these experimental procedures, but it should Another commonly used technique for estimating SOH is be emphasized that they can only be used offline in lab Ah method. This method entails keeping note of the number circumstances. of Ah that are charged or discharged as the battery is being charged or discharged. The battery’s remaining capacity can B. MODEL BASED METHODS be calculated by tracking the transferred Ah. The esti- 1) DATA FITTING mation of SOH is calculated using Equation (7), the measured capacity Qnom and the maximum available capacity Qmax. Resistance measurement is a valuable data acquisition method for estimating the SOH of a battery. To achieve a detailed fitting of internal resistance (IR), a characteristic Qmax SOH = (7) map is proposed, which calculates the IR at various SOC Qnom VOLUME 11, 2023 105785  R. Ranjith Kumar et al.: Advances in Batteries, Battery Modeling, Battery Management System However, the Ah counting method has some drawbacks. in detail in Section. V, Kalman filters have been employed in It requires a high capacity for storing the counted Ah, which for battery state and parameter estimation. Specifically, can be time-consuming. Additionally, the method is sensitive the battery internal resistance is estimated, allowing for accu- to precision due to the accumulation of errors over time. The rate prediction of the SOH. coulomb counting method is still popular due to its simplicity, despite these drawbacks and its minimal dependency on other 2) OBSERVER parameters such as Depth of Discharge (DOD), temperature In order to estimate SOH, observers have also been used as or C-rate which often have a stronger impact on other estima- an adaptive identification technique. A sliding mode observer tion methods. is used in to calculate the SOH and SOC of a Li-ion battery. The technique exhibits good accuracy and resistance 3) PARITY-RELATION METHOD to modelling error and temperature changes. This method is used to assess and compare the effectiveness of batteries, allowing for the assessment of their desired 3) LEAST SQUARE-BASED FILTERS functionality, as demonstrated in. This method involves Another widely used approach in adaptive filtering is the use analyzing the battery dynamics during cranking using a bat- of Least Square-based algorithms, as discussed in and tery model. The analysis reveals that the residual integrates. RLS algorithms, in particular, have gained attention information about the State of Health (SOH) provided by both due to their simple implementation and accuracy. These algo