Battery Management System for Electric Vehicles: Comprehensive Review of Circuitry Configuration and Algorithms

Abstract

Electric vehicles (EVs) are the fastest-growing type of transport. Battery packs are a key component in EVs. Modern lithium-ion battery cells are characterized by low self-discharge current, high power density, and durability. At the same time, the battery management system (BMS) plays a pivotal role in ensuring high efficiency and durability of battery cells and packs. The BMS monitors and controls the battery charge and discharge to ensure EV safety and optimum operation. This paper is devoted to analyzing BMS circuitry configurations and algorithms. The analysis includes circuit solutions and algorithms for implementing the main BMS functions, such as parameter monitoring, protection, cell balancing, state estimation, charging and discharging management, communication, and data logging. The paper provides insights into the recent research literature on BMS, and the advantages and disadvantages of methods for implementing BMS functions are compared. The paper also discusses the application of artificial intelligence technologies and aspects of further work on next-generation BMS technologies.

1. Introduction

Energy storage system technologies based on electrochemical battery cells have been used since the mid-19th century. Accumulator battery technologies have been consistently improved over the last 100 years; new types of cells have appeared, and energy density, service life, and working reliability have been enhanced. An important advantage of electrochemical accumulator batteries is the simplicity of scaling, i.e., the possibility to integrate a certain number of battery cells into a battery pack with the required parameters (capacity, power, voltage). Currently, battery manufacturing is one of the fastest-developing sectors of the energy industry. Lithium-ion (Li-ion) battery cells are the most common type of battery cell [1]. Compared to other battery cell types, Li-ion battery cells offer high energy density and cell voltage, which make them the most attractive choice for different technological systems [2], such as renewable energy source systems, standby power sources, and portable devices.

Li-ion battery technology is an important concept for the development of electric vehicles (EVs) [3]. Hybrid and EV battery packs are composed of series and parallel configurations of lithium-ion cells.

However, lithium technology is vulnerable and highly susceptible to catastrophic failures [4]. There are a variety of inherent and extrinsic problems with batteries that concern safe working conditions, temperature maintenance, and correct charging and discharging [5]. Thus, in order to efficiently and safely operate energy storage systems based on battery packs, it is necessary to carefully manage their charge and discharge processes.

A battery management system (BMS) is a crucial component in battery management. The BMS plays a pivotal role in regulating and controlling the charging and discharging of the battery pack to ensure safe and optimum operation [6]. The main components of the BMS are sensors, actuators, and controllers (Figure 1).

The BMS performs the following key functions [7,8]:

• Battery cell parameter monitoring—the BMS mainly focuses on monitoring voltage, current, and temperature.
• Battery cell protection—the BMS must ensure protection against battery system hazards (charge and discharge control; overcurrent).
• Cell balancing—the BMS must use a passive or active equalization method, minimizing the irregularity of cells.
• State estimation (SoC, SoH) and fault diagnosis (insulation)— the BMS estimates and predicts the state of charge (SoC) and the state of health (SoH); the BMS is also responsible for detecting faults, such as fires, thermal runaways, and explosions, and for minimizing the consequences of fault effects.
• Charging and discharging management—to ensure a long service life for the battery pack, the BMS must sustain the corresponding SoC and provide the most efficient method for charging and discharging procedures.
• Communication and data logging—the BMS must govern and filer battery pack data, as well as accumulate crucial information.

Generally, direct measurements of battery cell voltage, current, and temperature are used in the BMS. The data obtained are necessary for the operation of all BMS functions (Figure 2).

BMS techniques are being constantly improved. Their peculiarity is that, on the one hand, technology development for each BMS function can be considered a separate avenue. However, BMS functions are interconnected both on software and hardware levels. Thus, in order to obtain the required result and boost BMS operating efficiency, BMS techniques’ improvement should be integrated and should combine achievements in different areas. With this in mind, a crucial task for BMS developers is the choice of proper algorithms and circuitry, BMS characteristics, and functional capabilities, considering the existing range of implementation variations for each BMS function and modern trends in development. Proper architecture, functional blocks, and advanced circuitry can extend battery life.

Many BMS review papers are available in the research literature, considering various construction techniques for management systems (

Table 1. Previous works related to BMS technologies.

Reference	Year	Description
Vaideeswaran et al. [8]	2019	Overview of the main BMS functions
Darwish et al. [4]	2021	Overview of diagnostic functions and charging algorithms
Gabbar et al. [9]	2021	Analysis of BMS structures used in EV and stationary energy storage
Mishra et al. [3]	2021	Overview of the BMS functions; lithium-ion battery modeling analysis
Spoorthi et al. [10]	2022	Overview of BMS balancing and diagnostic functions
Long et al. [6]	2023	Overview of BMS technological improvement directions
Devi et al. [11]	2023	Overview of BMS functions
Bhat et al. [12]	2024	BMS modeling
Vijaychandra et al. [13]	2024	Methods to improve battery safety

).

Paper [8] provides a comprehensive overview of BMS technologies, such as monitoring, state estimation, charging and discharging control, temperature control, fault analysis, data acquisition and protection schemes, to improve the performance of batteries for EV applications. However, this review was performed in 2019. A description of diagnostic functions and charging algorithms is presented in paper [4]. A description of aspects of BMSs, covering testing, functionalities, topology, operation, architecture, and safety, is given in [9]. Paper [3] gives a review on strategies like battery modeling, state estimation and prediction. Paper [10] is devoted to the description of battery parameter monitoring, cell balancing and state estimation functions. Paper [6] provides an overview of cell balancing technologies, state-of-charge detection and the use of IoT for BMSs. Many of a BMS’s functions are discussed in paper [11]. However, this review does not cover data logging methods and artificial intelligence technologies. In paper [12], the main focus is on state estimation and fault diagnosis technologies. Paper [13] discusses battery protection and diagnostic functions.

Table 2. Comparison table between surveys that are related to BMSs.

Reference	Battery Parameter Monitoring	Battery Protection	Cell Balancing	State Estimation and Fault Diagnosis	Charging and Discharging Management	Communication and Data Logging	Artificial Intelligence and Big Data Technologies
[8]	∨	∨	∨	∨	∨	∨	-
[4]	-	-	∨	∨	∨	-	-
[9]	∨	∨	-	∨	-	∨	-
[3]	-	-	-	∨	-	∨	-
[10]	∨	-	∨	∨	-	-	-
[6]	-	-	∨	∨	-	∨	-
[11]	∨	∨	∨	∨	-	-	-
[12]	-	-	-	∨	-	-	∨
[13]	-	∨	-	∨	-	-	∨

shows a comparison between the BMS review papers.

This paper provides a comprehensive review of the BMS literature in recent years. The purpose of the paper is to provide a comprehensive understanding of the development of both hardware and software components of BMSs. The main focus is on the BMS functions shown in Figure 2. The paper presents approaches to the implementation of BMS hardware, in particular for measurement, protection, and balancing systems. The already applied and promising algorithms that are used in the functions under consideration are analyzed. In addition, the paper provides a brief discussion of the limitations and problematic aspects, as well as promising research directions, related to artificial intelligence technologies. The reader is introduced to battery cell parameter monitoring and battery cell protection in Section 2 and Section 3, respectively. Cell balancing (Section 4) and state estimation (Section 5) follow. Artificial intelligence and Big Data technologies are presented in Section 6. Then, charging and discharging management (Section 7) and communication and data logging (Section 8) are explained and analyzed. Finally, Section 9 discusses some aspects of future work on next-generation BMS technologies.

The paper can be a useful resource for researchers who are interested in the application of EV battery management systems and related topics.

2. Battery Cell Parameter Monitoring

Devices for measuring battery cell parameters can be divided into two groups: measurers of electrical and non-electrical parameters. Measuring only one parameter type is not enough to precisely estimate a battery cell’s state.

The basic minimal sensor set within a BMS to measure cell parameters comprises three sensor types: a temperature sensor, voltage sensor, and current sensor. This set of sensors is used in most series-produced BMSs. Here, only one current sensor is usually used for the battery pack. Individual voltage sensors are used for each battery cell, while the number of temperature sensors compared to voltage sensors is 4-6 times lower (1 temperature sensor per 4-6 battery cells).

Measuring non-electrical parameters (except temperature) is a prospective area for the development of monitoring battery cell state. However, the methods of measuring non-electrical parameters can be realized only within the framework of laboratory investigations.

2.1. Voltage Measurement

In order to ensure safe and efficient EV operation, it is necessary to monitor the voltage of each battery cell in the battery pack. Data of measured battery cell voltage values is necessary for most protection and diagnostics algorithms’ operation.

Figure 3 shows the assessment method of circuit architecture options to measure the battery cell voltage.

The simplest way to measure voltage is measuring with bias voltage circuits based on voltage dividers followed by digitalization (Figure 3a). Despite the low battery cell voltage (less than 5 V), this method is used exclusively for low-voltage battery packs due to the absence of galvanic decoupling. In addition, the measurement accuracy deteriorates when using bias voltage circuits [14].

When operating high-voltage battery pack, galvanic decoupling between the BMS and the battery cells is a necessary condition. In this case, optical or electromagnetic converters can be used, allowing one to eliminate galvanic coupling between the high- and low-voltage sides (Figure 3b) [15]. Employing an individual galvanic decoupling microcircuit and an individual analog–digital converter (ADC) allows one to obtain a high measurement rate. There are several drawbacks here, such as increased energy consumption for the measuring circuits, a large number of electronic components, and, consequently, greater sizes of printed circuit assemblies.

In order to reduce the energy consumption, number of components, and printed circuit assembly dimensions, multiplexers are used, allowing one to carry out consecutive measurements of battery cell voltage. The basic method of integrating a multiplexer into a voltage measurement system involves mounting it on the low-voltage side (Figure 3d). In this case, the number of ADC discrete microcircuits decreases. However, the number of galvanic-decoupling microcircuits remains the same.

Currently, a great number of ADCs with integrated multiplexers are produced in-series, allowing a decrease in the size of printed circuit assemblies. Examples of such microcircuits include the following:

• MCP3008, which enables one to measure the voltage of eight battery cells;
• LTC6802, which is a fully featured device for monitoring battery cell parameters [16,17].

Alongside that, despite the advantages of these devices, the necessity to connect digital isolators remains unsolved.

In order to eliminate the need for individual galvanic decoupling circuits, high-voltage multiplexers are used (Figure 3c) [18]. These measuring circuits’ architecture allows one to minimize the number of microcircuits. The main issue of using high-voltage multiplexers is their reliability in long-term operation.

As far as voltage measurement is concerned, the considerable progress seen in battery cell construction is noteworthy. The employment of standard electrodes to allow for measuring voltage on cathode and anode circuits independently could be seen as an example of this [19].

2.2. Current Measurement

Current measurement is necessary to estimate battery pack service life, as well as to ensure protection from overcurrents and overload. Real-time current value monitoring allows one to determine the accumulated energy volume for further SoC estimation. Data from current sensors gives an insight into the energy consumed by the battery pack, helping to determine the EV’s remaining usable lifespan. In addition, real-time-measured current values can be used by safety systems for system malfunction notifications [20].

Currently, the following types of current sensors can be singled out as those typically used in battery packs: current shunt sensors; Hall effect sensors; magnetoresistive sensors (XMR); and fiber-optic sensors (fiber Bragg grating (FBG)).

According to structural design, the simplest method of current measurement is the shunt resistance one. The main advantages of current-measuring shunts are their low cost and wide range of operational temperatures. However, this type of sensor is characterized by high electric losses and poor measurement accuracy.

Hall effect sensors have the following advantages: galvanic decoupling, low energy consumption, and the wide range of the current measured [21,22]. Alongside this, this sensor type has a number of serious drawbacks: high cost, necessity for temperature compensation, and sensitivity to external magnetic fields.

Magnetoresistive sensors, including anisotropic magnetoresistance (AMR), giant magnetoresistance (GMR), and tunnel magnetoresistance (TMR), are characterized by high measuring accuracy [23]. Current meters based on XMR are a prospective area for development. However, their application is limited by the problem of nonlinearity in measurements [24].

Fiber-optic sensors are another prospective avenue for developing current sensors [25]. Such sensors are practically insensitive to temperature changes and environmental impacts [26]. However, the complexity of their construction is a serious drawback limiting their wide application.

2.3. Temperature Measurement

Measuring battery cells’ temperature is an essential BMS function that is responsible for the battery pack’s safe operation. One of the main considerations is the temperature detector location. Bearing in mind that the difference between the inner and outer parts of the battery cell can be considerable, the best option is to locate the detector inside the battery cell [27]. However, due to aggressive media and possible battery cell deformation, such a method of mounting temperature detectors can be difficult.

According to their operating principle, temperature detectors can be divided into resistive temperature detectors (RTDs), thermocouples, or fiber-optic detectors [27]. Infrared imaging devices can be used to study battery pack temperature modes. However, in BMSs, such devices are seldom applicable due to their high cost, large amount of data, and resulting complexity in data processing [28].

A set of sensors can be used to obtain a comprehensive picture of the battery cell’s temperature distribution. Here, resistive temperature detectors are usually used [29]. Due to their simplicity of construction and sufficient accuracy to facilitate the functional ability of protection algorithms, resistive detectors are used in many temperature measurement systems. However, their considerable nonlinearity narrows the range of effective temperature measurement [30].

Thermocouples are most often used for temperature measurement within wide ranges [31]. However, due to their low measurement accuracy in practice, a thermocouple (with temperature fixation over a wide range) combined with an RTD (with temperature fixation in nominal mode) is used in order to monitor battery cell temperature [32].

Fiber-optic Bragg gratings for precise temperature measurement (with an accuracy in the range of ±0,12 °C) are a prospective area for development [33]. Such systems are immune to electromagnetic interference [34]. However, the complexity of their construction is a limiting factor for the wide use of this type of temperature detector.

2.4. Strain Measurement

When battery cells operate in the area of maximum admissible limits, excessive thermal load causes considerable mechanical stress and expansion. Taking this into consideration, the detection of bubbling in a battery cell when performing the complex monitoring of its state appears to be a critical task. In addition, it is possible to circumstantially estimate the battery cell’s state of charge on the basis of measuring its geometrical dimensions.

Publications from various research teams present methods employing optical and fiber-optic sensors [35,36]; acoustic devices to, inter alia, estimate a battery cell’s inner state [37]; and electromechanical sensors [38]. However, most methods are difficult to apply in an EV BMS structure due to their high cost and dependence on external conditions.

A prospective solution could be to use a strain sensor based on carbon nanotubes (CNTs) [39]. This sensor’s operation is based on measuring the sensor resistance at its deformation. Experiments have shown positive results when measuring battery cell strain [40]. However, the linearity of resistance change within narrow ranges and CNTs’ temperature dependence limit the mass production of this sensor type.

2.5. Analysis Results

All BMS functions’ implementation efficiency depends on the accuracy of measuring battery cell parameter values.

Table 3. Analysis results of monitoring methods.

Method	Measurement Accuracy	Measurement Linearity	Problems for Integrating into EVs	Cost
Voltage measurement
Voltage dividers	Depends on resistors	High	Impossible due to low operating voltage	Low
Individual galvanic decoupling	Depends on galvanic decoupling microcircuits	High	Absent	High
Low-voltage multiplexors	Depends on galvanic decoupling microcircuits and multiplexors	High	Absent	High
High-voltage multiplexors	Depends on galvanic decoupling microcircuits	high	Difficulty concerning high-voltage multiplexor reliability	Medium
Current measurement
Shunt resistance	Medium	High	Absent	Low
Hall effect sensors	Medium	Medium	Absent	Low
Magnetoresistance effect-based sensors	Depends on individual magnetoresistance effects	Medium	Absent	High
Fiber-optic sensors	Low	Low	Construction complexity	Medium
Temperature measurement
Thermoresistors	High within a narrow range	Low	Absent	Low
Thermocouples	Medium	Medium	Absent	Medium
Fiber-optic sensors	High	Medium	Construction complexity	High

shows comparative analysis results of methods for measuring battery cell parameter values.

For EVs, the main indicators are measurement accuracy, linearity, and the cost of implementation. Table 1 does not contain data on strain sensors due to the complexity of integrating them into EVs.

3. Battery Cell Protection

Battery cell protection involves the detection of and response to potentially dangerous operating conditions, such as overvoltage, undervoltage, overcurrent, and overtemperature. In addition, protective equipment can also be implemented at the site of battery cell deformation (for example, for EV accidents).

The differences between battery cell protection systems lie in the algorithms serving as their operating principles. Scientific research is mainly focused on developing and improving protection algorithms.

3.1. Protection from Overcurrent and Short Circuits

Current protection is a necessary component in energy storage systems. Apart from primary current protection, fuses can be additionally used to enhance system safety [7].

The following failure scenarios are possible for EVs: fault inception in the battery circuit; fault inception in the charger circuit; fault inception in the invertor circuit; or fault inception in the load circuit [41]. Detecting the place of failure is an important function of modern current protection systems. Such systems are more difficult to implement, but they are able to protect not only the battery pack but also the electric drive and charging infrastructure’s key elements.

A comparator circuit is the simplest example and is used to determine overcurrent and battery pack disconnection [42]. The main disadvantage of such circuits is the high probability of false responses.

Special microcircuits are used in modern BMSs, providing the possibility of multi-level protection [41]. Thus, a microcircuit can have three levels with a fixed delay time for each (for example, two overcurrent levels and one short-circuit current level).

In order to realize time-current protection algorithms, relay protection methods are applied [43]. Battery pack protection systems based on such methods can be realized on the basis of microcontrollers, and alongside the current itself, this allows one to take into consideration other key operating parameters (for example, the temperature of the battery cells and ambient medium).

A promising research avenue is the adaptation of protection principles from the electrical energy system field to energy storage systems [44]. Thus, controlled reactors in protection systems not only allow one to switch off a faulty section but also preserve its operability at the nominal current. However, due to these reactors’ large dimensions, such methods are not yet applicable for EVs.

3.2. Overvoltage and Undervoltage Protection

Overvoltage and undervoltage in battery cells have a negative impact on battery pack safety and service life.

In most cases, battery cell protection algorithms are based on setting voltage threshold values [45]. Voltage threshold values are well known for common battery cell types. The main problem with such realizing this protection is connected with the error when determining the real voltage values of battery cells, which takes place due to voltage drops in their non-core elements. It is necessary to take into consideration charge/discharge current values in order to determine voltage values more precisely.

Undervoltage protection is aimed at preventing battery cell overdischarge, which can lead to a sharp drop in the battery pack’s service life and a loss of its capacity. When the minimum admissible voltage is reached, at least one battery cell of the battery pack needs to be cut off from the load [46]. There are also algorithms that allow one to reconfigure the battery pack at a low voltage in its individual sections [47]. However, such systems are characterized by a large number of semiconductor switches and have not become widespread due to this fact.

Overvoltage protection prevents battery cell overcharge. Thermal runaway may occur at considerable overcharge, leading to hazardous situations (all the way up to battery cell explosion). Practically all battery cell protection means are based on setting voltage cut-off values; when these are exceeded, the charger switches off.

Currently, microcircuits are widely used to provide protection from both undervoltage and overvoltage [48]. Charger protection systems can also be used [49]. Such means of protection are additional to the main battery cell protection realized by a BMS. Protection duplication allows one to avoid emergency situations where one or several BMS protection elements fail.

Varistors can also be used in protection systems [50]. They make it possible to realize overvoltage protection at a comparatively low rate of voltage change (in standard charge mode). However, varistors are inefficient in short-circuit events.

3.3. Temperature Protection

The reasons for battery cell overtemperature can be the following: high ambient temperature, overcharge, thermal management system failure, or EV damage.

Battery cells overheating can lead to increased consumption of electrolytes, cathode breakage, or an increase in resistance. These processes lower the battery pack’s service life and capacity [51].

Battery cell temperature protection methods can be classified as passive or active. Passive ones comprise current interruption devices (CIDs) and temperature fuses [52]. Passive methods are not the most basic ones for EV battery packs because when they are actuated, maintenance work is necessary to bring the battery pack into operation once again.

Active protection methods imply battery temperature regulation, including time periods when the battery pack is not used or only prepared for operation. Battery temperature management systems (BTMSs) are responsible for temperature regulation in BMSs [53]. The methods of temperature regulation applied in a BTMS can be divided into passive (using natural air cooling, heat pipes, or phase transfer materials), active (using ventilators, liquid cooling, or thermoelectric coolers), and hybrid [54]. Passive battery pack cooling methods are cheaper compared to active ones, but they do not provide temperature control or eliminate thermal runaway. When designing a cooling system, it is important to take into consideration heat carrier parameters, such as the material’s melting temperature, thermal conductivity, mass, and distance between the cells [55]. Machine learning technologies and digital twins can be used to carry out the tasks of battery pack temperature forecasting and subsequent temperature regulation [56].

At low temperatures, battery capacity considerably decreases. Low-temperature preliminary heating is used to solve this problem. Preliminary heating methods are classified as external (air or liquid heating and using heat pipes), internal (discharge current heating, excitation current heating, and self-heating batteries), or hybrid [57]. External heating methods are simpler but require detailed designs and special materials to ensure safety. Internal heating methods provide higher efficiency and uniform temperature distribution. However, the implementation of these methods requires a more complex control system. In addition, internal heating methods’ long-term impact on the battery pack’s life and operational safety has not been researched.

3.4. Analysis Results

The protection function is critical as it allows one to avoid emergency scenarios while the battery pack is in operation. When permissible current values are exceeded or the voltage values deviate, it is necessary to decrease the load power or switch off the battery pack.

Table 4. Analysis results of protection methods.

Method	Response Rate	Selectivity	Difficulties of Integrating into EV	Cost
Voltage measurement
Comparator circuit	High	Absent	Absent	Low
Multi-level protection	High	Present	Absent	Low
Time-current protection	Medium	Present	Absent	Medium
Controllable reactors	Low	Present	Weight–size parameters	High
Voltage protection
Comparator circuit	High	Present	Absent	Low
Varistors	High	Absent	Absent	Medium
Temperature protection
Passive protection	High	Absent	Re-implementation requirement	Low
Passive management	Not assessed	Present	Absent	Medium
Active management	Not assessed	Present	Absent	High

shows comparative analysis results of protection methods.

When designing a protection system, temperature regulation is particularly emphasized as temperature values directly depend on load parameters. Both cooling and heating systems are used for temperature regulation. They allow the improvement of battery pack operational efficiency and safety.

4. Cell Balancing

Cell balancing methods are divided into two categories: passive cell balancing and active cell balancing. Each category comprises its own set of methods. Figure 4 shows a general breakdown of cell balancing methods.

Both active and passive balancing methods are of interest when applied to EVs. A possible growth area for cell balancing methods is developing modified and hybrid balancing methods.

4.1. Passive Balancing

The passive balancing principle implies leveling off all battery cells’ state of charge via dissipating excessive charge using a passive element [58]. Here, fixed resistor balancing and switched resistor balancing can be singled out.

The concept behind fixed resistor balancing is battery cell resistor shunting, where the battery cell is up to 100% charged but the charging of the whole battery pack is not yet completed. In this case, there are continuous currents passing through shunt resistors [59]. This method’s advantages are its simplicity and the low cost of its realization [60]. However, it is characterized by energy wastage and a low rate of balancing. Switched resistor balancing involves an additional switch connected to each shunt resistor [59]. If one or several cells are fully charged earlier than others, the switches connected to them will be turned on. Compared to fixed resistor balancing, this method is characterized by a higher efficiency and rate of balancing, though the cost of its realization is also higher.

Despite these disadvantages, EV manufacturers apply both passive balancing methods due to their reliability and simplicity to implement.

4.2. Active Balancing

The active balancing principle lies in the redistribution of energy from cells with a higher state of charge to ones with a lower charge state [61]. Active balancing methods can be applied during battery pack charging and discharging. Active balancing methods can be classified according to the active devices used to transfer the energy. Such devices include capacitors, inductors, transformers, and switching converters [62].

Capacitive cell balancing is also known as a charge-stabilizing method, in which capacitors store the charge and return it to the least charged cells [63,64]. The following basic capacitive cell balancing methods can be applied to EV battery packs: basic switched capacitors; single switched capacitors; and double-tiered switched capacitors.

Basic switched capacitors are the simplest method. According to their balancing principle, each cell contains a surge capacitor and the charge can be transferred only through adjacent cells [61]. Single switched capacitors are an enhanced variant of switched capacitors, requiring a single capacitor for balancing. In this method, the charge can be transferred directly from one cell to another through a surge capacitor within one module [65]. Double-tiered switched capacitors are based on double capacitor tiers for shuttling the charge between two cells. This requires n capacitors for a double-tiered capacitor for n cells, and 2n switches [61]. An intelligent control system is required to control the switches.

In inductor cell balancing, excess energy from overcharged cells is stored in the inductor, which is then used to charge the undercharged cells [66]. Three basic inductor balancing methods can be singled out: single-inductor balancing; multi-inductor balancing; and chain structure multi-inductor balancing.

For single-inductor balancing, a single inductor is used to balance the cells by controlling different switches [67]. The advantage of this method is its relatively high efficiency. However, the method’s disadvantage is its management complexity [68]. Multiple inductors are used to balance cells in multi-inductor balancing. Each pair of adjacent cells in the battery pack is balanced by an inductor. This allows for balancing multiple couples of cells simultaneously [66]. Compared to the single-inductor balancing method, this one allows one to reduce the balancing time and the number of switches needed. However, one notable drawback is the larger number of inductors required when more cells are connected in series [69]. In chain structure multi-inductor balancing, a capacitor is used in the multi-inductor balancing circuit. The capacitor provides an additional path for the current, which reduces the path distance between the first and last cells [66]. A chain structure cell balancing circuit with coupled-inductor-based modules is detailed in [70].

Transformer balancing methods provide energy transfer from one cell to another via various transformer types. The following transformer balancing types can be distinguished: single-winding transformer balancing; multiple-winding transformer balancing; and multiple-transformer balancing.

Single-winding transformer (switched transformer) balancing is based on transferring energy from the battery pack to the switching transformer and transferring energy to the weakest cell by means of switches [71]. One transformer with one primary and several secondary windings is used in multiple-winding transformer balancing. The multiple-transformer balancing method utilizes several transformers, where all the primary windings are connected in parallel, and each of the secondary windings is connected to a separate cell via a diode. The primary winding is connected across the pack voltage via a switch, and power is transferred from the pack to the cells by switching to a 50% duty cycle [72]. Unlike the multiple-winding transformer topology, this method allows connecting additional cells without changing the controller.

In contrast to other methods, switching-converter cell balancing can control the flow of power in any way that the BMS commands, allowing more flexibility in managing the cells’ SoC [72]. The following basic topologies can be singled out: Cuk converter balancing; buck-boost converter balancing; flyback converter balancing; and full-bridge converter balancing.

The Cuk converter balancing topology is detailed in [73]. A Cuk converter connects two adjacent cells. The cell voltage differences determine the control of the switches so as to control the energy flow between two adjacent cells. A bilateral buck-boost converter is used in buck-boost converter balancing as a bridge for the energy transfer between two adjacent cells [74]. There are several topologies of buck-boost converters which can be used for balancing circuits. One of the topology types is given in [62]. Flyback converter balancing is characterized by a high rate of balancing and is one of most common methods, including when a high-power battery cell is used [75]. Various topology types of this method can be found in the scientific literature: a conventional single-transistor converter [76], two-transistor converter [77], active clamp converter [78], bidirectional converter [79], flyback converter with multiple windings [80], and others. An example of the full-bridge converter balancing topology is given in [58]. Here, a sensing circuit senses the voltage across the battery cell, and a control signal is generated to operate the switch so that energy is transferred from one cell to another to maximize the charge capacity [81].

4.3. Development Trends of Cell Balancing Methods

The development trends of balancing methods imply the development of improved and upgraded modifications of the above methods, aimed at eliminating their deficiencies, as well as developing hybrid methods which combine several methods’ advantages.

Modifications of active balancing methods are extensively presented in the scientific literature. For example, a modification of the single-switched capacitor-based balancing method, which uses 2n switches to shorten the balancing time, is detailed in [58]. An improved balancing strategy for an inductor-based balancing circuit, which increases the remaining charge of the battery pack after balancing, reduces losses, and shortens the balancing time compared to the original balancing strategy, is presented in [82].

A number of hybrid methods combining active and passive balancing principles have been suggested. An active- and passive-based hybrid balancing method is proposed in [83]. It combines a conventional switched capacitor circuit with a switched resistor passive balancing circuit. The circuit consists of two parts, namely, a switched capacitor circuit and switched resistor circuit. The research results show that the proposed circuit allows increasing the balancing speed significantly compared to the conventional switched capacitor circuit and passive balancing circuit. A similar hybrid balancing method is shown in [84]. The authors suggested a hybrid balancing technique for Li-ion batteries using capacitor- and converter-based balancing. Capacitor-based balancing is used between two consecutive cells when slow balancing is required, while converter-based balancing is employed for other situations to transfer charge to or from the cell in order to encounter the imbalance as fast as possible. A hybrid balancing method comprising inductor-based and resistor-based balancing techniques is shown in [85]. By leveraging the strengths of both techniques, the proposed method aims to achieve optimal cell balancing while minimizing energy loss and balancing time.

Thus, hybrid methods allow combining the advantages of active and passive balancing principles, thereafter improving the balancing process’s efficiency.

4.4. Analysis Results

Table 5. Analysis results of cell balancing methods.

Method	Advantages	Disadvantages
Fixed resistor balancing	Simplicity and low cost of realization	Low efficiency; low rate of balancing
Switched resistor balancing	Higher rate of balancing and efficiency compared to fixed resistor balancing	Higher realization cost compared to fixed resistor balancing
Basic switched capacitor	Simplicity of operation	Big number of switch keys, low efficiency, low rate of balancing
Single switched capacitor	Simplicity of operation, high efficiency	Low rate of balancing
Double-tiered switched capacitor	Relatively high rate of balancing; simplicity of operation	Large number of switch keys; high realization cost
Single-inductor balancing	Relatively high rate of balancing; high efficiency	Complexity of operation; high realization cost
Multi-inductor balancing	High rate of balancing	The number of inductors increases when the cells are connected in series; high realization cost
Chain structure multi-inductor balancing	High rate of balancing	The circuit size and realization cost are higher
Single-winding transformer (switched transformer) balancing	Relative compactness	High realization cost
Multiple-winding transformer balancing	Relative compactness	Number of cells limited by the number of secondary windings; low efficiency
Multiple-transformer balancing	High rate of balancing	Low efficiency, large dimensions, and high realization cost
Cuk converter balancing	High rate of balancing and efficiency	Complexity of operation; relatively large dimensions
Buck-boost converter balancing	High rate of balancing and efficiency	Complexity of operation, relatively large dimensions, and high realization cost
Flyback converter balancing	High rate of balancing	Transformer needed
Full-bridge converter balancing	High rate of balancing	Complexity of operation, relatively large dimensions, and high realization cost

presents the analysis results of the advantages and disadvantages of cell balancing methods.

The comparison of balancing methods shows that there is no single best method. Each balancing method demonstrates different combinations of efficiency, balancing rate, complexity of the control system and operation, dimensions, and realization cost. The switching-converter cell balancing method can be considered to have more potential if the dimensions and realization cost are reduced.

5. State Estimation

State of charge (SoC), state of health (SoH), and state of function (SoF) estimation are the primary criteria in the assessment of battery pack utilization efficiency. The values of these criteria are used in monitoring, protection, and control algorithms [86].

Currently, cloud storage and artificial intelligence are considered prospective methods in assessing battery pack state. These allow us to lower the requirements for computational resources in BMS hardware [87].

5.1. Determining SoC

SoC estimation methods are divided into three groups: direct measurement methods; the Coulomb method; and different variations of adaptive filters (Figure 5).

The Coulomb method is the simplest. Its source calculation data are current values, temporary data, and initial state of charge values [88,89,90]:

$$\mathit{SoC}(t)=\mathit{SoC}(0)-{\displaystyle \frac{1}{Ahnom}}\int Itdt$$

(1)

However, the Coulomb method does not allow one to precisely estimate the SoC due to the change in battery pack capacity during the process of degradation.

The SoC estimation method that involves measuring open-circuit voltage (OCV) is the most common direct measurement method. In order to apply this method, the measurement of real-time battery pack voltage values and volt–charge characteristics is necessary [91]. However, SoC estimates based on this method may lead to overestimated values of instant capacity upon battery cell degradation.

Adaptive filters usually involve applying a Kalman filter variation [92]. The methods of this group are not independent; rather, they complement the other ones, allowing the SoC estimation accuracy to be improved. The main disadvantage of adaptive filters is the vast amount of calculations needed, which are difficult to realize on a BMS microcontroller.

Developing hybrid SoC estimation methods considering SoH is a promising area, which allows one to compensate for basic methods’ disadvantages and improve accuracy.

5.2. Determining SoH

SoH estimation is used to forecast remaining battery pack service life and prevent premature battery pack failure. SoH is also used to adjust charge/discharge algorithms in order to extend battery pack service life [93]. A battery pack’s actual and nominal capacity is source data for SoH estimation:

$$\mathit{SoH}={\displaystyle \frac{Qavailable}{Qnominal}}$$

(2)

SoH estimation methods can be divided into experimental and model-based ones (Figure 6).

Experimental SoH estimation methods are based on direct measurements and indirect analysis of battery pack parameters [94]. The disadvantage of experimental methods is that the possibility of dynamic SoH measurement is not always available.

Apart from determining the remaining capacity, SoH model-based estimation also allows one to forecast battery pack capacity change during service [95,96]. Such methods require information about parameters, taking into account operational activity, which leads to large data bases being necessary for correct operation [97].

5.3. Determining RUL

The remaining useful life (RUL) of battery packs is one of the criteria for battery pack state estimation and is determined as the time in which the first failure takes place [98,99]. RUL estimation methods can be divided into three categories: artificial intelligence-based; adaptive filters; and stochastic (Figure 7).

Artificial intelligence-based methods involve using battery pack degradation data bases for further RUL estimation [87]. However, battery pack degradation depends on a number of factors, which are difficult to combine in a single algorithm. This is the reason for the decreased accuracy of RUL estimation.

Kalman filters are distinguished among the adaptive methods [100]. However, RUL estimation accuracy based on these methods will deteriorate with limited data sets.

Stochastic methods represent a promising trend in RUL estimation [101]. However, it is necessary to take into consideration the changing parameters during battery pack operation to achieve high RUL estimation accuracy.

Thus, determining RUL is a complicated task, with complex battery parameter information during the battery’s operation required to solve it. This results in large amounts of processed data. In addition, RUL estimation algorithms require vast computational resources. With this in mind, currently, dynamic RUL estimation is practically not used in BMSs.

5.4. Determining SoF

SoF reflects the battery pack’s ability to provide the electrical receiver with power [102]. SoF is a digital parameter and depends on both SoC and SoH. SoF estimation is relevant when feeding devices with high start-up power consumption, as well as for devices operating at overload. Two main approaches are used to estimate SoF: required power estimation and the estimation of minimal voltage on the battery pack clamps [103].

When estimating SoF based on the required power, the instantaneous voltage value and residual capacity value, necessary to determine the battery pack momentary power, are used as source data [104].

In the second approach, actual voltage values on the battery pack clamps are compared with minimum admissible values to determine the SoF [105]:

$$\mathit{SoF}=\left\{ \begin{array}{c}1 if Vmin\ge Vlimit\\ 0 if Vmin<Vlimit\end{array}\right.$$

(3)

There are also generalized methods allowing one to estimate SoC, SoH, and SoF simultaneously [106].

5.5. Determining EIS

Electrochemical impedance spectroscopy (EIS) allows one to diagnose the battery pack state and estimate SoC, SoH, and other parameters. The essence of EIS lies in sending test signals to the battery pack clamps followed by measuring the current and voltage values [107]. The measurement results at different frequencies allow one to determine the battery pack equivalent circuit parameters and forecast parameter changes [108].

There are three major problems when developing EIS systems: the velocity of measurement [109], measurement multichanneling [110], and the possibility of integrating the hardware of EIS measurement systems into existing BMSs [111]. The latter problem poses an impediment to applying EIS for EVs as it requires complex hardware for precise measurements.

5.6. Analysis Results

Table 6. Analysis results of battery state diagnostic methods.

Method	Accuracy	Disadvantages	Requirements for Computational Resources	Cost
Determining SoC
Coulomb method	High	Initial point required	Low	Low
OCV	Medium	Disregards loss of capacity	Low	Low
Adaptive filters	High	Preliminary calculations required	Medium	Medium
Determining SoH
Direct measurements	Medium	Experiment required	Low	Low
Adaptive algorithms	High	Big data required	Medium	Medium
Digital twins	High	Big data and computation capacities required	High	High
Determining RUL
AI	High	Big data and computation capacities required	High	High
Adaptive filters	Medium	Data required during whole operational phase	Medium	Medium
Stochastic methods	Medium	Data required during whole operational phase	Medium	Medium

shows comparative analysis results of battery pack diagnostic methods.

The problem of diagnosing the battery state is a complex one, as there are no universal circuits, models, or sets of equations to very accurately describe the battery state during its whole life. Currently, classical methods to determine the battery state are still used. These methods are complemented by adaptive filters, AI technologies, and digital twin technologies in order to improve the accuracy of determining the battery state and forecast the battery parameters’ changes.

6. Artificial Intelligence and Big Data Technologies

Currently, great attention is paid to artificial intelligence and big data technologies, which are intensively researched relative to BMS operation. The previously described BMS functions can be complemented by such technologies to increase their efficiency.

6.1. Enhancing Cell Balancing Efficiency

The modern trend in improving balancing-process efficiency is towards machine learning technologies, which can be used for both passive and active balancing methods.

Ref. [112] proposes cell balancing methods that utilize backpropagation neural network (BPNN), radial basis neural network (RBNN), and long short-term memory (LSTM) models to select an optimal resistor for passive battery balancing. Each model’s parameters are based on the SOC, temperature rise, balancing time, and C-rate. Ref. [113] presents a reinforcement learning method that targets passive cell balancing using a cell simulation that contains a dynamic thermal model.

An example of using machine learning technologies to enhance active balancing efficiency is given in [114]. The authors developed and evaluated a model predictive controller (MPC) for active cell balancing to extend the range capability of EV batteries.

6.2. Improving State Estimation Efficiency

Machine learning and digital twin technologies are used to improve the accuracy of SoC and SoH estimation, as well as to forecast RUL and internal resistance and measure battery temperature.

In some publications, different variations of machine learning have been used to estimate SoC [115]. The input data were the following: current, voltage, and battery temperature. Employing artificial neural networks (ANNs) [116], support vector machine (SVM), [117], and fuzzy logic (FL) [118] has also been considered.

ANN and FL allow the determination of battery parameters in online mode with little sensitivity to disturbances while taking into consideration data nonlinearity. However, this method requires significant computational capacity and big training data. SVM requires less training data, though data within the whole range of parameter changes is necessary for learning. The data quality also greatly impacts accuracy. GA is characterized by high accuracy, but the solution obtained via this algorithm does not always bring about the optimal result.

The efficiency of long short-term memory (LSTM), a convolutional neural network (CNN) and LSTM in combination, a deep neural network (DNN), and an extreme learning machine (ELM) was assessed in several other publications [119]. In paper [120], machine learning allowed the linking of battery temperature with EIS, independently of SoC or SoH, thus lowering the computational load.

Digital twin technologies are used to monitor and estimate battery state [121]. During actual battery operation, the measured data are used by a digital twin. Thus, users can view the battery’s running state in real time through a virtual model. In addition, digital twins are employed to simulate batteries’ operating conditions, which are difficult to reproduce in practice [122]. The results obtained can be used to forecast the battery’s subsequent states and to shape relevant management and control strategies.

6.3. Real-World Data

Large amounts of information on battery parameters are necessary for machine learning algorithms and digital twins to operate efficiently [123]. However, with respect to EVs, additional parameters, such as mileage, can be used to specify the data on battery state [124]. Along with this, the relevant issue here is how to obtain such data.

A number of research papers have used data obtained from the National Monitoring and Management Center of New Electric Vehicles (NMMC-NEV) [125]. These data are collected from 11 million electric vehicles. The data summary comprises current, voltage, battery temperature, speed, mileage, and other parameters. These data allow the building of effective models considering real battery parameter changes during operation.

In addition, cloud services (e.g., Amazon Web Services, Google Cloud, IBM Cloud, or Microsoft Azure) compile data on rides, allowing one to identify a motorist’s driving habits and features, as well as information from the BMS, which is necessary to estimate the accumulators’ aging factors. One combined analysis of these data was focused on forming specific recommendations on driving style and optimizing charging processes [126].

7. Charging and Discharging Management

The management of charging and discharging modes allows one to extend a battery pack’s service life and enhance its operation efficiency, and provides the possibility to forecast the battery pack’s durability. A review of charging and discharging mode management methods is given below. These methods are widely used and hold potential for EVs and other electrical means of transportation.

7.1. Charging and Discharging Management Considering Battery PACK Lifespan

In contemporary battery storage systems, disparities in the physical and chemical properties of individual battery cells, arising during manufacturing, often lead to variations in SoC and depth of discharge (DoD). These variations critically impact the overall performance and lifespan of the battery pack. A dynamic strategy of planning the battery pack charge and discharge processes considering SoC and DoD is used to address and manage these variations effectively. During the charging process, battery packs with a low state of charge are prioritized, while during discharge, priority is given to the ones with a high state of charge. This allows the balancing of the battery pack’s operation and reduces its runout [127]. A sample algorithm of managing battery pack charging and discharging that is applicable to EVs is given in [127].

7.2. Step-by-Step Charging and Discharging with Direct Current

The optimum charging current is determined from the battery pack’s state of charge; next, direct-current charging is carried out step-by-step so that the charge curve is close to the optimum battery pack charge curve. This method allows increasing the charging efficiency and decreasing battery cell polarization and gas emission [128].

Applying this method to charging EV battery packs allows one to significantly increase their service life and optimize the charging process.

7.3. Managing Charging and Discharging in an SMES/Battery Hybrid Energy Storage System

Ref. [129] details a hybrid energy storage system (HESS) comprising superconductive magnetic energy storage (SMES) and a lithium-ion battery pack, which allows one to decrease the number of high-frequency discharge cycles and manage the battery pack’s SoC via an optimized integration circuit, which is based on the approach of power distribution.

7.4. Analysis Results

Table 7. Analysis results of battery charge and discharge management methods.

Method	Advantages	Disadvantages	Cost	Energy Efficiency
Management considering battery life	Reduces depreciation of individual cells; service life extension	Complex management system; high computation load	Medium (requires advanced BMS)	High
Step-by-step direct-current charging	Increases charging process efficiency; reduces polarization and gas evolution; overheating protection	Precise analysis of battery state required; extended time of algorithm implementation	Medium (requires precise current control)	High
SMES/battery hybrid system	Reduces the number of battery charge/discharge cycles; more uniform load distribution; service life extension	Construction and management complexity; high equipment cost; limited applicability outside specialized systems	High (complex technologies and materials)	Extra high

shows comparative analysis results of battery charge/discharge management methods.

As applied to EVs, management options that consider battery service life and step-by-step direct-current charging have become the most popular and widespread. These options are well suited for public, commercial, and personal transport. Hybrid SMES/battery systems are largely used for urban transport.

8. Communication and Data Logging

8.1. Data Storage Methods in BMS

One of the major aspects of BMS operation is effective data transfer and storage management: information on the state of charge, temperature, voltage, current, internal resistance, and other parameters. Basic data storage methods used in modern BMS are given in

Table 8. Results of cell-balancing-method analysis.

Memory Classification		Characteristics
Volatile memory	SRAM	High read and record rate, high cost, and is basically used in low cache memory
	DRAM	Low read and record rate, needs regular updates, and often used in high-capacity RAM
Non-volatile memory	PROM	Memory can be programmed only once and cannot be changed after programming
	EPROM	Electrically erasable programmable read-only memory only for constant use; can perform multiple programming processes
	EEPROM	Multiple programming support and high data-erasing speed
	PCRAM	Non-volatility, high read rate, low static power consumption, and byte addressability
	FLASH (divided into Nor Flash and Nand Flash)	High read and record efficiency, stability, packet writing and deletion support, and is often used for external memory expansion

[130].

The types of memory are distinguished by their ability to operate in the absence of a power supply and by their architecture. Most BMSs are based on microcontrollers containing built-in memory for temporary or persistent data storage: SRAM (Static Random Access Memory) and flash memory.

SRAM is used for operating storage while the system is running. It is quick but volatile, so data is lost upon power shutdown. SRAM is used for the storage of actual parameters, such as cell voltage, temperature, and current. Flash memory is used for persistent data storage, such as calibration coefficients, historical data, and configuration parameters. It is non-volatile, which makes it ideal for data storage, even when the power supply is absent. However, flash memory is limited by the number of write/delete cycles, which must be taken into consideration when designing the system.

External memory is used for systems which require storing large volumes of data (for example, those for analyzing battery pack state and predictive diagnostics).

EEPROM (Electrically Erasable Programmable Read-Only Memory) allows for multiple data re-records and storing this data without a power supply. This memory type is often used for storing settings and statistic data, such as the number of charge/discharge cycles and total battery pack operating time.

SD cards make it possible to store large volumes of data, such as test results or monitoring data, over a long period of time. They are particularly practical in the BMSs used in EVs, where the acquisition of a large volume of information is required for further analysis.

Contemporary BMSs are increasingly integrated with cloud services for remote monitoring and data analysis. Cloud storage allows the storage of data on battery pack operation over a long time period, which is especially important for failure forecasting and the optimization of operation [131,132].

Cloud storage has the following advantages: remote data access from any place in the world; scalability to process a large amount of information; the ability to use big data [133]; and the ability to be integrated with computer-assisted learning to forecast SoH, SoC, and RUL [134]. Meanwhile, cloud storage has a number of limitations: the necessity for a constant uninterruptable internet connection and potential data security problems [135].

Some cutting-edge BMS configurations use distributed ledger technology (or blockchain) to store data on the battery pack state and to improve security when operating energy systems. Blockchain ensures a high security level and permanency of records, which are especially significant for EVs [136]. Blockchain in BMSs allows the storage of historic data on battery pack usage, due to which it is possible to assess the battery pack for secondary use.

8.2. Methods of Data Transfer in BMSs

Reliable communication between system components is necessary for correct BMS operation, which requires various methods of data transfer. Depending on the field of application and operation conditions, both wired and wireless technologies are used. The basic methods of data transfer and protocols providing data exchange are given below.

In the BMSs of EVs, wired interface types are the most common due to their reliability, security, and immunity to electromagnetic interference. The basic types of wire communication channels are CAN, I²C, SPI, and UART [137,138]. CAN (Controller Area Network) is widely used in the BMSs of EVs due to its high noise immunity, low realization cost, and the ability to connect several devices into a single network. The CAN protocol supports a speed of up to 1 Mbit/s, allowing the quick transfer of battery pack state data. I²C (Inter-Integrated Circuit) is a simple economical interface for internal connection between the microcontroller and BMS sensors. However, its application is limited by short distances and relatively low data transfer rates (up to 400 kbit/s). SPI (Serial Peripheral Interface) provides a higher data transfer rate compared to I²C, but requires a larger number of conductors. This interface is usually used to directly connect the controller to the sensors or signal converters. UART (Universal Asynchronous Receiver-Transmitter) is applied for simple interaction points between BMS components. Though it does not require synchronization, its range and speed are limited, which makes it less practical for complex systems.

Ethernet is becoming increasingly popular in modern EVs due to its high broadband capacity (up to 10 Gbit/s) and the ability for complex networked architecture support. Ethernet is used to transfer large amounts of data, such as real-time detailed battery pack state diagnostics, especially in the context of autonomous vehicles. It is compatible with contemporary Internet of Things (IoT) solutions.

While battery packs require more capacity and voltage, it appears necessary to increase the number of cells. In turn, this leads to increasing the number of wires and ports in the system. In spite of wired solutions being effective, they considerably complicate construction, increase weight and cost along with lowering system flexibility. These problems appear particularly acute in limited-space conditions and when scaling is necessary [139].

According to [140], about 8% of electric components’ total cost in EV power units are taken up by the ports. In addition, such systems are more susceptible to physical breakdown incidents, e.g., wire breakages and poor contact of the ports, especially in constant-vibration conditions. These factors make it necessary to implement reliable wireless technologies in modern accumulator system platforms, allowing an increase in their efficiency, durability, and maintainability [141,142,143]. Figure 8 shows the general architecture types of wired BMSs and Wireless Battery Management Systems (wBMSs).

Wired BMSs (Figure 8a) include cell management units (CMUs), each connected to a group of battery cells for their monitoring and control. The Microcontroller Unit (MCU) provides all system control through wire communication channels by interacting with the CMU.

Unlike wired connection, the wBMS architecture (Figure 8b) completely eliminates physical connections, enabling the CMU to exchange data via a wireless network. This approach presents certain advantages, including greater flexibility when packing battery modules, connection process simplification, a reduction in the system weight and volume, and the cutting of maintenance costs [144]. The absence of wire harnesses speeds up battery module disassembly, which is crucial for their reutilization and failed-battery-cell disposal. Decreasing the number of wires minimizes connection failure risks, enhancing the overall system safety compared to traditional wired solutions.

wBMS integration opens up new possibilities for creating intelligent energy systems. In an IoT context, wireless control systems enable the real-time monitoring of battery packs and their remote control via different devices, which not only increases the efficiency of EV battery packs but also extends their service life [145]. Thus, wBMSs are a crucial step forward in creating modern, flexible, and immune energy solutions.

The following wireless technologies are used in wBMSs: Bluetooth Low Energy (BLE), Zigbee, Near-Field Communication (NFC), Wi-Fi, and cellular networks [146,147]. BLE is used to transfer data over short distances: up to 40 m, at a 2 Mbit/s data transfer rate. Zigbee is a low-energy connection technology with MESH network support that is suitable for multi-module BMSs. It provides a stable connection over distances of up to 100 m. Wi-Fi provides a high bandwidth and the possibility to connect to cloud services for remote monitoring and data analysis. However, its high energy consumption makes it less practical for autonomous systems. NFC is used for express settings and small volumes of data exchange over very short distances. It is especially practical for battery module identification or firmware upgrades. 4G/5G LTE cellular communication consumes more energy but is necessary for applications requiring high data transfer rates (up to several Mbit/s) and large distances (several kilometers), such as connection to the cloud for data analysis and remote control.

Despite holding good potential, the spread of wBMS application is limited by a number of problems, including data safety, signal interferences, and normative and standardized barriers, as well as competition from wireline BMSs [148]. Developing wBMSs is aimed at solving these problems. Data safety can be improved via reliable encryption and authentication methods. Distributed ledger technologies (blockchain) present one more solution to enhance information protection from cyber-threats, especially when wBMSs are used together with IoT and cloud technologies. Signal interferences could be minimized by advanced signal processing algorithms, and battery case construction and BMS optimization could be facilitated by shielding from external electromagnetic fields. The absence of standardization calls for cooperation between field representatives and regulatory authorities in order to develop universal standards for wireless BMSs [141].

At the EV scale, considering the required battery pack length of up to 5 m, all the above techniques except NFC provide data transfer. BLE, Wi-Fi, and cellular networks are capable of performing this task at data transfer rates exceeding that of CAN.

Table 9. Analysis results of data transfer technologies.

Technology	Advantages	Disadvantages	Cost	Energy Efficiency/Interference Immunity/Scalability
CAN	Multipoint connection; wide application in the automotive industry	Limited bandwidth (up to 1 Mbit/s)	Low	High/extra high/medium
I2C	Simplicity, economic efficiency, and is a good communication method between sensors inside a module	Short distances, low bandwidth (up to 400 Kbit/s), and the limited number of nodes	Low	Extra high/medium/low
SPI	High data transfer rate	Many wires required, inapplicable to long connections, and no addressing is provided	Medium	Medium/medium/medium
UART	Connection simplicity; asynchronous mode	Speed and coverage range limitations; inapplicable to complex networks	Low	High/low/low
Ethernet	Very high bandwidth (up to 10 Gbit/s); IoT and cloud technology support	High energy consumption; complexity of implementation	High	Low/high/extra high
BLE	Wireless connection within short distances (up to 40 m)	Limited bandwidth (up to 2 Mbit/s)	Low	Extra high/low/medium
Zigbee	Stable connection up to 100 m	Difficult network setup; limited bandwidth (up to 250 Kbit/s)	Low	Extra high/medium/high
Wi-Fi	High bandwidth (up to 600 Mbit/s), access to the cloud, and remote monitoring	High energy consumption; sensitive to interferences	Medium	Low/low/extra high
NFC	Very simple identification of modules	Very small range of coverage (~10 cm); limited volume of data	Low	Extra high/high/low
Cellular networks (4G/5G LTE)	High rate (up to several hundred Mbit/s) and coverage range (up to several kilometers), access to the cloud, and remote control	High energy consumption, Depends on coverage, Expensive traffic	High	Low/medium/extra high

shows comparative analysis results of data transfer technologies.

On the whole, both wired and wireless BMSs have their own advantages and disadvantages. Wired BMSs are characterized by technological maturity, reliability, and lower energy consumption. On the other hand, wireless BMSs have a lower weight, provide flexibility and scalability, and simplify maintenance, making them a promising avenue for future developments. However, problems such as signal interference, data security, and standard issues need to be solved.

8.3. Data Acquisition

In modern EV lithium-ion battery cells, the process of measuring electric parameters such as voltage and current, as well as physical characteristics including ambient temperature and pressure, is called data acquisition. This process is realized by a data acquisition system (DAQ). A DAQ is a complex solution which combines hardware and software to enable the recording, processing, and analysis of data.

A DAQ consists of several key components, each performing a certain function. Sensors are used to measure physical values, e.g., current, temperature, and pressure, converting them into electrical signals. A signal conditioner provides the preprocessing of these signals and prepares them for further conversion. An ADC converts analog signals into digital form, making it possible to further process them using software.

Depending on the priority, various data transfer technologies can be used in DAQs: electric parameters can be transferred continuously via fast and reliable links (CAN, Ethernet), while SoH, SoC, and other parameters can be transferred through slow and wireless data transmission technologies (BLE, Zigbee).

Thus, data acquisition systems are an indispensable part of modern lithium-ion battery management technologies. Their data monitoring, processing, and analysis resources not only allow control of the battery pack state but also prevent potentially hazardous situations, enhancing energy systems’ safety, reliability, and efficiency, including for EVs [8].

8.4. Summarizing the Results

The choice of data transfer method in the BMS of an EV depends on specific requirements, such as the transmission range, data transfer rate, energy consumption, and realization cost. Wired interfaces remain the preferred ones for critical systems where high reliability and safety are necessary [149]. Wireless technologies are becoming increasingly popular. They provide high data transfer rates over sufficient ranges within the electric vehicle’s dimensions with minimum energy consumption, as well as reducing the number of wires, cords, and links in the electric vehicle, which positively impacts its reliability. However, these solutions’ safety requires additional research and development. For example, cyber-attacks carried out through a wireless network might lead to a change in the battery pack parameter data, resulting in system operation failures, overheating, breakdown, and inflammation.

9. Prospective Research Avenues

This section discusses some aspects of future work about next-generation BMSs.

Advanced approaches are associated with the implementation of artificial intelligence and big data technologies, which can significantly improve the safety and efficiency of EV battery packs. The use of these technologies provides the following key capabilities: the comprehensive monitoring and analysis of battery parameters; the adjustment of battery charging strategies; high battery reliability under various operating conditions; the identification of potential battery performance issues; and the generation of predictive maintenance recommendations.

The use of artificial intelligence and big data technologies is not yet mandatory. However, continuous technological advances and cost reductions should lead to the widespread use of intelligent BMSs in the future, including the use of digital twins. Some of the main limitations that slow down the implementation of artificial intelligence methods in BMSs are the high requirements for computing resources and large volumes of transmitted information. A promising direction is to use external servers for storing and processing information. Ensuring high data transfer rates is a key task here.

An extensive review of the technologies for using artificial intelligence and big data in BMSs seems relevant for a more complete understanding of the limitations of these methods and ways to overcome them.

For EVs, an important issue is ensuring safety in the event of the possible deformation of batteries. Even a small deformation of the battery due to a minor accident or damage to the EV can lead to serious consequences. A promising direction is fiber-optic sensors that record changes in the geometric dimensions of the battery pack. It is possible to perform a comprehensive assessment of the battery pack’s condition based on linear dimensions (assessment of the SoC and remaining resources). Fiber-optic sensor systems could be used to obtain information about several battery pack parameters in the future. However, research into the use of fiber-optic sensors is limited to laboratory conditions. It is relevant to study the effectiveness of this method in real EV operating conditions.

Another direction for the development of BMS functions that requires study is the use of online impedance spectroscopy technologies. Impedance spectroscopy methods allow us to determine important battery parameters such as internal resistance. In turn, information about the battery’s internal resistance allows one to estimate the SoC and SoH with high accuracy. However, the implementation of online impedance spectroscopy methods in EVs is limited by the need for an additional source in the BMS.

10. Conclusions

The BMS is one of the key elements of an EV battery pack. The safety and durability of battery pack operation depend on the BMS’s functional capabilities and efficiency. When choosing the circuitry configurations and algorithms for constructing a BMS with the required functional capabilities and characteristics, it is necessary to determine the optimum combinations of efficiency, degree of technical complexity, overall dimensions, computational resources, and cost.

In this article, we considered the circuitry configurations and algorithms of BMSs when applied to providing such functions as battery cell parameter monitoring, battery cell protection, cell balancing, state estimation and fault diagnosis, and charging and discharging management. A review and comparative analysis were carried out concerning already-applied and prospective solutions of each avenue.

The paper forms an idea of both the implementation of BMS hardware and the software implementation of battery pack control algorithms. The results of the review will be interest to developers when solving problems of choosing how to implement the main BMS functions. Based on the key criteria, comparative tables were formed that can be useful when choosing BMS components.

The most promising technologies, the development of which will significantly increase the efficiency and reliability of EV battery packs, were identified. Such technologies primarily include artificial intelligence and big data. The implementation of these technologies will stimulate the transition to adaptive battery management and predictive maintenance. At the same time, the use of fiber-optic sensors and impedance spectroscopy technology can also be considered as promising areas for the development of BMSs.