A Comprehensive Review of Equalization Techniques for Reconfigured Second-Life Battery Systems

Abstract

As the demand for second-life lithium-ion battery applications continues to grow, efficient cell equalization has become essential to mitigate parameter inconsistencies and extend system longevity. Owing to their diverse origins and varying aging paths, second-life batteries exhibit significant parameter dispersion, which poses distinct challenges. In light of these issues, this paper presents a comprehensive review of passive, active, and dynamic equalization technologies. It analyzes the circuit topologies and control strategies associated with each method, with a particular focus on their applicability to second-life battery systems. Furthermore, emerging trends toward intelligent, modular, and adaptive equalization are discussed.

1. Introduction

Lithium-ion (Li-ion) batteries have become the dominant energy storage solution in electric vehicles (EVs), stationary energy storage systems (ESSs), and portable electronic devices, owing to their high energy density, long cycle life, and environmental compatibility [1]. However, as their deployment scales up globally, the challenges associated with long-term use—particularly performance degradation and capacity fade—are becoming increasingly prominent [2]. In China alone, it is estimated that the cumulative volume of retired power batteries will exceed 380 GWh by 2030 [3]. Notably, most retired Li-ion batteries still retain approximately 70–80% of their original capacity [4], which makes them highly suitable for secondary applications. Repurposing these end-of-life batteries presents a promising strategy to alleviate the growing demand for new batteries while simultaneously mitigating the environmental and resource pressures associated with battery recycling.

Compared to direct recycling, reusing retired Li-ion batteries reduces resource waste and carbon emissions, eases raw material supply pressures, and improves economic returns. These second-life batteries are increasingly applied in low-power, less demanding scenarios such as residential energy storage, grid peak shaving, telecom backup systems, and low-speed electric vehicles [5]. However, the inconsistency of second-life batteries in key parameters such as capacity, internal resistance, state of charge, and thermal behavior is considerably greater than that of newly manufactured batteries. New batteries undergo stringent quality control processes during production to ensure uniform performance, thereby minimizing parameter variation from the outset. In contrast, second-life batteries are sourced from a wide range of origins. They may come from different electric vehicle manufacturers, be used by consumers with diverse charging habits, or operate under varying thermal and usage conditions. These differences result in highly heterogeneous aging patterns [6]. Additionally, variations in manufacturing batches and original cell models further amplify discrepancies in critical parameters. Such significant inconsistencies reduce system efficiency and lifespan and may cause safety issues such as localized overcharge or over-discharge [7], posing a major challenge to efficient reuse.

To improve the efficient reuse of retired Li-ion batteries, a widely adopted technical route involves a “screening–classification–reassembly” process. In this approach, battery cells or modules with acceptable performance are reorganized into new systems tailored to specific application scenarios [8,9,10]. This process requires not only matching electrical characteristics but also ensuring the controllability and maintainability of the reassembled system. As the design of such systems advances, configurations are evolving from simple series-parallel connections to more structured architectures with enhanced management capabilities. However, cell inconsistency does not disappear after reassembly; instead, it continues to evolve during long-term operation. This ongoing divergence in parameters such as voltage and current response can reduce system reliability and energy utilization efficiency [11].

To mitigate these effects, it is essential to implement effective equalization management strategies. By leveraging appropriate topological designs and control methods, energy flow can be dynamically regulated and redistributed, enabling adaptive operation and improved performance at the system level [12]. This study investigates the reassembly of retired Li-ion batteries, with an emphasis on equalization management during system integration. It provides a systematic review of representative topologies and control strategies across three major categories of equalization technologies: passive, active, and dynamic. Their applicability, technical advantages, and associated challenges in second-life battery systems are critically evaluated. The insights presented herein offer both theoretical foundations and practical guidance to support the large-scale deployment of retired batteries, promoting the development of safer and more efficient energy storage systems.

2. Battery Reconfiguration and Equalization Management

2.1. Battery Reconfiguration System

A battery reconfiguration system is formed by retired cells that have been screened and regrouped according to defined performance criteria. It typically consists of three key components: the cell layer, the electrical interconnection layer, and the battery management system (BMS), as illustrated in Figure 1.

The cell layer and the electrical interconnection layer together form the hardware topology, which defines the internal energy transfer pathways and basic electrical characteristics of the system. The cell layer comprises all screened, reusable retired cells and serves as the system’s primary energy storage component. The interconnection layer arranges these cells through series, parallel, or more complex combinations to meet the target voltage and capacity requirements. In practical applications, considering the risk of circulating currents, series configurations are generally preferred for reassembled systems. Due to the constraints of engineering simplicity and cost control, most existing systems still rely on fixed connection structures [13]. As the control hub of the entire system, the BMS is responsible for critical functions including state monitoring, charge/discharge control, and communication with external devices [14]. Notably, even after screening, second-life cells still exhibit significant differences in capacity, internal resistance, and aging trajectories. Cells with a history of frequent fast charging or low-temperature operation show severe capacity fading and substantial internal resistance growth, while lightly used cells maintain performance close to that of new cells. This heterogeneity presents prominent challenges to the estimation of state-of-charge (SOC) and state-of-health (SOH). The estimation of SOC and SOH in second-life battery packs faces significant challenges. On one hand, cells come from diverse sources with distinct service histories and performance variations, leading to high dispersion in initial SOC values and SOH baselines. On the other hand, the non-uniform aging of cells brings complex nonlinear behaviors such as voltage hysteresis and nonlinear internal resistance growth, making a single estimation method unable to accurately capture their true states.

Given the diverse sources of retired Li-ion batteries and the complex, non-uniform nature of their aging processes, future reconfiguration systems must possess enhanced state awareness and adaptability to accommodate the ongoing evolution of cell performance. Developing advanced system architectures with capabilities for dynamic monitoring, structural flexibility, and operational adjustment will be essential for improving overall performance and ensuring long-term stability.

2.2. Equalization Management Technology

In conventional battery systems, large numbers of cells are typically connected in series to form battery strings. Due to the barrel effect inherent in series configurations, the usable capacity of the entire string is constrained by the weakest cell, i.e., the one with the lowest capacity, as illustrated in Figure 2. This structural limitation not only reduces capacity utilization but also introduces risks like localized overcharging or over-discharging.

To address this issue, equalization management has been introduced. By regulating energy flow among individual cells, these mechanisms improve system coordination, thereby enhancing usable capacity and extending cycle life. Based on the method of energy transfer and circuit implementation, equalization strategies can generally be categorized into three types: passive, active, and dynamic, as summarized in

Table 1. Comparison of three types of battery equalization technologies.

Category	Passive Equalization	Active Equalization	Dynamic Equalization
Energy Regulation	Excess energy dissipated as heat	Energy transferred from high-charge to low-charge cells	Cell charge/discharge controlled via topology switching
Energy Efficiency	Low	Medium to high	High
Circuit Complexity	Low	Medium (requires inductors, capacitors, or DC-DC circuits)	High (requires switch matrix and state monitoring)
Control Complexity	Low	Medium (real-time voltage/SOC feedback and modulation)	High (state estimation and topology reconfiguration needed)
Suitable Scenarios	Low-cost, small systems with minimal efficiency demand	EVs, portable ESS, medium-scale systems	Large ESS, reconfigured systems with high inconsistency
Key Challenges	High energy loss, thermal management burden	Costly, accuracy depends on state estimation	High complexity in communication, computation, and coordination

. Passive equalization works by dissipating excess energy from higher-capacity cells as heat, offering simplicity but lower efficiency [15]. Active equalization enables direct energy transfer from higher-charge to lower-charge cells, improving energy efficiency but requiring more complex circuitry [16]. Dynamic equalization, by contrast, adjusts the charge and discharge behavior of individual cells in real time to maintain balance, offering high adaptability at the cost of increased control complexity [17]. It is important to note that, unlike active equalization, which relies on a fixed series connection between cells, dynamic equalization allows for real-time reconfiguration of the cell interconnection topology. While active equalization transfers energy directly between cells with differing charge levels (for instance, from high-voltage to low-voltage cells), dynamic equalization enables energy management during power delivery to external loads. In dynamic equalization, high-voltage cells may discharge more, while low-voltage cells discharge less or are bypassed, adjusting the balance without necessarily transferring energy directly between cells. This flexibility allows for more adaptive and efficient energy distribution, optimizing the system’s performance based on real-time conditions.

These three approaches reflect different trade-offs between control sophistication, energy efficiency, and system flexibility, and represent key technical pathways for meeting the diverse demands of second-life battery reconfiguration. The following sections will provide a detailed analysis and comparison of their respective topological structures and control strategies.

3. Battery Reconfiguration System with Passive Equalization

3.1. Topology Structure

In battery systems equipped with passive equalization circuits (PEC), resistors are connected in parallel with individual cells to dissipate excess energy and align cell voltages. The typical topologies for such systems are categorized into fixed shunt and switched shunt configurations [18], as illustrated in Figure 3.

Figure 3a,b show the circuit structure of a fixed shunt equalization system, where voltage balancing is achieved by adjusting the resistance values across each cell. However, because resistor values cannot be modified in real time and the discharge process is uncontrollable, this topology is rarely used in modern battery systems [3]. The switched shunt equalization circuit, shown in Figure 3b, improves on this design by adding a controllable switch in series with each shunt resistor. During operation, only the switch corresponding to the target cell is activated, while others remain off. This configuration is structurally simple, allows flexible selection of target cells for equalization, and supports uninterrupted system operation, making it widely adopted in early industrial applications [19].

Despite advantages such as ease of implementation and low cost, passive equalization suffers from significant energy loss, as all excess energy is dissipated as heat and cannot be recovered, resulting in low energy efficiency [20]. Additionally, to avoid overheating the shunt resistors, the balancing current is typically limited to tens or hundreds of milliamps, much lower than the charge/discharge currents that may reach several amps. This leads to slow voltage equalization among cells within the pack [21]. Given the pronounced inconsistency among retired Li-ion cells, passive equalization alone is generally inadequate for battery reconfiguration systems composed of large quantities of second-life cells.

3.2. Control Strategy

In passive equalization schemes, fixed shunt circuits are limited by their continuous energy dissipation and lack of control flexibility, making real-time operation infeasible. Consequently, switched shunt structures with controllable switches are more commonly used in practice. The control logic for such systems is illustrated in Figure 4 [22].

The process begins with the system acquiring voltage data from all individual cells. When a cell’s voltage exceeds a predefined equalization threshold, the controller closes the corresponding switch, enabling current to flow through the parallel resistor and dissipate the excess energy. This continues until the voltages of all cells converge or meet a specified balancing criterion. Control strategies can be implemented either centrally—activating all relevant switches simultaneously—or in a distributed manner, where each switch is evaluated and operated independently. Both approaches offer basic real-time equalization capabilities during system operation.

4. Battery Reconfiguration System with Active Equalization

4.1. Topology Structure

Active equalization systems employ an energy transfer element to rapidly redistribute energy among individual cells, avoiding the resistive loss inherent in passive methods. As a result, they typically deliver higher energy-utilization efficiency and faster balancing. According to the type of transfer element, common topologies are grouped into inductor-, capacitor-, transformer-, and power-electronic-converter-based designs [23,24], as shown in Figure 5.

4.1.1. Inductor-Based Active Equalization System

Inductor-based active equalization circuits (AECs) leverage the non-instantaneous nature of the inductor current to control energy flow between cells according to their voltage differences, enabling energy transfer in the form of magnetic fields [16]. This method maintains relatively high equalization efficiency even when voltage differences are small. Depending on the number of inductors used, such topologies are generally classified into single and multiple switched inductor (SI) circuits.

As shown in Figure 6a, the single SI-based AEC connects multiple series cells through a switching array that shares a single inductor [25]. This configuration is simple, compact, and easy to integrate due to its minimal component count. However, its equalization speed is limited by the single magnetic flux path, resulting in longer balancing times, especially in large-scale battery systems.

In contrast, the multiple SI-based topology places an inductor between each pair of adjacent cells, forming closed energy transfer loops through independently controlled switches, as illustrated in Figure 6b [26]. While this structure enables localized balancing between neighboring cells, it cannot efficiently transfer energy between distant cells. Longer transfer paths reduce equalization efficiency when cell pairs are far apart.

To address this issue, a parallel equalization architecture based on SIs has been proposed [27,28,29], as shown in Figure 6c. By increasing the number of equalization layers, this design maintains a high balancing speed even for long-distance energy transfers. Such improved multi-inductor topologies offer good scalability and are well suited for modular design. However, as the number of cells increases, the total number of components grows rapidly, significantly increasing circuit complexity and hardware cost.

4.1.2. Capacitor-Based Active Equalization System

Capacitor-based AECs utilize the property that the voltage across a capacitor cannot change instantaneously. By controlling the switching states of connecting devices, cells with different voltages are sequentially linked with the capacitor to form a closed loop for energy transfer [6]. Depending on the number and configuration of capacitors and switches, these circuits can be categorized into several types, including single switched capacitor (SC), single-layer SC, configurable SC, and resonant SC topologies.

The single SC equalization circuit represents the simplest configuration, as shown in Figure 7a [30]. It employs only one capacitor as the energy transfer element, allowing energy balancing between just two cells at a time. This leads to relatively slow balancing speed when applied to large battery strings. To improve performance, the single-layer SC topology introduces multiple capacitors, with each pair of adjacent cells connected via a switch-controlled capacitor, as illustrated in Figure 7b. This design simplifies control and improves balancing speed compared to the single SC-based structure. However, its efficiency is still affected by the spatial position of cells within the string.

Building upon the single-layer structure, researchers have proposed a range of advanced SC-based configurations by modifying the geometric layout of capacitors. These include chain, series-parallel, dual-layer, delta, star, and mesh topologies. As shown in Figure 7c, the chain structure adds extra capacitors between the cells at both ends of the battery string to reduce the long energy transfer paths and accelerate edge-cell balancing [31]. Nevertheless, as the number of cells increases, balancing speed remains limited and efficiency continues to decline.

To address this, a series-parallel SC equalizer has been proposed in ref. [32], enabling direct energy transfer from high-voltage to low-voltage cells by alternating between series and parallel capacitor configurations. This structure, as depicted in Figure 7d, enhances balancing speed in long strings. However, in such topologies, the newly introduced switching devices must withstand the full string voltage, increasing voltage stress and component reliability requirements.

Figure 7e presents a dual-layer SC topology, where an additional capacitor is placed between every two balancing capacitors. This design reduces voltage stress on the switches and shortens the balancing time by nearly fourfold compared to the single-layer counterpart [33].

To further improve energy routing flexibility, several novel SC-based AEC topologies have been developed, including delta, star, and mesh structures, as shown in Figure 7f–h. Among them, the delta-structured topology enables efficient energy transfer between any two cells without significantly increasing circuit or control complexity. In four-cell systems, its balancing efficiency has been reported to reach 94.5% [34]. The star-structured topology establishes direct balancing paths between any two cells while reducing the number of required switches and capacitors [35]. However, since all current paths converge at the central node, the electrical and thermal stress is concentrated there, making it more suitable for small- to medium-scale systems requiring fast global balancing. Mesh-structured SC circuits offer a richer set of energy transfer paths, allowing the selection of optimal routes based on specific imbalance states [36,37]. To further enhance performance, a bridged mesh structure has been proposed in ref. [38], where each cell is paralleled with an H-bridge capacitor converter. This design creates multiple optimal paths and improves balancing efficiency beyond the standard mesh configuration, albeit at the cost of increased redundancy, complexity, and system cost. It is best suited for applications with stringent fault tolerance requirements. In addition, modular SC-based topologies have been introduced in refs. [39,40], using standardized circuit modules for clear structure and scalability. This makes them particularly well suited for large-scale battery systems [41].

In high-frequency equalization applications, frequent switching of power devices leads to significant switching losses. To address this issue, LC series resonant circuits have been employed as the energy transfer medium to enable zero-current switching (ZCS), thereby reducing equalization power consumption [42,43,44]. A representative resonant-converter equalizer in ref. [45] employs a series LC path to shuttle energy between the most imbalanced cells, achieving fast equalization with simple magnetics. Building on this approach, resonant-based SI and SC equalization circuits have been proposed in refs. [46,47]. The resonant SC circuit introduced in ref. [46] can ensure zero voltage difference between the target cells and achieve both zero-current turn-on and turn-off of the switching devices, thus improving energy efficiency and reducing electromagnetic stress. To minimize the number of resonant components, an LC-L-based SC topology has been proposed in ref. [48] that directly transfers energy between the most imbalanced cells. An additional inductor buffer loop is incorporated to suppress current spikes and enhance equalization stability. Moreover, the accuracy of voltage measurements can be affected by cell impedance variation and polarization effects, which in turn degrades equalization performance. To mitigate this, a direct cell-to-cell equalization scheme based on a switched-matrix single-capacitor converter has been proposed [49]. This design integrates additional current sensors and an optimal pairing algorithm, allowing it to maintain stable performance and high energy efficiency under varying initial conditions.

4.1.3. Transformer-Based Active Equalization System

Transformer-based AECs employ magnetic coupling to transfer energy between cells via transformer windings, providing galvanic isolation and thus offering higher safety and electromagnetic interference immunity compared to non-isolated SI-based or SC-based methods. Based on winding configurations, transformer-based AECs can be broadly classified into single-winding and multi-winding topologies.

The single-winding transformer-based topology features a central winding connected to each cell through dedicated switches, as illustrated in Figure 8a [50,51]. This structure is relatively simple; however, energy transfer must be scheduled sequentially among cells, resulting in limited balancing speed due to the single transfer path.

In contrast, as shown in Figure 8b, multi-winding transformer-based AECs utilize multiple secondary windings wound around a common magnetic core, with each winding corresponding to an individual cell [52]. This configuration enables direct energy transfer between any pair of cells via the shared magnetic core, providing high energy transfer efficiency. Depending on the control scheme, multi-winding transformers operate in either forward or flyback mode. A hybrid approach presented in ref. [53] divides the windings into two groups with opposite polarities, enabling forward-mode conversion within each group and flyback mode conversion across groups, thereby achieving automatic magnetic core reset. Further advancements include the design of a half-bridge multiport transformer equalizer [54,55], which incorporates both active and passive energy paths and uses boundary-voltage-based control strategies to enhance balancing speed [54]. A planar transformer design was demonstrated to support equalization of up to 24 cells, achieving efficiency exceeding 97% [55]. In pursuit of low-cost solutions for second-life battery reuse, dual-layer equalization architectures have been proposed in refs. [56,57,58] which reuse existing DC-DC circuitry and multi-winding transformers. This approach adds only one diode and one inductor per cell, eliminating the need for extra switching devices and significantly reducing system cost. To address the modularization demands of long battery strings, a phase-shift-modulated modular equalizer has been introduced in ref. [59], combining LC resonance with a multi-winding transformer. Energy is transferred directionally between modules via phase-shift control, and the design supports both soft switching and volume-efficient resonant operation. In addition, related integrated implementations based on parallel transformers further improve scalability and reduce the switch count for high-voltage strings [60].

Although transformer-based AECs enable high-current, high-speed balancing, their application is constrained by the volume and weight of magnetic components such as coils and cores. As such, they are best suited for large-scale battery systems where installation space is sufficient.

4.1.4. Power Electronic Converter-Based Active Equalization System

In these active equalization systems, DC-DC power electronic converters are employed as intermediate modules to transfer energy between cells, leveraging their ability to flexibly regulate input and output voltages. Based on the topology and operational characteristics of the converters, converter-based equalization schemes can be categorized into two main types: non-centralized and centralized configurations. Among them, non-centralized schemes are further divided into adjacent cell-to-cell, string-to-cell, and energy bus-based architectures. Representative topologies are illustrated in Figure 9.

Among these, the adjacent cell-to-cell scheme in Figure 9a is the most widely adopted. It typically utilizes non-isolated DC-DC circuits such as Buck, Boost, Buck-Boost, and Cuk converters [61,62]. Buck and Boost converters feature simple structures but support only unidirectional energy flow, making them unsuitable for bidirectional equalization. Buck-Boost converters offer bidirectional energy transfer capability and improved flexibility, but since energy must be temporarily stored before release, their efficiency is generally lower than that of single-stage Buck or Boost converters. The Cuk converter, an extension of the Buck-Boost design, introduces a coupling capacitor to achieve continuous input and output currents with reduced ripple. However, its structure with two inductors and one capacitor leads to higher energy losses. When applied to large-scale battery systems, traditional non-isolated converter-based AECs require a large number of components, resulting in increased structural complexity and system cost. These limitations highlight the need for further optimization in high-density reconfiguration scenarios.

To reduce component count and improve integration, string-to-cell equalization schemes enhance traditional adjacent cell-to-cell architectures by sharing certain circuit elements [63,64], as shown in Figure 9b. A modified Cuk converter proposed in ref. [61] replaces discrete inductors with a coupled inductor, reducing the number of switches by approximately 50%. However, energy transfer is restricted to adjacent cells, resulting in slow balancing. Building on this, a hybrid structure in ref. [65] combines Buck-Boost and Cuk converters, further minimizing component usage. In the domain of multilevel converters, half-bridge and resonant converters have been integrated with voltage or current multipliers, eliminating the need for multi-winding transformers and offering high flexibility in adapting to battery systems with varying capacities and voltage levels [66,67,68,69].

Energy bus-based equalization schemes in Figure 9c enable energy redistribution among cells via a shared bus, offering strong modular scalability and efficient balancing performance [70,71,72]. For example, a ZCS bidirectional Cuk equalizer has been developed in ref. [73], in which a resonant tank enables zero-current switching for all semiconductor devices, significantly reducing high-frequency switching losses. However, the introduction of the resonant network increases both component count and circuit complexity. In a further optimization, a capacitor-coupled ZETA-derived topology has been proposed in ref. [74]. The front-end DC/AC stage generates an AC bus, while the back-end diode rectifier network autonomously distributes current based on cell voltage differences. This approach significantly reduces control complexity. However, since the number of converters scales linearly with the number of series-connected cells, system cost increases considerably for large-scale battery configurations.

Given that the cost, volume, and complexity of non-centralized equalization schemes increase rapidly with the number of cells, their scalability and practical reliability are limited. In contrast, centralized equalization architectures, characterized by structural simplicity, concentrated power handling, and flexible control, are better suited to multi-cell battery systems. The typical structure is depicted in Figure 9d. Centralized DC-DC equalization systems typically consist of a bidirectional converter and a switch matrix, with overall balancing performance primarily determined by the converter design [75,76]. In ref. [77], an isolated bidirectional DC-DC converter has been combined with quasi-resonant zero-voltage switching (ZVS), which minimizes hard-switching losses while ensuring high-frequency operation with excellent conversion efficiency and high-voltage-side isolation. Since excessive current ripple can elevate the surface temperature of battery cells, adversely affecting both cycle life and overall system efficiency, it is essential for converters to maintain a stable and constant equalization current. Addressing current quality concerns, the equalization converter with a ripple cancelation scheme has been proposed in ref. [78], which theoretically achieves zero-ripple equalization current while incorporating both ZVS and high voltage gain capabilities. Although this topology increases circuit complexity and component count, it represents a comprehensive and high-performance solution in centralized converter-based equalization systems.

4.2. Control Strategy

Compared to passive equalization, which relies solely on energy dissipation, active equalization technology redistributes charge intentionally, thereby achieving intra-pack balancing with greater efficiency and adaptability. The control method for such systems must be tailored to the specific circuit topology employed. In general, active equalization control strategies depend on high-precision state measurement and estimation. They offer significant advantages in adaptability and energy efficiency but come with increased complexity in both control algorithms and hardware implementation. As a result, careful trade-offs are required among cost, reliability, and performance. The following sections analyze active equalization control strategies from two key perspectives: control variables and control algorithms, as illustrated in Figure 10.

4.2.1. Control Variable

Control variables carry essential information about cell status, and their accuracy directly affects equalization effectiveness and system performance. Active equalization strategies are commonly categorized based on the selected control variable, including voltage-based, SOC-based, and capacity-based approaches [79].

Voltage-based control is the most widely used method. The controller monitors terminal voltages of all cells and initiates energy transfer when a cell’s voltage exceeds a preset threshold or deviates significantly from others [80]. However, due to internal resistance differences, terminal voltage does not always reflect actual energy state. In systems with large resistance variation, this may lead to significant control errors and limit available capacity utilization [81].

SOC-based control offers a balance between accuracy and responsiveness. It provides a more reliable measure of residual energy and supports efficient energy scheduling. When a cell’s SOC is significantly higher than the average, the controller adjusts the operation of the equalization circuit to transfer energy to lower-SOC cells or to a shared bus, repeating the process until SOC values converge or meet the desired threshold [82]. Compared with voltage-based methods, SOC control reduces the risk of misjudgment caused by resistance mismatch or transient voltage fluctuations, enabling better energy utilization [83]. However, its performance depends heavily on the accuracy of SOC estimation, which may be affected by temperature, aging, and load variations. High-precision estimation models and correction mechanisms are often required in practice. This strategy offers a strong balance of precision, complexity, and adaptability, making it well suited for applications such as electric vehicles and grid storage systems.

Capacity-based control introduces actual usable capacity information to further improve control accuracy and system efficiency. Instead of relying solely on voltage or SOC, this method estimates each cell’s available capacity and compares it to its rated value to evaluate cell health and prioritize equalization. The controller allocates current based on capacity differences, allowing higher-capacity cells to discharge more and reducing the charge burden on lower-capacity ones, minimizing risks of overcharging or over-discharging. As reported in refs. [84,85], this approach is especially effective in addressing capacity degradation over long-term operation and is suitable for large-scale storage systems with noticeable cell-to-cell capacity variation. However, it relies on precise capacity estimation algorithms, increasing system complexity and computational load, and is therefore more appropriate for high-end applications where energy efficiency and service life are critical.

4.2.2. Control Algorithm

Control algorithms for active equalization can be broadly categorized into autonomous and governed types. Autonomous algorithms rely on local decision-making without centralized coordination. They exploit the inherent properties of circuit topology to enable self-regulated energy transfer [86], typically based on voltage differences or other simple sensor inputs. This approach often uses hardware self-triggered mechanisms to perform the balancing task. In contrast, governed algorithms depend on external controllers to monitor battery states, make decisions, and dynamically manage power flow [87,88]. These algorithms typically consider global information, such as SOC or capacity, to perform more refined dynamic adjustments, achieving more precise equalization behavior.

Autonomous algorithms have been applied across inductor-, capacitor-, and transformer-based systems. For example, in SI-based topologies, complementary PWM signals with fixed duty cycles have been employed to drive energy transfer [27]. Voltage differences between adjacent cells naturally guide current from high- to low-voltage cells, while synchronous rectification reduces conduction loss. The current amplitude is self-adjusted based on voltage mismatch and parasitic parameters. In SC-based equalization systems [35,36,38,39,89], two complementary fixed-frequency PWM signals have been used to control different capacitor configurations. In star-structured topologies, one PWM signal controls all odd-numbered switches, and the other controls even-numbered switches, achieving bidirectional energy transfer via alternating conduction [35]. For transformer-based systems [52], a single PWM signal synchronously drives all switches in a symmetrical multi-winding structure. Voltage differences across windings naturally facilitate energy transfer from high- to low-voltage cells. Autonomous algorithms feature simple control logic and ease of implementation, as they do not require a central controller or complex algorithms. This makes them well-suited for low-cost applications where balancing precision is not critical. However, the reliance on voltage differences or basic sensor inputs limits their effectiveness in more complex systems, where factors such as SOC or capacity variation need to be considered. As a result, while they are ideal for simple battery management systems in devices like power tools or low-power storage systems, they may not be suitable for more demanding applications such as electric vehicles or large-scale energy storage systems, where more precise and adaptable balancing is required.

Governed algorithms are primarily applied in DC-DC converter-based active equalization systems. Their core mechanism involves using a central processor to continuously acquire cell voltage data and estimate each cell’s SOC. The deviation between individual SOC values and the average SOC of the pack serves as the basis for equalization decisions [54,59]. For example, in a quasi-resonant ZVS isolated converter system, when a cell exceeds the threshold for SOC deviation, the controller adjusts the converter’s operating mode based on the deviation’s direction and regulates the duty cycle and switching sequence to maintain a stable target equalization current [77]. In the ECRC converter design, a two-degree-of-freedom phase-shift PWM scheme is adopted; the inner loop determines the duty cycle based on voltage gain requirements, while the outer loop fine-tunes power transfer through continuous phase-shift adjustment, enabling precise current control [78]. Overall, governed algorithms integrate SOC-based closed-loop control, priority scheduling, current-level regulation, and soft-switching techniques to establish a safe, efficient, and intelligent energy redistribution framework.

5. Battery Reconfiguration System with Dynamic Equalization

In both passive and active equalization schemes, hardwired configurations are commonly used, with balancing circuits added on top of fixed battery strings. A fundamental limitation of this approach is that the scale of the equalization circuit grows linearly with the number of cells. Under constraints of efficiency and cost, these circuits often suffer from limited current-handling capacity [23]. As cell capacity and power ratings continue to increase, and inconsistency among cells becomes more dynamic during operation, traditional equalization mechanisms struggle to meet the demand for faster balancing, leading to reduced usable capacity and declining system economics.

To address these limitations, reconfigurable battery systems (RBS)—based on low-voltage power electronic switches—have emerged as a promising solution. RBS dynamically alters the electrical interconnections (series/parallel) among cells through a switching network, enabling voltage restructuring and current redistribution at the pack level. With proper control algorithms, the system can regulate charge/discharge behavior across cells, balance SOC and temperature distribution, and significantly extend runtime and cycle life [90,91]. Faulty cells can be isolated at the module level in real time, enhancing fault tolerance without interrupting system operation [92]. RBS also offers high compatibility, accommodating cells with different aging levels, models, or even chemistries. Given the heterogeneous origins and aging paths of second-life batteries, this reconfigurable architecture presents a highly viable approach for their efficient reuse.

5.1. Topology Structure

In general, the basic building block of an RBS is the energy storage unit (ESU), which can be either a single cell or a module. However, using cell-level ESUs provides the greatest flexibility in system configuration and maximizes energy utilization [93]. Based on the ESU connection approach, RBS architectures have evolved along two main technical paths: switch array-based and multilevel converter-based topologies, as shown in Figure 11.

5.1.1. Switch Array-Based Dynamic Equalization System

In switch array-based RBS, each cell is equipped with multiple controllable switches. By coordinating the on/off states of these switches in real time, various operating modes can be achieved. As these circuits typically lack additional energy storage components, they do not offer voltage regulation and generally operate on a control timescale of several seconds [94]. A representative switch array-based RBS topology is illustrated in Figure 12.

In the single-switch RBS topology, each cell is managed individually through a dedicated switch, enabling flexible interconnection to scale system capacity, as shown in Figure 12a. A bypass switch is placed across each parallel module and activates only when all intra-module cell switches are turned off, providing a redundant protection mechanism [94]. This structure maintains fine-grained control while enhancing reliability through modular bypass design and is also referred to as a digital energy exchange system [95].

The dual-switch RBS topology, shown in Figure 12b, adopts a half-bridge configuration per ESU and is also known as a cascaded half-bridge RBS [96]. Two complementary switches control the transition between active and bypass states. A high-power reconfigurable storage system based on this design demonstrated broader output voltage range and higher efficiency compared to traditional hardwired packs combined with DC/DC modules [97].

Building on these basic forms, researchers have proposed multi-switch RBS topologies by increasing the number of switches and optimizing connection schemes. Figure 12c–e illustrate three-switch topologies, which equip each cell with three switches to support more advanced functions. For example, the configuration in Figure 12d enables not only series-parallel reconfiguration but also precise fault isolation, effectively preventing failure propagation [98,99,100].

Higher-order structures such as the four-switch RBS in Figure 12f and five-switch RBS in Figure 12g [92,101] allow more complex interconnection patterns and intelligent control capabilities. However, these come at the cost of increased system complexity and hardware expense. As such, three-switch and higher configurations are generally reserved for specialized applications and are rarely used in large-scale battery systems. In this context, a recent modular multi-switch RBS, proposed in ref. [102], realizes flexible series–parallel reconfiguration and fast balancing without additional equalizers, offering a practical solution for module-level reconfiguration.

5.1.2. Multilevel Converter-Based Dynamic Equalization System

Compared to switch array-based RBS, multilevel converter-based RBS is derived from high-voltage cascaded energy storage system topologies. It introduces new voltage-level capabilities while enabling cell-level reconfiguration without the need for dedicated equalization circuits. Based on output characteristics, multilevel RBS architectures are categorized into AC-type and DC-type systems. It is important to note that the distinction between AC and DC is not directly related to the equalization mechanism itself, but rather refers to the overall system output type. This is because the converter functions both as the interface between the battery system and the external grid or DC bus and as the equalization circuit.

AC-type RBS directly delivers AC output, eliminating the centralized DC-AC conversion stage required in switch array-based systems. This simplifies the system and improves overall efficiency. The two most common AC-type RBS architectures are based on cascaded H-bridge (CHB) and modular multilevel converter (MMC) designs. The CHB-RBS topology [103] integrates power conversion and balancing functions within each full-bridge submodule. Coordinated control strategies allow simultaneous cell switching and dynamic energy balancing. In MMC-RBS systems, each submodule adopts a half-bridge configuration [104]. Although each MMC submodule requires fewer switches than CHB, the dual-arm structure offsets this advantage, making the overall device count comparable between the two.

DC-type RBS utilizes cascaded half-bridge submodules to form a modular DC-DC converter architecture with high system efficiency [105]. Unlike AC-type systems, it avoids low-frequency ripple current on the battery side caused by unbalanced three-phase AC, thereby enhancing battery stability and lifespan. Moreover, for a given voltage level, DC-type architectures require fewer switching devices, which helps reduce cost and control complexity. However, when interfacing with the AC grid, an additional DC-AC converter is required to support energy conversion and grid connection.

5.2. Control Strategy

Dynamic equalization systems offer flexible cell-level access control, enabling localized isolation of faulty cells to ensure uninterrupted system operation. At the same time, their scheduling mechanisms support charge state balancing across cells. To achieve these objectives, the control strategy must coordinate real-time adjustments of cell status, load demands, and topological reconfiguration.

5.2.1. Switch Array-Based Dynamic Equalization System

In ref. [17], a control strategy integrates state monitoring with model predictive control. It first determines the optimal output voltage and current based on cell-level voltage, current, and SOC data, along with external power demand. The required number of series-connected cells is then calculated, excluding faulty units to form a healthy operating path. By prioritizing discharge from high-SOC cells or charging of low-SOC cells, intra-pack voltage balance is achieved. This process is iteratively executed at fixed intervals, improving output stability, fault tolerance, and energy efficiency.

Building on this, a control strategy introduced in ref. [106] is focused on global path optimization. Using a dynamic programming algorithm, the controller selects the optimal configuration in each cycle from all feasible topologies, ensuring compliance with voltage-current constraints while minimizing SOC consumption. This enhances system capacity utilization and operational runtime.

However, control strategies for switch array-based RBS largely depend on real-time state monitoring and topological reconfiguration. While they offer adaptability and fault tolerance, their computational complexity scales rapidly with system size, posing challenges for real-time control. Moreover, as these methods rely on open-loop control, output voltage regulation remains weak, often requiring additional DC-DC or DC-AC stages for voltage and current stabilization.

5.2.2. Multilevel Converter-Based Dynamic Equalization System

For AC-type multilevel RBS, distinct SOC-based dynamic balancing strategies have been proposed in refs. [107,108] from the perspectives of hierarchical control and cell-level coordination. In particular, for the layered distributed control architecture adopted in ref. [107], each controller acquires real-time average SOC and deviation information from its subordinate layer. Voltage references are then dynamically allocated to enable inter-layer duty cycle adjustment and SOC balancing. This method supports priority switching across layers, improving system-wide balancing while reducing communication complexity and control latency.

In contrast, fine-grained SOC equalization at the cell level has been studied in ref. [108]. By adjusting the position of each cell within the output waveform, current distribution is differentiated to promote balancing. A pseudo-open-circuit voltage measurement technique is also introduced to enhance SOC estimation accuracy. Both strategies leverage the degrees of freedom inherent in multilevel converter control to enable high-precision closed-loop balancing, making them suitable for large-scale systems with complex structures or significant cell inconsistency.

For DC-type multilevel RBS, a shared modulation control strategy has been proposed in ref. [105]. It dynamically generates the bus voltage through fast reconfiguration of battery modules, reducing the modulation burden of the main inverter, lowering switching losses, and enabling load-adaptive energy balancing across modules. Addressing current ripple in modular multilevel DC-DC converters, a harmonic suppression strategy has been introduced in ref. [109] based on Fourier series modeling. By applying segmented modulation and phase-shifted control, harmonic components are selectively canceled, improving current balance and output stability under varying conditions, and enhancing overall system performance.

While these strategies perform well in small to medium-scale systems, the centralized control architecture becomes a bottleneck as the number of submodules increases, due to communication bandwidth and computational limitations. This restricts scalability in large-scale energy storage applications.

6. Technical Challenges and Future Directions

6.1. Comparison of Typical Equalization Approaches

A comprehensive analysis of passive, active, and dynamic equalization systems reveals distinct differences in circuit architecture, control complexity, energy efficiency, and system scalability, each suited to specific application requirements. To clarify the core characteristics of various equalization topologies,

Table 2. Comparative summary of equalization topologies and representative studies.

Topology	Core Components	Reported Operating Range	Efficiency	Balancing Speed	References
Switched Shunt PEC	One resistor and one switch per cell	Low voltage (e.g., 3–12 cells in series)	Extremely low, with all the energy dissipated as heat	Slow	[20,22]
SI-based AEC	One inductor and 2N switches for N cells, or at least one inductor and one switch per cell	Medium to high voltage (e.g., 8–100 cells in series)	About 90–95%, depending on control and design	Medium to fast	[26,27]
SC-based AEC	One capacitor and 2N switches for N cells, or at least one capacitor and two switches per cell	Low to medium voltage (e.g., 4–32 cells in series)	About 90%, limited by capacitor losses	Medium to fast	[34,35,36]
Transformer-based AEC	One transformer and at least N switches for N cells	High voltage (up to 800 V for large strings)	Typical more than 95%, with transformer integration	Medium to fast	[52,54,56]
Converter-based AEC	Oner convertor or N convertors with multiple switches for N cells	Wide voltage range	Up to 95%, for optimized converter designs	Fast	[67,77,78]
RBS (Dynamic Equalization Circuit)	At least one switch per cell, or a multilevel converter for N cells	Flexible with dynamic reconfiguration	Varies with reconfiguration strategy (typically more than 90%)	Medium to fast	[105,108,109]

summarizes typical topologies from the literature, highlighting key indicators such as core components, operating range, efficiency, and balancing speed. Since the balancing speed is influenced by factors like cell count, initial SOC difference, and equalization current, the table provides only relative qualitative descriptions.

Passive equalization relies on resistive dissipation to balance cell voltages. Its simple circuit architecture uses a single shunt resistor and switch per cell, with no additional energy storage components, making it cost-effective and widely used in small-scale or cost-sensitive systems. However, its low energy efficiency, slow equalization speed, and thermal burden due to continuous resistive discharge make it unsuitable for second-life batteries with poor consistency and large-scale energy storage.

Active equalization redistributes energy between cells using components like inductors, capacitors, transformers, or DC-DC converters, improving energy efficiency and balancing resolution. Various active equalization topologies differ in characteristics. SI-based designs, such as the coupled inductor topology in ref. [27], suit medium-to-high-voltage scenarios (8–100 series cells) with efficiency of 88–95% and equalization time of 30–60 min. Capacitor-based designs, like the star structure in ref. [35], excel in low-to-medium-voltage scenarios (4–32 series cells), with efficiency of 90–93%. Transformer-based topologies, designed for high-voltage large-scale systems (up to 800 V [52]), achieve efficiencies over 95%, though they are complex. Converter-based designs offer wide voltage adaptability and high efficiency [77,78], providing adaptability across various scenarios. However, active equalization faces challenges of high hardware costs, large circuit volume, and high control complexity, especially in SOC- or capacity-based balancing, which requires precise real-time state estimation.

Dynamic equalization, as implemented in RBS, introduces dynamic reconfiguration of inter-cell connections through power electronic switching networks, with components like reconfigurable switch arrays or multilevel converters. This approach maintains energy redistribution capabilities while enabling fault isolation (e.g., bypassing faulty cells) and system-level adaptability [105]. Its flexible voltage range and modular design suit heterogeneous second-life batteries, offering a new direction for future energy storage systems. However, dynamic equalization requires sophisticated scheduling algorithms and fast-response control frameworks. As the number of switching elements increases, scalability and computational overhead become more prominent, requiring a balance between modular design and control complexity.

In summary, the optimal choice of equalization strategy depends on a combination of system scale, cell consistency, energy efficiency targets, and cost-performance trade-offs. Rather than a one-size-fits-all solution, future deployment of second-life battery systems will benefit from a hybrid or scenario-specific approach, integrating different equalization technologies based on the operational and economic constraints of the target application.

6.2. Current Limitations and Technical Challenges

In the context of second-life battery applications, equalization technologies face several domain-specific challenges that go beyond those in new battery systems.

First, state estimation remains a core limitation. Equalization strategies often rely on parameters such as terminal voltage, SOC, or capacity deviation. However, in second-life systems, aging inconsistency, sensor drift, temperature variation, and load fluctuations collectively degrade estimation accuracy. This can result in persistent balancing errors, directly impacting system stability, energy utilization, and safety.

Second, circuit complexity and integration difficulties are amplified in reconfigured second-life systems. Active and dynamic equalization topologies typically require numerous switches, magnetic components, or converter modules. When deployed at scale, the increased communication demand and computational overhead exceed the capabilities of conventional embedded hardware, making real-time control difficult.

Moreover, reliability and robustness remain insufficiently addressed. Long-term operation introduces concerns such as thermal imbalance, electromagnetic interference, and component aging, especially in systems where second-life cells have diverse degradation profiles. Fault isolation, degradation tracking, and adaptive control are still in early research stages. Currently, the field lacks standardized, cost-effective, and easy-to-deploy solutions tailored to the heterogeneity and unpredictability of retired battery modules.

6.3. Emerging Trends in Second-Life Battery Equalization Technologies

To support the large-scale reuse of retired Li-ion batteries, equalization technologies must evolve toward greater adaptability, intelligence, and architectural integration, as shown in Figure 13.

Data-driven intelligent control will be central to future equalization strategies. Advanced SOC and capacity estimation models, informed by machine learning and multi-source sensor fusion, will enable dynamic optimization of balancing behavior. This is critical for managing the performance variability inherent in second-life cells.

Modular and integrated hardware design will help reduce cost and implementation complexity. Embedding equalization within combined BMS–DC/DC architectures, and adopting distributed or hierarchical control frameworks, will improve scalability in grid storage and second-life EV battery systems. Application-specific configurations will also become more common—tailored to the needs of residential, commercial, or grid-level deployments.

RBS will play a key role in enabling flexible equalization. The integration of dynamic topology reconfiguration with real-time balancing will allow systems to adapt to evolving cell characteristics, isolate faults, and extend useful life. This architecture supports both cell-level flexibility and system-level resilience, making it a promising foundation for sustainable second-life energy storage platforms.

In summary, the future of second-life battery equalization lies in smart, scalable, and structure-aware solutions that account for the unique variability and degradation dynamics of repurposed cells.

7. Conclusions

With the rapid growth in electric vehicle deployment and energy storage applications, large volumes of lithium-ion batteries are approaching end-of-life. These second-life batteries, though no longer suitable for their original applications, still retain substantial usable capacity and present significant potential for reuse in less demanding scenarios. However, their performance inconsistency, arising from aging pathways, usage histories, and design variations, poses major challenges to system reliability, energy utilization, and safety. Effective equalization mechanisms are therefore essential for ensuring the stable and efficient integration of second-life batteries into reconfigured systems.

This paper provides a comprehensive review of equalization technologies for second-life battery applications, covering passive, active, and dynamic approaches. The structural characteristics and control strategies of each method are analyzed in detail. Passive methods offer simplicity and low cost but suffer from high energy loss and limited adaptability. Active strategies—based on inductors, capacitors, transformers, or DC-DC converters—achieve higher efficiency and control flexibility but introduce implementation complexity and BMS dependency. Dynamic equalization, particularly in the form of RBS, provides superior adaptability and cell-level scheduling capabilities, making it especially suitable for second-life scenarios with high inconsistency. However, its technical maturity and control demands remain key barriers to large-scale deployment.

Furthermore, this paper discusses control algorithms ranging from localized autonomous schemes to centralized governed strategies, highlighting the trade-offs in complexity, real-time performance, and precision. Special attention is given to multilevel converter-based RBS, which integrates equalization with power conversion, offering a promising direction for high-performance, scalable energy storage systems.

Looking forward, the advancement of second-life battery equalization will require a synergistic evolution of intelligent control, modular system design, and adaptive topology reconfiguration. Addressing technical challenges such as real-time estimation, communication constraints, and heterogeneous degradation will be crucial. Ultimately, robust and cost-effective equalization solutions will play a foundational role in enabling the sustainable reuse of lithium-ion batteries and supporting the development of resilient, low-carbon energy infrastructures.