In modern energy storage systems, secondary batteries such as lithium-ion cells are widely used in various applications, from small-scale low-voltage systems to large-scale implementations like electric vehicles, uninterruptible power supplies, and renewable energy generation. Due to the low nominal voltage of individual energy storage cells, multiple cells are connected in series to meet higher system voltage requirements. However, during repeated charging and discharging cycles, inconsistencies in capacity, internal resistance, self-discharge rates, and environmental temperatures among series-connected energy storage cells lead to voltage imbalances. Without balancing measures, this can result in overcharging or over-discharging of weaker cells, reducing overall system capacity, shortening lifespan, and potentially causing safety hazards such as fires or explosions. Thus, balancing circuits are essential in energy storage systems.
Balancing techniques are generally categorized into passive and active types. Passive balancing dissipates excess charge through resistors, while active balancing transfers charge between cells or between cells and the battery pack using components like capacitors, inductors, or transformers. Although passive balancing is simple and low-cost, it reduces usable capacity and complicates thermal management. Active balancing, though more efficient and effective, often faces challenges in complexity and cost. Among active methods, multi-winding transformer-based balancing circuits offer a good balance of simplicity, cost-effectiveness, and reliability due to their magnetic isolation and automatic voltage balancing without requiring voltage measurement circuits. However, their scalability is limited, which can be addressed through modular approaches.
Traditional energy storage systems typically employ separate charging/discharging converters and balancers. This paper proposes an integrated converter that combines the voltage balancer with the charging/discharging converter, reducing component count, system complexity, and cost. The proposed converter automatically equalizes voltages across series-connected energy storage cells without the need for voltage sensing circuits. Two types are introduced: Type A uses one active switch, one n-winding inductor, and n diodes, while Type B replaces the diodes with n active switches to improve efficiency and enable bidirectional power flow. The relationship between Type A and Type B is analogous to that between asynchronous and synchronous Buck converters. This work focuses on analyzing Type B and validating it through experimental prototyping.
The proposed converter leverages a multi-winding coupled inductor to achieve automatic voltage balancing. In charging mode, the control strategy involves complementary PWM signals for the switches, ensuring zero-voltage switching (ZVS) conditions to enhance efficiency. The operational principles are analyzed under continuous conduction mode (CCM), with key waveforms and modes illustrating the current distribution and balancing mechanism. During the freewheeling period, cells with lower voltages receive higher currents, naturally equalizing the voltages over time. The steady-state analysis confirms that the voltage gain matches that of a Buck converter, given by the sum of cell voltages equaling the input voltage multiplied by the duty cycle. The balancing current depends on voltage differences among cells, leakage inductances, and switching parameters, allowing control through frequency and duty cycle adjustments.
Experimental validation was conducted on a Type B prototype with three series-connected lithium-ion energy storage cells. The setup included a bidirectional programmable power supply, electronic load, and microcontroller for generating complementary PWM signals. Tests under unbalanced and balanced conditions demonstrated the converter’s ability to automatically equalize cell voltages. For instance, in a constant-current-constant-voltage (CC-CV) charging scenario, the standard deviation of cell voltages reduced from 25.3 mV to 3.6 mV after 7200 seconds. Similarly, in a constant-current (CC) charging test with a larger initial imbalance, the standard deviation decreased from 213.6 mV to 10.6 mV after 5200 seconds. The input-output characteristics aligned with theoretical predictions, and efficiency reached 94.7% at 13.5 W output power.
A comparative analysis with existing integrated balancing topologies highlights the advantages of the proposed converter in terms of component count and cost. For n=3 energy storage cells, Type A and Type B require fewer magnetic components and diodes compared to alternatives, resulting in lower estimated costs. The trade-offs between simplicity, efficiency, and bidirectional capability are discussed, guiding selection based on application requirements.
In summary, this paper presents a novel converter that integrates voltage balancing into charging/discharging operations for series-connected energy storage cells. The design eliminates the need for voltage monitoring, reduces system complexity, and maintains high efficiency. Future work will explore modular scalability for larger battery strings and optimization for various operating conditions.
Topology and Operating Principles
The proposed converter with automatic voltage balancing capability is depicted in two variants: Type A and Type B. Both types utilize a multi-winding inductor with n windings, each coupled to an energy storage cell. For analysis, we assume identical turns for all windings, and each winding has an inductance Lm. The leakage inductance for each winding is denoted as Lki, where i=1,2,…,n.
Type A, as shown in Figure 3(a), consists of one active switch S0, one n-winding inductor, and n diodes D1 to Dn. Each energy storage cell Bi is connected to a diode Di and the corresponding winding Li. This configuration is unidirectional, suitable only for charging, and operates with simple control but lower efficiency due to diode losses.
Type B, illustrated in Figure 3(b), replaces the diodes with n active switches S1 to Sn. This allows bidirectional power flow and higher efficiency, albeit with slightly more complex control. In charging mode, S0 is turned on before S1-Sn, and in discharging mode, S0 is turned off after S1-Sn. This paper focuses on Type B in CCM for charging operations.
The key waveforms for the charging process are shown in Figure 4, and the operational modes are detailed in Figure 5. The switching period is Ts, and the duty cycle for S0 is D. The voltage across the inductor is uLm(t), the current into cell Bi is iBi(t), and the average current is iave(t). The differential current ΔiBi(t) represents the deviation from the average, driving the balancing action.
Mode 1 [t0 ≤ t < t1]: S0 is on, and S1-Sn are off. The inductor current increases linearly, and the current into each cell is equal, given by:
$$ i_{Bi}(t) = i_{Bi}(t_0) + \frac{V_{in} – \sum_{j=1}^{n} V_{Bj}}{n L_m + \sum_{j=1}^{n} L_{kj}} (t – t_0) $$
The voltage across each winding is:
$$ V_{Lm\_1} = – \frac{n L_m (V_{in} – \sum_{j=1}^{n} V_{Bj})}{n L_m + \sum_{j=1}^{n} L_{kj}} $$
Mode 2 [t1 ≤ t < t2]: S0 turns off, and the parasitic capacitances of S1-Sn are discharged until their body diodes conduct, enabling ZVS turn-on. This interval is short and negligible for current changes.
Mode 3 [t2 ≤ t < t3]: S1-Sn turn on, and energy is redistributed among cells. The equivalent circuit in Figure 6 shows that cells with lower voltages draw higher currents. The current into cell Bi is:
$$ i_{Bi}(t) = i_{Bi}(t_2) + \frac{V_{Lm\_3} – V_{Bi}}{L_{ki}} (t – t_2) $$
The current through the equivalent inductor Le is:
$$ i_{Le}(t) = n i_{Bi}(t) – \frac{V_{Lm\_3}}{L_m} (t – t_2) $$
Applying Kirchhoff’s current law:
$$ i_{Le}(t) = \sum_{j=1}^{n} i_{Bj}(t) $$
Solving for VLm_3:
$$ V_{Lm\_3} = \frac{ \sum_{j=1}^{n} \left( V_{Bj} \frac{L_m}{L_{kj}} \right) }{ \sum_{j=1}^{n} \frac{L_m}{L_{kj}} + 1 } $$
Since Lm >> Lki, VLm_3 approximates the average cell voltage when cells are balanced.
Mode 4 [t3 ≤ t < t4]: S1-Sn turn off, and after a dead time, S0 turns on at t4.
Mode 5 [t4 ≤ t < t0]: S0 is on, and currents equalize rapidly. Modes 2, 4, and 5 are brief and have minimal impact on energy transfer.
Type A lacks Modes 2, 4, and 5, operating only in Modes 1 and 3. The balancing mechanism primarily occurs during Mode 3, where current differences correct voltage imbalances automatically.
Steady-State Characteristics
Input-Output Characteristics
Ignoring the brief modes, the volt-second balance yields:
$$ \left( V_{in} – \sum_{j=1}^{n} V_{Bj} \right) D T_s = \sum_{j=1}^{n} V_{Bj} (1 – D) T_s $$
Simplifying, the voltage gain is:
$$ \sum_{j=1}^{n} V_{Bj} = V_{in} D $$
This matches the Buck converter gain, confirming that the proposed converter maintains familiar input-output relationships while integrating balancing functionality.
Balancing Characteristics
The balancing current ΔiBi(t) during Mode 3 is derived from the difference between cell current and average current:
$$ \Delta i_{Bi}(t) = i_{Bi}(t) – \frac{1}{n} \sum_{j=1}^{n} i_{Bj}(t) $$
Substituting expressions and assuming nLm >> Lki, the average balancing current over a cycle is:
$$ \Delta I_{Bi} = \frac{ (V_{Lm\_3} – V_{Bi}) (1 – D) T_s }{2 L_{ki}} $$
Using the approximation for VLm_3, we get:
$$ \Delta I_{Bi} = \frac{ \left( \sum_{j=1}^{n} \frac{V_{Bj}}{L_{kj}} – V_{Bi} \sum_{j=1}^{n} \frac{1}{L_{kj}} \right) (1 – D) T_s }{2 \sum_{j=1}^{n} \frac{1}{L_{kj}} } $$
Key insights from this analysis include:
– Balancing current arises automatically when cell voltages differ.
– The magnitude depends on voltage disparities, leakage inductances, duty cycle, and switching frequency.
– Larger leakage inductances reduce balancing current, and inconsistencies in Lki affect current distribution.
– Control parameters D and Ts can be adjusted to regulate balancing speed.
Simulation studies under varying leakage inductances validate these findings, showing that higher Lki values decrease peak balancing currents, and imbalances in Lki distort current sharing.
Experimental Validation
A Type B prototype was built to verify the converter’s performance. The test platform included three series-connected lithium-ion energy storage cells (NCR21700T), a bidirectional DC power supply, electronic load, oscilloscope, and STM32 microcontroller for PWM generation. Components and parameters are listed in Table 1.
| Component | Specification |
|---|---|
| Switches (S0-S3) | STP220N6F7, Ron = 2.40 mΩ |
| Output Capacitor (C0) | 22 μF MLCC × 3 |
| Magnetizing Inductance (Lm) | 12.2 μH |
| Switching Frequency (fs) | 50-250 kHz |
Control involves complementary PWM signals for S0 and S1-Sn, as shown in Figure 9. No individual cell voltage sensing is needed; only the module voltage is monitored.
Charging current waveforms were measured under balanced and unbalanced conditions. For example, with initial voltages VB1=VB2>VB3, the average currents during freewheeling were Iave1=1.56 A, Iave2=1.56 A, Iave3=1.67 A, indicating higher current into the lower-voltage cell. Similarly, when VB1=VB2<vb3, 1.43="" a,="" a.="" all="" automatic="" balanced="" balancing="" behavior.
Input-output特性测试 used capacitors instead of cells. With D=0.7, the output voltage matched Vin D, consistent with theoretical predictions.
Voltage balancing performance was evaluated in two scenarios:
– CC-CV charging: Initial voltages VB1=3.220 V, VB2=3.250 V, VB3=3.282 V (std. dev. 25.3 mV). After 7200 s, voltages reached VB1=4.105 V, VB2=4.109 V, VB3=4.114 V (std. dev. 3.6 mV).
– CC charging: Initial voltages VB1=3.234 V, VB2=3.121 V, VB3=2.735 V (std. dev. 213.6 mV). After 5200 s, voltages were VB1=4.109 V, VB2=4.092 V, VB3=4.083 V (std. dev. 10.6 mV).
Efficiency tests at Vin=15 V showed a peak of 94.7% at 13.5 W output.
Comparative Analysis
Table 4 compares the proposed converter with existing integrated balancing topologies in terms of component count and cost for n=3 energy storage cells. Cost estimates are based on typical component prices: MOSFET ($0.2), driver IC ($0.8), diode ($0.15), capacitor ($0.2), winding ($0.2), and core ($0.9).
| Topology | Active Switches | Magnetic Components (Core, Windings) | Diodes | Capacitors | Cost (n=3) |
|---|---|---|---|---|---|
| Multiplier Boost [40] | 1 | 1 (1,1) | 2n | n | $2.8 |
| Forward-Flyback [41] | 1 | 1 (1,2) | 2n | n+1 | $3.2 |
| Multi Stacked Buck-Boost [42] | 1 | n+1 (n+1,n+1) | n | n | $5.65 |
| Multi-winding Buck [33] | 1 | 1 (1,n+1) | n+1 | 0 | $2.5 |
| Multi-winding Boost [24] | 1 | 1 (1,n+1) | n+1 | 0 | $2.5 |
| Proposed Type A | 1 | 1 (1,n) | n | 0 | $2.15 |
| Proposed Type B | n+1 | 1 (1,n) | 0 | 0 | $2.3 |
Type A offers the lowest cost and simplicity but is unidirectional and less efficient. Type B provides bidirectional operation and higher efficiency at a slightly higher cost. Both proposed types reduce magnetic components compared to alternatives, lowering volume and cost. The choice depends on application needs: Type A for cost-sensitive, charging-only systems, and Type B for high-efficiency, bidirectional applications.
Conclusion
This paper presents a novel converter that integrates voltage balancing into charging/discharging operations for series-connected energy storage cells. The design eliminates the need for voltage monitoring circuits, reduces component count, and simplifies system architecture. Type A uses a single switch and diodes for unidirectional operation, while Type B employs multiple switches for bidirectional power flow and improved efficiency. Steady-state analysis confirms Buck-like voltage gain and automatic balancing currents dependent on voltage differences and leakage inductances. Experimental results demonstrate effective voltage equalization and high efficiency. The proposed converter offers a compelling solution for energy storage systems, with future work aimed at modular expansion and optimization for diverse applications.