In the field of distributed energy storage battery systems, the series connection of numerous cells often leads to state-of-charge (SOC) inconsistencies due to manufacturing tolerances, self-discharge variations, and environmental effects. This voltage imbalance reduces the usable capacity of the energy storage battery and accelerates aging. Traditional passive balancing strategies, such as minimum voltage balance strategy (MVBS) and average voltage balance strategy (AVBS), are widely deployed in commercial energy storage battery cabinets. However, these methods suffer from limited effective balancing time and insufficient capacity recovery, especially for lithium iron phosphate (LFP) chemistry with a flat voltage plateau. To address these limitations, we propose a novel Self-Calibration Balance Strategy (SCBS) for energy storage battery systems. The SCBS leverages the dynamic voltage characteristics at the end of charge and discharge to achieve real-time SOC self-calibration, then performs full-time adaptive balancing to ensure that the cell with the minimum capacity is fully utilized. This paper presents the methodology, experimental setup, and comparative results on three 200 kWh LFP energy storage battery cabinets.
1. System Topology and Balancing Circuit
The distributed energy storage battery system under study consists of 224 LFP cells (280 Ah nominal, 3.2 V) connected in series. The total nominal voltage is 716.8 V, and the rated energy is 200 kWh. The system operates in a two‑charge–two‑discharge daily cycle with a power rating of 100 kW (0.5P). Charge and discharge are cut off when any cell voltage reaches 3.6 V or 2.8 V, respectively.
The passive balancing circuit uses a switched shunt resistor topology. Each cell is connected in parallel with a 33 Ω resistor controlled by a MOSFET. The balancing current is approximately 80 mA. Voltage sampling and balancing are interleaved to avoid interference. The control timing is shown in Figure 5 of the original work. The balancing efficiency θ is the duty cycle of active balancing within each sampling period.
2. Proposed Self-Calibration Balance Strategy (SCBS)
2.1 Dynamic SOC Self-Calibration at End of Charge/Discharge
LFP cells exhibit a nearly linear voltage curve in the last 5%–10% of SOC (95%–100% during charging, 0%–5% during discharging) under constant current. This property allows real‑time SOC calibration using the actual terminal voltage of the first cell that hits the cutoff. For each charge/discharge cycle, we record voltage and coulomb counting (Ah integral) after a certain threshold (e.g., voltage > 3.4 V for charging, < 3.1 V for discharging). The cell that reaches the cutoff first provides a reference voltage‑SOC mapping.
Let t be the current time. During discharge, the first cell that reaches 2.8 V defines the reference SOC ref_dis_SOC(t):
$$ref\_dis\_SOC(t) = \frac{Th_{Ah}(t_1) – Th_{Ah}(t)}{Q_e} \times 100\%$$
where t1 is the moment when SOC=0%, ThAh is the cumulative Ah integral (starting from system power‑on), and Qe is the rated capacity. Similarly, during charge, the reference SOC ref_cha_SOC(t) is:
$$ref\_cha\_SOC(t) = \left(1 – \frac{Th_{Ah}(t_2) – Th_{Ah}(t)}{Q_e}\right) \times 100\%$$
where t2 is the moment when SOC=100%. Using these reference curves, each cell’s SOC at the cutoff moment (SOCend_cha and SOCend_dis) can be interpolated from its voltage. The actual dischargeable capacity of cell i is then:
$$Q_a(i) = \frac{Q_{dis}}{SOC_{end\_cha}(i) – SOC_{end\_dis}(i)}$$
where Qdis is the total discharge capacity from full charge to empty. The minimum Qa among all cells, denoted Qcell,min, represents the maximum usable capacity of the energy storage battery system.
Table 1 compares the SOC accuracy of the proposed dynamic self-calibration against an offline dynamic voltage‑SOC curve for a cabinet that has aged by 10%. The dynamic calibration significantly reduces error, especially in the flat voltage region.
| Phase | Voltage (mV) | True SOC (%) | Offline SOC (%) | Offline Error (%) | Dynamic SOC (%) | Dynamic Error (%) |
|---|---|---|---|---|---|---|
| Charge (0.2P) | 3445 | 94.7 | 98.7 | 4.0 | 94.4 | 0.3 |
| 3488 | 96.7 | 99.4 | 2.7 | 96.5 | 0.2 | |
| 3513 | 97.7 | 99.6 | 1.9 | 97.5 | 0.2 | |
| 3542 | 98.7 | 99.8 | 1.1 | 98.5 | 0.2 | |
| 3570 | 99.7 | 99.9 | 0.2 | 99.5 | 0.2 | |
| Discharge (0.5P) | 2988 | 7.8 | 2.5 | 5.3 | 7.7 | 0.1 |
| 2957 | 5.7 | 1.9 | 3.9 | 5.6 | 0.1 | |
| 2921 | 3.7 | 1.3 | 2.4 | 3.6 | 0.1 | |
| 2889 | 2.6 | 0.9 | 1.8 | 2.4 | 0.2 | |
| 2821 | 0.6 | 0.2 | 0.4 | 0.5 | 0.1 |
2.2 Full-Time Adaptive Balancing Control Based on SOC
The balancing objective is to make all cells reach the same SOC at the end of charge (within 1% deviation) and to fully utilize the cell with the minimum capacity. For a given charge/discharge cycle, when the highest SOC at the end of charge exceeds 95% and the lowest SOC at the end of discharge is below 5%, the balancing calculation is triggered. The baseline SOC (SOCbase) is defined as the SOC of the cell that has the lowest SOC at the end of discharge (the “weakest” cell). All cells with higher SOC at the end of charge are selected for balancing. The required balancing time for cell i is:
$$T(i) = \frac{(SOC_{ref\_cha}(t_1(i)) – SOC_{base}) \times Q_e – \Delta Q}{\theta \times I_b \times 100}$$
where θ is the balancing efficiency (typically ~0.85 for interleaved sampling), Ib is the balancing current (80 mA), and ΔQ is a reserve capacity margin (1 Ah, about 0.36% of rated capacity) to prevent ineffective balancing near the end. All selected cells are balanced simultaneously. If the AFE chip temperature exceeds 75°C, we switch to an odd‑even interleaved mode; if it exceeds 80°C, we pause balancing until the temperature drops.
The balancing process is iterative. After each cycle, a new reference voltage‑SOC curve is generated, the actual capacities are recalculated, and the remaining balancing times are updated. This feedback adaptive control mechanism gradually eliminates SOC calibration biases and component aging effects. Figure 6 in the original work illustrates the closed‑loop control flow.
3. Experimental Setup and Unbalanced State Construction
Three 200 kWh energy storage battery cabinets (rated capacity 280 Ah each) that had been in operation for six months were selected. The initial discharge capacities and voltage differences at the cut-off were measured (Table 2).
| Cabinet No. | Initial Discharge Capacity (Ah) | Max Voltage Difference at Charge Cut-off (mV) |
|---|---|---|
| 1 | 277.9 | 145.7 |
| 2 | 278.5 | 141.2 |
| 3 | 277.4 | 148.2 |
To create an unbalanced baseline, we deliberately discharged certain cells continuously for several days: cell 11 in each pack was never balanced; cell 10 was balanced for 10 days continuously; all other cells were balanced for 5 days continuously. After this treatment, the voltage distribution at the end of charge and discharge became highly scattered, as shown in Figure 7 of the original work. This constructed imbalance allowed a fair comparison of different balancing strategies.
4. Results and Comparison of Balancing Strategies
Three strategies were implemented on the three cabinets: MVBS (minimum voltage balance, activated when the voltage difference exceeds 50 mV and deactivated when below 30 mV, only during charging), AVBS (average voltage balance, activated when cell voltage exceeds the average by 50 mV, deactivated when below 30 mV, active in all modes), and the proposed SCBS. The experiments lasted 12 days under the same operating schedule. The key results are summarized in Table 3.
| Strategy | Effective Balancing Time (h) | System Capacity Before (Ah) | System Capacity After (Ah) | Capacity Improvement (%) | Max Voltage Difference Before (mV) | Max Voltage Difference After (mV) | Voltage Difference Reduction (mV) |
|---|---|---|---|---|---|---|---|
| MVBS | 5.7 | 262.4 | 262.8 | 0.2% | 192.5 | 191.5 | 1.5 |
| AVBS | 34.9 | 263.1 | 266.2 | 1.2% | 212.4 | 205.1 | 7.3 |
| SCBS | 220 | 261.5 | 277.1 | 5.9% | 211.2 | 150.5 | 60.7 |
The SCBS achieved the highest capacity improvement (5.9%), restoring the cabinet to its original capacity before the imbalance construction. The effective balancing time was 220 hours, far exceeding that of AVBS (34.9 h) and MVBS (5.7 h). The voltage difference at charge cut-off was reduced by 60.7 mV, compared to only 7.3 mV for AVBS and 1.5 mV for MVBS.
We also extended the AVBS test to 36 days. The capacity eventually stabilized at 269.5 Ah (still 8 Ah below the theoretical maximum), because the average voltage criterion cannot effectively trigger balancing for a few cells with severely lower voltage windows. In contrast, the SCBS directly targets the SOC deviation and iteratively corrects it.
Figure 12 from the original work shows the evolution of charging cut-off voltage and system capacity during SCBS balancing. After 10 days, the capacity reached a plateau, indicating that the weakest cell had been fully utilized. The final charging cut-off voltage dispersion was reduced to 150.5 mV, close to the original factory state.
5. Energy Loss and Long-Term Stability
Passive balancing dissipates energy as heat. However, in a voltage‑limited energy storage battery system, the energy consumed is exactly the energy that would otherwise be wasted due to cell inconsistency (the high‑SOC cells cannot be fully discharged). Therefore, the overall round‑trip efficiency is not significantly reduced. The balancing resistors (33 Ω) dissipate about 0.2 W each at 3.2 V. The AFE chip temperature reached a maximum of 73°C in a 24‑hour full‑balancing test, well below the 105°C safe limit. For long‑term deployment, we adopted an odd‑even interleaving mode to distribute heat.
After the initial 12‑day SCBS balancing, we monitored cabinet No. 3 for another 120 days. The capacity and voltage difference remained stable (Figure 14 of the original work), confirming that the balanced state is maintained without drift.
6. Scalability of SCBS
The SCBS strategy is applicable to different energy storage battery chemistries (e.g., NMC) as long as the terminal voltage near the cut‑offs is approximately linear. It works for various series configurations. We verified its performance on 8‑cell, 128‑cell, and 416‑cell strings, as shown in Table 4.
| Series Cells | Cell Capacity (Ah) | Test Duration (h) | Voltage Diff. Before (mV) | Capacity Before (Ah) | Voltage Diff. After (mV) | Capacity After (Ah) |
|---|---|---|---|---|---|---|
| 8 | 314 | 96 | 176.2 | 275.4 | 100.3 | 281.1 |
| 128 | 314 | 120 | 162.3 | 273.5 | 66.8 | 280.6 |
| 416 | 280 | 96 | 117.5 | 272.1 | 50.2 | 277.5 |
One limitation is that if the cell capacities within an energy storage battery pack are highly inconsistent, the weakest cell will experience deeper cycles, accelerating its aging. In such cases, the SCBS should be combined with a limited depth‑of‑discharge (DOD) operation and periodic full charge/discharge calibration to balance both performance and lifetime.
7. Conclusion
We have developed and validated a Self-Calibration Balance Strategy (SCBS) for series‑connected energy storage battery systems. The strategy uses dynamic voltage–SOC mapping at the end of charge and discharge to periodically self‑calibrate each cell’s actual capacity and SOC. A feedback‑adaptive balancing controller then selects the cells requiring equalization and adjusts their balancing times after every cycle until the weakest cell’s capacity is fully utilized. Experiments on three 200 kWh LFP cabinets demonstrate that the SCBS achieves a 5.9% capacity recovery in 12 days, far outperforming conventional passive strategies (MVBS 0.2%, AVBS 1.2%). The method is robust against cell aging, temperature variations, and resistor degradation, and it maintains stable long‑term performance. The SCBS provides an efficient, low‑cost, and scalable solution for improving the usable capacity of distributed energy storage battery systems in real‑world applications.