In this work, we present a novel adaptive balancing framework for series-connected battery energy storage systems, termed the Self-Calibration Balance Strategy (SCBS). This strategy addresses the critical challenge of State of Charge (SOC) inconsistency and capacity underutilization in large-scale battery energy storage systems. Our approach integrates a dynamic SOC self-calibration method based on terminal voltage characteristics during charging and discharging, coupled with a full-time adaptive control that aims to fully exploit the cell with minimum capacity. We validated the SCBS on three 200kWh lithium iron phosphate (LFP) energy storage cabinets operating under real-world two-cycle-per-day conditions. Comparative experiments against conventional Minimum Voltage Balance Strategy (MVBS) and Average Voltage Balance Strategy (AVBS) demonstrate that SCBS achieves superior balancing efficiency and capacity recovery, with a 5.9% capacity improvement over 12 days of operation, nearly five times that of AVBS. The strategy also exhibits robust adaptability to cell aging and parameter variations, making it highly suitable for practical battery energy storage systems.
1. Introduction
Battery energy storage systems have become a cornerstone of modern distributed energy resources, offering high energy utilization efficiency, flexible deployment, and enhanced grid stability. Among various chemistries, lithium iron phosphate (LFP) batteries are widely adopted in commercial and industrial battery energy storage systems due to their safety, long cycle life, and cost-effectiveness. However, series-connected LFP battery packs inevitably suffer from cell-to-cell variations in capacity, self-discharge rates, and internal resistance, which are exacerbated by manufacturing tolerances, welding inconsistencies, and temperature gradients. Over time, these discrepancies cause certain cells to reach voltage limits prematurely during charging and discharging, leading to substantial underutilization of the system’s total capacity. This issue is particularly pronounced in large-scale battery energy storage systems where hundreds of cells are connected in series.
Balancing strategies are essential to mitigate voltage and SOC divergence. Active balancing circuits, which transfer energy between cells using inductors or capacitors, offer high efficiency and fast balancing speed but are often limited by high hardware costs, reliability concerns, and thermal management issues. In contrast, passive balancing—shunting excess energy through resistors—is widely deployed in commercial battery energy storage systems due to its simplicity, low cost, and acceptable energy losses over the system lifetime. Despite its prevalence, conventional passive balancing algorithms, such as those based on instantaneous minimum voltage during charging or real-time average voltage across all operating states, suffer from limited effective balancing windows and insufficient correction capacity, especially under the flat voltage plateau characteristic of LFP chemistry. Moreover, SOC estimation relying on offline OCV-SOC curves or fixed dynamic voltage-SOC mappings degrades over time due to aging and changing operating conditions.
To overcome these limitations, we propose the Self-Calibration Balance Strategy (SCBS), which continuously redefines the voltage-SOC relationship at the end of charge and discharge using the actual cell that first reaches the cutoff condition. By leveraging the near-linear voltage behavior in the high and low SOC regions, we dynamically calibrate individual cell SOC and actual capacity without requiring extensive offline testing. Based on these calibrated values, we compute exact equalization durations for cells with excess SOC, and we implement a full-time adaptive balancing process that iteratively refines the strategy based on feedback. This paper details the design, implementation, and experimental validation of SCBS in three 200kWh battery energy storage systems.
2. System Description and Passive Balancing Circuit
2.1 System Architecture
Our battery energy storage systems consist of 224 LFP cells connected in series, each with a nominal capacity of 280 Ah and nominal voltage of 3.2 V. The entire system is rated at 200 kWh and operates with a bidirectional power conversion system capable of 0.5C (100 kW) charge/discharge. The system uses voltage-based cutoff: charging stops when any cell reaches 3.6 V, and discharging stops when any cell drops to 2.8 V. Key parameters are summarized in Table 1.
| Parameter | Value |
|---|---|
| Rated energy (kWh) | 200 |
| Rated power (0.5P) (kW) | 100 |
| Cell chemistry | LFP |
| Nominal cell voltage (V) | 3.2 |
| Nominal cell capacity (Ah) | 280 |
| Number of series cells | 224 |
| Total nominal voltage (V) | 716.8 |
| Round-trip efficiency (%) | >95 |
| PCS conversion efficiency (%) | >98 |
2.2 Passive Balancing Circuit
Each cell is equipped with a shunt resistor (33 Ω) connected in parallel via a MOSFET switch, controlled by the battery management unit (BMU). The balancing current is approximately 80 mA, and the resistor is physically located on the pack-level sampling board, which is thermally isolated from the cells. The AFE (Analog Front End) sampling and balancing operations are time-multiplexed to avoid interference. The duty cycle of balancing within a sampling period defines the effective balancing efficiency $\theta$, as shown in Figure 5 (described in the proposed strategy section).
3. Proposed SCBS Strategy
3.1 Dynamic SOC Self-Calibration
The core innovation of SCBS is the on-the-fly calibration of the voltage-SOC relationship using the terminal voltage of the first cell that reaches the cutoff limit during each charge or discharge event. For LFP batteries, the voltage exhibits a steep, near-linear change in the high-SOC region (95%–100%) during charging and in the low-SOC region (0%–5%) during discharging. This property allows us to calibrate the SOC at these boundaries with high accuracy.
During discharge, when the cell with the lowest voltage falls below 3.1 V, we begin logging voltage and cumulative ampere-hour increments at fixed intervals. The cell that first reaches 2.8 V is designated as the reference cell, and its reference SOC at any time $t$ is computed as:
$$
SOC_{\text{ref\_dis}}(t) = \frac{Th_{\text{Ah}}(t_1) – Th_{\text{Ah}}(t)}{Q_e} \times 100\%
$$
where $t_1$ is the moment the system reaches 0% SOC (i.e., discharge cutoff), $Th_{\text{Ah}}$ is the accumulated ampere-hour count (calculated via ampere-hour integration from the initial time $t_0$), and $Q_e$ is the system nominal capacity (280 Ah). Similarly, during charging, when the highest-voltage cell exceeds 3.4 V, we track the reference cell that first hits 3.6 V. Its reference SOC is:
$$
SOC_{\text{ref\_cha}}(t) = \left(1 – \frac{Th_{\text{Ah}}(t_2) – Th_{\text{Ah}}(t)}{Q_e}\right) \times 100\%
$$
where $t_2$ is the moment of full charge (SOC = 100%). The resulting curves $C_{\text{ref\_dis}}$ and $C_{\text{ref\_cha}}$ provide a dynamic voltage-to-SOC mapping that adapts to the specific state of the battery energy storage systems at the time of calibration.
Using these reference curves, we can interpolate the SOC at the cutoff moments for every other cell, denoted as $SOC_{\text{end\_cha}}(i)$ and $SOC_{\text{end\_dis}}(i)$. The actual maximum discharge capacity of each cell $Q_a(i)$ is then:
$$
Q_a(i) = \frac{Q_{\text{dis}}}{SOC_{\text{end\_cha}}(i) – SOC_{\text{end\_dis}}(i)}
$$
where $Q_{\text{dis}}$ is the total discharged capacity from full charge to empty, measured by the system. The minimum of $Q_a(i)$, denoted as $Q_{\text{cell\_min}}$, determines the maximum capacity that the battery energy storage systems can utilize after balancing.
3.2 Full-Time Adaptive Equalization Control
The equalization objective is to ensure that at the end of charging, the maximum difference between SOC values of any two cells is less than 1%, and that the cell with the smallest capacity can be fully utilized. Equalization is triggered when two conditions are met: (a) the highest SOC at charge cutoff exceeds 95%, and (b) the lowest SOC at discharge cutoff falls below 5%. We first identify the cell with the minimum $SOC_{\text{end\_dis}}$ at the discharge cutoff. The baseline SOC for judging which cells need equalization is then the $SOC_{\text{end\_cha}}$ of that same cell at the charge cutoff, denoted as $SOC_{\text{base}}$. For every cell $i$ whose $SOC_{\text{end\_cha}}$ is above $SOC_{\text{base}}$, we calculate the equalization time $T(i)$ as:
$$
T(i) = \frac{(SOC_{\text{ref\_cha}}(t_1) – SOC_{\text{base}}) \times Q_e – \Delta Q}{\theta \times I_b \times 100}
$$
Here $\theta$ is the balancing efficiency (duty cycle of active balancing within a sampling period), $I_b$ is the balancing current (80 mA), and $\Delta Q$ is a reserved capacity margin (1 Ah) to avoid over-equalization near the end. The timing diagram for AFE sampling and balancing is given in Figure 5 (conceptual).
All candidate cells begin equalization simultaneously. If the AFE chip temperature exceeds 75°C, the system switches to a parity-based interleaved equalization mode; if it exceeds 80°C, balancing is paused until the temperature drops. After each charge/discharge cycle that again satisfies the calibration conditions, the reference curves are updated, and the residual equalization times are recalculated based on the latest state. This feedback loop ensures convergence even in the presence of parameter drift (e.g., resistor aging, cell degradation). Figures 9–12 in the original work illustrate the iterative improvement.
3.3 Comparison with Conventional Strategies
We implemented two common passive balancing strategies for comparison:
- MVBS (Minimum Voltage Balance Strategy): During charging, any cell whose voltage exceeds the lowest cell voltage by more than 50 mV is balanced until the difference drops below 30 mV.
- AVBS (Average Voltage Balance Strategy): Regardless of operating state (charge, discharge, idle), any cell with voltage above the system average by more than 50 mV is balanced until the deviation is below 30 mV.
4. Experimental Setup and Results
4.1 Creation of Unbalanced State
We selected three 200kWh battery energy storage systems (Cabinets No.1–3) that had been in operation for six months. Their initial discharge capacity and maximum voltage deviation at charge cutoff are listed in Table 2. To accelerate imbalance, we intentionally discharged specific cells within each cabinet: cell No.11 in every 14-cell pack was never balanced; cell No.10 was balanced continuously for 10 days; all other cells were balanced for 5 days. This created a pronounced voltage divergence at both charge and discharge cutoffs, as shown in Figure 7 (conceptual).
| Cabinet | Initial Discharge Capacity (Ah) | Max Voltage Deviation at Charge Cutoff (mV) |
|---|---|---|
| 1 | 277.9 | 145.7 |
| 2 | 278.5 | 141.2 |
| 3 | 277.4 | 148.2 |
4.2 SCBS Performance on Cabinet No.3
Cabinet No.3 was equipped with the SCBS strategy. Figures 8–10 in the original paper depict the voltage evolution and capacity calibration results. The calibrated cell capacities are shown in Figure 10 (conceptual), indicating a minimum cell capacity of 275.4 Ah while the system’s actual discharge capacity was only 261.5 Ah, leaving a 5.3% headroom for improvement. Figure 11 (conceptual) shows the calculated equalization times for each candidate cell. Over 12 days of continuous operation with the SCBS, the maximum charge-cutoff voltage difference dropped from 211.2 mV to 150.5 mV, and the discharge capacity rose from 261.5 Ah to 277.1 Ah, essentially restoring the initial state before the artificial imbalance. The capacity stabilized after 10 days, indicating that the balancing objective was met.
4.3 Comparative Evaluation of Three Strategies
Cabinets No.1 and No.2 were operated under the MVBS and AVBS strategies respectively, under identical daily cycling conditions. Table 3 summarizes the 12-day performance comparison. The SCBS achieved a capacity improvement of 5.9%, whereas AVBS improved by only 1.2% and MVBS by a negligible 0.2%. The effective balancing time (hours of actual current flow through the shunt resistors) was 220 h for SCBS, 34.9 h for AVBS, and only 5.7 h for MVBS. The SCBS also reduced the charge-cutoff voltage deviation by 60.7 mV, compared to 7.3 mV for AVBS and 1.5 mV for MVBS.
| Strategy | Effective Balancing Time (h) | Initial Capacity (Ah) | Final Capacity (Ah) | Capacity Improvement (%) | Initial Voltage Deviation (mV) | Final Voltage Deviation (mV) | Voltage 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 |
We further extended the AVBS test to 36 days to evaluate its saturation point. Figure 13 (conceptual) shows that AVBS eventually reached a plateau at 269.5 Ah with a residual voltage deviation of 199 mV, unable to fully utilize the minimum-capacity cell. This limitation arises because the average voltage reference fails to capture deeply unbalanced low-SOC cells, preventing adequate balancing activation.
4.4 Energy Loss, Thermal Management, and Long-Term Stability
Since SCBS dissipates energy only from cells that are already at a higher SOC than the baseline, the discharged energy would otherwise be inaccessible during normal operation due to voltage constraint. Thus, the balancing process does not reduce overall system round-trip efficiency. Thermal impact is minimal: the 33 Ω shunt resistors generate about 0.2 W per cell during balancing, and the AFE chip temperature remains below 73°C even under worst-case continuous balancing of all 224 cells—well within the 105°C limit. After 12 days of SCBS, the system was monitored for another 120 days, showing stable capacity and voltage deviation without any degradation recurrence, confirming the strategy’s long-term effectiveness.
4.5 Scalability of SCBS
We validated SCBS on different series configurations (8 cells, 128 cells, and 416 cells) with varying cell capacities. The results in Table 5 demonstrate consistent reduction in voltage deviation and capacity recovery, confirming the applicability of SCBS across diverse battery energy storage systems scales.
| Series Configuration | Cell Capacity (Ah) | Test Duration (h) | Initial Voltage Deviation (mV) | Initial Capacity (Ah) | Final Voltage Deviation (mV) | Final Capacity (Ah) |
|---|---|---|---|---|---|---|
| 8-cell | 314 | 96 | 176.2 | 275.4 | 100.3 | 281.1 |
| 128-cell | 314 | 120 | 162.3 | 273.5 | 66.8 | 280.6 |
| 416-cell | 280 | 96 | 117.5 | 272.1 | 50.2 | 277.5 |
5. Conclusion
We have developed and experimentally validated a Self-Calibration Balance Strategy (SCBS) for series-connected LFP battery energy storage systems. By dynamically calibrating the voltage-SOC relationship at charge/discharge endpoints using the first-cutoff cell, SCBS provides accurate per-cell SOC and capacity estimates without relying on aging-prone offline curves. The full-time adaptive equalization then ensures that the cell with minimum capacity is fully utilized, while all other cells are brought to a consistent SOC at the charge cutoff. Comparative tests on three 200kWh battery energy storage systems demonstrate that SCBS outperforms conventional MVBS and AVBS strategies, achieving 5.9% capacity recovery in 12 days—approximately five times better than AVBS. The strategy is robust to cell aging, resistor drift, and temperature variations, and its scalability across different series configurations makes it highly suitable for a wide range of battery energy storage systems. Future work will extend SCBS to active balancing topologies and explore advanced state-of-health tracking to further improve lifecycle management of battery energy storage systems.
We believe that the SCBS framework provides a practical, low-cost solution to the persistent challenge of SOC imbalance in large-scale battery energy storage systems, enabling higher capacity utilization, extended system life, and improved economic returns for stationary storage applications.