Battery energy storage systems are essential for integrating renewable energy sources into modern power grids. However, traditional battery energy storage systems suffer from the “shortboard effect” due to fixed series-parallel connections, which leads to safety risks and reduced capacity utilization. Dynamic reconfigurable battery networks (DRBN) offer a promising solution by using power electronic switches to flexibly change the topology of battery modules. This allows for real-time balancing of state-of-charge (SOC), fault isolation, and improved safety. Nevertheless, the interaction between the DRBN and the power conversion system (PCS) introduces new challenges regarding power matching and operational stability. In this paper, we propose a bidirectional interactive cooperative operation strategy that coordinates the control of DRBN and PCS under six typical operating conditions. By exchanging information between the two subsystems and adjusting both the retention factor and the power reference, our strategy achieves optimal performance in terms of SOC balancing and power support. We validate the effectiveness of the proposed method through detailed simulations in MATLAB/Simulink.
1. Introduction
With the increasing penetration of renewable energy sources, battery energy storage systems have become crucial for smoothing power fluctuations and improving power quality. However, conventional battery energy storage systems are built with fixed series-parallel configurations, where the weakest cell limits the overall performance. This “shortboard effect” reduces the usable capacity and accelerates aging. Moreover, thermal runaway events in lithium-ion batteries raise serious safety concerns. To address these issues, dynamic reconfigurable battery networks have been proposed. In a DRBN, each battery module is connected to a MOSFET switch, allowing the system to dynamically change the series-parallel topology. This enables active SOC balancing, fault isolation within milliseconds, and high energy utilization even with highly heterogeneous cells. Despite these advantages, most existing research focuses on the DRBN alone, treating the PCS as a separate entity. The coupling between the DRBN and the PCS is often neglected, leading to potential power mismatch or voltage instability when the DRBN reconfigures. In this work, we aim to bridge this gap by proposing a cooperative operation strategy that simultaneously considers the DRBN’s balancing needs and the PCS’s power tracking requirements.
2. System Structure of Dynamic Reconfigurable Battery Energy Storage Systems
The studied battery energy storage systems consist of a DRBN and a PCS. The DRBN topology we adopt is the classical N-parallel, M-series configuration, as shown in the control diagram (not illustrated). Each battery module is connected in series with a MOSFET switch, and each row also has a bypass switch. The switch matrix \(\mathbf{S}\) describes the state of each switch:
$$ \mathbf{S} = \begin{bmatrix} S_{11} & \cdots & S_{1N} \\ \vdots & S_{ij} & \vdots \\ S_{M1} & \cdots & S_{MN} \end{bmatrix} $$
where \(S_{ij}=1\) means the switch is closed and the corresponding battery module is connected. The DC-link voltage \(U_{dc}\) and current \(I_{dc}\) of the DRBN are given by:
$$ U_{dc} = \sum_{i=1}^{M} \frac{\sum_{j=1}^{N} (1 – S_i) S_{ij} U_{ij}}{\sum_{j=1}^{N} S_{ij}} $$
$$ I_{dc} = P / U_{dc} $$
Here, \(S_i\) is the bypass switch state of row \(i\), and \(U_{ij}\) is the terminal voltage of the battery module at position \((i,j)\). The PCS is controlled in either grid-following (GFL) or grid-forming (GFM) mode, depending on the grid strength. In GFL mode, the dq-axis current references are derived from the power loops:
$$ i_{dref} = \left( k_p + \frac{k_i}{s} \right) (P_{ref} – P) $$
$$ i_{qref} = \left( k_p + \frac{k_i}{s} \right) (Q_{ref} – Q) $$
In GFM mode, a virtual synchronous generator (VSG) control is used. The motion equation and virtual excitation are:
$$ J\frac{d\omega}{dt} = \frac{P_{ref}-P}{\omega_0} + D(\omega_0 – \omega), \quad \theta = \int \omega dt $$
$$ E = \left( k_p + \frac{k_i}{s} \right) \left[ K_q (Q_{ref} – Q) + K_u (U_N – U) \right] $$
The following table summarizes the key parameters used in our simulation.
| Parameter | Value |
|---|---|
| Topology | 4 parallel × 14 series |
| Battery module voltage | 60 V |
| Battery module capacity | 20 Ah |
| Max charge/discharge rate | 1C |
| PCS DC voltage range | 650 V – 1000 V |
| Switching frequency | 20 kHz |
| Filter inductance | 0.5 mH |
| Filter capacitance | 50 μF |
| Grid voltage (AC) | 380 V |
| Grid frequency | 50 Hz |
3. Flexibility Analysis of DRBN Reconfiguration
3.1 Electrical Constraints
The DRBN must operate within safe boundaries. The DC-link voltage must stay within the allowable range of the PCS, and the current through each battery module must not exceed its rated value:
$$ U_{\min} \le U_{dc} \le U_{\max}, \quad I_{ij,\min} \le I_{ij} \le I_{ij,\max} $$
The maximum output power of the battery energy storage systems under these constraints is:
$$ P_{\max} = U_{dc} \cdot \min \left( \sum_{j=1}^{N} I_{ij,\max} \right), \quad i = 1,2,\dots,M $$
3.2 Comparison of Reconfiguration Methods
Two reconfiguration methods are compared: the conventional method and the improved method. In the conventional method, the number of modules \(k\) connected per row is determined by:
$$ k = \lceil N \times \frac{|P_0|}{P_{\max}} \rceil $$
where \(\lceil \cdot \rceil\) is the ceiling function. This method only balances modules within the same row and cannot handle row-to-row imbalances. To address this, we introduce a retention factor \(\alpha\) and allow entire rows to be bypassed. The decision rule for discharging is:
$$ S_{ij} = \varepsilon (SOC_{\text{ave},i} – \alpha \cdot SOC_{\text{ave},\text{all}}) $$
$$ S_i = 1 – S_{ij} $$
where \(\varepsilon\) is the unit step function, and \(\alpha\) is adjusted adaptively based on the SOC extreme deviation \(r\):
$$ r = SOC_{\max} – SOC_{\min} $$
The optimal \(\alpha\) is designed as a function of \(r\):
$$ \alpha = \begin{cases} \alpha_{\max} & r > 0.1 \\ \alpha_{\min} + \frac{r – 0.02}{0.08}(\alpha_{\max} – \alpha_{\min}) & 0.02 < r < 0.1 \\ \alpha_{\min} & r < 0.02 \end{cases} $$
The following table compares the balancing performance of the three methods (no reconfiguration, conventional, improved) after discharging until any module reaches 20% SOC, starting from an initial extreme deviation of 33% and variance of 48.16.
| Method | Final SOC Extreme Deviation | Final SOC Variance |
|---|---|---|
| No reconfiguration | 18% | 31.80 |
| Conventional DRBN | 5% | 1.20 |
| Improved DRBN (proposed) | 4% | 0.62 |
The improved method achieves better balancing because it can handle both inter-row and intra-row differences. Although the retention factor slows down the balancing speed, the overall uniformity is superior, especially under high-power conditions where the conventional method must connect all modules and loses balancing capability.
4. Proposed Cooperative Operation Strategy
4.1 Evaluation Metrics
We define two metrics to evaluate the performance: the SOC variance \(\sigma^2\) and the SOC extreme deviation \(r\):
$$ \sigma^2 = \frac{1}{MN} \sum_{i=1}^{M} \sum_{j=1}^{N} (SOC_{ij} – SOC_{\text{ave},\text{all}})^2 $$
Power matching is achieved when the PCS power command \(P_0\) falls within the DRBN’s output power range:
$$ -P_{\max} < P_0 < P_{\max} $$
4.2 Six Operating Conditions
Based on whether the retention factor \(\alpha\) and the power command \(P_0\) are adjustable, we classify the operation into six conditions. The following table summarizes these conditions and the corresponding regulation methods for the discharging case.
| Condition | \(\alpha\) Adjustable | \(P_0\) Adjustable | Power Matched? | Regulation Method |
|---|---|---|---|---|
| 1 | No | No | Yes | No regulation needed (both optimal) |
| 2 | No | Yes | After regulating \(P_0\) | Adjust \(P_0\) to match DRBN power; module-balancing optimal |
| 3 | No | Yes | No | Adjust \(P_0\) to limit extent; module-balancing optimal, power at limit |
| 4 | Yes | No | After regulating \(\alpha\) | Adjust \(\alpha\) to match power; power command optimal |
| 5 | Yes | No | No | Adjust \(\alpha\) to limit extent; power optimal, module at limit |
| 6 | Yes | Yes | Yes or No | Adjust both \(\alpha\) and \(P_0\); priority depends on SOC deviation |
4.3 Cooperative Control Flow
The overall strategy is as follows. First, the DRBN determines the optimal retention factor based on current SOC deviation. Then it calculates the achievable power range after reconfiguration. If the PCS power command \(P_0\) is within this range, condition 1 is satisfied. If not, the system checks adjustability. For condition 2 or 3, the PCS modulates its power command within its allowable margin (e.g., ±5%). If power matching remains impossible, the DRBN reduces its retention factor to allow more modules to participate, transitioning to conditions 4, 5, or 6. In condition 6, the priority is given to the parameter that causes less sacrifice: if SOC deviation is large, keep \(\alpha\) as high as possible and adjust \(P_0\); if SOC deviation is small, adjust \(\alpha\) first to ensure power tracking.
5. Simulation Validation
5.1 Balancing Performance Under Different Retention Factors
We simulated the discharging process with three fixed retention factors (\(\alpha=0.95,0.90,0.85\)) when the power command varied. The following table shows the SOC extreme deviation and variance at 1000 s for each case, along with the operating conditions encountered as the power command increased.
| \(\alpha\) | SOC Extreme Deviation (%) | SOC Variance | Conditions Encountered |
|---|---|---|---|
| 0.95 | 16.10 | 10.89 | 1 → 3 |
| 0.90 | 15.38 | 10.44 | 1 → 2 → 3 |
| 0.85 | 18.55 | 13.85 | 1 → 2 |
Results show that a larger \(\alpha\) yields better balancing but narrower power range, causing the system to enter condition 3 (power mismatch) earlier. A smaller \(\alpha\) provides wider power range but worse balancing. This trade-off confirms that the retention factor should be adjusted based on the SOC deviation.
5.2 Verification of Cooperative Strategy
In another test (2000 s onward), we allowed \(\alpha\) to be adjustable. With \(P_0\) fixed, the system started in condition 1. When a larger power command was issued, the DRBN reduced \(\alpha\) to connect more modules and matched the power (condition 4). A further increase led to \(\alpha\) hitting its minimum and the power command still not satisfied, entering condition 5. With \(P_0\) adjustable, the system could also raise the power command to match the DRBN’s maximum output, operating in condition 6. These scenarios demonstrate that the proposed cooperative strategy can effectively handle all six conditions and maintain power tracking while preserving SOC balancing as much as possible.
6. Conclusion
In this paper, we have presented a comprehensive cooperative operation strategy for dynamic reconfigurable battery networks and power conversion systems in battery energy storage systems. By introducing an improved reconfiguration method based on the retention factor, we enhanced SOC balancing without sacrificing power capability. The six-condition classification provides a systematic way to coordinate the adjustment of both the retention factor and the power command. Simulation results confirm that the proposed strategy enables the battery energy storage systems to operate optimally under various scenarios, balancing the need for cell equalization and power grid support. Future work will focus on real-time optimization of the retention factor using machine learning and experimental validation on a hardware platform.