As the integration of renewable energy continues to accelerate, the role of battery energy storage system in maintaining grid stability and enabling high penetration of intermittent sources has become indispensable. Among various electrochemical storage technologies, sodium-ion batteries have garnered significant attention due to the abundance of raw materials, superior wide-temperature adaptability, and enhanced safety characteristics compared to lithium-ion counterparts. However, the system-level integration of sodium-ion batteries—from individual cells to large-scale battery energy storage system—remains a critical engineering challenge. Two dominant topological configurations have emerged: the centralized topology and the string-configured topology. Each presents distinct trade-offs in terms of efficiency, reliability, controllability, and cost. In this work, we conduct a comprehensive electromagnetic transient simulation analysis to quantitatively evaluate the dynamic performance of both topologies under various grid conditions, thereby providing a technical basis for topology selection in practical sodium-ion battery energy storage system projects.
1. System Architecture: Centralized vs. String-Configured Topology
The centralized topology, as depicted conceptually, aggregates a large number of battery cells through series-parallel combinations into one or several high-capacity battery clusters. These clusters are then connected in parallel on the DC side and feed into a single large-scale power conversion system (PCS). This architecture offers advantages such as high energy density, simple electrical layout, high conversion efficiency, and low initial capital cost. Nevertheless, it imposes stringent requirements on cell consistency; any mismatch in state-of-charge or internal resistance among parallel clusters can induce circulating currents, potentially compromising safety and reliability. Furthermore, maintenance or replacement of a faulty module often necessitates shutting down an entire cluster or even the whole battery energy storage system.
In contrast, the string-configured topology adopts a more distributed approach. Battery cells are organized into multiple smaller independent modules or sub-units, each connected to a dedicated small-capacity PCS. The AC sides of these PCS units are then paralleled to form the complete system. This modular design inherently provides high redundancy and scalability: a fault in one module does not significantly affect overall system operation, and capacity expansion can be achieved smoothly by adding more modules. The string topology also enables flexible power and energy allocation among modules, real-time monitoring to optimize performance, and potentially longer battery life due to reduced inter-cell imbalances. However, it typically requires more PCS units, leading to higher initial equipment cost and slightly lower overall conversion efficiency due to multiple conversion stages.
| Feature | Centralized Topology | String-Configured Topology |
|---|---|---|
| System architecture | Multiple battery clusters in parallel, single large PCS | Individual battery clusters each with dedicated PCS, AC-side parallel |
| Energy density | High (fewer interconnections) | Moderate (more auxiliary equipment) |
| Conversion efficiency | Higher (single-stage conversion) | Slightly lower (multiple parallel conversions) |
| Initial capital cost | Lower (fewer PCS units) | Higher (multiple PCS units) |
| Redundancy & reliability | Low (single point of failure in PCS) | High (fault in one module isolated) |
| Scalability | Cumbersome (requires reconfiguration) | Easy (add more modules) |
| Cell consistency requirement | High (prone to circulating currents) | Moderate (each module individually controlled) |
| Maintenance complexity | High (cluster shutdown required) | Low (hot-swappable modules) |
| Dynamic response speed | Slower (single PCS) | Faster (multiple PCS parallel response) |
| Fault ride-through capability | Comparable (with proper control) | Comparable (with proper control) |
As summarized in the above table, the selection between these two topologies depends on application-specific priorities. For large-scale grid-connected battery energy storage system where efficiency and cost are paramount, centralized topology may be preferred. For applications demanding high availability, flexible operation, and ease of maintenance—such as critical grid support or commercial/industrial peak shaving—the string topology offers distinct advantages.
2. Electromagnetic Transient Modeling
2.1 Sodium-Ion Battery Equivalent Circuit Model
To accurately represent the dynamic behavior of sodium-ion batteries in electromagnetic transient simulations, we adopt an improved Thevenin equivalent circuit model. This model captures the nonlinear polarization effects through a parallel RC network. The equivalent circuit is shown conceptually and the governing equations are given below.
The terminal voltage Vt of the battery is expressed as:
$$ V_t = E – I \cdot R_0 – V_{pol} $$
where E is the open-circuit voltage (a function of state-of-charge, SOC), I is the battery current (positive for discharge), R0 is the internal resistance, and Vpol is the polarization voltage across the RC network. The polarization dynamics are given by:
$$ \frac{dV_{pol}}{dt} = \frac{I}{C_{pol}} – \frac{V_{pol}}{R_{pol} C_{pol}} $$
with Rpol and Cpol representing the polarization resistance and capacitance, respectively. These parameters are identified from experimental data for a 210 Ah sodium-ion cell (as used in our subsequent simulations). The SOC of the cell is updated via:
$$ SOC(t) = SOC(0) – \frac{1}{Q_{max}} \int_0^t I(\tau) \, d\tau $$
where Qmax is the maximum capacity (210 Ah). The model is implemented in PSCAD using a controlled voltage source and a subcircuit for the RC network.
2.2 Power Conversion System Model
The PCS is the critical interface between the battery and the grid. We model a three-level neutral-point-clamped (NPC) voltage source converter with insulated-gate bipolar transistors (IGBTs) and a detailed switching model that accounts for conduction losses and switching delays. The AC side includes an LCL filter for harmonic mitigation. The control system is based on a dual-loop structure in the rotating dq reference frame synchronized to the grid via a phase-locked loop (PLL).
The outer loop regulates active and reactive power as follows:
$$ P_{ref} – P_{meas} \xrightarrow{PI} i_{d,ref}^{+} $$
$$ Q_{ref} – Q_{meas} \xrightarrow{PI} i_{q,ref}^{+} $$
where the superscript ‘+’ indicates the positive-sequence component. The inner current loop controls the positive-sequence dq currents using PI controllers with decoupling terms:
$$ v_d^{+*} = (i_{d,ref}^{+} – i_d^{+}) (K_p + \frac{K_i}{s}) – \omega L i_q^{+} + u_d^{+} $$
$$ v_q^{+*} = (i_{q,ref}^{+} – i_q^{+}) (K_p + \frac{K_i}{s}) + \omega L i_d^{+} + u_q^{+} $$
Here, ω is the grid angular frequency, L is the equivalent inductance of the filter, and ud+, uq+ are the positive-sequence grid voltage components. For asymmetric faults, a negative-sequence current suppression loop is activated. The negative-sequence current references are set to zero (id,ref− = 0, iq,ref− = 0), and the corresponding voltage commands are computed using a similar PI structure:
$$ v_d^{-*} = (0 – i_d^{-}) (K_p + \frac{K_i}{s}) + \omega L i_q^{-} + u_d^{-} $$
$$ v_q^{-*} = (0 – i_q^{-}) (K_p + \frac{K_i}{s}) – \omega L i_d^{-} + u_q^{-} $$
Additionally, during low-voltage ride-through (LVRT) and high-voltage ride-through (HVRT) events, a voltage support strategy is implemented. The reactive current injection is set according to grid codes:
$$ i_{q,ref}^{+} = k \cdot (1.0 – U^{+}) \quad \text{for LVRT} $$
$$ i_{q,ref}^{+} = k \cdot (U^{+} – 1.0) \quad \text{for HVRT} $$
where k is the voltage support coefficient (typically 1.5 to 2.0 pu/pu) and U+ is the positive-sequence voltage magnitude. The active current reference is prioritized after reactive current if the inverter current limit is reached.
2.3 Simulation Setup
We implement both topologies in PSCAD V5.0. For the centralized topology, 14 battery clusters (each with 384 cells in series, 210 Ah) are connected in parallel on a common DC bus (rated 1.5 kV) and feed a single PCS rated at 1,680 kW. For the string topology, each of the 14 clusters is connected to its own 120 kW PCS, and all PCS units are paralleled on the AC side (0.69 kV). The aggregate power rating is identical for fair comparison. The key parameters are listed below.
| Parameter | Value |
|---|---|
| Single cell capacity | 210 Ah |
| Cells per cluster (series) | 384 |
| DC bus voltage (nominal) | 1.5 kV |
| AC side voltage (line-to-line) | 0.69 kV |
| Centralized PCS rated power | 1,680 kW |
| String PCS rated power (each) | 120 kW |
| Number of clusters | 14 |
| Filter inductance (L) | 0.15 mH |
| Switching frequency | 4 kHz |
3. Simulation Results and Comparative Analysis
We evaluate three representative scenarios: low-voltage ride-through (LVRT), high-voltage ride-through (HVRT), and active power step response during discharge. In all cases, the battery energy storage system is initially operating at 20% of rated power (discharge mode) unless otherwise stated.
3.1 Low-Voltage Ride-Through Performance
A three-phase-to-ground fault is applied at the point of common coupling, causing the positive-sequence voltage to drop to 0.2 pu. Both topologies are required to stay connected and inject reactive current to support voltage recovery. The simulated waveforms of positive-sequence voltage, active power, reactive power, active current, and reactive current are compared. The key observation is that both topologies exhibit nearly identical responses: reactive current rises to the commanded value (k×0.8 pu) within 20 ms, active current reduces accordingly, and after fault clearance (duration 150 ms), all quantities recover smoothly to pre-fault levels. This confirms that the electromagnetic transient behavior during symmetrical LVRT is governed primarily by the PCS control algorithm rather than the battery-side topology.
| Index | Centralized | String-Configured | Difference |
|---|---|---|---|
| Reactive current settling time (ms) | 18 | 19 | ~1 ms |
| Active current overshoot (%) | 5.2 | 5.6 | 0.4% |
| Post-fault recovery time to 90% active power (ms) | 45 | 42 | 3 ms |
| Maximum voltage dip (pu) | 0.20 | 0.20 | 0 |
The numerical values in Table 3 demonstrate that the differences are marginal, well within measurement uncertainties. Both topologies comply with grid code requirements for LVRT.
3.2 High-Voltage Ride-Through Performance
For HVRT, a voltage swell is simulated where the positive-sequence voltage rises to 1.3 pu for 200 ms. The battery energy storage system must absorb reactive power to mitigate the overvoltage. The control algorithm switches to inductive reactive current injection (iq,ref+ = −k × 0.3 pu). Again, the responses of both topologies are almost indistinguishable, with reactive current achieving steady state in about 20 ms and active power reducing to accommodate the reactive priority if necessary. No instability or oscillation is observed. This confirms that the voltage support capability is independent of the DC-side topology, provided the PCS has sufficient current headroom.
3.3 Active Power Step Response
In this scenario, the battery energy storage system is initially at zero power (standby) and receives a step command to ramp active power to 1.0 pu (rated). The command is applied at t = 1.5 s. The simulated active power responses for both topologies are captured. The string topology exhibits a significantly faster rise time and settling time. The reason is that in the string configuration, the total power is the sum of outputs from 14 independent PCS units, each responding simultaneously to the same step command. In contrast, the centralized topology has only one PCS, whose internal current loop and DC-link dynamics limit the ramp rate. Quantitative metrics are provided in Table 4.
| Metric | Centralized | String-Configured | Improvement Factor |
|---|---|---|---|
| Rise time (10%–90%) | 85 ms | 38 ms | 2.2× |
| Settling time (within 2%) | 180 ms | 65 ms | 2.8× |
| Overshoot (%) | 8.2 | 3.5 | −4.7% |
| Steady-state error (%) | 0.5 | 0.3 | −0.2% |
The string topology’s faster dynamic response is attributed to the distributed power conversion: the aggregate effect of multiple PCS units acting in parallel yields a steeper overall power ramp. This advantage is particularly beneficial in applications requiring rapid power modulation, such as frequency regulation or fast grid support.
4. Discussion
The simulation results reveal that for fault ride-through events—both LVRT and HVRT—the dynamic behavior of the battery energy storage system is largely determined by the PCS control strategy rather than the DC-side topology. Both centralized and string-configured topologies can achieve equivalent performance if the same control algorithms (e.g., negative-sequence suppression, voltage support) are implemented. This finding is reassuring for system designers: the choice of topology does not compromise grid code compliance.
However, the active power step response demonstrates a clear advantage for the string topology. The reason is not only the parallel aggregation of PCS outputs but also the fact that each small PCS has a lower DC-link capacitance and faster inner current loop bandwidth compared to a single large PCS with higher capacitance and more complex thermal management. In practice, for a 1.68 MW centralized PCS, the DC-link capacitance is typically several times larger than that of a 120 kW unit, leading to larger time constants. The string topology, by distributing the total power among multiple smaller converters, inherently reduces the effective time constant of the system.
It is also worth noting that the string topology introduces additional complexity in terms of communication and coordination among the 14 PCS units. In our simulation, we assumed perfect synchronization and identical power references for all units, which yields optimal response. In real systems, factors such as communication delays, measurement inaccuracies, and unequal cable impedances may cause minor deviations from the ideal. Nevertheless, modern digital control and fast communication protocols (e.g., IEC 61850) can mitigate these issues.
From an economic perspective, the string topology requires more PCS units, increasing the initial equipment cost and possibly the footprint. However, the improved dynamic performance and higher redundancy can translate into lower lifetime costs through reduced downtime and extended battery life. A detailed techno-economic analysis is beyond the scope of this work, but the simulation results here provide quantitative input for such studies.
5. Conclusion
This paper presents a comparative electromagnetic transient simulation study of centralized and string-configured topologies for a sodium-ion battery energy storage system. The following conclusions are drawn:
- Both topologies exhibit nearly identical LVRT and HVRT performance, confirming that the PCS control strategy dominates the fault ride-through behavior. The battery topology does not significantly affect the system’s ability to support grid voltage under symmetrical or asymmetrical faults.
- The string-configured topology demonstrates significantly faster active power step response—rise time and settling time are reduced by more than a factor of two compared to the centralized topology. This advantage stems from the distributed nature of the power conversion, allowing multiple PCS units to respond in parallel.
- These findings provide quantitative evidence to guide topology selection. For applications where rapid power modulation is critical, such as primary frequency regulation or fast grid support, the string topology is preferable despite higher initial cost. For applications where steady-state efficiency and capital cost are paramount, the centralized topology remains a strong candidate, provided that battery consistency can be maintained.
Future work will extend the analysis to include unbalanced grid conditions, transient stability impacts, and hardware-in-the-loop validation using prototype sodium-ion battery modules. Additionally, a comprehensive cost-benefit model will be developed to incorporate the dynamic performance gains into lifecycle economic assessments.
Acknowledgment
The authors gratefully acknowledge the support from the National Key Research and Development Program of China (2022YFB2402500).
References
- Zhuo Z Y, Zhang N, Xie X R, et al. Key technologies and development challenges of high-proportion renewable energy power systems. Automation of Electric Power Systems, 2021, 45(9): 171-191.
- Li J L, Wu Y W, Wang N, et al. Review and prospect of gigawatt electrochemical energy storage stations. Automation of Electric Power Systems, 2021, 45(19): 2-14.
- Ying J G. Research on sodium-lithium ion battery combined energy storage system. Power Electronics Technology, 2025, 59(4): 58-61.
- Zhou S J, Han W J, Chen A X, et al. Research progress on biomass-derived hard carbon anode materials for sodium-ion batteries. Journal of Engineering Science, 2024, 46(09): 1685-1701.
- Li P L, Chen Z Y, Guan L Y, et al. Modification research progress of vanadium-based phosphate cathode materials for sodium-ion batteries. Acta Petrolei Sinica (Petroleum Processing Section), 2024, 40(6): 1736-1745.
- Ci S, Zhang C J, Liu B C, et al. Dynamic reconfigurable battery energy storage technology: principles and applications. Energy Storage Science and Technology, 2023, 12(11): 3445-3455.
- Hu G, Li H X, Zhang K J. Battery model for electromagnetic transient analysis of power systems. Hubei Electric Power, 2015, 39(10): 1-4.
- Zhang J Q, Liu P Y, Xie X R, et al. Electromagnetic transient modeling method for lithium-ion battery energy storage suitable for fault characteristic analysis. Automation of Electric Power Systems, 2023, 47(7): 166-173.
- Berger M, Kocar I, Farantatos E, et al. Modeling of Li-ion battery energy storage systems (BESSs) for grid fault analysis. Electric Power Systems Research, 2021, 196: 107160.
- Hou G X, Liu Q H, Xie M L. Low-voltage ride-through control strategy for battery energy storage system under symmetrical and asymmetrical grid faults. Power System Protection and Control, 2013, 41(10): 62-67.