In the context of evolving power systems towards greater sustainability and resilience, energy storage technologies have become pivotal. Among these, the battery energy storage system stands out due to its rapid response, high power density, and flexibility in installation. However, traditional low-voltage battery energy storage systems often require multiple parallel units and step-up transformers to connect to medium- or high-voltage grids, leading to increased complexity and losses. To address this, the high-voltage direct-connection battery energy storage system utilizing cascaded H-bridge converters has emerged as a promising solution, offering modularity, high-voltage direct grid integration, large single-unit capacity, and multi-level output. Despite its advantages, there is a lack of real-time simulation models that incorporate battery external characteristics and support control strategy validation for such systems. In this study, I developed a real-time simulation model for a high-voltage large-capacity battery energy storage system based on CPU-FPGA co-simulation, the first of its kind globally. Using the OPAL-RT RT-LAB platform, I built a corresponding real-time simulation system for a typical 35 kV battery energy storage system, enabling closed-loop testing of control and protection strategies. Various operational scenarios were tested to validate the model and control approaches, providing technical support for the design and optimization of high-voltage direct-connection battery energy storage systems.
The topology of the high-voltage direct-connection battery energy storage system is based on cascaded H-bridge converters, as illustrated below. This configuration allows for direct connection to the high-voltage grid without a transformer, enhancing efficiency and reliability. Each phase consists of multiple submodules connected in series, with a Y-type connection to the grid. This modular design facilitates scalability and maintenance, making it ideal for large-scale energy storage applications. The battery energy storage system leverages this topology to achieve high power ratings and improved performance.
Each submodule within the battery energy storage system includes a battery pack, an H-bridge converter, an LC filter, and a pre-charging circuit. The H-bridge converter enables bidirectional power flow, allowing the battery energy storage system to operate in both charging and discharging modes. The LC filter mitigates voltage and current ripples, protecting the battery and extending its lifespan. The pre-charging circuit ensures safe startup by limiting inrush currents. This detailed submodule design is crucial for the overall performance and safety of the battery energy storage system.
To analyze and control the battery energy storage system, a mathematical model is derived. In the three-phase stationary coordinate system, the voltage balance equations are expressed as:
$$L\frac{di_a}{dt} = -Ri_a + e_a – v_a + v_{NO}$$
$$L\frac{di_b}{dt} = -Ri_b + e_b – v_b + v_{NO}$$
$$L\frac{di_c}{dt} = -Ri_c + e_c – v_c + v_{NO}$$
where \(i_a, i_b, i_c\) are the grid currents, \(e_a, e_b, e_c\) are the grid voltages, \(v_a, v_b, v_c\) are the output voltages of the battery energy storage system, \(v_{NO}\) is the potential difference between the neutral points, \(L\) is the filter inductance, and \(R\) is the filter resistance. Assuming balanced grid conditions, these equations simplify. For control purposes, transformation to the dq rotating coordinate system is performed, leading to:
$$L\frac{di_d}{dt} = -Ri_d + \omega L i_q + e_d – V_{dc}S_d$$
$$L\frac{di_q}{dt} = -Ri_q – \omega L i_d + e_q – V_{dc}S_q$$
Here, \(i_d\) and \(i_q\) represent the active and reactive current components, \(e_d\) and \(e_q\) are the grid voltage components, \(V_{dc}\) is the DC-side voltage of the submodules, and \(S_d\) and \(S_q\) are the switching functions in the dq frame. This model forms the basis for designing power control strategies for the battery energy storage system.
The power control strategy aims to regulate active and reactive power exchange between the battery energy storage system and the grid. When the d-axis aligns with the grid voltage vector, \(e_q = 0\), and the active power \(P\) and reactive power \(Q\) are given by:
$$P = \frac{3}{2}e_d i_d$$
$$Q = -\frac{3}{2}e_d i_q$$
Thus, controlling \(i_d\) and \(i_q\) directly controls \(P\) and \(Q\). To achieve decoupled control, the control inputs are designed as:
$$V_{dc}S_d = (i_d^* – i_d)H(s) + \omega L i_q + e_d$$
$$V_{dc}S_q = (i_q^* – i_q)H(s) – \omega L i_d + e_q$$
where \(i_d^*\) and \(i_q^*\) are the reference currents, and \(H(s)\) is a PI controller transfer function. This approach ensures fast and accurate power tracking for the battery energy storage system.
For modulation in the cascaded H-bridge battery energy storage system, the nearest level modulation (NLM) method is adopted due to its efficiency and simplicity for high-voltage applications with many submodules. The number of activated submodules per phase is determined by rounding the reference voltage to the nearest integer level. To address state-of-charge (SOC) imbalance among submodules within a phase, an enhanced NLM strategy is employed, where sorting is based on a composite parameter \(M_{ki}\) that incorporates both capacitor voltage and SOC deviation:
$$M_{ki} = u_{ki}(1 + \epsilon \Delta S_{ki}^{SOC})$$
Here, \(u_{ki}\) is the capacitor voltage of the \(i\)-th submodule in phase \(k\), \(\epsilon\) is a balancing coefficient, and \(\Delta S_{ki}^{SOC}\) is the deviation of the submodule’s SOC from the phase average. This strategy promotes SOC balancing during charging and discharging cycles, enhancing the longevity and performance of the battery energy storage system.
The battery model is critical for accurate simulation of the battery energy storage system. I used a second-order RC equivalent circuit to represent the lithium-ion batteries, capturing dynamic characteristics such as polarization effects. The state-space equations are:
$$\frac{dV_1}{dt} = -\frac{1}{R_1 C_1}V_1 + \frac{1}{C_1}I_{bat}$$
$$\frac{dV_2}{dt} = -\frac{1}{R_2 C_2}V_2 + \frac{1}{C_2}I_{bat}$$
where \(V_1\) and \(V_2\) are the voltages across the RC branches, and \(I_{bat}\) is the battery current. The terminal voltage \(V_{bat}\) is given by:
$$V_{bat} = V_{oc} – V_1 – V_2 – I_{bat}R_0$$
\(V_{oc}\) is the open-circuit voltage, which varies with SOC. Parameters \(R_0, R_1, C_1, R_2, C_2\) are identified experimentally for different SOC levels, enabling real-time updates during simulation to reflect battery behavior accurately in the battery energy storage system.
The real-time simulation platform was built using RT-LAB software and OP5707XG hardware from OPAL-RT. The model employs CPU-FPGA co-simulation: the grid and controller run on the CPU with a 35 μs time step, while the detailed battery energy storage system model runs on the FPGA with a 1 μs time step. This setup allows for high-fidelity simulation of the power electronics and battery dynamics. The table below summarizes key parameters of the grid and battery energy storage system used in the simulations.
| Parameter | Value |
|---|---|
| Grid Voltage | 35 kV |
| Grid Frequency | 50 Hz |
| Filter Inductance | 20 mH |
| Filter Resistance | 0.001 Ω |
| Battery Type | Lithium-ion |
| Battery Rated Voltage | 846 V |
| Battery Rated Capacity | 85 Ah |
| Submodules per Phase | 46 |
| System Rated Power | 10 MW |
| Initial SOC | 50% |
To validate the real-time model and control strategies, I conducted several tests under different operational scenarios for the battery energy storage system. The first test involved pre-charging of the submodule capacitors. Initially, the AC-side switches were open, and the battery pre-charge circuit was activated. The capacitor voltage rose smoothly to the battery voltage level, with current limited by the pre-charge resistor. The results matched theoretical expectations, demonstrating the model’s accuracy in simulating startup dynamics of the battery energy storage system.
In the passive inverter test, the battery energy storage system was disconnected from the grid, and the controller generated modulation waves synchronized with the grid voltage using a phase-locked loop. The output voltage of the battery energy storage system exhibited a multi-level staircase waveform with a peak value of approximately 42.8 kV, consistent with calculations based on 46 submodules and capacitor voltages around 930.5 V. This confirmed the proper functioning of the NLM strategy in the battery energy storage system.
For closed-loop operation, the battery energy storage system was connected to the grid, and power references were applied. Initially, both active and reactive power references were set to zero. At t=5 s, the active power reference was changed to -10 MW, commanding the battery energy storage system to discharge at rated power. The grid currents quickly adjusted, reaching an amplitude of about 233.28 A, as calculated from:
$$I_{norm} = \frac{2P_{norm}}{3U_{norm}}$$
where \(P_{norm} = 10\) MW and \(U_{norm} = 35\) kV. The battery current per submodule was around 77.87 A, with ripple within expected limits. At t=10 s, the active power reference was switched to 10 MW, causing the battery energy storage system to charge at rated power. The current reversed direction seamlessly, showing effective power control in the battery energy storage system.
To evaluate SOC balancing, I initialized five submodules in phase A with different SOC values (49.5% to 50.5%) while keeping others at 50%. Without balancing, all submodules discharged at similar rates during rated power discharge. When the balancing strategy was enabled, submodules with higher SOC discharged faster, and those with lower SOC charged faster during the charging phase, leading to convergence of SOC values over time. The composite parameter-based NLM successfully balanced SOC within the phase, enhancing the uniformity and efficiency of the battery energy storage system.
The real-time simulation model also allowed for detailed analysis of ripple characteristics. During rated power discharge, the battery current ripple ratio was measured at 15.15%, and the capacitor voltage ripple ratio was 1.84%, both aligning with theoretical predictions. SOC calculations using ampere-hour integration matched the simulation results, validating the battery model’s accuracy in the battery energy storage system.
In conclusion, this study presents a comprehensive real-time simulation framework for high-voltage large-capacity battery energy storage systems based on cascaded H-bridge converters. The CPU-FPGA co-simulation model accurately captures the dynamics of power electronics and battery behavior, enabling hardware-in-the-loop testing of control and protection strategies. Tests under pre-charging, passive inverter, closed-loop power control, and SOC balancing scenarios demonstrated the model’s validity and the effectiveness of the proposed control approaches. This work provides a valuable tool for the design, optimization, and validation of battery energy storage systems, contributing to the advancement of grid-connected energy storage technologies. Future work could explore fault scenarios, advanced battery management, and integration with renewable energy sources to further enhance the battery energy storage system’s role in modern power systems.
The development and validation of such real-time models are essential for accelerating the deployment of large-scale battery energy storage systems. By enabling thorough testing without physical prototypes, costs and risks are reduced, and system reliability is improved. As the demand for energy storage grows, innovations in simulation and control will continue to drive the evolution of battery energy storage systems towards higher efficiency, scalability, and grid support capabilities.