In the field of photovoltaic (PV) energy storage systems, the parallel operation of solar inverters is a key technology to expand capacity and improve system reliability. However, the circulating current caused by parameter mismatches and the dynamic instability during grid-connected to off-grid switching remain challenging problems. In this paper, I propose a comprehensive control strategy based on circulating current characteristic analysis and multi-mode coordination. By constructing an equivalent model of the parallel circulating current, quantifying the influence of voltage amplitude, phase, and line impedance differences on the circulating current, and designing an adaptive circulating current suppression algorithm combined with a seamless grid-connected and off-grid switching control strategy, the stable operation of solar inverters under all working conditions is achieved. The experimental results show that this strategy significantly outperforms traditional methods in suppressing current distortion, improving power distribution balance, and shortening switching response time. This work provides a practical solution for the design of high-reliability photovoltaic energy storage systems.
1. Introduction
Photovoltaic energy storage systems integrate solar power generation with energy storage devices to provide clean and stable power support to the grid. The parallel technology of solar inverters can effectively expand system capacity, but it encounters problems such as current imbalance, voltage fluctuation, and mode switching impact, which seriously affect system stability. Existing research, such as the adaptive current prediction model method and the impedance reshaping-based resonance suppression method, has improved the parallel performance of solar inverters to some extent. However, these methods suffer from insufficient robustness to parameter differences among multiple inverters and dynamic response lag during grid-connected to off-grid switching, making them inadequate for stable control in high-power-density scenarios. To address these issues, I propose a strategy that integrates circulating current feature extraction with multi-mode coordinated control. By optimizing no-load voltage consistency control and designing seamless switching logic, the full-operating-condition stability of parallel solar inverters is realized.
2. Circulating Current Generation Mechanism in Parallel Solar Inverters
When photovoltaic energy storage solar inverters operate in parallel, the essence of circulating current generation is the non-load-demand current caused by differences in output voltage and line impedance among inverters. If the output voltage amplitudes differ, a potential difference between inverters drives current flow between them. Even if the amplitudes are identical, phase differences in output voltage create instantaneous voltage differences that also induce circulating current. Furthermore, inconsistency in line impedance exacerbates the circulating current effect because different impedances offer varying resistance to the circulating current, leading to nonlinear characteristics in magnitude and direction. From the equivalent circuit perspective, each solar inverter connects to a common load through its line impedance. When inverters or line impedances are mismatched, Kirchhoff’s laws inevitably produce circulating current. This circulating current causes current distortion, inductor saturation, and uneven power distribution, severely affecting system stability.
| Factor | Impact on Circulating Current |
|---|---|
| Voltage amplitude difference (ΔV) | Creates potential difference, induces fundamental-frequency circulating current |
| Voltage phase difference (Δθ) | Generates instantaneous voltage difference, causes reactive circulating current |
| Line impedance mismatch (ΔZ) | Varies the current-sharing path, enhances circulating current nonlinearity |
3. Stability Requirements and Control Objectives for Solar Inverters
For stable parallel operation of photovoltaic energy storage solar inverters, the output voltage amplitude, frequency, and phase must be consistent. This consistency eliminates the root causes of circulating current induced by parameter differences. The control objectives of parallel solar inverter operation include:
- Suppressing circulating current to balance power distribution.
- Realizing impact-free transition during grid-connected and off-grid mode switching.
- Improving the system’s anti-interference capability.
These objectives are achieved through multi-dimensional optimization to ensure system stability.
4. Control Strategy for Parallel Stability of Solar Inverters
4.1 Adaptive Suppression Strategy Based on Circulating Current Characteristics
The adaptive suppression strategy based on circulating current characteristics uses real-time monitoring and dynamic compensation as its core. By quantifying the influence of voltage amplitude, phase, and line impedance differences on circulating current, a closed-loop control system is constructed. First, the output voltage and current signals of each solar inverter are sampled to extract the no-load voltage amplitude deviation and phase difference. A virtual impedance compensation algorithm is introduced, which calculates the compensation amount based on line impedance differences and adjusts the reference voltage. This ensures that the output voltages of all inverters are consistent under no-load conditions, thereby weakening the potential difference that triggers circulating current at the source. Then, a circulating current dynamic detection link is built. The measured circulating current signal is separated into high-frequency and low-frequency components through a band-pass filter, and a proportional-resonant (PR) regulator is used to achieve real-time compensation for amplitude and phase deviations. For low-frequency circulating current, the fundamental resonant characteristics of the PR regulator suppress periodic deviations; for high-frequency circulating current, a combination of high-pass filtering and proportional control quickly attenuates harmonic components.
The equivalent model of the parallel solar inverter system can be expressed as:
$$
I_{\text{circ}} = \frac{\Delta V \angle \Delta \theta}{Z_1 + Z_2}
$$
where \( \Delta V \) is the voltage amplitude difference, \( \Delta \theta \) is the phase difference, and \( Z_1, Z_2 \) are the line impedances of two inverters. The control law for virtual impedance compensation is given by:
$$
V_{\text{ref},i} = V_{\text{base}} – Z_{\text{virt},i} \cdot I_{o,i}
$$
where \( V_{\text{ref},i} \) is the reference voltage for the i-th inverter, \( V_{\text{base}} \) is the common base voltage, \( Z_{\text{virt},i} \) is the virtual impedance, and \( I_{o,i} \) is the output current.
4.2 Droop Control Strategy Based on Grid Voltage Synchronization
Droop control simulates the frequency-active power and voltage-reactive power droop characteristics of synchronous generators to achieve power sharing and voltage/frequency regulation among distributed sources. The frequency-active power droop relationship is:
$$
f = f_0 – m_p (P – P_0)
$$
where \( f_0 \) is the rated frequency, \( P_0 \) is the rated active power, and \( m_p \) is the frequency droop coefficient. When the system frequency drops, the solar inverter increases its active power output; when the frequency rises, it reduces active power. The voltage-reactive power droop relationship is:
$$
V = V_0 – m_q (Q – Q_0)
$$
where \( V_0 \) is the rated voltage amplitude, \( Q_0 \) is the rated reactive power, and \( m_q \) is the voltage droop coefficient. When the voltage amplitude decreases, the solar inverter increases its reactive power output; when the voltage increases, it reduces reactive power. This droop control ensures that multiple solar inverters share the load proportionally without requiring communication.
| Parameter | Symbol | Value |
|---|---|---|
| Rated frequency | \( f_0 \) | 50 Hz |
| Rated voltage | \( V_0 \) | 380 V |
| Frequency droop coefficient | \( m_p \) | 0.001 Hz/W |
| Voltage droop coefficient | \( m_q \) | 0.002 V/VAr |
5. Experimental Verification
5.1 Experimental Platform and Parameter Setup
An experimental platform was built using three solar inverters, each with a capacity of 40 kW, powered by a photovoltaic energy storage system. The main circuit voltage, current, and the grid voltage provided by the parallel AC source were sent to an analog-to-digital converter for sampling to achieve real-time data monitoring. The DC voltages of the inverters were set as differentiated parameters: inverter 1 had a DC voltage (Ud1) of 800 V, inverter 2 had Ud2 = 700 V, and inverter 3 had Ud3 = 600 V. The line impedances were set to RL1 = 0.1 + j0.4 Ω, RL2 = 0.2 + j0.8 Ω, and RL3 = 0.3 + j1.2 Ω. The filter inductance (L) was 2 mH, the filter capacitance (C) was 10 μF, the equivalent series resistance of the filter inductor (R) was 0.01 Ω, the grid line voltage (Ug) was 380 V, and the load active power (P) was 10 kW with reactive power (Q) of 5 kVar. Different DC voltages and line impedance parameters were set to simulate non-ideal conditions in real scenarios. The performance of the proposed control strategy was compared with the adaptive current prediction model method and the impedance reshaping method in terms of voltage and current stability and switching response.
5.2 Experimental Results
5.2.1 Circulating Current Suppression Effect
During the experiment, when two solar inverters were sequentially connected at 0.30 s and 0.40 s, the adaptive current prediction model method, which relies only on a zero-sequence circulating current equivalent model to control common-mode voltage, exhibited a circulating current peak of 120 A when inverter 2 was connected at 0.32 s, with a current waveform distortion rate of 5.2%. After inverter 3 was connected at 0.40 s, the circulating current experienced a secondary fluctuation due to line impedance differences, with a peak value maintained at around 95 A. The impedance reshaping method, although it reshapes virtual impedance through harmonic voltage control, still had a circulating current peak of 80 A after 0.38 s under multi-inverter parameter differences, with a current waveform distortion rate of 4.1%. This method could not effectively suppress the amplitude deviation circulating current caused by DC voltage differences. In contrast, the proposed strategy with no-load voltage consistency control kept the output voltage deviation among inverters below 1.0%. Combined with the dual suppression of fundamental and harmonic components by the PR regulator in the circulating current dynamic feedback control, when inverter 2 was connected at 0.34 s, the circulating current peak was only 48 A, which is 60.0% lower than the adaptive current prediction model method. When inverter 3 was connected at 0.40 s, the circulating current peak further stabilized at 35 A, and the current waveform distortion rate dropped from 5.2% to 1.8%.
| Method | Circulating Current Peak (A) at 0.34 s | Circulating Current Peak (A) at 0.42 s | Current Distortion Rate (%) |
|---|---|---|---|
| Adaptive current prediction model | 120 | 95 | 5.2 |
| Impedance reshaping | 80 | 80 | 4.1 |
| Proposed strategy | 48 | 35 | 1.8 |
5.2.2 Grid-Connected to Off-Grid Switching Performance
In the test simulating grid anomalies (e.g., a 20% grid voltage sag at 0.50 s), the adaptive current prediction model method required a parameter identification process of 0.30 s, resulting in a switching response time of 0.62 s. The voltage during switching dropped from 380 V to 325 V, a fluctuation of 14.5%. Moreover, due to the lack of a pre-magnetization link, the filter capacitor charging current caused a secondary voltage dip, and the recovery time exceeded 1.00 s. The impedance reshaping method, although capable of suppressing some harmonics through virtual impedance adjustment, lacked a secondary frequency modulation algorithm, causing the frequency to drop from 50.0 Hz to 48.2 Hz during switching, with a voltage amplitude fluctuation of 12.3% and a response time of 0.58 s. These methods could not meet the requirements of sensitive loads.
The proposed control strategy, upon detecting a grid anomaly (voltage deviation exceeding ±10%), triggered the off-grid command within 0.02 s. Through pre-synchronization control, the phase difference between the inverter output voltage and the load terminal voltage was kept within ±10%. Simultaneously, the secondary frequency modulation algorithm stabilized the output frequency within 49.8 Hz to 50.2 Hz in 0.05 s. The pre-magnetization technology brought the filter capacitor voltage to 98% of the rated value within 0.03 s. Combined with a state observer for dynamic matching of load parameters (estimated equivalent resistance of 25 Ω and inductance of 8 mH), the final voltage fluctuation during switching was only 2.8% (380 V to 369 V), the response time was 0.12 s, and the voltage recovered to 99% of the rated value within 0.08 s after switching.
| Method | Response Time (s) | Voltage Fluctuation (%) | Frequency Deviation (Hz) | Recovery Time (s) |
|---|---|---|---|---|
| Adaptive current prediction model | 0.62 | 14.5 | 1.8 | >1.00 |
| Impedance reshaping | 0.58 | 12.3 | 1.8 | 0.90 |
| Proposed strategy | 0.12 | 2.8 | 0.2 | 0.08 |
5.2.3 Power Distribution Balance
Under the experimental conditions with differentiated DC voltages and line impedances among the three solar inverters (Ud1 = 800 V, Ud3 = 600 V; RL1 = 0.1+j0.4 Ω, RL3 = 0.3+j1.2 Ω), the adaptive current prediction model method only relied on the zero-sequence circulating current model to regulate common-mode voltage and did not dynamically compensate for line impedance and DC voltage differences. As a result, the active power distribution error reached 5.3% (inverter 3 measured 8.7 kW vs. theoretical 10.0 kW), and the reactive power error reached 8.1% (inverter 2 measured 4.1 kVar vs. theoretical 5.0 kVar). The impedance reshaping method, although it adjusted virtual impedance through harmonic voltage control, lacked consistency control of no-load voltage amplitude. When the DC voltage difference between inverter 1 and inverter 3 reached 200 V, the active power error was still 4.8% and the reactive power error was 7.5%. The power imbalance problem was especially severe after inverter 3 was connected due to its larger line impedance.
With the proposed strategy, the virtual impedance compensation algorithm adjusted the reference voltage of each solar inverter in real time. Combined with the dynamic matching of power-frequency characteristics and voltage-reactive power characteristics in droop control, when inverter 2 was connected, the droop coefficients were adjusted, and the circulating current dynamic feedback corrected the current deviation in real time. The active powers of inverters 1 to 3 were 10.10 kW, 9.92 kW, and 9.98 kW, respectively, with errors all below 2.0%. The reactive powers were 5.05 kVar, 4.93 kVar, and 4.98 kVar, respectively, with errors below 3.0%. This demonstrates excellent power sharing balance.
| Method | Active Power Error (%) | Reactive Power Error (%) |
|---|---|---|
| Adaptive current prediction model | 5.3 | 8.1 |
| Impedance reshaping | 4.8 | 7.5 |
| Proposed strategy | <2.0 | <3.0 |
6. Conclusion
This paper addresses the issues of circulating current suppression and mode switching stability in parallel operation of photovoltaic energy storage solar inverters. I propose a control strategy based on circulating current characteristic analysis and multi-mode coordination. By constructing a circulating current equivalent model to quantify the impact of parameter differences, designing virtual impedance compensation and circulating current dynamic feedback for suppression, and employing pre-synchronization control, secondary frequency modulation, and pre-magnetization technology for seamless grid-connected/off-grid switching, the strategy achieves significant improvements. Experimental results confirm that the proposed method reduces the circulating current peak by 60% compared to traditional methods, lowers the current waveform distortion rate to 1.8%, shortens the switching response time to 0.12 s with voltage fluctuation below 3.0%, and controls active and reactive power distribution errors within 2.0% and 3.0%, respectively. This work greatly enhances the stability and dynamic response capability of parallel solar inverter systems, offering a practical solution for high-reliability photovoltaic energy storage applications.