In modern photovoltaic (PV) energy storage systems, the integration of solar power generation with energy storage devices provides clean and stable electrical support to the grid. Parallel operation of solar inverters is a key technique to expand system capacity, but it introduces challenges such as current imbalance, voltage fluctuations, and mode switching impacts, which severely affect system stability. Existing methods, including adaptive current prediction models and impedance reshaping for resonance suppression, have improved parallel performance to some extent. However, they suffer from limited robustness against multi-inverter parameter differences and dynamic response delays during grid-connected to off-grid transitions, making them unsuitable for high-power-density scenarios. To address these issues, we propose a control strategy that integrates circulating current characteristic analysis with multi-mode coordination. By optimizing no-load voltage consistency control and designing seamless switching logic, our approach ensures full-operating-condition stability for parallel-connected solar inverters.
The core of our strategy lies in a comprehensive analysis of circulating currents in parallel solar inverters. Circulating currents arise due to differences in output voltage amplitude, phase, and line impedance among inverters, leading to non-load-demand currents that circulate between units. When multiple solar inverters operate in parallel, voltage amplitude discrepancies create potential differences, driving currents between inverters. Similarly, phase differences cause instantaneous voltage imbalances, even if amplitudes match. Line impedance inconsistencies further exacerbate circulating currents, as varying impedance paths result in nonlinear current distribution. From an equivalent circuit perspective, each solar inverter connects to a common load through line impedances. According to Kirchhoff’s laws, any mismatch in inverter parameters or line impedances inevitably generates circulating currents, which can cause current distortion, inductor saturation, and uneven power distribution, compromising system stability.
For stable parallel operation, solar inverters must maintain consistent output voltage amplitude, frequency, and phase. This eliminates the primary sources of circulating currents induced by parameter differences. The control objectives for parallel solar inverter systems include suppressing circulating currents to balance power distribution, achieving seamless transitions between grid-connected and off-grid modes, and enhancing system anti-interference capability. Through multi-dimensional optimization, these goals ensure robust system performance under varying conditions.
Adaptive Circulating Current Suppression Strategy Based on Characteristic Analysis
Our adaptive suppression strategy focuses on real-time monitoring and dynamic compensation by quantifying the effects of voltage amplitude, phase, and line impedance differences on circulating currents. We establish a closed-loop control system that begins with sampling output voltage and current signals from each solar inverter. By extracting no-load voltage amplitude deviations and phase differences, we apply a virtual impedance compensation algorithm. This algorithm calculates compensation values based on line impedance variations to adjust reference voltages, ensuring no-load output voltage consistency across all solar inverters and mitigating the root cause of circulating currents.
Next, we implement a dynamic circulating current detection mechanism. Measured circulating current signals are processed through band-pass filters to separate high-frequency and low-frequency components. A proportional-resonant (PR) regulator is introduced to provide real-time compensation for amplitude and phase deviations. For low-frequency circulating currents, the PR regulator’s fundamental resonance characteristics suppress periodic errors. For high-frequency components, a combination of high-pass filtering and proportional control rapidly attenuates harmonic distortions. The control law for the PR regulator can be expressed as:
$$ G_{PR}(s) = K_p + \frac{2K_r\omega_c s}{s^2 + 2\omega_c s + \omega_0^2} $$
where \( K_p \) is the proportional gain, \( K_r \) is the resonant gain, \( \omega_c \) is the cutoff frequency, and \( \omega_0 \) is the fundamental frequency. This approach ensures effective suppression of circulating currents across a wide frequency range.
Grid Voltage Synchronization-Based Droop Control Strategy
Droop control emulates the frequency-active power and voltage-reactive power characteristics of synchronous generators to enable power sharing and voltage/frequency regulation in distributed solar inverter systems. The fundamental equations are as follows:
Frequency-active power droop: When system frequency decreases, distributed solar inverters increase active power output; when frequency increases, they reduce output. The relationship is given by:
$$ f = f_0 – m_p (P – P_0) $$
where \( f_0 \) is the rated frequency, \( P_0 \) is the rated active power, and \( m_p \) is the frequency droop coefficient.
Voltage-reactive power droop: When voltage amplitude drops, solar inverters increase reactive power output; when voltage rises, they decrease output. The expression 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. These droop characteristics enable decentralized control without requiring communication between solar inverters, enhancing system reliability.
Experimental Validation and Performance Analysis
To validate our strategy, we constructed an experimental platform with three 40 kW solar inverters powered by a PV energy storage system. Voltage, current, and grid voltage signals were sampled via analog-to-digital converters for real-time monitoring. Key parameters were set to simulate non-ideal conditions, as summarized in Table 1.
| Parameter | Inverter 1 | Inverter 2 | Inverter 3 |
|---|---|---|---|
| DC Voltage (V) | 800 | 700 | 600 |
| Line Impedance (Ω) | 0.1 + j0.4 | 0.2 + j0.8 | 0.3 + j1.2 |
| Filter Inductance (mH) | 2 | ||
| Filter Capacitance (μF) | 10 | ||
| Equivalent Series Resistance (Ω) | 0.01 | ||
| Grid Line Voltage (V) | 380 | ||
| Load Active Power (kW) | 10 | ||
| Load Reactive Power (kVar) | 5 | ||
We compared our strategy with traditional methods: an adaptive current prediction model and an impedance reshaping approach. The results demonstrated significant improvements in circulating current suppression, switching performance, and power sharing.
Circulating Current Suppression Effectiveness
During parallel operation, inverters were sequentially connected at 0.30–0.40 s. The adaptive current prediction method, which relies solely on a zero-sequence circulating current model, exhibited a peak circulating current of 120 A at 0.32 s when inverter 2 was connected, with a current distortion rate of 5.2%. After inverter 3 connection at 0.40 s, circulating currents persisted at 95 A due to line impedance differences. The impedance reshaping method reduced peaks to 80 A but could not fully suppress amplitude deviation-induced currents, resulting in a 4.1% distortion rate.
In contrast, our strategy maintained output voltage deviations below 1.0% through no-load voltage consistency control. Combined with PR regulator-based dynamic feedback, circulating currents peaked at 48 A at 0.34 s (60.0% reduction) and stabilized at 35 A after inverter 3 connection. The current distortion rate dropped to 1.8%, showcasing superior performance. The voltage and current stability curves are visualized below:
Grid-Connected to Off-Grid Switching Performance
We simulated grid abnormalities by introducing a 20.0% voltage sag at 0.50 s. The adaptive current prediction method required 0.30 s for parameter identification, leading to a switching response time of 0.62 s and a voltage drop to 325 V (14.5% fluctuation). The impedance reshaping method showed a response time of 0.58 s with frequency drops to 48.2 Hz and voltage fluctuations of 12.3%.
Our strategy triggered off-grid commands within 0.02 s of detecting voltage deviations exceeding ±10.0%. Pre-synchronization control minimized phase differences to ±10%, while secondary frequency regulation stabilized output frequency at 49.8–50.2 Hz within 0.05 s. Pre-magnetization technology charged filter capacitors to 98.0% of rated voltage in 0.03 s. A state observer dynamically matched load parameters (estimated resistance 25 Ω, inductance 8 mH), limiting voltage fluctuations to 2.8% (380 V → 369 V) and achieving a response time of 0.12 s. Voltage recovered to 99% of rated value within 0.08 s post-switching.
Power Sharing Balance
Under differentiated DC voltages and line impedances, the adaptive current prediction method exhibited active power errors of 5.3% and reactive power errors of 8.1%. The impedance reshaping method showed errors of 4.8% and 7.5%, respectively. Our approach, incorporating virtual impedance compensation and droop control, ensured balanced power distribution. Active power outputs for inverters 1–3 were 10.10 kW, 9.92 kW, and 9.98 kW (errors <2.0%), while reactive powers were 5.05 kVar, 4.93 kVar, and 4.98 kVar (errors <3.0%). Table 2 summarizes the comparative results.
| Metric | Adaptive Current Prediction | Impedance Reshaping | Proposed Strategy |
|---|---|---|---|
| Peak Circulating Current (A) | 120 | 80 | 48 |
| Current Distortion Rate (%) | 5.2 | 4.1 | 1.8 |
| Switching Response Time (s) | 0.62 | 0.58 | 0.12 |
| Voltage Fluctuation (%) | 14.5 | 12.3 | 2.8 |
| Active Power Error (%) | 5.3 | 4.8 | 2.0 |
| Reactive Power Error (%) | 8.1 | 7.5 | 3.0 |
The mathematical foundation of our power sharing optimization can be extended using the following generalized droop equations for multiple solar inverters:
$$ P_i = P_{0,i} + \frac{1}{m_{p,i}} (f_0 – f) $$
$$ Q_i = Q_{0,i} + \frac{1}{m_{q,i}} (V_0 – V) $$
where \( i \) denotes the inverter index. By dynamically adjusting \( m_{p,i} \) and \( m_{q,i} \) based on real-time circulating current feedback, we achieve precise power distribution even under parameter disparities.
Conclusion
Our proposed control strategy, based on circulating current characteristic analysis and multi-mode coordination, effectively addresses the stability challenges in parallel solar inverter systems. By constructing an equivalent circulating current model and quantifying parameter influences, we designed adaptive suppression algorithms incorporating virtual impedance compensation and dynamic feedback. Seamless switching between grid-connected and off-grid modes was achieved through pre-synchronization, secondary frequency regulation, and pre-magnetization. Experimental results confirmed that our approach reduces circulating current peaks by 60.0%, decreases current distortion to 1.8%, shortens switching response time to 0.12 s with less than 3.0% voltage fluctuation, and limits active and reactive power errors to 2.0% and 3.0%, respectively. These advancements significantly enhance the stability and dynamic response of solar inverter parallel operations in photovoltaic energy storage systems, providing a reliable solution for high-power-density applications.