The global energy landscape is undergoing a profound transformation, accelerating towards decarbonization and clean energy integration. Within this paradigm, photovoltaic (PV) power generation stands as a pivotal renewable source, witnessing a continuously increasing penetration rate in power systems worldwide. As the core interface connecting the DC side of PV arrays to the AC grid, the performance of solar inverters is critical. Specifically, the quality of their output current directly dictates grid-tied power quality and system stability. This becomes particularly challenging during Low Voltage Ride-Through (LVRT) events, where grid disturbances and necessary control strategy adjustments can cause significant fluctuation and even exceedance of current harmonic content, posing substantial risks to safe system operation.
In this context, this article systematically investigates the issue of output current harmonic control for solar inverters under LVRT conditions. It analyzes the mechanisms of harmonic generation, reviews existing mitigation strategies, and proposes a novel, integrated optimization approach. A practical case study of a 1 MW PV plant demonstrates the effectiveness of the proposed method in significantly reducing current Total Harmonic Distortion (THD) during severe voltage sags, thereby enhancing grid adaptability.
1. Origins and Impacts of Output Current Harmonics in Solar Inverters
Under normal grid-connected operation, the output current harmonics of a solar inverter primarily originate from several non-ideal characteristics inherent to the system. A comprehensive breakdown is provided in the table below:
| Harmonic Source | Primary Mechanism | Dominant Harmonic Orders | Key Influencing Factors |
|---|---|---|---|
| Power Semiconductor Switching | Non-ideal switching transitions (finite rise/fall times, switching losses). | High-frequency bands near switching frequency multiples. | Switching frequency, device technology (Si, SiC, GaN). |
| Dead-Time Effect | Intentional delay between turning off one switch and turning on its complement in a leg to prevent shoot-through. | Low-order harmonics (5th, 7th, 11th, 13th). | Dead-time duration, load current polarity, modulation index. |
| Control System Limitations | Digital control delays (computation, PWM update), sampling errors, quantization noise. | Broadband spectrum, can alias higher frequencies. | Processor speed, sampling frequency, ADC resolution. |
| Grid Voltage Disturbances | Unbalanced or distorted grid voltages induce negative-sequence and harmonic currents through the controller. | Harmonics present in the grid voltage (e.g., 3rd, 5th). | Grid impedance, severity of unbalance/distortion, controller bandwidth. |
The Pulse Width Modulation (PWM) strategy is a fundamental contributor. If the switching frequency is set too low, the associated high-frequency harmonic components may not be adequately attenuated by the output LCL or L filter, leading to an elevated THD. The dead-time effect is particularly detrimental, causing waveform distortion around current zero-crossings and generating characteristic low-order harmonics. This effect can be modeled as a voltage error $V_{err}$ injected in series with the ideal inverter output voltage:
$$
V_{err} = \text{sign}(i_{abc}) \cdot \frac{T_d}{T_s} \cdot V_{dc}
$$
where $T_d$ is the dead-time, $T_s$ is the switching period, $V_{dc}$ is the DC-link voltage, and $i_{abc}$ is the output phase current.
During LVRT, the problem is exacerbated. Grid voltage imbalances or sudden dips can cause transient overcurrents. The current control loop, which may have limited bandwidth or be operating at its saturation limit to provide reactive current support, can struggle to track its reference accurately. This lag can amplify pre-existing harmonics or even introduce new inter-harmonic components. The consequences of elevated harmonics are severe: increased power losses and device thermal stress, potential mal-operation of sensitive protection relays, interference with communication systems, and overall degradation of power quality for connected loads.
2. Existing Control Strategies for Low Voltage Ride-Through
2.1 Conventional LVRT Control Strategies
2.1.1 Reactive Current Compensation
This is the cornerstone strategy mandated by most grid codes. During a voltage dip, the solar inverter is required to inject reactive current ($I_q$) to support grid voltage recovery. The reference is typically a piecewise-linear function of the voltage sag depth. For example, for a voltage dip $V_{g}$ (in p.u.) below 0.9 p.u., the reference is often set as:
$$
I_{q,ref} = K \cdot (0.9 – V_{g}) \cdot I_{n}, \quad \text{for } 0.2 \le V_{g} < 0.9
$$
where $K$ is a gain (usually between 2.0 and 4.0 as per standards), and $I_n$ is the rated current. The implementation is typically done in the synchronous reference frame (dq-frame), where the q-axis current controller tracks $I_{q,ref}$. The primary advantage is direct support for grid stability. The major drawback is the potential conflict between the elevated reactive current demand and the inverter’s current limit ($I_{max} = \sqrt{I_d^2 + I_q^2}$), which can force a reduction in active current ($I_d$), distort the current waveform, and increase harmonic content.
2.1.2 Maximum Power Point Tracking (MPPT) Control
During LVRT, maintaining MPPT is often secondary to grid support. However, some strategies aim to manage the DC-link voltage and power balance. A common approach is to switch from normal MPPT to a constant power or current-limiting mode to prevent overvoltage on the DC bus when the active power injection is curtailed. The challenge lies in the interaction between the DC-link dynamics and the AC current control. An oscillating or misjudged MPPT point under rapidly changing grid conditions can introduce low-frequency oscillations that translate into current harmonics. The control period for MPPT algorithms (e.g., Perturb & Observe) is usually an order of magnitude slower (10-100 ms) than the inner current loop, making it a potential source of low-frequency disturbance.
2.2 Advanced Optimization Strategies and Theoretical Analysis
2.2.1 Improved PWM Modulation Techniques
Standard Space Vector PWM (SVPWM) or Sinusoidal PWM (SPWM) concentrate harmonic energy at well-defined frequencies ($mf_s$, where $m$ is an integer). Under LVRT, these concentrated harmonics can excite resonances with the grid impedance. Improved modulation strategies aim to spread this energy. Random PWM (RPWM) is a prominent example, where a random variable modulates the switching frequency or the pulse position. By making the harmonic spectrum continuous, the peak amplitudes at specific frequencies are reduced. The switching function for a carrier-based RPWM can be expressed with a randomized carrier frequency $f_s(t)$:
$$
f_s(t) = f_{s0} + \Delta f \cdot r(t)
$$
where $f_{s0}$ is the center switching frequency, $\Delta f$ is the maximum frequency variation, and $r(t)$ is a bounded random signal. While effective for electromagnetic interference (EMI) reduction, its direct impact on low-order harmonic reduction under LVRT is limited unless combined with other techniques.
2.2.2 Virtual-Impedance-Based Harmonic Suppression
This method is highly effective for damping resonances and suppressing selective harmonics. The core idea is to emulate a virtual impedance ($Z_v(s) = R_v + sL_v$) in the control loop without dissipating real power. A high-pass or band-pass filter is used to extract the harmonic components of the current ($i_{h}$). A voltage drop proportional to this harmonic current is then subtracted from the modulation reference:
$$
V_{comp} = -Z_v(s) \cdot i_{h}
$$
For example, to damp high-frequency resonances, a virtual resistor ($R_v$) is effective: $Z_v(s) = R_v$. To suppress a specific low-order harmonic like the 5th, a virtual impedance tuned to be high at 250 Hz can be used, such as a virtual LC trap filter: $Z_v(s) = \frac{sL_v}{1 + s^2 L_v C_v}$. The parameters $L_v$ and $C_v$ are chosen so that the resonant frequency is $\omega_r = 1/\sqrt{L_v C_v} = 2\pi \cdot 250$ rad/s. This method provides targeted harmonic compensation but requires careful tuning to avoid affecting the fundamental component control stability.
2.2.3 Application of Intelligent Algorithms in Harmonic Optimization
The complex, non-linear dynamics of a grid-tied solar inverter during LVRT, coupled with varying grid impedances, make fixed-parameter controllers suboptimal. Intelligent algorithms like Particle Swarm Optimization (PSO) and Genetic Algorithms (GA) can be employed for online or offline tuning of control parameters to minimize a defined cost function, such as THD. In an online PSO implementation, a population of “particles” (each representing a set of controller parameters like $K_p$, $K_i$, $R_v$) moves through the parameter space. Each particle’s position $x_i(k)$ and velocity $v_i(k)$ at iteration $k$ are updated based on its own best experience ($pbest_i$) and the swarm’s global best ($gbest$):
$$
\begin{aligned}
v_i(k+1) &= \omega(k) \cdot v_i(k) + c_1 r_1 (pbest_i – x_i(k)) + c_2 r_2 (gbest – x_i(k)) \\
x_i(k+1) &= x_i(k) + v_i(k+1)
\end{aligned}
$$
Here, $\omega(k)$ is an inertia weight, $c_1$, $c_2$ are learning factors, and $r_1$, $r_2$ are random numbers. The fitness function $J$ for harmonic optimization could be:
$$
J = \alpha \cdot THD_i + \beta \cdot \int |e_i(t)| dt + \gamma \cdot \text{Penalty}(I_{peak})
$$
where $e_i(t)$ is the current tracking error, and the penalty term prevents overcurrent. This allows the solar inverter to autonomously adapt to changing grid conditions for optimal harmonic performance.
3. Practical Case Study: Implementation and Validation
3.1 Experimental Background and Problem Identification
A case study was conducted on a distributed PV power plant with a total capacity of 1 MW, comprised of multiple 50 kW solar inverters. The inverters were equipped with standard LVRT capabilities per relevant grid codes. Data logging during various operational scenarios revealed a significant challenge: while operating stably under normal conditions, the inverters exhibited a sharp increase in output current THD during grid voltage dips, particularly severe sags. The baseline performance data is summarized below:
| Parameter | Rated Operation | Mild LVRT (0.8 p.u.) | Severe LVRT (0.2 p.u.) |
|---|---|---|---|
| Output Active Power (kW) | 1000 | 800 | 200 |
| Output Reactive Power (kvar) | 0 | 100 | 300 |
| Current THD (%) | 2.5 | 5.2 | 8.7 |
The THD of 8.7% during a 0.2 p.u. voltage sag exceeded the typical grid code limit of 5%, identifying a clear need for an optimized harmonic control strategy.
3.2 Proposed Integrated Harmonic Optimization Control Strategy
To address the identified issue, a multi-layered control strategy was developed and implemented on the inverter’s digital signal processor (DSP). The strategy synergistically combines the techniques previously analyzed.
Step 1: Improved Randomized Modulation. An RPWM scheme with a fixed center frequency of $f_{s0}=10$ kHz and a randomized offset bounded by $\Delta f = \pm 2$ kHz was implemented. This provided the first layer of defense by dispersing high-frequency harmonic energy, reducing the risk of exciting high-frequency resonances with the grid.
Step 2: Targeted Virtual Impedance Compensation. A selective harmonic virtual impedance loop was added in parallel to the fundamental current controller. Multiple Second-Order Generalized Integrators (MSOGIs) were used as adaptive filters to extract the 5th and 7th harmonic components ($i_{5h}$, $i_{7h}$) from the measured output currents. The dynamics of an SOGI for a frequency $\omega’$ are:
$$
\begin{aligned}
\varepsilon &= i – i’ \\
\frac{d}{dt}\begin{bmatrix} i’ \\ q i’ \end{bmatrix} &= \begin{bmatrix} 0 & \omega’ \\ -\omega’ & 0 \end{bmatrix} \begin{bmatrix} i’ \\ q i’ \end{bmatrix} + \begin{bmatrix} k \omega’ \\ 0 \end{bmatrix} \varepsilon
\end{aligned}
$$
where $i’$ is the extracted harmonic component, and $q i’$ is its quadrature signal. Virtual resistor actions were then applied specifically to these extracted harmonics: $V_{comp,5} = -R_{v5} \cdot i_{5h}$ and $V_{comp,7} = -R_{v7} \cdot i_{7h}$. The compensation voltages were added to the modulation signals.
Step 3: Online Parameter Optimization via PSO. A lightweight PSO algorithm was embedded to run periodically or triggered by a significant change in grid voltage. The particle dimension included key control parameters: the PI controller gains for the current loop ($K_p$, $K_i$), and the virtual resistor values for the 5th and 7th harmonics ($R_{v5}$, $R_{v7}$). The fitness function $J$ was designed as:
$$
J = w_1 \cdot THD_{estimated} + w_2 \cdot \sigma_I + w_3 \cdot \max(0, I_{peak} – I_{lim})
$$
where $\sigma_I$ is the standard deviation of current error (measuring tracking performance), and the last term penalizes peak current violations. A swarm size of 40 particles was used with coefficients $c_1=c_2=1.5$, and inertia $\omega$ decaying from 0.8 to 0.3 over 20 iterations.
The implementation flow ensured that the RPWM provided broad-spectrum smoothing, the virtual impedance offered precise, targeted damping of the most problematic low-order harmonics, and the PSO adapted all parameters in real-time to the specific LVRT event and grid conditions.
3.3 Results Analysis and Empirical Insights
The proposed strategy was deployed and tested under controlled LVRT conditions. The results demonstrated a marked improvement in the harmonic performance of the solar inverters. The key outcome is presented in the following comparison table, showing the optimized parameters and the resulting performance metric.
| Parameter | Baseline Value | PSO-Optimized Value | Performance Change |
|---|---|---|---|
| Current Loop $K_p$ | 8.5 | 12.7 | +49.4% |
| Current Loop $K_i$ | 420 | 683 | +62.6% |
| Virtual Resistor $R_{v5}$ (p.u.) | 0.05 | 0.28 | +460% |
| Current THD @ 0.2 p.u. Sag | 8.7% | 4.6% | -47.1% |
The reduction of THD from 8.7% to 4.6% is significant, bringing the inverter well within the standard compliance limit. Spectral analysis confirmed a drastic reduction in the 5th and 7th harmonic amplitudes, by approximately 62% and 58% respectively, with minimal impact on the fundamental component. The dynamic response time, measured as the settling time for the current waveform after the voltage dip, was also improved to within one grid cycle.
The case study offers several crucial insights for the engineering of robust solar inverters:
- Layered Defense is Effective: Combining a modulation-level strategy (RPWM) with a control-loop strategy (virtual impedance) addresses both broad-spectrum and targeted harmonic issues.
- Adaptability is Key: Fixed parameters are insufficient for the wide range of grid impedances and fault conditions encountered in practice. Online optimization algorithms like PSO provide the necessary adaptability to maintain optimal performance across diverse LVRT scenarios.
- Selective Compensation is Efficient: Extracting and compensating only for the dominant low-order harmonics (5th, 7th) reduces computational burden compared to full-spectrum harmonic elimination methods, making it more feasible for real-time implementation on standard inverter hardware.
The principles validated in this centralized plant are transferable to other scenarios like commercial & industrial systems or weak rural grids. In weaker grids with higher impedance, the virtual impedance parameters would naturally be optimized by the PSO to larger values to provide more damping. Integration with energy storage systems would involve adding state-of-charge (SOC) management and power dispatch terms to the optimization fitness function.
4. Conclusion
This investigation has thoroughly examined the critical issue of output current harmonic control for grid-tied solar inverters during Low Voltage Ride-Through operations. The analysis traced harmonic origins to fundamental limitations in switching devices, PWM techniques, and control interactions with the disturbed grid. A review of conventional and advanced strategies highlighted their respective strengths and shortcomings, particularly the trade-off between reactive support and harmonic distortion.
The core contribution lies in proposing and validating an integrated optimization framework. By fusing an improved randomized PWM technique for spectral spreading, a virtual-impedance-based scheme for targeted harmonic damping, and an online PSO algorithm for autonomous parameter tuning, the strategy addresses the harmonic problem holistically. The empirical results from a 1 MW plant case study are compelling, demonstrating a reduction in current THD from 8.7% to 4.6% under severe voltage sags, thereby ensuring compliance with stringent grid codes.
The proposed method enhances the dynamic response and harmonic immunity of solar inverters, directly increasing their adaptability and reliability in complex, modern power grids with high renewable penetration. This work provides a feasible and effective technical pathway for improving the power quality contribution of PV systems during grid disturbances. Future research directions include extending the framework to parallel operations of multiple inverters to address harmonic resonance risks at the plant level, and integrating more advanced machine learning techniques for predictive and even more robust adaptive control.