The rapid global integration of renewable energy sources presents significant challenges to the stability and reliability of modern power grids. Among these sources, photovoltaic (PV) generation has seen exponential growth. However, the inherent intermittency and power-electronic interface of large-scale solar plants introduce new dynamics. A critical requirement, mandated by grid codes worldwide, is the ability of solar inverters to remain connected and provide support during grid faults, specifically during periods of severe voltage depression or even a complete loss of voltage—a capability known as Zero-Voltage Ride-Through (ZVRT). This article delves into the challenges of ZVRT under asymmetric grid faults and presents a detailed exposition of an advanced control strategy based on a Model Predictive Modulation Function (MPMF), contrasting it with conventional methods.
The core challenge during a grid fault, particularly an asymmetric one (e.g., a single-phase fault), is the injection of unbalanced currents. If unmitigated, this can lead to overcurrent protection tripping, causing the solar inverter to disconnect precisely when grid support is most needed. Furthermore, grid codes stipulate that during voltage dips, solar inverters must not only stay connected but also dynamically inject reactive current to aid in voltage recovery. The response time for this reactive current injection is typically required to be under 30 ms. Therefore, the control system for solar inverters must achieve multiple objectives simultaneously: fast and accurate current tracking, suppression of negative-sequence currents, adherence to reactive power mandates, and maintenance of current harmonics within acceptable limits.
The foundation of any high-performance control strategy is an accurate mathematical model of the system. For a three-phase, two-level voltage source inverter (VSI) used in solar applications, the dynamics in the stationary αβ-frame are given by:
$$ L \frac{d\mathbf{i}_{\alpha\beta}}{dt} = \mathbf{v}_{\alpha\beta} – \mathbf{e}_{\alpha\beta} – R \mathbf{i}_{\alpha\beta} $$
where \( \mathbf{i}_{\alpha\beta} = [i_\alpha, i_\beta]^T \) is the inverter output current vector, \( \mathbf{v}_{\alpha\beta} = [v_\alpha, v_\beta]^T \) is the inverter output voltage vector, \( \mathbf{e}_{\alpha\beta} = [e_\alpha, e_\beta]^T \) is the grid voltage vector, \( L \) is the filter inductance, and \( R \) is the parasitic resistance. Under normal, balanced conditions, transforming these equations into a synchronous rotating dq-frame simplifies control as AC quantities become DC. However, under asymmetric faults, the grid voltage contains both positive- and negative-sequence components:
$$ \mathbf{e}_{abc} = \mathbf{e}_{abc}^+ + \mathbf{e}_{abc}^- $$
Transforming these unbalanced components into a single dq-frame results in oscillatory signals at twice the grid frequency, rendering simple PI control ineffective.
The solution is to model the system in dual rotating frames: a positive-sequence dq+ frame rotating counter-clockwise and a negative-sequence dq- frame rotating clockwise. In these frames, the sequence components appear as DC quantities. The model for solar inverters in these frames is:
$$
\begin{aligned}
v_d^+ &= L \frac{di_d^+}{dt} + R i_d^+ – \omega L i_q^+ + e_d^+ \\
v_q^+ &= L \frac{di_q^+}{dt} + R i_q^+ + \omega L i_d^+ + e_q^+ \\
v_d^- &= L \frac{di_d^-}{dt} + R i_d^- + \omega L i_q^- + e_d^- \\
v_q^- &= L \frac{di_q^-}{dt} + R i_q^- – \omega L i_d^- + e_q^-
\end{aligned}
$$
While this representation is accurate, the cross-coupling terms (e.g., \( -\omega L i_q^+ \)) complicate direct control. Advanced strategies often calculate current references in these sequence frames but perform the fast control action in the stationary αβ-frame to avoid this complexity.
The current references for solar inverters during a voltage dip are dictated by grid codes. Typically, the reactive current reference \( i_{q}^{+,ref} \) is proportional to the voltage dip depth, while the active current reference \( i_{d}^{+,ref} \) is reduced to limit the total apparent power and prevent overcurrent. The negative-sequence current references \( i_{d}^{-,ref} \) and \( i_{q}^{-,ref} \) are usually set to zero to prevent unbalanced current injection. These sequence-domain references are then transformed back to the αβ-frame to obtain \( i_{\alpha}^{ref}(k) \) and \( i_{\beta}^{ref}(k) \).
Conventional approaches like dual-PI control in sequence frames or Proportional-Resonant (PR) control in the αβ-frame have limitations. PI controllers require careful decoupling and tuning, and their response can be sluggish. PR controllers are sensitive to frequency variations. Finite Control Set Model Predictive Control (FCS-MPC) has emerged as a powerful alternative for solar inverters due to its fast dynamic response and intuitive handling of constraints. However, FCS-MPC applies only one voltage vector per control period, leading to high harmonic distortion, and requires significant computational effort for evaluating all possible switching states.
The proposed Model Predictive Modulation Function (MPMF) control for solar inverters seeks to combine the benefits of predictive control—fast dynamics—with the superior harmonic performance of continuous modulation schemes like SPWM. Instead of predicting discrete switch states, it predicts the continuous modulation function \( \mathbf{m}_{\alpha\beta} = [m_\alpha, m_\beta]^T \), which is the duty cycle command for the SPWM modulator, defined as \( \mathbf{v}_{\alpha\beta} = \mathbf{m}_{\alpha\beta} \cdot V_{dc} \).
The first step is to discretize the plant model for prediction. Using a forward-Euler approximation, the current at the next sampling instant \( k+1 \) is predicted from measurements at instant \( k \):
$$
\mathbf{i}_{\alpha\beta}(k+1) = \mathbf{i}_{\alpha\beta}(k) + \frac{T_s}{L} \left( \mathbf{m}_{\alpha\beta}(k) V_{dc} – \mathbf{e}_{\alpha\beta}(k) – R \mathbf{i}_{\alpha\beta}(k) \right)
$$
where \( T_s \) is the sampling period. To compensate for computational delay, a two-step prediction is used. The modulation function \( \mathbf{m}_{\alpha\beta}(k+1) \) computed at time \( k \) is applied during the interval \( [k+1, k+2] \). Therefore, we predict the current at \( k+2 \):
$$
\mathbf{i}_{\alpha\beta}(k+2) = \mathbf{i}_{\alpha\beta}(k+1) + \frac{T_s}{L} \left( \mathbf{m}_{\alpha\beta}(k+1) V_{dc} – \mathbf{e}_{\alpha\beta}(k+1) – R \mathbf{i}_{\alpha\beta}(k+1) \right)
$$
The grid voltage at \( k+1 \) can be estimated using a simple rotation:
$$ \mathbf{e}_{\alpha\beta}(k+1) = \begin{bmatrix} \cos(\omega T_s) & -\sin(\omega T_s) \\ \sin(\omega T_s) & \cos(\omega T_s) \end{bmatrix} \mathbf{e}_{\alpha\beta}(k) $$
The control objective for the solar inverter is to minimize the tracking error. Thus, a cost function \( J \) is defined as the sum of squared errors between the predicted currents and their references:
$$ J = \left[ i_\alpha(k+2) – i_\alpha^{ref}(k+2) \right]^2 + \left[ i_\beta(k+2) – i_\beta^{ref}(k+2) \right]^2 $$
The optimal modulation function \( \mathbf{m}_{\alpha\beta}(k+1) \) is the one that minimizes this cost function. Unlike FCS-MPC, which performs an exhaustive search, MPMF finds the optimum analytically by taking the derivative of \( J \) with respect to \( \mathbf{m}_{\alpha\beta}(k+1) \) and setting it to zero:
$$ \frac{\partial J}{\partial m_\alpha(k+1)} = 0, \quad \frac{\partial J}{\partial m_\beta(k+1)} = 0 $$
Solving these equations yields the optimal modulation functions that will minimize the future current error for the solar inverter:
$$
\begin{aligned}
m_\alpha(k+1) &= \frac{L}{V_{dc}T_s} i_\alpha^{ref}(k+2) – \frac{L}{V_{dc}T_s} \left[ i_\alpha(k+1) + \frac{T_s}{L}\left( – R i_\alpha(k+1) – e_\alpha(k+1) \right) \right] \\
m_\beta(k+1) &= \frac{L}{V_{dc}T_s} i_\beta^{ref}(k+2) – \frac{L}{V_{dc}T_s} \left[ i_\beta(k+1) + \frac{T_s}{L}\left( – R i_\beta(k+1) – e_\beta(k+1) \right) \right]
\end{aligned}
$$
These \( m_\alpha \) and \( m_\beta \) values are then fed to a standard SPWM block to generate the gate signals for the solar inverter. This approach maintains a fixed switching frequency, leading to predictable and lower harmonic distortion compared to FCS-MPC, while retaining the fast, model-based response of predictive control.
To validate the performance of the MPMF strategy for solar inverters, a detailed simulation study is conducted and compared against traditional PI-based control and FCS-MPC. The system parameters for a representative medium-scale solar inverter are summarized below:
| Parameter | Value |
|---|---|
| Rated Grid Voltage | 35 kV |
| Grid Frequency | 50 Hz |
| Filter Inductance (L) | 0.12 mH |
| DC-Link Voltage (V_dc) | 480 V |
| Switching Frequency | 2500 Hz |
| Rated Power | 250 kW |
| Rated Current (I_N) | 6 A |
| Overcurrent Protection | 7.2 A |
| Sampling Frequency | 10 kHz |
The scenario involves a severe asymmetric fault: a single-phase-to-ground fault where the voltage on one phase drops to zero, lasting for 150 ms as per ZVRT requirements. The solar inverter must stay connected. Under the MPMF control, the inverter currents remain balanced and are successfully limited below the overcurrent threshold. More importantly, the dynamic reactive current injection meets the grid code specification, tracking its reference within approximately 15 ms. For a symmetric three-phase fault to zero voltage, the performance is equally robust, with balanced currents and fast reactive support.
A direct comparison of the dynamic response highlights the advantages of the predictive approach for solar inverters. The following table summarizes the key performance indicators for the three control methods during the asymmetric ZVRT event:
| Performance Metric | PI Control | FCS-MPC | MPMF Control |
|---|---|---|---|
| Reactive Current Settling Time | ~25 ms | ~18 ms | ~15 ms |
| Current THD (Steady-State during fault) | 9.07% | 4.86% | 1.56% |
| Current Overshoot | Significant | Moderate | Minimal |
| Computational Load | Low | Very High | Moderate |
The superiority of the MPMF strategy for solar inverters is evident. It achieves the fastest dynamic response, crucial for meeting strict grid code deadlines. Furthermore, by utilizing a continuous modulation scheme, it achieves a significantly lower current Total Harmonic Distortion (THD) compared to both PI control (which suffers from poor tracking of oscillatory references) and FCS-MPC (which applies discrete voltage vectors). The THD of 1.56% is well within typical harmonic standards for grid-connected equipment. While computationally more intensive than PI control, the MPMF method avoids the exhaustive search of FCS-MPC, making it a practical and efficient choice for modern digital signal processors used in solar inverters.
In conclusion, the Model Predictive Modulation Function control presents a highly effective solution for achieving robust Zero-Voltage Ride-Through in solar inverters, particularly under challenging asymmetric grid faults. By merging the principles of model predictive control with the benefits of fixed-frequency pulse-width modulation, it addresses the core limitations of prior methods. For solar inverters, this translates to guaranteed grid code compliance through ultra-fast reactive current injection, superior power quality with low harmonic distortion, and maintained system stability during the most severe voltage dips. As grid penetration of photovoltaic systems continues to rise, advanced control strategies like MPMF will be indispensable for ensuring these solar inverters act as reliable and supportive assets to the power grid, enhancing overall resilience and facilitating the transition to a sustainable energy future.