In recent years, the rapid growth of the global economy has been accompanied by increasingly prominent environmental issues. To meet the escalating demand for electricity while reducing environmental pollution, distributed renewable energy generation technologies have emerged. In the field of renewable energy generation, the grid-connected inverter is regarded as a core device, and its performance directly determines the quality of grid-connected power. Therefore, researching efficient control strategies for grid-connected inverters holds significant practical importance. Simultaneously, the topological structure of inverters has attracted considerable attention. Compared to traditional two-level inverters, the diode Neutral Point Clamped (NPC) three-level grid-connected inverter offers advantages such as lower voltage stress on power switching devices, reduced output current ripple, and lower switching losses, making it widely used in high-power grid-connected inverters.
As inverter technology advances, their topological structures have become increasingly complex. The unique topology of the NPC three-level inverter imposes higher demands on control strategies, and traditional control methods are now often inadequate to meet the system’s control requirements efficiently. In this context, Model Predictive Control (MPC) has emerged as an ideal choice for controlling NPC three-level inverters due to its straightforward implementation, rapid response capability, and excellent performance in multivariable joint control. Among these, Finite Control Set Model Predictive Control (FCS-MPC) strategy has further propelled development in this field. FCS-MPC is characterized by its lack of a modulation unit, multi-objective function optimization, fast response, and strong robustness, making it effective in handling nonlinear problems in grid-connected inverter control. Consequently, this method is widely applied in grid-connected inverter control. However, the FCS-MPC method requires online traversal optimization, necessitating the calculation of all possible switching state combinations to output voltages, with the goal of minimizing system parameter errors through a value function. This results in a substantial computational load for the algorithm. Given that three-level inverters have a large number of switching vectors, employing this method not only increases the computational burden on the controller but also leads to significant switching losses. This limits the application of FCS-MPC in multilevel inverter domains.
To address the issues of high computational load and redundant switching behavior in FCS-MPC systems, numerous improved methods have been proposed by scholars worldwide. Some studies have introduced model-free predictive control algorithms to minimize switching losses in permanent magnet synchronous motor systems. By refining the value function, optimal voltage sequences are obtained, avoiding switching actions when phase currents are at their maximum. Although effective, this method is relatively complex and does not resolve the issue of high computational load. To reduce computational load, a simplified FCS-MPC control strategy has been proposed, which divides the space vector into hexagonal sectors and defines the sector containing the reference voltage as the optimal switching subset, thereby reducing calculations. However, this method omits optimal voltage vectors, potentially affecting system control performance. Other research has partitioned voltage vectors in FCS-MPC, determining the partition of the reference voltage vector to eliminate the need for predictive judgment of 27 voltage vectors, thus reducing computational load and achieving fast dynamic response, but at the cost of increased switching losses. To reduce switching losses, a control strategy based on a constrained value function has been proposed, effectively lowering switching losses and improving system response performance by introducing a switching frequency penalty term. However, this approach faces challenges in weighting factor allocation. In other works, an event-triggered mechanism has been integrated into active disturbance rejection model predictive control. By triggering control actions at critical moments, adaptive system regulation and disturbance suppression are achieved, reducing computational load and improving control effectiveness without the complexity of weighting factor allocation. Successful validation of event-triggered model predictive control in two-level inverters has demonstrated reduced switching losses, but research on three-level inverters remains insufficient. Given the widespread use of three-level inverters in high-voltage applications, particularly their switching loss issues, studying the applicability of event-triggered model predictive control in such topologies is of great significance.
To address the problems of high computational load and losses in NPC three-level FCS-MPC, this paper proposes an Event-Triggered Simplified Model Predictive Control (ET-S-MPC) strategy. This strategy integrates deadbeat control with FCS-MPC, simplifying the model prediction and value function optimization processes to reduce computational burden. An LCL-type filter is introduced to enhance stability, and a “two-step prediction” method is used to compensate for delays in FCS-MPC, with a vector angle compensation method applied to correct reference values. Experimental results validate the correctness and effectiveness of the theoretical analysis.
Mathematical Model of the Grid-Connected Inverter
The topology of the NPC three-level LCL-type inverter is illustrated in the figure above. The DC-side voltage of this inverter is $V_{dc}$, with two capacitors $C_1$ and $C_2$. The neutral point current on the DC side is $i_m$. The inductances on the inverter side and grid side are $L_1$ and $L_2$, respectively, each with parasitic resistances $R_1$ and $R_2$. The three-phase currents on the inverter side are $i_{1a}$, $i_{1b}$, and $i_{1c}$, and on the grid side are $i_{2a}$, $i_{2b}$, and $i_{2c}$. The capacitance of the LCL filter is $C$, and the three-phase grid voltages are $e_a$, $e_b$, and $e_c$. Points $M$ and $N$ are the neutral points of the DC side and grid side, respectively.
A switching function is defined as:
$$S_x = \begin{cases}
1 & \text{for state P (S_{x1}, S_{x2} on; S_{x3}, S_{x4} off)} \\
0 & \text{for state O (S_{x2}, S_{x3} on; S_{x1}, S_{x4} off)} \\
-1 & \text{for state N (S_{x3}, S_{x4} on; S_{x1}, S_{x2} off)}
\end{cases}$$
where $x = a, b, c$.
For this topology, the terminal voltage output from the DC side is obtained by combining different switching states. Thus, the voltage from the inverter AC output side (phases a, b, c) relative to the DC-side neutral point ($M$) can be expressed as:
$$u_{xM} = S_x \frac{V_{dc}}{2}, \quad x = a, b, c$$
In the NPC three-level topology, each bridge arm has three states: P, O, and N. With three bridge arms, there are $3^3 = 27$ possible states. However, due to six redundant vectors, there are actually 19 distinct switching states, distributed in the voltage space vector diagram.
Based on the grid-connected inverter topology and Kirchhoff’s laws, the voltage balance equations for the inverter side and grid side are derived. After Clark transformation, the mathematical model in the two-phase stationary coordinate system is obtained:
$$L_1 \frac{di_{1\alpha\beta}}{dt} = u_{C\alpha\beta} – u_{\alpha\beta} – R_1 i_{1\alpha\beta}$$
$$L_2 \frac{di_{2\alpha\beta}}{dt} = e_{\alpha\beta} – u_{C\alpha\beta} – R_2 i_{2\alpha\beta}$$
$$C \frac{du_{C\alpha\beta}}{dt} = i_{2\alpha\beta} – i_{1\alpha\beta}$$
where $i_{1\alpha\beta}$ is the complex vector of inverter-side current, $i_{2\alpha\beta}$ is the complex vector of grid-side current, $u_{\alpha\beta}$ is the complex vector of inverter output voltage, $e_{\alpha\beta}$ is the complex vector of grid voltage, and $u_{C\alpha\beta}$ is the complex vector of filter capacitor voltage in the $\alpha\beta$ coordinate system.
Using the forward difference formula, discretize the above equations to obtain the circuit state equation at discrete time step $k$:
$$i_{1\alpha}(k+1) = i_{1\alpha}(k) + \frac{T_s}{L_1} [u_{C\alpha}(k) – u_{\alpha}(k) – i_{1\alpha}(k) R_1]$$
$$i_{1\beta}(k+1) = i_{1\beta}(k) + \frac{T_s}{L_1} [u_{C\beta}(k) – u_{\beta}(k) – i_{1\beta}(k) R_1]$$
where $i_{1\alpha}(k+1)$ and $i_{1\beta}(k+1)$ are the predicted inverter-side current vectors at time $k+1$, $i_{1\alpha}(k)$ and $i_{1\beta}(k)$ are the current inverter-side current vectors, $u_{C\alpha}(k)$ and $u_{C\beta}(k)$ are the current filter capacitor voltage vectors, $u_{\alpha}(k)$ and $u_{\beta}(k)$ are the current grid voltage vectors, and $T_s$ is the sampling period.
In NPC three-level grid-connected inverters, the two split capacitors on the DC side cause neutral point potential fluctuations during charging and discharging. This imbalance can degrade the output power quality of the inverter, reduce the lifespan of power switches, and affect system grid connection. Therefore, controlling the neutral point potential is a critical aspect. Let $u_{C1} – u_{C2} = \Delta u$, introduce the switching vector $S_x (x = a, b, c)$, and use forward difference to write:
$$\Delta u(k+1) = \Delta u(k) + \frac{T_s}{C} \sum i_{1x} |S_x(k)|$$
where $\Delta u(k+1)$ is the feedback value of the DC-side neutral point voltage at time $k+1$.
Active Damping Strategy for LCL-Type Grid-Connected Inverters
To reduce harmonics in the grid current, high-frequency filters are introduced to mitigate high-frequency switching harmonics in the inverter output voltage. The LCL filter is superior to the L-type filter, as it uses a Second Order Generalized Integrator (SOGI) to extract the fundamental frequency component of the capacitor voltage. The high-frequency portion of the inverter output voltage is extracted, converted into high-frequency current for control, optimizing MPC feedback control of the capacitor voltage to suppress harmonic peaks.
Since the resistances $R_1$ and $R_2$ are small and negligible, the active damping control adds a virtual resistor to the system control block diagram. The final reference current for the inverter side is obtained by adding the high-frequency component current to the inverter-side current reference value, ensuring balanced current control performance and active damper performance to meet system requirements.
In the model predictive control algorithm, Park inverse transform is applied to obtain the final reference currents $i_{ref11\alpha}$ and $i_{ref11\beta}$ in the $\alpha\beta$ coordinate system.
Simplified FCS-MPC Strategy
In NPC three-level FCS-MPC, equations are computed 27 times per control cycle, requiring substantial calculations. To improve efficiency, a simplified FCS-MPC is adopted, drawing on deadbeat ideas, using single prediction and vector partitioning to simplify model prediction and value function calculations. After incorporating active damping control, the influence of the active damper on model predictive current control must be considered as an additional input term. Using deadbeat control principles, combined with the mathematical model, we derive:
$$i_{ref11\alpha}(k+1) = i_{1\alpha}(k) + \frac{T_s}{L_1} [u_{C\alpha}(k) – u_{\alpha}(k) – i_{1\alpha}(k) R_1]$$
$$i_{ref11\beta}(k+1) = i_{1\beta}(k) + \frac{T_s}{L_1} [u_{C\beta}(k) – u_{\beta}(k) – i_{1\beta}(k) R_1]$$
To overcome delay issues in FCS-MPC, a two-step predictive control method is employed for compensation. Specifically, the inverter output current is split into secondary prediction terms and linear increment terms, and interpolation is used to predict the inverter-side reference current value $i_{ref11\alpha\beta}(k+2)$ at time $k+2$. Thus, combining the mathematical model, the inverter output current and neutral point voltage in the system can be expressed as:
$$i_{ref11\alpha}(k+2) = i_{1\alpha}(k+1) + \frac{T_s}{L_1} [u_{C\alpha}(k+1) – u_{\alpha}(k+1) – i_{1\alpha}(k+1) R_1]$$
$$i_{ref11\beta}(k+2) = i_{1\beta}(k+1) + \frac{T_s}{L_1} [u_{C\beta}(k+1) – u_{\beta}(k+1) – i_{1\beta}(k+1) R_1]$$
$$\Delta u(k+2) = \Delta u(k+1) + \frac{T_s}{C} \sum i_{1x} |S_x(k+1)|$$
To avoid output level imbalance and high grid harmonics, the DC-side neutral point potential must be balanced. Combining the equations, the value function $g$ is:
$$g = |i_{ref11\alpha}(k+2) – i_{1\alpha}(k+2)|^2 + |i_{ref11\beta}(k+2) – i_{1\beta}(k+2)|^2 + \lambda \Delta u(k+2)^2$$
where $\lambda$ is the weighting coefficient for DC-side neutral point voltage balance, which can be obtained through trial and error.
In simplified FCS-MPC control, sector division requires calculating the reference voltage vector. Solving and rearranging the equations yields:
$$V_{ref\alpha}(k+1) = u_{C\alpha}(k+1) – i_{1\alpha}(k+1) R_1 – \frac{L_1}{T_s} [i_{ref11\alpha}(k+2) – i_{1\alpha}(k+1)]$$
$$V_{ref\beta}(k+1) = u_{C\beta}(k+1) – i_{1\beta}(k+1) R_1 – \frac{L_1}{T_s} [i_{ref11\beta}(k+2) – i_{1\beta}(k+1)]$$
In the $\alpha\beta$ coordinate system, sectors are divided by lines $V_\alpha = 0$, $V_\beta = \sqrt{3} V_\alpha / 3$, and $V_\beta = -\sqrt{3} V_\alpha / 3$. Let the angle between the reference voltage vector $V_{ref}$ and the $\alpha$-axis be $\theta$. Define the small quadrilateral sector number $M$ in relation to $\theta$:
$$M = \text{ceil}\left(\frac{\theta + 30^\circ}{60^\circ}\right)$$
where $\theta \in [0^\circ, 360^\circ)$ and $\text{ceil}(n)$ rounds $n$ up to the nearest integer.
Based on the calculated $M$ value, the sector and pre-selected voltage vectors $V_{ref}$ are determined, as summarized in the table below.
| M | Sector | Pre-selected Vector Set |
|---|---|---|
| 1 | I | $V_0, V_1, V_7, V_{12}, V_{13}$ |
| 2 | II | $V_0, V_2, V_7, V_8, V_{14}$ |
| 3 | III | $V_0, V_3, V_8, V_9, V_{15}$ |
| 4 | IV | $V_0, V_4, V_9, V_{10}, V_{16}$ |
| 5 | V | $V_0, V_5, V_{10}, V_{11}, V_{17}$ |
| 6 | VI | $V_0, V_6, V_{11}, V_{12}, V_{19}$ |
After determining the reference voltage $V_{ref}$, the system identifies its small quadrilateral sector and selects eight basic voltage vectors. These vectors are then substituted into the delay-compensated value function for optimization to choose the switching state that minimizes the value function while outputting voltage.
Derivation and Determination of Event-Triggered Conditions
To derive the triggering conditions, rewrite the state equation as:
$$\dot{x} = f(x, u) = A x + B u$$
Define $u = [u_{C\alpha} – u_{\alpha}, u_{C\beta} – u_{\beta}]^T$, $x = [i_{1\alpha}, i_{1\beta}]^T$, $A = \begin{bmatrix} 1 – R_1/L_1 & 0 \\ 0 & 1 – R_1/L_1 \end{bmatrix}$, and $B = \begin{bmatrix} 1/L_1 & 0 \\ 0 & 1/L_1 \end{bmatrix}$.
Define the state error as:
$$e(t) = x(t_i) – x(t), \quad t \in [t_i, t_{i+1})$$
According to literature, when a closed-loop system is input-to-state stable, there exists a continuously differentiable function satisfying:
$$\frac{dV[x(t)]}{dx(t)} f[x(t), u(t)] \leq -\alpha(\|x(t)\|) + \gamma(\|e(t)\|)$$
where $\alpha$ and $\gamma$ are $K_\infty$ functions, and $V$ is a composite voltage vector.
Since static event triggering is related to system state and sampling error, consider the inequality relation from literature. The condition can be set as:
$$\|e(t)\| \leq \gamma^{-1}(\sigma \alpha(\|x(t)\|))$$
where $\sigma$ is an adjustment coefficient, with $\sigma \in [0,1)$.
By defining the state error and combining it with the state-space model, derive:
$$\frac{d\|e(t)\|}{dt} \leq \left\| \frac{de(t)}{dt} \right\| \leq \|A x(t)\| + \|B\| \cdot \|u(t)\|$$
To ensure satisfactory system performance, the relationship between grid line voltage $U_y$ and DC-side voltage $U_{dc}$ can be defined as:
$$U_{dc} \simeq \sqrt{3} U_y, \quad y = ab, bc, ca$$
Using equal amplitude transformation and combining equations, obtain $\|u(t)\|$ as:
$$\|u(t)\| = |u_1| + |u_2| \leq |u_1|_{\text{max}} + |u_2|_{\text{max}} = \frac{(8 – 6) U_{dc}}{6} \triangleq \delta$$
Define:
$$\phi(t) = \|e(t)\| + \frac{\|A x(t_i)\| + \delta \|B\|}{\|A\|}$$
Substitute into the inequality to further obtain:
$$\phi(t) \leq \|A\| \phi(t_i), \quad t \in [t_i, t_{i+1})$$
where $\phi(t_i) = \frac{\|A x(t_i)\| + \delta \|B\|}{\|A\|}$. Replace $\phi(t)$ to get:
$$\|e(t)\| \leq \frac{\|A x(t_i)\| + \delta \|B\|}{\|A\|} (\exp(\|A\|(t – t_i)) – 1)$$
Substitute $t_{i+1} = t_i + n t_{ET}$ (where $n \in \mathbb{N}^+$, $t_{ET}$ is the event time) into the equation. Thus, when trigger time $t_{i+1}$ arrives:
$$\|e(t_{i+1})\| \leq \frac{\|A x(t_i)\| + \delta \|B\|}{\|A\|} (\exp(\|A\| n t_{ET}) – 1)$$
To set the triggering condition more strictly, ignore $\|A x(t_i)\|$ in the equation and set $n = 1$. To make the number of switching actions adjustable in practical operation, introduce a variable coefficient $K_{ET}$:
$$\|e(t_{i+1})\| \leq K_{ET} \delta \|B\| \cdot \|A\|^{-1} (\exp(\|A\| t_{ET}) – 1)$$
When the state error exceeds the triggering condition, event-triggered simplified model predictive control will be executed.
ET-S-MPC Strategy
The algorithm flow for event-triggered FCS-MPC in NPC three-level LCL-type grid-connected inverters is illustrated in the flowchart. In the simplified FCS-MPC method, due to the randomness of voltage vector selection, there are often simultaneous commutations of multiple bridge arm switching devices. In practical applications, high-frequency switching actions typically result in higher switching losses, necessitating a reduction in these losses. Cost function optimized MPC reduces switching times by adding a penalty term for switching frequency in the cost function. However, this introduces challenges in weighting coefficient allocation. Therefore, event-triggered control is integrated into FCS-MPC to reduce redundant switching actions, thereby lowering switching frequency.
The event-triggered mechanism is fused with the simplified FCS-MPC strategy. By setting a clear boundary threshold to control error, once this threshold is exceeded, a control event is immediately triggered. This mechanism is applied to the control process of NPC three-level LCL-type grid-connected inverters, forming the ET-S-MPC strategy structure.
Experimental Validation and Analysis
To validate the effectiveness of the ET-S-MPC method, an RCP2000 rapid prototype is used as the controller to build a hardware-in-the-loop experimental setup for the NPC three-level grid-connected inverter, with a TMS320F28335 from TI serving as the成型 controller. Through an upper computer, the control algorithm debugged via a Matlab simulation model is directly downloaded into the rapid prototype controller. Grid connection experiments are conducted at the grid access point in the microgrid laboratory, equipped with circuit breakers and contactors to ensure grid safety. The parameters of the experimental system are listed in the table below, and control parameters for the three methods are also provided.
| Parameter | Value |
|---|---|
| DC-side voltage $V_{dc}$ | 600 V |
| DC-side capacitors ($C_1 = C_2$) | 4700 μF |
| Grid voltage RMS $e_n$ | 220 V |
| Grid fundamental frequency $f$ | 50 Hz |
| Inverter-side filter inductance $L_1$ | 3.3 mH |
| LCL filter capacitance $C_f$ | 9.4 μF |
| Grid-side filter inductance $L_2$ | 100 μH |
| Parasitic resistance of $L_1$, $R_1$ | 0.05 Ω |
| Event-triggered time $t_{ET}$ | 50 μs |
| Variable coefficient $K_{ET}$ | 0–1 |
| Inverter sampling period $T_s$ | 50 μs |
| Control Method | Parameter | Value |
|---|---|---|
| PI-SVPWM | Voltage outer loop proportional coefficient $K_{Vp}$ | 0.335 |
| Voltage outer loop integral coefficient $K_{Vi}$ | 20 | |
| Current inner loop proportional coefficient $K_{Ip}$ | 0.005 | |
| Current inner loop integral coefficient $K_{Ii}$ | 5 | |
| Cost Function Optimized MPC | Neutral point voltage balance weight $\lambda_1$ | 0.10 |
| Switching frequency limit weight $\lambda_2$ | 0.6 | |
| ET-S-MPC | Event-triggered variable coefficient $K_{ET}$ | 0.6 |
Steady-State Performance Comparison
The steady-state performance of three different control methods is compared. Current waveforms are sampled by an oscilloscope, saved as CSV files, and then imported into Matlab’s Workspace. Subsequently, using Simulink’s From Workspace and Scope modules, these data are converted into current waveforms, and Fast Fourier Transform analysis is performed to obtain spectral information and calculate Total Harmonic Distortion (THD).
The results show that under Proportional-Integral Space Vector Pulse Width Modulation (PI-SVPWM) control, the grid-connected current meets basic grid standards but has high harmonic content, with THD approximately 4.33%, indicating that this control strategy has shortcomings in reducing harmonics and improving current control accuracy, requiring further optimization.
Under cost function optimized MPC control, the grid-connected current waveform shows significant improvement, but the weighting coefficients for neutral point balance and switching action limitation are difficult to determine optimally, resulting in THD around 3.96%.
Under ET-S-MPC control, current ripple content is significantly reduced, and the system’s grid-connected current THD is only about 3.03%. This indicates that the ET-S-MPC control strategy is more effective in reducing harmonics and improving current quality.
To compare the computational time of cost function optimized MPC and ET-S-MPC algorithms, the proportion of control program calculation time in the entire sampling period is obtained by monitoring GPIO port state changes. Results show that the cost function optimized MPC algorithm takes about 13.8 μs, while the ET-S-MPC algorithm requires only 8.6 μs. This indicates that the ET-S-MPC algorithm reduces computational burden while maintaining current control performance, offering higher real-time performance and computational efficiency.
Based on experimental results, the drive pulse signals for power switch $S_{a1}$ under the three control methods are compared, verifying that ET-S-MPC has longer inactive switching times, significantly reducing current ripple and switching frequency.
The neutral point potential balance control in NPC three-level inverters is demonstrated. By controlling the inverter’s switching states and PWM modulation techniques, regulation of each level voltage can be achieved. The voltages of the upper and lower capacitors on the DC side are $u_{C1}$ and $u_{C2}$, respectively, with $\Delta u$ being the capacitance difference obtained by subtracting them. The fluctuation range shows that $\Delta u$ varies within ±5 V, proving the effectiveness of neutral point potential control.
Experimental results show the waveforms of AC-side phase voltage $u_a$ and output current $i_a$ for the grid-connected inverter under steady-state conditions. Observing the phase relationship between $u_a$ and $i_a$ at zero-crossing points, the power factor of the grid-connected inverter is approximately 0.99. This indicates that the grid-connected inverter operates with high power factor under steady-state conditions, effectively delivering power to the grid and synchronizing with it.
Dynamic Response Performance Comparison
To verify the dynamic response effects of cost function optimized MPC and ET-S-MPC strategies, the same parameter settings as in steady-state simulation are used. The sampling frequency is set to 50 kHz, and the grid current reference value is stepped from 10 A to 13 A at 0.4 s. The response effects of grid-side current under the two strategies are shown.
Results show that traditional MPC strategy has good transient response effects. After selecting an appropriate event-triggered variable coefficient, ET-S-MPC responds faster to current reference changes without generating additional inrush current during the response process, exhibiting superior dynamic response.
Robustness Performance Comparison
To verify the impact of the event-triggered mechanism on MPC control effectiveness, first, the switching actions of traditional FCS-MPC are limited through trial and error to obtain an optimal weighting coefficient of 0.95. Then, with weighting coefficient and sampling period as independent variables, and THD and switching count as dependent variables, a three-dimensional coordinate graph is created in Matlab to visually present the impact of the event-triggered mechanism on model predictive control effectiveness.
Experimental results show that the adjustment coefficient of the event-triggered mechanism can be stably maintained within an acceptable range through appropriate parameter settings, unaffected by significant deviations from changes in sampling period. From the three-dimensional coordinates, it is evident that improving sampling accuracy increases switching count. To reduce switching frequency and grid-connected current THD, MPC parameters are adjusted each sampling period. However, multiple simulations show that limiting the weighting coefficient to the range (0,1] (with sampling period 40–50 μs) maintains system stability, achieving the required control accuracy and switching frequency. When the weighting coefficient is small, current changes slowly, making it difficult to meet accuracy; when too large, switching becomes frequent, easily inducing grid current harmonics and causing system instability.
From the graphs, shortening the sampling period increases switching frequency. However, through the event-triggered mechanism and adjusting the event-triggered coefficient, ET-S-MPC achieves good frequency reduction effects. Within the sampling accuracy range of 40–50 μs, the system adjustment coefficient remains in the range (0.5,1], maintaining stability.
By comparing the adjustment ranges of the two control strategies, it is found that the number of switching state transitions in the cost function optimized MPC method is affected by changes in sampling period, with weighting coefficient and grid-connected current THD also varying. In contrast, in event-triggered control, the event-triggered period aligns with the sampling period. When the sampling period changes, both the event-triggered boundary conditions and error function change, creating a dual effect, resulting in its coefficient’s effective adjustment range being unaffected by significant changes in sampling period.
Conclusion
Addressing the issues of high computational load and switching losses in LCL-type NPC three-level grid-connected inverters using FCS-MPC, the effectiveness of a simplified model predictive control strategy based on an event-triggered mechanism is proposed and validated. Research results show that ET-S-MPC reduces computational requirements and switching losses, enhancing system efficiency. Additionally, while optimizing current tracking performance, it improves grid-connected current quality. Moreover, this strategy responds rapidly to disturbances, exhibits excellent dynamic performance, and simplifies weighting coefficient settings through the event-triggered mechanism, ensuring the stability of the control strategy across various sampling periods. This approach is particularly beneficial for grid-connected inverters in renewable energy systems, where efficiency and reliability are paramount. The integration of event-triggered mechanisms with model predictive control represents a significant advancement in the control of high-power grid-connected inverters, contributing to the broader adoption of sustainable energy solutions. Future work may focus on extending this strategy to other multilevel inverter topologies and exploring its applicability in larger-scale grid-connected systems.