Investigation on GaN HEMTs Based Three-Phase STATCOM with Hybrid Control Scheme

The modern trend of decarbonization has encouraged intensive research on renewable energy (RE)-based distributed power generation (DG) and smart grid, where advanced electronic power interfaces are necessary for connecting the generator with power grids and various electrical systems. On the other hand, modern technologies such as Industry 4.0 and electrical vehicles (EV) have higher requirements for power converters than that of conventional applications. Consequently, the enhancement of power interfaces will play an important role in the future power generation and distribution systems as well as various industrial applications. It has been discovered that wide-bandgap (WBG) switching devices such as gallium nitride (GaN) high electron mobility transistors (HEMTs) and silicon carbide (SiC) metal-oxide-semiconductor field-effect transistors (MOSFETs) offer considerable potential for outperforming conventional silicon (Si) switching devices in terms of breakdown voltage, high temperature capability, switching speed, and conduction losses. This paper investigates the performance of a 2kVA three-phase static synchronous compensator (STATCOM) based on a GaN HEMTs-based voltage-source inverter (VSI) and a neural network-based hybrid control scheme. The proportional-integral (PI) controllers along with a radial basis function neural network (RBFNN) controller for fast reactive power control are designed in synchronous reference frame (SRF). Both simulation and hardware implementation are conducted. Results confirm that the proposed RBFNN assisted hybrid control scheme yields excellent dynamic performance in terms of various reactive power tracking control of the GaN HEMTs-based three-phase STATCOM system.


Introduction
The static synchronous compensator (STATCOM) is a power electronic-based, shunttype flexible AC transmission system (FACTS) device whose main functions include voltage regulation and reactive power flow control for power transmission, distribution, and industrial power supply systems. The STATCOM has become increasingly important because of the modern trend of renewable energy (RE) based distributed power generation (DG), where unpredictable fluctuations due to various weather conditions are unavoidable. Moreover, modern technologies such as Industry 4.0 and electric vehicles (EV) have various fast compensating requirements for their power utilization systems, including reactive power, unbalanced and harmonic currents, etc., which can be optimally dealt with using power electronic-based compensators [1][2][3][4].
In recent years, the investigation of STATCOM has always been very popular. Elserougi et al. [5] proposed a three-phase STATCOM based on a hybrid full-bridge (FB)/half-bridge (HB) 9-level boost modular converter with the aim of reducing the number of FB sub-modules. An interesting yes/no algorithm was proposed and its effectiveness was verified. In [6], a three-phase modular cascaded multilevel STATCOM consisting of conventional voltage source inverters (VSIs) was proposed with three independent DC (PCB) layout in order to obtain the best tradeoff between safety and losses [22,23]. For mid-to high-power applications, the requirements for driving WBG devices include the following: (1) high drive strength, (2) enough isolation between driving and power circuits (by using isolators or isolated drivers), (3) gate voltage oscillation damping (with high enough turn-on impedance and separated turn-on and turn-off paths), (4) gate voltage spike limiting (by using voltage clamps), fast turn-off (with small turn-off impedance and negative turn-off voltage) and (5) dead time optimization. For PCB layout, it is necessary to minimize parasitic inductance and capacitance. In other words, minimized trail lengths, device packages with small parasitic inductance, and minimized overlapping between paths are required [24][25][26].
Considering that there are currently no published articles discussing the use of GaN HEMTs for STATCOM design and applications, this paper aims to demonstrate a GaN HEMTs-based three-phase STATCOM and a hybrid control scheme for the first time. To explore the potential of improving dynamic control performance, the proposed GaN HEMTs-based STATCOM is based on a voltage-source inverter (VSI) topology and controlled with a dual-loop hybrid control scheme. Background knowledge of STATCOM and GaN HEMTs is briefly addressed in the first section. Section 2 establishes the proposed STATCOM system, including the architecture and relevant parameters. Section 3 presents the design of the proposed dual loop hybrid control scheme, including inner loop current controllers and outer loop DC link voltage and reactive power controllers. In this paper, a hybrid control structure using radial basis function neural network (RBFNN) controller alongside the PI controller is used in the reactive power control loop so that the dynamic performance of the GaN HEMTs-based STATCOM used as a fast reactive power tracking controller can be assessed. Simulation studies and experimental tests on a 2 kVA prototype are respectively presented in Sections 4 and 5. A brief discussion and conclusion are given in Section 6.

System Description
To focus on the control performance of the proposed GaN HEMTs-based STATCOM incorporated with different control schemes, the test system used in this study is simple, as shown in Figure 1. It consists of a healthy and balanced three-phase grid, a load, and a three-phase GaN HEMTs-based STATCOM. The three-phase STATCOM adopts a voltagesource inverter (VSI) architecture because of the simplicity, as shown in Figure 2, where V dc represents DC link voltage, C dc represents DC link capacitor, A, B, and C are the switching points of phase legs A, B, and C, respectively, N is the reference point of the VSI voltages, L f represents the filter inductor, I a , I b , and I c represent inductor currents, C f represents the filter capacitor, V c f ,a , V c f ,b and V c f ,c represents filter capacitor voltages, I grid_a , I grid_b and I grid_c represent grid currents, V grid_a , V grid_b , and V grid_c represent grid voltages, and n represents grid grounding point. Relevant parameters are listed in Table 1. In this paper, the main control function of the GaN HEMTs-based three-phase STATCOM is to provide fast and precise reactive power regulation.

Controller Design
In theory, the reactive power output of a STATCOM can be controlled with its voltage phase shifts or various reactive current control schemes. Direct reactive current control is normally chosen because of the simplicity and easy to implement. To achieve a fast dynamic control feature, the control strategy of the proposed STATCOM is based on the dq-axis decoupling in SRF and a dual-loop architecture, as shown in Figure 3, where the inner loop consists of dq-axis type-II current controllers, and the outer loop consists of

Controller Design
In theory, the reactive power output of a STATCOM can be controlled with its voltage phase shifts or various reactive current control schemes. Direct reactive current control is normally chosen because of the simplicity and easy to implement. To achieve a fast dynamic control feature, the control strategy of the proposed STATCOM is based on the dq-axis decoupling in SRF and a dual-loop architecture, as shown in Figure 3, where the inner loop consists of dq-axis type-II current controllers, and the outer loop consists of two PI controllers for both the reactive power (d-axis) and DC link voltage (q-axis). For comparison purposes, an RBFNN controller and a PI controller are designed for the reactive power control loop in which sinusoidal pulse width modulation (SPWM) is used to switch the GaN HEMTs.  Figure 3. The proposed control architecture for GaN HEMTs-based three-phase STATCOM.

Design of Inductor Current Controllers
For the design of inductor current controllers, the mathematical models of the inductor currents are obtained based on Figure 2 and Kirchhoff's voltage law:

Design of Inductor Current Controllers
For the design of inductor current controllers, the mathematical models of the inductor currents are obtained based on Figure 2 and Kirchhoff's voltage law: where v AN , v BN , and v CN represent switching point voltages, and v nN represents the voltage difference between grid ground and VSI neutral point. Assuming that the three phases of the VSI are balanced, we get the following: As a result, v nN can be expressed as follows: which allows (1)-(3) to be expressed as the following: Since pulse width modulation (PWM) technique is used in the control, the three-phase modulation signals v cona , v conb and v conc are compared with carrier signal v tri respectively to trigger the switches of the switching legs. As a result, v AN , v BN , and v CN can be expressed as follows: Substituting (7)-(9) into (6) and letting V dc 2V tri = K pwm yield the following: Next, (10) can be converted into dq0-axis equivalents using synchronous reference frame (SRF) theory: where v cond , v conq , and v con0 represent dq0-axis modulation signals, v grid_d , v grid_q , and v grid_0 represent grid dq0-axis voltages, and I d , I q , and I 0 represents inductor dq0-axis currents. Then, block diagrams of dq-axis inductor current controllers can be graphed according to (11), as shown in Figures 4 and 5, where type II controllers are used, i d and i q represent sensed dq-axis currents, i * d and i * q represent dq-axis current commands, and v * grid_d and v * grid_q represent qd-axis grid voltage commands. Then, we can obtain current loop transfer function, controller transfer function, and loop gain as follows: Micromachines 2021, 12, x 7 of 29 Then, block diagrams of dq-axis inductor current controllers can be graphed according to (11), as shown in Figures 4    In practical applications, the crossover frequency of a type II controller is designed below 1/4 of the switching frequency. In this paper, the crossover frequency is designed at 1/10 of the switching frequency, the zero is designed at 1/5 of the crossover frequency, and pole is designed at 36.15 kHz: Then, block diagrams of dq-axis inductor current controllers can be graphed according to (11), as shown in Figures 4 Figure 5. Control loop of q-axis current controller.
In practical applications, the crossover frequency of a type II controller is designed below 1/4 of the switching frequency. In this paper, the crossover frequency is designed at 1/10 of the switching frequency, the zero is designed at 1/5 of the crossover frequency, and pole is designed at 36.15 kHz: In practical applications, the crossover frequency of a type II controller is designed below 1/4 of the switching frequency. In this paper, the crossover frequency is designed at 1/10 of the switching frequency, the zero is designed at 1/5 of the crossover frequency, and pole is designed at 36.15 kHz: Next, current loop gain and controller gain at crossover frequency can be obtained: The required gain compensation at crossover frequency can then be calculated: As a result, controller transfer function is expressed as follows: The designed proportional and integral gains are 63.9257 and 8.033188, respectively. Figure 6 displays the Bode plot of the inductor current control loops, where the phase margin is 63 degrees.

DC Link Voltage Controller
The DC link voltage control loop balances the active power between AC and DC sides of the VSI. By neglecting steady-state operating point, we can obtain an equivalent circuit of the voltage loop as illustrated in Figure 7.

DC Link Voltage Controller
The DC link voltage control loop balances the active power between AC and DC sides of the VSI. By neglecting steady-state operating point, we can obtain an equivalent circuit of the voltage loop as illustrated in Figure 7.

DC Link Voltage Controller
The DC link voltage control loop balances the active power betw of the VSI. By neglecting steady-state operating point, we can obtain of the voltage loop as illustrated in Figure 7. According to Figure 7, the instantaneous AC power at AC side as follows: Mapping AC side signals onto the DC side and assuming that t we obtain the following: According to Figure 7, the instantaneous AC power at AC side in SRF can be defined as follows: Mapping AC side signals onto the DC side and assuming that the inverter is lossless, we obtain the following: 1.5v grid_q I q = V dc I dc .
As a result, the relationship between DC side current and AC side current is as follows: According to (25) and (26), we can obtain the transfer function of DC side voltage: where k dc represents the conversion factor from AC side to DC side. This relationship yields the block diagram of DC link voltage control loop, as shown in Figure 8, where PI controller is used, and only q-axis current is controlled. Then, we can obtain DC link voltage loop transfer function, controller transfer function, and loop gain as follows: The main purpose of the PI controller is to reduce possible interference affecting the DC link voltage when the STATCOM regulates reactive power. Usually, the crossover frequency of the outer loop is designed below 1/3 of that of the inner loop. The crossover frequency is set at 1/50 of that of the current loop, and the zero is designed at 1/5 of DC loop crossover frequency: (31) The main purpose of the PI controller is to reduce possible interference affecting the DC link voltage when the STATCOM regulates reactive power. Usually, the crossover frequency of the outer loop is designed below 1/3 of that of the inner loop. The crossover frequency is set at 1/50 of that of the current loop, and the zero is designed at 1/5 of DC loop crossover frequency: Next, DC link voltage loop gain and controller gain at crossover frequency can be obtained: The required gain compensation at crossover frequency can then be calculated: As a result, controller transfer function is expressed as follows: The designed proportional and integral gains are 10.3673 and 0.02591825, respectively. Figure 9 displays the Bode plot of the inductor current control loops, where the phase margin is 78 degrees.

Reactive Power Controller
The reactive power control loop regulates the reactive power output of the VSI. Instantaneous reactive power can be defined as follows: (37) As a result, the block diagram of reactive power control loop can be graphed as shown in Figure 10, where PI controller is used, only d-axis current is controlled, and q and q * represent sensed reactive power and reactive power command, respectively. In this paper, an RBFNN controller is designed and added in the reactive power control loop. It is designed to improve the dynamic performance of the PI controller in fast tracking control of reactive power. By switching SW1 and SW2, the NN and PI controllers can be activated simultaneously or separately. It is assumed that the bandwidth of current control loop is much wider than that of the active/reactive power control loop. The addition of the RBFNN controller can help the linear PI controller deal with transients, but it naturally increases complexity and cost. We can obtain reactive power loop transfer function, controller transfer function, and loop gain as follows: The required gain compensation at crossover frequency can then be calculated: (36) The designed proportional and integral gains are 10.3673 and 0.02591825, respectively. Figure 9 displays the Bode plot of the inductor current control loops, where the phase margin is 78 degrees.

Reactive Power Controller
The reactive power control loop regulates the reactive power output of the VSI. Instantaneous reactive power can be defined as follows: As a result, the block diagram of reactive power control loop can be graphed as shown in Figure 10, where PI controller is used, only d-axis current is controlled, and q and q * represent sensed reactive power and reactive power command, respectively. In this paper, an RBFNN controller is designed and added in the reactive power control loop. It is designed to improve the dynamic performance of the PI controller in fast tracking control of reactive power. By switching SW1 and SW2, the NN and PI controllers can be activated simultaneously or separately. It is assumed that the bandwidth of current control loop is much wider than that of the active/reactive power control loop. The addition of the RBFNN controller can help the linear PI controller deal with transients, but it naturally increases complexity and cost. We can obtain reactive power loop transfer function, controller transfer function, and loop gain as follows: For the PI parameters, the crossover frequency is set at 1/150 of that of the current control loop, and the zero is designed at 10 times the crossover frequency: For the PI parameters, the crossover frequency is set at 1/150 of that of the current control loop, and the zero is designed at 10 times the crossover frequency: Next, reactive power loop gain and controller gain at crossover frequency can be obtained: The required gain compensation at crossover frequency can then be calculated: As a result, controller transfer function is expressed as follows: The designed proportional and integral gains are 0.1191 and 0.0049888996, respectively. Figure 11 displays the Bode plot of the inductor current control loops, where the phase margin is 95.5 degrees.
The required gain compensation at crossover frequency can then be calculated: (47) The designed proportional and integral gains are 0.1191 and 0.0049888996, respectively. Figure 11 displays the Bode plot of the inductor current control loops, where the phase margin is 95.5 degrees. The RBFNN controller added in the reactive power controller loop is as shown in Figure 12. The RBFNN is a feedforward NN using Gaussian activation functions, and the weights of the RBFNN are determined adaptively. The RBFNN offers advantages such as structural and computational simplicity, high noise immunity, avoidance of local minima, The RBFNN controller added in the reactive power controller loop is as shown in Figure 12. The RBFNN is a feedforward NN using Gaussian activation functions, and the weights of the RBFNN are determined adaptively. The RBFNN offers advantages such as structural and computational simplicity, high noise immunity, avoidance of local minima, and fast online learning. The RBFNN is widely used in applications such as parameter identification and pattern recognition [27][28][29]. In this paper, 2, 4, and 1 node are designed for the input, hidden, and output layers, respectively, determined with try and error within the computational limitation of the adopted digital signal processor (DSP) TMS320F28335 by Texas Instruments. The adopted RBFNN controller is trained with online learning. and fast online learning. The RBFNN is widely used in applications such as parameter identification and pattern recognition [27][28][29]. In this paper, 2, 4, and 1 node are designed for the input, hidden, and output layers, respectively, determined with try and error within the computational limitation of the adopted digital signal processor (DSP) TMS320F28335 by Texas Instruments. The adopted RBFNN controller is trained with online learning.

Simulation Study
The simulation studies in this work were conducted using simulation software. The simulation environment is as shown in Figure 13. The control objective of the designed 2 kVA GaN HEMTs-based three-phase STATCOM is fast response and precise reactive power output based on a given reactive power command, as shown in Figure 14a. The command consists of three states: 0, 600, and −600 VAR, and each interval is 0.2 s long. For comparison of the designed reactive power controllers, two controller settings are arranged, as listed in Table 2. In the following results, the values of the control signals (command and feedback waveforms) are the actual values multiplied by their respective sensing factors and have been converted into voltage signals, as listed in Table 3. That is, the 600 VAR reactive power corresponds to 0.186V reactive power control signal, and the 3.15 A reactive current corresponds to 0.1575V reactive current control signal.

Simulation Study
The simulation studies in this work were conducted using simulation software. The simulation environment is as shown in Figure 13. The control objective of the designed 2 kVA GaN HEMTs-based three-phase STATCOM is fast response and precise reactive power output based on a given reactive power command, as shown in Figure 14a. The command consists of three states: 0, 600, and −600 VAR, and each interval is 0.2 s long. For comparison of the designed reactive power controllers, two controller settings are arranged, as listed in Table 2. In the following results, the values of the control signals (command and feedback waveforms) are the actual values multiplied by their respective sensing factors and have been converted into voltage signals, as listed in Table 3. That is, the 600 VAR reactive power corresponds to 0.186V reactive power control signal, and the 3.15 A reactive current corresponds to 0.1575V reactive current control signal.

Results of Simulation Case 1: PI Controller
In this case, only the PI controller is activated in the reactive power control loop. Simulation results are presented in Figures 14b-16b. Figures 14b and 15 show that the phases and amplitudes of three-phase currents are well controlled. In Figures 16a and 17, it can be observed that the rise and fall times for reactive power are 3.469 ms and 3.558 ms, respectively, and the overshoot and undershoot percentages are both 20%. In Figure  16b, it can be observed that the accumulated reactive power control error is 3.57m at the end of the simulation.

Results of Simulation Case 1: PI Controller
In this case, only the PI controller is activated in the reactive power control loop. Simulation results are presented in Figures 14b, 15b and 16b. Figures 14b and 15 show that the phases and amplitudes of three-phase currents are well controlled. In Figures 16a and 17, it can be observed that the rise and fall times for reactive power are 3.469 ms and 3.558 ms, respectively, and the overshoot and undershoot percentages are both 20%. In Figure 16b, it can be observed that the accumulated reactive power control error is 3.57m at the end of the simulation.

Results of Simulation Case 2: PI Controller + RBFNN Controller
In this case, the PI controller in the last subsection is joined with the RBFNN controller in the reactive power control loop. Simulation results are presented in Figures 18-20b. Figures 18 and 19 show that the phases and amplitudes of three-phase currents are well

Results of Simulation Case 2: PI Controller + RBFNN Controller
In this case, the PI controller in the last subsection is joined with the RBFNN controller in the reactive power control loop. Simulation results are presented in Figures 18, 19 and 20b. Figures 18 and 19 show that the phases and amplitudes of three-phase currents are well controlled. In Figures 20a and 21, it can be observed that the rise and fall times for reactive power are 3.2 ms and 3.5 ms, respectively, and the overshoot and undershoot percentages are improved to 5% and 6%, respectively. As can be seen, the main purpose of the RBFNN controller in this case is to suppress the overshoots and undershoots. In Figure 20b, it can be observed that the accumulated reactive power control error is 2.44 m at the end of the simulation.
Micromachines 2021, 12, x 1 controlled. In Figures 20a and 21, it can be observed that the rise and fall times for rea power are 3.2 ms and 3.5 ms, respectively, and the overshoot and undershoot percen are improved to 5% and 6%, respectively. As can be seen, the main purpose of the RB controller in this case is to suppress the overshoots and undershoots. In Figure 20b, i be observed that the accumulated reactive power control error is 2.44m at the end o simulation.

Hardware Prototype Implementation
In order to verify the simulated results of the proposed GaN HEMTs-based STAT-COM system and the control schemes, this section presents the hardware implementation

Hardware Prototype Implementation
In order to verify the simulated results of the proposed GaN HEMTs-based STATCOM system and the control schemes, this section presents the hardware implementation of the proposed system using previously simulated scenarios. The schematic block diagram of the hardware implementation is illustrated in Figure 22, where the related specifications are listed in Table 4. The proposed three-phase STATCOM was constructed using GaN HEMT TPH3207 by Transphorm (Goleta, CA, USA). The gate driver adopted for driving the GaN HEMTs was Si8271 by Silicon Labs (Austin, TX, USA). A programmable three-phase power source and a delta-Y 200V/146V transformer emulated the healthy balanced grid. DSP TMS320F28335 was used to provide efficiency and flexibility in controller design. A photograph of the constructed GaN-based STATCOM prototype is shown in Figure 23, while the numbered devices are listed in Table 5. For easier waveform observation, the reactive power command was prolonged to 2 s per interval, as shown in Figure 24, where t 1 = 2 s, t 2 = 4 s, t 3 = 6 s, t 4 = 8 s. Results of hardware implementation of two reactive power controller settings are presented in the following subsections. anced grid. DSP TMS320F28335 was used to provide efficiency and flexibility in controller design. A photograph of the constructed GaN-based STATCOM prototype is shown in Figure 22, while the numbered devices are listed in Table 5. For easier waveform observation, the reactive power command was prolonged to 2 s per interval, as shown in Figure  23, where t1 = 2 s, t2 = 4 s, t3 = 6 s, t4 = 8 s. Results of hardware implementation of two reactive power controller settings are presented in the following subsections.

Test Results of Implementation Case 1: PI Controller
In this case, only the PI controller is activated in the reactive power contro Measured results are presented in Figures 25-29. Figures 25 and 26 show that the and amplitudes of three-phase currents are well controlled. In Figures 27 and 28, it observed that the rise and fall times for reactive power were 4.3 ms and 4.68 ms, tively, and the overshoot and undershoot percentages were 24% and 14%, respecti Figure 29, it can be observed that the PI controller was able to control the reactive but the dynamic performance was not sufficient.

Test Results of Implementation Case 1: PI Controller
In this case, only the PI controller is activated in the reactive power control loop. Measured results are presented in Figures 25-29. Figures 25 and 26 show that the phases and amplitudes of three-phase currents are well controlled. In Figures 27 and 28, it can be observed that the rise and fall times for reactive power were 4.3 ms and 4.68 ms, respectively, and the overshoot and undershoot percentages were 24% and 14%, respectively. In Figure 29, it can be observed that the PI controller was able to control the reactive power but the dynamic performance was not sufficient.

Test Results of Implementation Case 2: PI Controller + RBFNN Controller
In this case, the PI controller in the last subsection is joined with the RBFNN controller in the reactive power control loop. Measured implementation results are presented in Figures 30-34. Figures 30 and 31 show that the phases and amplitudes of three-phase currents were well controlled. In Figures 32 and 33, it can be observed that the rise and fall times for reactive power were shortened to 3.52 ms and 4.64 ms, and the overshoot and

Test Results of Implementation Case 2: PI Controller + RBFNN Controller
In this case, the PI controller in the last subsection is joined with the RBFNN controller in the reactive power control loop. Measured implementation results are presented in Figures 30-34. Figures 30 and 31 show that the phases and amplitudes of three-phase currents were well controlled. In Figures 32 and 33, it can be observed that the rise and fall times for reactive power were shortened to 3.52 ms and 4.64 ms, and the overshoot and undershoot percentages were successfully improved to 4% and 2%, respectively. In Figure 34, it can be observed that the RBFNN controller successfully suppressed the overshoots and undershoots.

Conclusion and Discussion
The results of all simulated and 2kVA hardware implemented scenarios are highlighted in Table 6. It can be observed that combining the PI controller with RBFNN controller forming a hybrid, nonlinear control scheme can effectively improve the dynamic performance of reactive power control, especially in terms of regulation speed and overshoot/undershoot suppression. Therefore, we can conclude that the proposed GaN HEMTs-based three-phase STATCOM taking the advantages of better material features of WBG switching devices and the advanced control scheme on both trained RBFNN and conventional PI controllers is a highly effective design example. It is important to note that the need to improve the performance of various power converters has driven researchers around the world to investigate on new WBG switching devices and their applications. It has been well proved that the WBG switching devices, including GaN HEMTs and SiC MOSFETs, offer superior performance to that of conventional Si-based switching devices, including higher blocking voltage, current, switching frequency, efficiency, and operating temperature. However, in various design cases reported in open literature, GaN devices tend to dominate low-to mid-voltage and low-to

Discussion and Conclusions
The results of all simulated and 2kVA hardware implemented scenarios are highlighted in Table 6. It can be observed that combining the PI controller with RBFNN controller forming a hybrid, nonlinear control scheme can effectively improve the dynamic performance of reactive power control, especially in terms of regulation speed and overshoot/undershoot suppression. Therefore, we can conclude that the proposed GaN HEMTs-based three-phase STATCOM taking the advantages of better material features of WBG switching devices and the advanced control scheme on both trained RBFNN and conventional PI controllers is a highly effective design example. It is important to note that the need to improve the performance of various power converters has driven researchers around the world to investigate on new WBG switching devices and their applications. It has been well proved that the WBG switching devices, including GaN HEMTs and SiC MOSFETs, offer superior performance to that of conventional Si-based switching devices, including higher blocking voltage, current, switching frequency, efficiency, and operating temperature. However, in various design cases reported in open literature, GaN devices tend to dominate low-to mid-voltage and low-to mid-power converter applications, while SiC devices are suitable for higher-voltage and higher-power converter applications. In this aspect, this paper has demonstrated a GaNbased three-phase STATCOM functioning as a fast reactive power regulator. The proposed dual-loop control architecture using SRF consists of inner loop, type II current controllers and outer loop, DC link voltage, and reactive power PI controllers. In addition, an RBFNN controller is designed for constructing a hybrid reactive power control scheme to improve the tracking speed and dynamic performance of the GaN HEMTs-based STATCOM. Both simulation and measured hardware implementation results have verified the feasibility and effectiveness of the proposed design case.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.