High-Efficiency Isolated Photovoltaic Microinverter Using Wide-Band Gap Switches for Standalone and Grid-Tied Applications

Abstract

An isolated photovoltaic micro-inverter for standalone and grid-tied applications is designed and implemented to achieve high efficiency. System configuration and design considerations, including the proposed active-clamp forward-flyback resonant converter for the DC-DC stage and a dual-frequency full-bridge inverter for the DC-AC stage, are analyzed and discussed. A prototype microinverter system is built and tested. Experimental results verify the feasibility of the proposed system, which achieves 95% power conversion efficiency at full load.

1. Introduction

Photovoltaic (PV) energy systems have been extensively studied in research and industries as an alternative to fossil fuels to reduce greenhouse effects [1,2,3]. The concept of maximum power point tracking (MPPT) is used to extract maximum solar power from PV panels [4,5]. Several PV panels are typically connected in series/parallel to provide high voltage/current for high-power applications. However, PV panels frequently work under mismatched conditions due to partial shading and variations in panel orientation [6,7,8]. Thus, the power production of an entire PV string is reduced by this mismatching issue. One way to address the mismatch problem in large strings is to employ MPPT at the individual panel level using microinverters. Figure 1 shows the system diagram of the isolated PV microinverter mounted on each PV panel [9,10,11]. The pre-stage DC-DC converter provides isolation and controls the PV panel to extract its maximum power and provides a high-voltage DC bus for the post-stage DC-AC inverter [12,13,14]. The inverter converts the DC voltage into a sinusoidal AC output voltage that connects and supplies power to the AC-grid system [15].

One problem with conventional solar microinverters is their poor conversion efficiency mainly attributed the hard-switching operation of the power converter. The commercial success of a PV microinverter depends on its reliability and efficiency. However, high device stress and hard switching operation results in low conversion efficiency. A PV microinverter with a pseudo DC-link is advantageous because it can achieve high efficiency. However, the component count and control complexity are high such that the overall system reliability is decreased [16,17,18].

Wide band-gap (WBG) materials such as Silicon Carbide (SiC) and Gallium Nitride (GaN) have attracted increasing attention in switching power supply applications [19,20,21]. Compared with the current Silicon (Si) technology, WBG devices provide significant advantages of high operating temperature, fast switching speed, and low on-resistance. Thus, they can reduce switching losses to enable high-frequency operation and achieve high-efficiency performance.

This work presents a highly efficient grid-tied isolated PV microinverter that can individually extract the maximum solar power to transfer power to the AC utility network. This work focuses on implementing a two-stage topology. The active-clamp forward-flyback converter topology is adopted for the front-end of the system, which converts energy from the PV panel to a higher voltage for the post-stage inverter. The proposed active-clamp forward-flyback converter topology combines the concepts of forward and flyback converters to increase the utilization of the flyback transformer, while the resonant mechanism is added to the converter unit design. The resonant aspect brings the benefits of extending the zero-voltage-switching (ZVS) region and allowing the secondary diodes to achieve zero-current-switching (ZCS). Furthermore, the fixed frequency pulse-width modulation (PWM) can be achieved to control the converter by only changing the component value. It retains the advantages of high efficiency and easy of control. The microinverter is implemented with WBG devices for PV panel applications to achieve low harmonic distortion, small circuit size and high efficiency. The contributions of this work are: (1) the introduction and analysis of the resonant control strategy of the active-clamp forward-flyback converter to allow for ZVC and ZCS to reduce losses, and (2) the implementation and experimental validation of the interleaved active-clamp forward-flyback converter using WBG devices to achieve a highly-efficiency PV microinverter.

2. System Design and Considerations

The proposed system configuration consists of an interleaved active-clamped forward-flyback converter for the DC-DC stage and a dual-frequency full-bridge inverter for the DCAC stage, as shown in Figure 2. The inverter can operate in a standalone mode (also called grid forming mode), where the inverter is not connected to the AC electric grid, or in an on-grid mode (also called grid feeding mode), where PV power is supplied to the AC electric grid [22]. The control of each converter differs depending on the operation mode, which is further detailed in Section 2.2.

The detailed schematic of the interleaved active-clamped forward-flyback resonant converter and dual-frequency full-bridge circuit are shown in Figure 3. The first stage is the interleaved active-clamped forward-flyback resonant converter, whose characteristics are similar to those of forward resonant converter in the charging phase and the flyback resonant converter in the discharging phase. This topology has advantages of small size and low input current ripple. The active-clamped circuit is used to recycle energy stored by the leakage inductance of the coupled inductor and to reduce voltage stress on the switch. The DC-DC pre-stage circuit is operated in either single or interleaved mode depending on power conditions, which is used to improve light-load efficiency.

For the second stage, the dual-frequency full-bridge circuit is selected because it produces a wider output voltage range for the same input voltage and also reduces the current stress on the switch. The dual-frequency full-bridge circuit exhibits the important advantage of improved efficiency because of the single arm-chopped waveform action, as shown in Figure 4. This feature nearly eliminates the switching loss on one arm. Details of the proposed microinverter system and control strategy are covered in the next section.

2.1. Active-Clamp Forward-Flyback Resonant Converter

Under steady-state operation, the active-clamped forward-flyback resonant converter undergoes eight switching states. Figure 5a shows the circuit schematics and Figure 5b shows the theoretical waveforms.

As shown in Figure 6, the circuit principles of the DC-DC converter stage are analyzed and discussed in detail. A capacitor is placed in parallel with the PV panel such that the PV voltage is relatively constant over the full cycle of the period. Thus, V_in is assumed to be constant for the following analysis. Eight states are described for time t over a full cycle.

State 1 (t₀ − t₁): During this time interval, the auxiliary switch S_Hi is off. The main switch S_Lo is turned on with zero-voltage switching (ZVS) at t₀ because of the conduction of its body diode. At the secondary side, the diode D₁ is on and D₂ is off. Load power is supplied alone by the output capacitor C_out. The leakage inductance L_r resonates with capacitor C_p. The resonant current i_Lr can be represented as follows:

$$i_{Lr}(t)=i_{Lm}(t)+i_p(t)\cong i_{Lr}(t_0)+\frac{V_{in}}{L_m+L_r}(t-t_0)+i_p(t)$$

(1)

$$i_p(t)=\frac{V_{in}}{Z_0}\cdot \sin {\omega }_0(t-t_0)$$

(2)

where L_m is the magnetizing inductance and i_Lm is its current, i_p is the primary transformer current, and V_in is the input voltage. Resonant angular frequency ω₀ is defined as

$${\omega }_0=\frac{1}{\sqrt{L_r\cdot (C_p/n^2)}}$$

(3)

where n is the transformer turns ratio, and impedance Z₀ is defined as

$$Z_0=\sqrt{\frac{L_r}{(C_p/n^2)}}.$$

(4)

State 2 (t₁ − t₂): The transformer primary current reaches zero at t₁, and the diode D₁ is off with zero-current switching (ZCS). During this time interval, D₂ remains turned off and the load power is still supplied only by output capacitor C_out. The resonant current i_Lr can be represented as follows:

$$i_{Lr}(t)=i_{Lm}(t)=i_{Lr}(t_1)+\frac{V_{in}}{L_m+L_r}(t-t_1).$$

(5)

State 3 (t₂ − t₃): The main switch S_Lo is turned off at t₂, and the equivalent inductor (L_m + L_r) resonates with S_Lo D-S capacitor C_DSLo. The resonant current i_Lr and resonant voltage v_DSLo can be represented as follows:

$$i_{Lr}(t)=\frac{V_{in}}{Z_2}\cdot \sin {\omega }_2(t-t_2)+i_{Lr}(t_2)\cdot \cos {\omega }_2(t-t_2)$$

(6)

$$v_{DSLo}(t)\cong V_{in}\cdot [1-\cos {\omega }_2(t-t_2)]+i_{Lr}(t_2)\cdot Z_2\cdot \sin {\omega }_2(t-t_2)$$

(7)

where resonant angular frequency ω₂ is defined as

$${\omega }_2=\frac{1}{\sqrt{(L_m+L_r)\cdot C_{DSLo}}}$$

(8)

and impedance Z₂ is defined as

$$Z_2=\sqrt{\frac{(L_m+L_r)}{C_{DSLo}}}.$$

(9)

State 4 (t₃ − t₄): The resonant voltage v_DSLo reaches (V_in + v_C) at t₄, and the body diode of the auxiliary switch S_Hi and output diode D₂ are turned on. The resonant current i_Lr, the transformer primary current i_p and diode D₂ current i_D2 can be represented as follows:

$$i_{Lr}(t)=i_{Lm}(t)+i_p(t)\cong i_{Lm}(t_3)-\frac{v_{Lm}(t)}{L_m}+i_{Lr}(t_3)\cdot \cos {\omega }_3(t-t_3)$$

(10)

$$i_p(t)\cong i_{Lr}(t_3)\cdot \cos {\omega }_3(t-t_3)$$

(11)

$$i_{D2}(t)=n\cdot i_p(t)$$

(12)

where resonant angular frequency ω₃ is defined as

$${\omega }_3=\frac{1}{\sqrt{L_r\cdot C_{eq}}}$$

(13)

equivalent capacitance C_eq is defined as

$$C_{eq}=\frac{C_c·(C_p/n^2)}{C_c+(C_p/n^2)}$$

(14)

and C_c is the clamping capacitor.

State 5 (t₄ − t₅): During this time interval, the main switch S_Lo remains turned off. The auxiliary switch S_Hi is turned on with ZCS at t₄. The circuit equations are the same with those during State 4.

State 6 (t₅ − t₆): The transformer primary current reaches zero at t₅. The output diode D₂ is turned off with ZCS and the load power is supplied only by the output capacitor C_out. The resonant current i_Lr is as follows:

$$i_{Lr}(t)\cong i_{Lr}(t_5)\cos {\omega }_5(t-t_5)+\frac{v_{C_C}(t_5)}{Z_5}\cdot \sin {\omega }_5(t-t_5)$$

(15)

where resonant angular frequency ω₄ is defined as

$${\omega }_5=\frac{1}{\sqrt{(L_r+L_m)\cdot C_c}}$$

(16)

and impedance Z₅ is defined as

$$Z_5=\sqrt{\frac{(L_r+L_m)}{C_c}}.$$

(17)

State 7 (t₆ − t₇): The auxiliary switch S_Hi is turned off at t₆. During this time interval, the energy stored in S_Lo D-S capacitor C_DSLo is released by the resonance with the equivalent inductor (L_m + L_r). The resonant current i_Lr and voltage v_DSLo can be represented as follows:

$$i_{Lr}(t)=i_{Lr}(t_6)\cos {\omega }_6(t-t_6)+\frac{[V_{in}-v_{DSLo}(t_6)]}{Z_6}\cdot \sin {\omega }_6(t-t_6)$$

(18)

$$v_{DSLo}(t)=V_{in}-[V_{in}-v_{DSLo}(t_6)]\cos {\omega }_6(t-t_6)+i_{Lr}(t_6)\cdot Z_6\cdot \sin {\omega }_6(t-t_6)$$

(19)

where resonant angular frequency ω₆ is defined as

$${\omega }_6=\frac{1}{\sqrt{(L_r+L_m)\cdot C_{DSLo}}}$$

(20)

and impedance Z₆ is defined as

$$Z_6=\sqrt{\frac{L_r+L_m}{C_{DSLo}}}.$$

(21)

State 8 (t₇ − t₈): The resonant voltage v_DSLo is reduced to zero at t₇ and the body diode of the main switch S_Lo conducts to achieve ZVS. The resonant current i_Lr is as follows:

$$i_{Lr}(t)=i_{Lr}(t_7)+\frac{V_{in}}{L_r+L_m}(t-t_7).$$

(22)

After completing one switching period, the operation mode starts over from State 1. The detailed resonant waveforms are depicted in Figure 7. The ZCS conditions for diode D₁ is

$$T_{on,1}>1/2f_{r,1}=\pi \sqrt{L_r\left(\frac{C_p}{n^2}\right)}$$

(23)

and the ZCS conditions for D₂ is

$$T_{on,2}>1/2f_{r,2}=\pi \sqrt{L_r\left(\frac{C_c\ast C_p/n^2}{C_c+C_p/n^2}\right)}$$

(24)

where T_on,1 is the turn on time of S_Lo and f_r,1 is its associated resonant frequency, and T_on,2 is the turn on time of S_Hi and f_r,2 is its associated resonant frequency.

2.2. System Control

The microinverter used in this study has two operation modes: standalone and grid-tied operation. The system will automatically detect which mode to use. If the microinverter detects no voltage at the output terminal upon start-up, then it will operate in standalone mode. However, if a voltage is detected, then the microinverter will attempt to start the grid-tied mode operation. The control flowchart is shown in Figure 8.

In the standalone mode, the amount of power to be extracted from the solar panel is determined by the load demand. Therefore, the inverter and DC-DC stages will only be responsible for maintaining the correct voltage. The inverter will be in voltage-mode feedback, which ensures that the output voltage is a correctly-shaped sine wave with the correct amplitude. The voltage loop on the inverter stage uses proportional-resonant (PR) control to provide sufficient gain at 60 Hz output frequency. In this mode, it is particularly important to ensure good voltage regulation and low total harmonic distortion (THD). Meanwhile, the DC-DC stage regulates the bus voltage to ensure that the inverter has the correct voltage to work with. The block diagrams for the control in standalone mode for the DC-DC stage and DC-AC inverter stage are shown in Figure 9.

In the grid-tied mode, the unit decides how much power it will deliver to the grid. The duty of the DC-DC stage is to implement MPPT controller. The controller will attempt to deliver maximum power to the DC bus regardless of the bus voltage. Therefore, the inverter side has the task of maintaining the bus voltage. The inverter will have to deliver the same amount of power delivered to the bus to maintain the bus voltage at a constant value. The output voltage in grid-tied mode is dictated by the AC grid voltage; hence, the inverter will control the amount of current pumped into the grid. A feedback control loop is used for the inverter to achieve the current dictated by the reference. The grid-tied inverter requires that a sine wave with the same phase as the grid voltage be pumped into the grid. Thus, the input to the reference of the current loop is a numerically controlled oscillator (NCO). However, similar to the standalone mode, the gain at the AC-line frequency must be sufficient to ensure that the output follows the reference with accuracy. Thus, a PR control strategy is also implemented for the inverter’s current loop in grid-tied mode. The regulation requires that the output current signal must be in phase with the grid voltage. Therefore, certain mechanisms are necessary to produce a sine wave that is in phase with the output voltage. Accordingly, a phase-locked loop (PLL) that uses a phase detector is used to detect the phase difference between the output and input signals of the NCO. The phase difference is then fed through a controller, which will produce an output frequency signal given to the NCO. Consequently, a feedback loop in frequency is formed, which forces the NCO frequency to follow that of the grid voltage signal, as shown in Figure 10.

In addition to the PLL and current loop, the bus voltage must be kept at a constant value. This regulation task is handled by the inverter, which varies the current signal in order to control the voltage at the bus. A perturb and observe (P&O) method, which exhibits the benefits of simplicity and good performance, is adopted to realize MPPT for the DC-DC stage operated in grid-tied mode. The control black diagram for the DC-DC stage is also shown in Figure 10.

3. Results

3.1. Power Switch Comparison

GaN switches provide significant advantages over silicon (Si) metal oxide semiconductor field effect transistors (MOSFETs) with lower gate charge, smaller reverse recovery charge, and faster switching speeds. As outlined in

Table 1. Parameter of GAN HEMT and COOLMOS Devices.

Device	On-Resistance (Ω)	Gate-Charge (nC)	FOM
GaN HEMT TPH3006PS	0.15	6.2	0.93
CoolMOS 20N60CFD	0.19	95	18.05

, the GaN high electron mobility transistor (HEMT) presents much lower Figure of Merit (FOM), which is better than the conventional CoolMOS device. With much lower gate-charge Q_g, the switching speed of the GaN HEMT device is much faster. Figure 11 shows the measured switching waveforms of Infineon CoolMOS 20N60CFD and Transphorm GaN HEMT TPH3006PS with Silicon Labs gate driver Si8261 and a 0-Ω gate resistor. In this work, 20N60CFD and TPH3006PS were tested and compared as high-frequency switches of the inverter stage. Figure 12 shows the gate driver circuit for the GaN HEMT power switches in the experiments. Due to the consideration of a lower V_gs threshold voltage to turn the switch fully on, negative-voltage driving is provided during switch-off time to increase noise immunity. As GaN switches operate with low rise and fall times, parasitic inductances and capacitances in the printed circuit board (PCB) layout must be minimized to avoid excessive ringing in the circuit [20].

3.2. Experimental Verification

A 500-W microinverter was designed and implemented for both standalone and grid-tied applications. The specifications of the prototype are listed in

Table 2. Circuit Specifications for the microinverter experimental prototype.

Parameter	Symbol	Specification
Input voltage	Vin	18–40 V
Output voltage	Vo	240 Vrms
Output current	Io	2.1 Arms (rated)
Efficiency	ŋ	≥0.95 (at 375 W)
Voltage THD	THDv	≤5% (at nominal output)

. A photograph of the developed prototype system is shown in Figure 13. In the experimental setup, the PV input was a Chroma PV simulator 62150H-1000S.

In the prototype system, several wide band-gap devices were used and tested to improve the efficiency performance. For the interleaved active-clamped forward-flyback converter stage, the CREE SiC diode C3D10060 was used to reduce reverse-recovery loss on the secondary diodes. Infineon CoolMOS 20N60CFD and Transphorm GaN HEMT TPH3006PS were tested and compared in the experiments for high-frequency switches of the inverter stage. Approximately 1% improvement for CEC efficiency can be achieved using GaN HEMT, compared with the results using CoolMOS.

Figure 14 shows the measured gating signals for the DC-DC stage. The converter is operated in single mode at low power levels, as shown in Figure 14a, or interleaved mode, as shown in Figure 14b. The operation mode is changed during operation, depending on the power conditions. The resonant frequencies f_r,1 and f_r,2 are 66 kHz and 95 kHz, respectively. Figure 15 show the measured resonant current and soft-switching waveforms, which are consistent with the theoretical waveforms shown in Figure 5b. Figure 16 shows the voltage and current waveforms of the power switches, which show that the converter achieves ZVS and ZCS. The efficiency performance of the DC-DC stage, which reaches an efficiency of 97%, is shown in

Table 3. Experimental results for the DC-DC stage.

Mode	Po (W)	Efficiency (%)
single	100	97.10
interleaved	250	97.25
interleaved	375	97.10
interleaved	500	96.50

.

Figure 17 shows the switching waveforms of the dual-frequency full-bridge DC-AC inverter under a 250-W condition, which demonstrates the single-arm switching feature.

Table 4. Experimental results for the dual-frequency full-bridge inverter stage.

Vbus (V)	Vo (Vrms)	Po (W)	Efficiency (%)
380	240	100	96.65
380	240	250	97.80
380	240	375	98.00
380	240	500	98.15

shows the measured efficiency performance under different power conditions. The efficiency reaches 98% at full load.

Table 5. Experimental results for the full 500-W microinverter system.

Vpv (V)	Vo (Vrms)	Po (W)	Efficiency (%)
34	240	100	94.15
34	240	250	95.25
34	240	375	95.17
34	240	500	94.70

shows the measured overall efficiency of the 500-W microinverter system under an input voltage of 34 V and an output voltage V_o of 240 V_rms. In the grid-tied mode, experimental results showed that the maximum power point was tracked with an accuracy of approximately 99%. The power efficiency of the microinverter system reaches 95% and maintains high efficiency over the full power range.

4. Conclusions

An isolated PV microinverter was studied for standalone and grid-tied applications. An interleaved active-clamp forward-flyback resonant converter was used as the isolated DC-DC converter stage and dual-frequency full-bridge inverter was used as the DC-AC conversion stage was used to achieve high efficiency and reliability of the microinverter. The control algorithm of the microinverter was described for startup, standalone mode, and grid-tied mode. Wide band-gap devices were also used and compared with conventional Si-based devices in the experiments, and GaN switches were selected for the final design. The experimental verifications showed that the prototype system exhibited high efficiency and high MPPT accuracy, reaching 95% for the overall microinverter system.