Dual-mode magnetically integrated photovoltaic microconverter with adaptive mode change and global maximum power point tracking

This study proposes a high step-up galvanically isolated dc-dc converter based on the quasi-Z-source (qZS) network. The voltage gain of the converter can change in a wide range. This feature makes the converter suitable for applications with a wide input voltage range. The range of the dc gain is increased by the implementation of the combined energy transfer principle. A reconﬁgurable structure is used to combine the energy transfer by the isolation transformer and the coupled inductor of the impedance-source network. The adaptive mode change activates the energy transfer via an additional winding in the qZS network at lower voltages, which results in considerable efﬁciency improvement. Consequently, the input voltage range of the converter is extended to the values useful in photovoltaic applications impacted by partial shading. The proposed approach is veriﬁed in a photovoltaic (PV) microconverter, where the energy harvest from a PV module is possible during shading by enabling the global maximum power point tracking (GMPPT).

family of hybrid Z-source converters have been presented in [13]. A qZS network is combined with a traditional Z-source network to improve the boost ability of the converter. These converters provide a wide range of variation in voltage gain, but there is no isolation between the output and the input. Their voltage boosting ability can only be achieved by increasing the duty cycle. The qZS and Y-source networks were also applied to the isolated push-pull converter in [14] and [15], respectively. These converters can achieve high voltage gain, while their range is limited due to high losses at low input voltage. A qZS series resonant dc-dc converter has been proposed in [16] to extend the input voltage regulation range to 1:6 via multimode control. High efficiency was obtained by using a resonant tank in the output side without adding extra components. A qZS network with a three-winding coupled inductor allowing 1:10 input voltage regulation range has been presented in [17]. The proposed converter is suitable for wide input voltage range applications at sub-kW power levels. Its applicability is limited by the non-optimal operation of magnetics in a wide range of voltages, which results in their bulky design. To optimise the utilisation of magnetic components, a single-switch galvanically isolated qZS converter was proposed for cost-sensitive applications such as PV module-level power electronics in [18]. Both the transformer and the coupled inductor of the qZS network were used to meet the high step-up conditions along with reasonable power density. High efficiency, very high voltage gain, and a low number of semiconductor devices are the main advantages of that converter. Its input voltage range is still compromised due to limited control freedom of the shoot-through duty cycle, as in most of the converters reviewed above.
As mentioned previously, the galvanically isolated dc-dc converters are commonly used in renewable energy applications, especially in PV module-level power electronics. One of the most important issues around the PV system is the influence of environmental conditions. PV module operation is mostly defined by the weather conditions, that is, ambient temperature and solar irradiance. Partial shading effects are most prominent in small PV systems with distributed maximum power point tracking (MPPT) [19] and could be caused by various obstructions like trees, rooftop structures, neighbouring buildings, dust, or snow, influencing the operation of the PV system significantly [20,21]. Opaque shading is a severe type of partial shading that could be caused by fallen leaves, snow, or bird droppings, fully blocking sunlight to PV cell(s). Even one opaque shaded PV cell in a substring can block this substring entirely, causing the corresponding bypass diode to conduct. When one or two substrings are blocked, the PV module can potentially lose one-or two-thirds of its output voltage, respectively. To harvest the power under these conditions, an interface converter should typically have an ultra-wide input voltage regulation range with a lower boundary of the MPPT window below 10 V. A high step-up converter with high efficiency has been presented in [18]. In order to improve the operation of the converter and increase the efficiency in the range of low input voltages, it can be realised as a dual-mode converter. In low voltage conditions, such as partial shading, the high efficiency at high voltage gain can be achieved by applying adaptive mode change to the dual-mode operation like in [22,23].
This study proposes a dual-mode magnetically integrated PV microconverter with adaptive mode change. Depending on the panel voltage level, the new galvanically isolated qZS converter can change its operation modes adaptively to improve the converter efficiency. The operation principle of the proposed converter is explained in two operation modes along with the derivation of voltage gain expressions. The experimental results are provided to examine the practicality of the converter. The control system is synthesised and explained. Based on that, the global MPPT (GMPPT) is implemented. The efficiency is analysed to validate the performance of the proposed converter in a dual-mode operation. Finally, the conclusions are drawn.

DUAL-MODE MAGNETICALLY INTEGRATED DC-DC CONVERTER
The proposed topology ( Figure 1) is originated from the singleswitch qZS dc-dc converter [18]. It consists of a coupled inductor TX2, two capacitors (C qZS1 and C qZS2 ), switches S qZS and S 1 , an isolation transformer TX1, a blocking capacitor C b , voltage doubler rectifier (VDR) implemented with diodes D 1 , D 2 and capacitors C 1 , C 2 and a half-wave rectifier (HWR) formed by D 3 , S 2 , D 4 , and C 3 . The input inductor L 1 ensures continuous current from a PV module.
The main switch S 1 short circuits the output port of the qZS network to perform the voltage step-up. The isolation transformer TX1 is also connected to this port through the dc blocking capacitor C b . The diode D 3 and the switch S 2 constitute a unidirectional switch. The switch S 2 performs adaptive mode change by enabling or disabling the HWR connected to the secondary winding of TX2. If the HWR is disabled, the converter operates as the baseline single-switch qZS dc-dc converter. When it is enabled, the output voltage will be equal to the summed outputs of the HWR and VDR.
The dc gain of the converter depends on a duty cycle of S 1 as well as on a state of the switch S 2 , which can be set either on or off for a whole switching period. Switches S 1 and S qZS are controlled complimentarily with short dead-time. Hence, in the ideal case, the switching period T could be broken into two time intervals, that is, when the switch S 1 is conducting or is in the off-state.

GENERALISED OPERATION PRINCIPLE
As it was mentioned above, the switch S 2 operates as a static (on/off) switch, thus enabling or disabling the HWR. Therefore, the operation of the converter is analysed in two modes: With disabled HWR when S 2 is turned off and enabled HWR with S 2 turned on.

3.1
Converter operation with disabled HWR

Time interval t on
Within this time interval, the switch S 1 is turned on, and S qZS is turned off. The equivalent circuit is shown in Figure 2(a). Waveforms of the control signals, voltages and currents are presented in Figure 3. The energy transmission process in this time inter-

FIGURE 3
Idealised sketch of the control signals, and primary side voltages and currents val is such that the capacitors C qZS1 and C qZS2 are discharged while the inductors L qZS1 and L qZS2 are charged. Diodes D 1 and D 2 are reverse biased and conducting, respectively. The secondary winding of TX1 charges the capacitor C 1 through the diode D 2 , C 2 is discharged by the output current. The relationship between mutual inductance and leakage inductance is described as follows: where k 1 and k 2 are the coupling coefficients and k 1 = k 2 = 1 at the perfect coupling. The relationship between the primary and the secondary windings of the transformer and the coupled inductor is as follows: where v Np and v Ns are the voltages of the primary and secondary windings of TX1, respectively, v Nsec is the voltage of the secondary winding of TX2, and v LqZS is the voltage of L qZS2 . The voltage across the inductor L qZS1 is as follows: The current through the inductor L qZS1 is expressed as where I LVqZS1 is the minimum current through the L qZS1 . Considering Equation (6), during this time interval, the inductor L qZS1 is charged, and current through it has been increased, so that at t = D⋅T it reaches the maximum value. Applying t = D⋅T to Equation (6), the maximum current through the inductor is obtained as During this time interval, the voltage of the inductor L qZS2 is given by The current through the inductor L qZS2 is as follows: Similar to L qZS1 , the maximum current through the inductor L qZS2 is obtained as As the switch S 1 is turned on, the voltage across it equals zero. By considering Figure 2(a), the voltage of the primary winding of TX1 is given by Considering Equations (3) and (11), the voltage of the secondary winding of TX1 can be calculated as Applying Kirchhoff 's voltage law (KVL) to the secondary side of TX1, the following relation is obtained: where Considering Equations (12), (13), (14), and Figure 2(a), the voltage across the capacitor C 1 could be found by During this time interval, the current through the capacitors is defined by the currents of the qZS inductors:

Time interval t off
During this time interval, S 1 and S qZS are turned off and on, correspondingly. The equivalent circuit of the proposed converter is shown in Figure 2(b). Inductors L qZS1 and L qZS2 release part of their energy to charge up the capacitors C qZS1 and C qZS2 . The VDR diodes D 1 and D 2 are conducting and reverse biased, respectively. Through the conducting diode D 1 , the capacitor C 1 is discharged by the output current and the capacitor C 2 . The voltage of the inductor L qZS1 is given by The current through the inductor is as follows: During T off , the inductor is discharging; so its current is decreasing to I LVqZS1 at the end of T off . Applying t = (1-D)T to Equation (19), the minimum current through the inductor L qZS could be found: According to Figure 2(b), the voltage of the inductor L qZS2 is obtained as follows: Considering Equation (22), the current through the inductor L qZS2 is given by The voltage across the switch S 1 is defined by the voltage of the qZS capacitors: The voltage of the primary winding of TX1 is obtained as follows: The voltage of the secondary winding of TX1 is as follows: By applying KVL to the secondary side of TX1, the following equation is obtained: The following expression is derived from Equation (26) taking into account Equations (14), (24), (25), and Figure 2(b) as follows: (27) By replacing Equations (15) in (27), the capacitor voltage V C2 can be obtained: By applying the volt-second balance to the inductors L qZS1 and L qZS2 , and considering Equations (15) and (27), the voltages across the capacitors C qZS1 and C qZS2 are obtained as follows: Considering Equation (23), the boost factor of the proposed converter with disabled HWR could be found as Taking into account Equations (28) to (30) and Figure 2(b), the voltage gain of the proposed converter is calculated as follows: (32) If k 1 =1, the voltage gain of the converter is given by For the lossless converter, the dc input-to-output current conversion factor can be expressed as

Converter operation with enabled HWR
For enabling the HWR, the switch S 2 is turned on and the diode D 4 is reverse biased for a whole switching period. It should be mentioned that depending on the polarity of the secondary winding of TX2 (Figure 4), the HWR could rectify either the positive or negative half-wave of TX2 voltage. The procedure of the analysis is similar for both polarities. Here, the converter is analysed for the rectification of the positive half-wave. Similar to the previous section, the switching period is divided into two time intervals corresponding to the turn-on (T on ) and turn-off (T off ) states of the switch S 1 .

Time interval t on
The equivalent circuit of the proposed converter is shown in Figure 5(a). During this time interval, the diode D 3 is conducting, the capacitor C 3 is charging. Waveforms of the control signals, voltages, and currents are shown in Figure 6. The relations for the current and the voltage of the inductors and the capacitor are the same as those at the turned-off switch S 2 . According to Figure 5(a), the voltage of the capacitor C 3 is obtained as follows: Applying KVL to the secondary side of TX2, the following relation is obtained: where Considering Equations (37), (38), (39), and Figure 5(a), the voltage across the capacitor C 3 is as follows: Replacing Equations (29) in (40) yields:

Time interval t off
During this time interval, the diode D 3 will be reverse biased when the switch S 1 is turned off. The capacitor C 3 is discharging while supplying the output current. The equivalent circuit of the proposed converter is shown in Figure 5(b). The relations of the current and the voltage of the capacitor and inductors are the same as the converter operates as a single switch qZS converter. According to Figure 5(b), the output voltage is equal to the summed outputs of the HWR and VDR. Using Equations (29) to (31) and (42), the voltage gain of the proposed converter is calculated as follows: (42) Assuming k 1 = k 2 = 1, for the rectification of the positive half-wave of TX2, the voltage gain could be obtained as follows: The dc input-to-output current conversion factor for the rectification of the positive voltage half-wave is given by .
As it was mentioned, the procedure of the steady-state analysis when the negative half-wave is rectified is the same as for the positive half-wave. Therefore, only the final equation of the voltage gain and the dc input-to-output current conversion factor are given: ( Figure 6 shows the comparison of the theoretical voltage gains of the proposed converter for the operating modes with  Figure 7 shows the variations of the normalised voltage of the switch S 1 versus voltage gain for the unity turns ratio of TX1 and TX2. The curves are plotted based on the equation given by [24]. As can be seen, the voltage stress of the switch S 1 is lower for the rectification of the positive half-wave. High voltage gain with low voltage stress of the switch shows that the proposed converter has the best configuration when the HWR rectifies the positive half-wave of TX2.

EXPERIMENTAL VERIFICATION
This section provides the experimental results to validate the performance of the proposed converter. The experimental prototype with the peak power of 300 W was assembled following the schematics shown in Figure 1. The main specifications and types of semiconductor components used in the prototype are listed in Table 1. The converter was designed for MPPT and interfacing of the 60-and 72-cell PV modules into a dc bus with the nominal operating voltage of 400 V.

Steady-state waveforms
The steady-state waveforms of the experimental prototype are shown in Figures 8 and 9. In Figure 8, the converter is operating with disabled HWR, and to ensure the demanded dc gain, the duty cycle of S 1 was set to 0.27. Evidently, the current through the L lk2 equals zero when the HWR is disabled. Therefore, the current I LqZS is changing linearly. The output voltage is equal to the average voltage of the capacitor C 2 . For the turn-on state of the switch S 1 , the primary voltage of the TX1 is equal to the voltage of the capacitor C b . By turning on the switch S 2 , the HWR is activated to increase the voltage gain ( Figure 9). As a result, the duty cycle is decreased from 0.27 to 0.227 for the same operating point. The output voltage is equal to the sum of the voltages of C 2 and C 3 , which are 330 and 70 V, respectively. The voltage stress of the switch S 1 is 55 V, which is the voltage of C 2 divided by the turns ratio of TX1. The decrease in the duty cycle value is positively mirrored on the input current ripple, which is 25% smaller with an activated HWR. Also, the primary winding voltage of TX1 is lower than that for the operation with disabled HWR. Figures 8 and 9 show that due to the presence of the input inductor, the converter features the continuous input current, irrespective of the operating mode. To reduce the number of magnetic components in the circuit, the input inductor L qZS1 could be integrated into the magnetic core of TX2, as reported in [25]. However, due to the increased ripple of the core flux, such a three-winding coupled inductor would have a considerably increased volume of the magnetic core, which will negatively impact the power density of the converter.  Figure 10 shows the efficiency variations of the experimental prototype versus the input voltage. First, the efficiency was acquired within the input voltage range from 8 to 65 V, and at the input current of 2 A. As is seen from Figure 10(a), the activation of the HWR allows for extending the lower bound of the input voltage range from 15 to 5 V without any remarkable impact on the efficiency. Moreover, by using the HWR, the power conversion efficiency in the operating points with 15 and 25 V was improved by 5.8% and 2.2%, respectively. At the input current of 4 A, the activation of the HWR results in the efficiency rise from 1% to 3%, depending on the operating point ( Figure 10(b)). Figure 11 shows the comparison of the efficiency between two possible connection possibilities of the TX2 and HWR (positive and negative, see Figure 4), within the input voltage range from 5 to 30 V and at the input current of 10 A. It is evident that the operation of the converter with the rectification of the positive half-wave is more beneficial for the high step-up applications as it can give up to 3% of efficiency rise at the lower values of the input voltage. Figure 12 shows the efficiency profile of the converter operating with continuous power of 100 W in the range of input voltages from 8 to 65 V. For the input voltages below 25 V, the enabling of the HWR results in higher efficiency, especially in the case of the positive half-wave rectification by the HWR.

Efficiency analysis
In conclusion, the proposed approach with the possibility of mode change of the combined energy transfer significantly enhances the step-up performance of the baseline single-switch qZS dc-dc converter, allowing either for a wider input voltage regulation range or higher power conversion efficiency at large dc gain values.

Adaptive mode change implementation
For the proposed approach of the combined energy transfer, the optimal tradeoff between the dc gain range and the power conversion efficiency could be achieved by the proper selection of the turns ratios of TX1 and TX2. Another possibility of the dc gain extension without serious penalties to the efficiency is the implementation of the adaptive mode change (AMC) principle derived from the topology-morphing control [26,27]. In that case, the operating mode of the converter is automatically selected by the control system to change the dc voltage gain instead of regulating the duty cycle outside the favourable range. As a result, the duty cycle is constrained to a region of high efficiency despite variations of the input voltage in a wide range.

FIGURE 10
The efficiency of the experimental prototype measured at different input currents during the rectification of the positive half-wave: (a) input current of 2 A; (b) input current of 4 A Figure 13 shows the AMC realisation principle in the experimental converter operating at the constant power of 100 W. The transition between modes is realised at V in = 24 V by turning on or off the switch S 2 . This resulted in an efficiency rise of up to 4% at the lowest values of the input voltages ( Figure 14). Moreover, owing to AMC implementation, the converter now features a relatively flat efficiency of over 92% within a dc gain range from 8 to 26.

ЕXPERIMENTAL ЕVALUATION OF MPPT PERFORMANCE
In this section, we will demonstrate how the proposed approach could be used in a low-cost PV microconverter for improving its MPPT performance. More specifically, an extension in a dc gain variation range using AMC enables the implementation of a GMPPT based on a periodic P-V curve sweep [22], which was technically impossible with a baseline singleswitch qZS microconverter due to the dc gain limitation issues [10].

Implementation of GMPPT
The GMPPT method is examined under several operation conditions to show the performance of the converter. For a typical Si-based PV module with three sub-strings, the shading conditions on the sub-strings can make three MPPs: two local MPPs (LMPPs) and the global MPP (GMPP). As mentioned previously, in the partial shading or opaque shading conditions, the range of the input voltage is low. In these conditions, to achieve the GMPP, activation of the HWR at low voltages is needed to achieve acceptable efficiency and cover the demanded input voltage range.
Applying the method presented in [28] and using the reconfiguration ability of the proposed converter, GMPPT is enabled along with improved efficiency even at low irradiations. The global MPPT routine is implemented in two stages. First, the converter has to preset the operating point to the maximum achievable voltage of the PV module, which corresponds to the minimum allowed duty cycle of the switch S 1 . Second, the reference voltage is gradually decreasing down to the minimum allowed input voltage. During the voltage sweep, the control system measures the input voltage and current in tabular form with a certain voltage step. Implementation of these stages results in the identification of the actual GMPP position. The control system defines the position of the GMPP and gradually presets the converter towards the vicinity of this point. After approaching the actual position of the GMPP close enough, the local MPPT takes over and performs accurate tracking of the GMPP.
To achieve optimal power conversion efficiency during the GMPPT, the converter has to be switched between the operating modes employing AMC. Considering the experimental results in the previous section, there is an optimum transition point for the AMC where the efficiency curves of the converter with enabled and disabled HWR intersect. Theoretically, the AMC threshold voltage can be approximated using a polynomial expression as demonstrated in [28]. However, this imposes increased requirements for the measurement sensors as well as the main microcontroller which is an important issue in such cost-sensitive applications like PV module-level power electronic converters. Therefore, to simplify AMC implementation and, consequently, constrain the cost of the proposed technology, this study embraces low-cost realisation using a simple hysteresis transition around the input voltage of 20 V. This threshold voltage was identified as the most probable transition point for the AMC at high input currents, that is, highest losses and thermal loading of semiconductors, where the optimisation of the efficiency has the highest influence on the converter reliability. To improve converter dynamics, especially during transients, feedforward control is implemented.
Considering the conditions explained above, the control system shown in Figure 15 was synthesised. The MPPT block determines the reference input voltage. The reference input voltage is applied to the feedforward block. This feedforward control calculates the theoretical duty cycle, while in practice, the real value should be a little larger. This difference between the theoretical and the experimental duty cycle values results in the input voltage setting error applied to the PI controller. The PI controller compensates for this difference by the signal ΔD. Moreover, the regulation speed is improved as the PI controller needs to handle only a small portion of the control signal. The feedforward duty cycle D ff is calculated by the equations below, depending on the operation mode: Clearly, Equation (49) corresponds to the disabled HWR, while Equation (50) should be used when the HWR is enabled. The AMC threshold voltage V TH defines selection between these equations. Hence, the AMC is seamlessly integrated into the closed-loop control system. The resulting duty cycle that is applied to the modulator is obtained by adding the signal produced by the PI controller (∆D) to the feedforward term D ff and taking into account the relative duration of the dead-time D DT . The hysteresis block is used for stable AMC implementation to avoid frequent changes between the modes when the GMPP is close to the threshold voltage V TH = 20 V. In Figure 15, variables in red colour are either acquired from sensors or were predefined in the software.

Experimental validation of GMPPT
To evaluate the GMPPT performance of the proposed microconverter, different operating scenarios have been tested. This study considers flexible CIGS PV module MiPV-165W as this type of PV modules is rapidly gaining popularity in buildingintegrated PV, recreational vehicles, industrial uses, and so forth. These PV modules contain tens of bypass diodes as there is one in every cell, which results in countless shading scenarios possible.

FIGURE 15
Closed-loop control system comprising maximum power point tracking (MPPT) and adaptive mode change of the proposed converter intended for PV module integration into a stable dc bus

FIGURE 16
Test P-V profiles used in the evaluation of MPPT performance: PV 1 (red line) describes the standard test conditions, and PV 2 (green line) is used to model hard shading scenarios Figure 16 shows two P-V profiles that were implemented employing a solar array simulator Agilent E4360A controlled by the LabView software. Due to the limitations of the testbench, the PV module was represented by its equivalent sectioned into four substrings with four bypass diodes. The profile PV 1 with one MPP is for the PV module operation under nominal operating cell temperature (NOCT) of 45 • C and uniform irradiance of 800 W/m 2 without shading. The profile PV 2 corresponds to the operating conditions derived from the profile PV 1 where cells in three substrings are shaded down to 200 W/m 2 . This is a reason for having two MPPs. Figure 17 shows the experimental waveforms acquired from the implementation of MPPT. The waveforms show the trend of voltage and current changes from the GMPPT scanning start until the converter operates around the GMPP. To show the dual-mode operation, the signal S 2 measured at the gate driver output is shown. The converter was controlled by microcontroller STM32F334 utilising Cortex-M4 core and highresolution PWM periphery.
For profile PV 1 , the AMC performed the switching between the modes twice. First, during the GMPPT scanning where the input voltage was lower than the threshold voltage of 20 V. Second, the converter switches from operation with the HWR to the baseline single-switch topology near GMPP. For the profile PV 2 , the AMC results in the switch S 2 turnon during the GMPPT scanning. The GMPP was set for low input voltage because of shading conditions. The GMPP was reached at the best efficiency due to the implementation of the dual-mode AMC. This means that the HWR was active during the GMPPT scanning at the input voltage of below 20 V. Near the GMPP, it was disabled to increase the efficiency.

CONCLUSIONS
In this study, a dual-mode galvanically isolated dc-dc converter has been proposed. It can utilise voltage step-up by means of both an isolation transformer and a coupled inductor utilised with half-wave and a VDR, respectively. The proposed topology uses adaptive mode change to define whether the HWR is enabled. As a result, both magnetic elements provide voltage step-up to the output at low input voltages, while only a highly efficient transformer is used at higher voltages. Implementation of the adaptive mode change has resulted in the extension of the input voltage range. Also, the efficiency has been improved for the low range of the input voltage. The efficiency of the converter in different modes was investigated experimentally. According to the results, the converter FIGURE 17 Voltage and current waveforms acquired during start-up, GMPPT scanning and transition to local MPPT (LMPPT) using the proposed converter when the solar array simulator is configured to reproduce P-V profiles: (a) PV 1 , and (b) PV 2 provides the best efficiency at a low input voltage when the HWR is configured to rectify the positive half-wave of the coupled inductor output voltage. The efficiency analysis shows that the converter can operate at the input voltage of over 35 V and rated power only if the HWR is disabled. In the experimental study, the efficiency increase of over 5% was observed at low input voltages owing to the implemented adaptive mode change.
The capability of the proposed converter for PV applications was examined by the implementation of a GMPPT. The proposed control system seamlessly utilises the adaptive mode change principle to optimise the converter operation during the sweep of the P-V curves to harvest the energy at the highest efficiency. The experimental results have shown that the converter is capable of harvesting the maximum available power from a PV module even at severe shading conditions.