1. Introduction
In recent years, Photovoltaic (PV) holds a pivotal position in ever-increasing energy demand due to easy accessibility, easy installation, high return on investment, and low maintenance cost [
1]. A complete PV system consists of a PV array, DC–DC converter (optional), DC-link (DCL) capacitor, inverter, filter, and a grid. Therefore, based on the devices, the PV system configuration is categorized into two types i.e., single-stage and two-stage configuration systems. A single-stage system consists of a PV array, DCL capacitor, inverter, filter, and a grid. In this system, the weight and size of the system are considerably reduced but complexity is greatly increased due to the handling of different functionalities (Maximum Power Point Tracking (MPPT), current control, voltage control, and grid synchronization) by the inverter alone. Moreover, for some applications, the PV voltage needed to be increased to the desired level that cannot be achieved without the use of a DC–DC converter. Therefore, to reduce the system complexity and widen its applications range, a DC–DC converter is introduced in the two-stage configuration system [
2]. A schematic of the 2-stage three-phase (3Φ) grid-connected PV system used in this research work is shown in
Figure 1.
In the 1st stage of the PV system, a DC–DC boost converter is used to extract the maximum power from the PV panel through the maximum power point technique and lift-up the voltage level according to the desired application. Generally, the classical converters such as boost, Single Ended Primary Inductor Converter (SEPIC), and cuk, etc. are used to attain a High Voltage Gain (HVG). For this purpose, a converter is needed to be operated at a high duty cycle, which has some disadvantages, such as (a) increased duty cycle causing lower efficiency due to increased losses in parasitic resistances of a diode, capacitor, and inductor; (b) as the duty cycle increases, the voltage stress on the switch increases; and (c) increased conduction and switching losses [
3].
To overcome the above-mentioned limitations, numerous researchers have designed different converter topologies that can be categorized into isolated and non-isolated topologies based on coupling [
4]. A transformer is used in isolated topologies to attain a HVG; however, due to the heavy weight of the transformer and its core and winding losses, they are not feasible for low power PV applications. Consequently, the authors used non-isolated topologies to overcome the issues in isolated topologies [
5,
6,
7,
8,
9]. A coupled inductor-based Switch Inductor (SL) configured high lift-up converters for renewable application is proposed in [
5]. Due to the usage of coupled inductors, the efficiency of these topologies is low due to the leakage inductances of the windings. The authors proposed single inductor-based converter topologies that can achieve a HVG while maintaining good efficiency [
6,
7,
8,
9]. In [
6], eight different SL-based converter topologies are presented, in which SL is configured with traditional boost, cuk, and SEPIC. Although HVG is achieved by most of these topologies, usage of high component count increases the operational complexity and cost of the converters. An SL and Switch Capacitor (SC)-based improved cuk converter is presented in [
7]. This topology is very attractive in terms of HVG but the utilization of two switches and a high number of magnetizing components causes an increment in the control, complexity, and cost. A single switch SEPIC-based modified converter topology is proposed in [
8]. However, even though a high voltage gain is achieved for continuous input current, the converter initially suffers from high conduction losses as voltage stress on the switch is equal to the output voltage. An SL/SC-based cascaded boost converter proposed in [
9] can achieve HVG with reduced component count but is unable to attain high efficiency. To overcome the limitations in the above-discussed topologies, in this research work, an SL configured hybrid topology that combines the Voltage Doubler Circuit (VDC) with the SEPIC converter is proposed. The most prominent features include high voltage conversion ratio, fewer components, single switch, low voltage stress on semiconductor components, and contiguous input current.
In the second stage of the PV system, besides DC–AC inversion, the functionality of the inverter also involves grid synchronization, current control and protection, and DCL voltage regulation, etc. Therefore, for the smooth and stable performance of the PV system, the inverter control plays a prominent role and is usually implemented in the form of two cascaded loops i.e., outer Voltage Control Loop (VCL) and inner Current Control Loop (CCL). A VCL is responsible for maintaining a constant DCL voltage and generates a reference for the current loop. A CCL is responsible for grid synchronization and injection of a high-quality current [
2]. Generally, a Proportional Integral (PI) controller is implemented in the VCL, however, due to PV intermittency, a PI controller fails to maintain a constant DCL voltage due to its fixed gain parameters [
10]. Therefore, to enhance the system performance, an arbitrary real term is introduced in the PI controller through fractional calculus and named the Fractional Order Proportional Integral (FOPI) controller [
11]. Although the system performance is improved, high ripple contents and distortion from the reference are observed during uncertainties (PV intermittency). A Sliding Mode Controller (SMC) for DCL regulation in proposed in [
12]. This controller has a fast dynamic response but the waveforms suffer from chattering phenomena. To enhance the SMC performance and smoothen the shattering effect, the authors introduced an integral term in the SMC [
13]. The performance of the PV system is enhanced to a great extent by using fuzzy PI and fuzzy SMC controllers in [
14]. However, these controllers have complex architectures and failed to achieve high-quality waveforms. Due to the above discussed limitations in the DCL controller topologies, there is a need to develop a controller that adopts the system uncertainties and ensures the system stability. Therefore, such a controller is needed to be designed that adopts the system dynamics and regulates a DC-link voltage. Hence, in this research work, a Back Propagation (BP) algorithm-based Adaptive Proportional Integral (API) and Adaptive Fractional Order Proportional Integral (AFOPI) controllers are designed. The parameters of these controllers will be updated automatically, in case of any environmental variation or disturbance, and achieve a regulated voltage at a DC-Link.
To maintain a high-quality grid injected current, PI or Proportional Resonant (PR) controllers are usually applied in the inner current control loop. However, when a PI controller is applied in a stationary frame, it cannot track the sinusoidal reference accurately; therefore, the system variables need to be transformed into a synchronous frame that requires large calculations [
15]. On the other hand, a PR controller is highly suitable to track the sinusoidal reference and thus, making it an optimal solution for inverter-based grid-connected PV applications. It provides an infinite gain at the selected resonant frequency (at other frequencies there exists almost no gain), and thus, the zero steady-state error can be easily attained [
16]. Although it has a high tracking capability, the output current of the grid connected inverter is not invulnerable for harmonic content that usually occurs due to the non-linearities in converter and inverter. Therefore, to reduce or eliminate the harmonic contents, a selective Resonant Harmonic Compensator (RHC) can be added to a PR controller [
16]. The performance of the system is greatly improved by using a PR controller and can be found widely in the literature. However, most of the researchers did not explain the design and tuning procedures of their respective controller in detail, and the trial-and-error method is adopted to find the optimal gain values. Although some of the researchers described the design procedure in a continuous-time domain, most of the Grid Connected PV Inverter (GCPVI) applications involve digital implementation. Therefore, the authors in [
17] proposed a digital PR controller that has a simple control architecture and tuning methodology. Compared to the conventional methodology, which involves complex trigonometric equations, this work presents a systematic procedure to calculate proportional gain (
kp), resonant gain (
kir), and the coefficients of the resonant path. However, it shows an effective performance, although it is limited to 1Φ grid connected system and is developed by considering the dynamic behavior of LC filter. However, most of the grid connected power systems involves 3Φ operation and operates at a low frequency. At low frequency the size of L or LC filters becomes very bulky and expensive as a result LCL filters being introduced. In addition to the cost and size, LCL filter offers numerous other advantages such as: (a) smoothens the inverter output current, (b) low Total Harmonic Distortion (THD), (c) improved performance, and (d) and high attenuation [
18]. Due to these advantages, in this research work, a PR controller proposed in [
17] is modified by considering the dynamics of LCL filter and implemented in a 3Φ grid connected PV system.
The core contributions of this work are listed as:
A transformer-less high voltage gain DC–DC converter is presented that extracts the maximum power from the PV panel by using Perturb and Observe (P&O) MPPT technique and step-up the low generated PV voltage to a level applicable for grid connected PV inverter. In the proposed topology, a voltage doubler circuit and switch inductor cell are integrated with the SEPIC converter to attain a high voltage conversion ratio. The prominent features of the converter include single switch, simple control, low voltage stress, high efficiency, and fewer components.
Back propagation algorithm-based API and AFOPI controllers are designed to regulate a DC-link voltage in case of PV intermittency. This adaptive scheme enhances the system performance and handles the system uncertainties by dynamically updating the control law parameters. The proposed controllers have a faster dynamic response and easy implementation, which improves the system stability.
Based on the Naslin polynomial method, a PR controller is designed, considering the dynamic behavior of an LCL filter. In the proposed controller, no parameter is computed empirically or by trial-and-error method. To show the effectiveness of the controller, it is implemented in a 3Φ grid-connected PV system that results in low THD.
The rest of the manuscript is divided into the following sections as: A schematic and detailed analysis of the first stage of the PV configuration system, including MPPT algorithm and DC–DC converter, which are elaborated on in
Section 2, while the second stage that involves DC-link voltage and current controllers are discussed in
Section 3. The performance validation of the proposed system is described in
Section 4. Finally, this research work is concluded in
Section 5.
2. Analysis of DC–DC Converter
Due to the dependency of the Maximum Power Point (MPP) of the PV array on climatic parameters, the position of MPP varies indefinitely. Therefore, to precisely locate the MPP, numerous MPPT algorithms were reported in literature, that vary from each other in various aspects such as complexity, cost, type of sensors required, implementation, and range of effectiveness [
19]. In this research, a P&O technique is adopted due to its simplicity and easy implementation, as presented in
Figure 2a [
19]. According to the operating voltage and current, the P&O technique generates a Pulse Width Modulation (PWM) for the converter as shown in
Figure 2b. Generally, the PV system generates a low-level voltage that is needed for amplifying the appropriate operation of the inverter; therefore, a step-up converter is designed, which is discussed below. Moreover, the specifications of the PV panel and the gain parameters of the PI controller used to drive the converter switch are listed in
Table A1 and
Table A2, respectively.
2.1. Proposed Converter Topology and Operating Modes
A schematic circuitry of a single switch high lift-up DC–DC converter is presented in
Figure 3. It consists of an input voltage source (
UPV), switch (
S), diodes (
D1,
D2,
D3,
D4,
D5, and
D6), inductors (
L1,
L2, and
L3), and capacitors (
C1,
C2, and
CDC). In this converter, the VDC and SEPIC converter are arranged in such a manner that it enhances the voltage conversion ratio. Furthermore, the primary inductor of the SEPIC converter is replaced with the SL unit (
L1,
L2,
D1,
D2, and
D3) to further enhance the voltage conversion characteristics. The inductors
L1 and
L2 of SL unit have the same magnitude and are always connected to the input source, therefore, the converter input current will always be continuous in nature.
The proposed converter can operate in two operating modes, i.e., Continuous Conduction Mode (CCM) and Discontinuous Conduction Mode (DCM). The CCM can be divided into two switching intervals, while a DCM can be divided into three switching intervals. A detailed analysis of CCM and DCM are discussed below in detail. However, for simplicity, the following considerations are made in theoretical analysis: (a) the Equivalent Series Resistances (ESRs) value of all components used in converter are not taken into account during the analysis of operating modes, (b) the capacitors are large enough so that they act as a constant voltage source, (c) the leakage inductances have a low value and are not considered in analysis.
2.1.1. Continuous Conduction Mode
Based on the switch characteristics of the converter, its operation is discussed for two switching intervals, i.e., (
t0 ~
t1) (
S is ON) and (
t1 ~
t2) (
S is OFF). The equivalent circuitries of both switching intervals along with the direction of current flow (red arrows) are presented in
Figure 4a,b respectively, whereas its typical waveforms in steady state are sketched in
Figure 4c.
In this switching interval, switch (
S) is turned and diodes (
D1,
D3, and
D5) are in forward conduction mode, while the diodes (
D2,
D4, and
D6) are in reverse biased state as presented in
Figure 4a. The inductors (
L1 and
L2) will be charged from input source (
UPV) through
D1 and
D3. As
L1 and
L2 have the same magnitude, the same amount of current will therefore be flowing through them and is charged in parallel (i.e.,
UL =
UL1 =
UL2, and
iL =
iL1 +
iL2). The inductor (
L3) is also charge in this interval and
UC1 −
UC2 is applied to it. Moreover,
C2 discharges to charge
C1 and
L3. The capacitor
CDC also discharges, and its stored energy is used to feed the load.
During this interval, diodes (
D1,
D3, and
D5) and switch (
S) are turned OFF, while the diodes (
D2,
D4, and
D6) are in forward conduction mode as presented in
Figure 4b. As
D1 and
D3 are in reverse biased and
D2 is in forward biased,
L1 and
L2 therefore act as a series connected circuit (i.e.,
iL1 =
iL2). The stored energy in
L1 and
L2 are used to charge
C2 through
D4 and also transfer its energy to support the load through
C1 and
D6.
L3 also discharges in this interval to charge
CDC and support the load.
2.1.2. Discontinuous Conduction Mode
Based on the switch characteristics, the operation of the converter in DCM is divided in three switching intervals, i.e., (
t0 ~
t1) (
S is ON), (
t1 ~
t2) (
S is OFF), and (
t1 ~
t2) (
S is OFF). The typical waveforms of proposed converter in steady state operating in DCM are presented in
Figure 5a.
In this switching interval, switch (
S) and diodes (
D1,
D3, and
D5) are in forward conduction mode while all other diodes are in reverse biased as shown in
Figure 4a. The inductors (
L1, and
L2) stores energy by using input voltage source while
UC1 −
UC2 is applied across
L3. This mode ends when the current flows through switch S and diodes (
D1,
D3, and
D5) becomes zero at
t1.
In this interval, the diodes (
D2,
D4, and
D6) are in forward biased while all other semiconductor components are in reverse conduction mode as shown in
Figure 4b. The inductors discharges in this interval to charge the capacitors and supports the load. This mode ends when the current flowing through diodes (
D2,
D4, and
D6) becomes zero.
During this interval, switch (
S) remains in turned OFF state and all the diodes (
D1,
D2,
D3,
D4,
D5, and
D6) are in reverse conduction mode as presented in
Figure 5b. The voltage that appears across inductors (
L1,
L2, and
L3) becomes zero and the current flowing through inductors is continued as constant. During this mode, the load is supported by the output or DC-link capacitor (
CDC). This mode ends when the switch (
S) is turned ON and returns to the first operating mode.
2.2. Performance Evaluation of Converter
In this section, different characteristics such as voltage stress, voltage gain, design of inductors and capacitors, and loss analysis of the proposed converter in CCM mode are discussed. Moreover, a comparative analysis of proposed topology with other topologies is also presented.
2.2.1. Voltage Gain Derivation
The output voltage (
UO) of the converter is given as,
as it is assumed that the converter is operating in a CCM and therefore that the average inductor voltage at steady state condition is zero. Therefore, by applying inductor volt-second balance theorem over one switching cycle to
L (
L1 =
L2 =
L) we produce,
where
D is the duty cycle of the switch. By solving (2), we produce the capacitor voltage (
UC2) as,
Similarly, by applying inductor volt-second balance theorem to
L3 over one switching cycle we produce,
Solve Equation (4) for
UC1 we produce,
Put (3) and (5) in (1) and rearrange them, we produce a voltage gain of the proposed converter in CCM as,
2.2.2. Voltage Stress Calculation
The voltage stress on the semiconductor components can be calculated as,
From (7), it can be observed that the proposed topology exhibits lower voltage stress compared to the classical converters, thus results in lower conduction losses.
2.2.3. Design of Inductors and Capacitors
The design of inductors is based on the value of current ripples (Δ
iL), and is considered between 5–10% of the nominal output current to achieve optimal inductor size [
20]. Therefore, the inductances can be calculated as:
where
fsw is the switching frequency and
iL is the inductor current and can be calculated as:
where
iO is the output current of the converter. The capacitors (
C1 and
C2) are designed by using the voltage ripples (Δ
UC), which are assumed to be 5–10% of the output or DC-link voltage of the capacitor [
20]. Hence, the capacitances can be calculated as,
where voltage ripples (Δ
UC) is given as,
The main functionality of the DCL capacitor is to store or supply the surplus power demand during transience and to lower the DC ripples [
13]. It is designed according to [
21] as,
where
PDC is the DCL power, Δ
UDC is the peak-to-peak voltage ripples, and
fg is the grid frequency.
2.2.4. Loss Analysis
Equivalent circuitries of proposed converters, including the parasitic resistances in switching interval (
t0 ~
t1) and (
t1 ~
t2) are presented in
Figure 6a,b respectively. In the
Figure 6a,b, the parasitic resistances of the components are considered as; (a)
rONS is the resistance of the switch in turn ON state, (b)
rD1,
rD2,
rD3,
rD4,
rD5, and
rD6 are the ON-resistances of diodes
D1,
D2,
D3,
D4,
D5, and
D6, respectively, (c)
UD1,
UD2,
UD3,
UD4,
UD5, and
UD6 are the threshold voltages of
D1,
D2,
D3,
D4,
D5, and
D6, respectively, (d)
rL1,
rL2, and
rL3 are the equivalent ESRs of inductors
L1,
L2, and
L3, respectively, and (e)
rC1,
rC2, and
rCDC are the ESR of capacitors
C1,
C2, and
CDC, respectively.
The copper loss of inductor is given as,
where
iL−A and
iL3−A are the average current of
L (
L1 =
L2 =
L hence
rL1 =
rL2 =
rL) and
L3, respectively. Solving (17) would yield us to:
The power loss in the capacitor mainly depends on the Root Mean Square (RMS) value of the current flowing through capacitor and ESR of capacitor. Moreover, as
C1 and
C2 are of same rating thus
rC1 =
rC2 =
rC. The power loss can be calculated as,
Solving (19) in term of duty cycle would give us:
There are two types of switch losses, i.e., conduction loss and switching loss and can be given as:
The switching loss depends on the turn ON and OFF time, peak current (
iS−P), switching frequency, and voltage appeared across switch (
US). Hence, the switching power loss can be given as:
Putting values in (22) and solve in term of duty cycle would give us:
The conduction loss of the switch can be calculated as:
where
iS−R is the RMS current of
S.
The forward voltage drop and the conduction loss are the two types of losses associated with the diode.
The forward voltage drop loss is given as:
where
iD1−A to
iD6−A are the average current flows through diodes
D1 to
D6, respectively. Solving (26) would give us:
The conduction loss of diodes is calculated as,
where
iD1−R to
iD6−R are the RMS current of diodes
D1 to
D6, respectively, where
rD is the equivalent ON-resistance of diodes. Solving (28) in term of duty cycle would yield us to:
The total loss of the proposed converter can be calculated as:
The efficiency of the converter is obtained as:
2.2.5. Comparative Analysis
A comparative analysis of the proposed converter with the conventional and newly developed topologies on the basis of some main indicators such as capacitors, inductors, diodes, and switches count, are listed in
Table 1. Moreover, the comparative analysis of the voltage conversion ratio and voltage stress on switch are sketched in
Figure 7a,b, respectively. A topology presented in [
6] has used one more component than the proposed converter but it cannot attain a high voltage gain as at 90% duty cycle its voltage conversion ratio is 18 and its switch also suffers from high voltage stress. A topology presented in [
8] attains a high conversion ratio (i.e., at 90% duty the conversion ratio is 27) as compared to [
6] but low compared to the proposed topology, despite it using one more inductor. However, the voltage stress across the switch is lower compared to the proposed converter. A converter proposed in [
22] attains a low voltage conversion ratio and a switch suffers from a very high voltage stress. Moreover, the topologies proposed in [
23,
24,
25] attain a high voltage gain compared to the proposed one, i.e., at 90% duty cycle they all attain a voltage gain of 37, while the proposed converter attains a conversion ratio of 36.1. However, these topologies have used more components and the switch suffers from high voltage stress. The topology in [
26] has low component count and voltage stress on a switch compared to the proposed converter but it is unable to attain a high voltage conversion ratio. From the above discussion and from
Figure 7a,b, it can be observed that most of the topologies have low voltage conversion ratio and high voltage stress on a switch as compared to the proposed topology.
5. Conclusions
In this research work, a single switch non-isolated hybrid DC–DC converter is proposed. It attains a high voltage conversion ratio without the use of a transformer or extreme high duty cycle. The usage of a single switch makes its control simple hence the cost is considerably reduced. Moreover, due to hybrid structure and low voltage stress on semiconductor components, a proposed converter attains high efficiency compared to classical converters at high duty cycle. The operating principle, steady-state and comparative analysis of the proposed converter are also discussed. From the comparative analysis and simulation results, it is concluded that the proposed converter attains a high step-up conversion ratio with a low duty cycle, and reduced component count.
In the 2nd stage of the PV system, API and AFOPI controllers are employed in the outer voltage control loop to regulate the DC-link voltage in case of PV intermittency. Compared to the conventional PI and FOPI (having fixed parameters), the proposed controllers ensure (a) high tracking accuracy and efficiency, (b) fast dynamic response, (c) reduction in steady-state error, (d) reduced fluctuation in DC-link voltage during the transient state, and (e) less sensitivity to sudden disturbances. Furthermore, a PR controller is used in the inner control loop. In the proposed PR controller, there is a unique equation for every parameter, hence no parameter is computed empirically or by a trial-and-error method such as a conventional PR controller. The Bode diagram of the controller is plotted according to pre-indicated desired parameters. From the result, it is concluded that the controller accurately follows its reference with zero steady-state error, unpredictable behavior, and results in high-quality grid injected current.