1. Introduction
Due to the increasing global demand for energy, environmental concerns, air pollution from fossil fuel use, and rapid depletion of such fuel, renewable resources have attracted attention. Solar energy attracted a great deal of research attention as an unlimited, safe, CO
2-free, and clean alternative energy resource. For example, in [
1], some grid mix scenarios with increased photovoltaic (PV) capacity were assessed that improved energy use and reduced greenhouse gas emissions. In [
2], life-cycle analysis (LCA) was used to assess the environmental aspect of future PV systems. As an effective way to increase extracted power from PV panels, the MPPT approach and an appropriate control approach are required.
MPPT algorithms are essentially categorized based on two distinct operating conditions, i.e., rapidly changing irradiance and partial shading conditions. Under rapidly changing irradiance, a single MPP exists which is tracked by the MPPT algorithm. On the other hand, during partial shading conditions, two or more MPPs exist and the role of the MPPT algorithm is to track the GMPP. In the proposed study, the authors have developed a MPPT algorithm for rapidly changing irradiance conditions.
The configuration, control method, and characteristics of different MPPT methods were examined in [
3], some of which are as follows. The constant voltage method is the simplest MPPT method and in each time instant it compares PV voltage with the reference voltage obtained under standard test conditions to change the duty ratio for MPPT [
4]. Its disadvantage is not paying attention to radiation and temperature. The open-circuit voltage method compensates for the shortcomings of the constant voltage method. Since MPP voltage falls within a certain range of the open-circuit voltage, the MPP voltage is approximated based on the open-circuit voltage [
5]. The short circuit current method is another MPPT method that approximates the MPP current using the characteristic that the MPP current is generally present between 78% and 92% of the short circuit current [
6]. The load is separated to measure the open-circuit voltage and the short circuit current is measured by shorting the load according to a certain period or condition. The main disadvantage of the last two methods is reducing the efficiency because of opening or shorting the load periodically. The constant voltage, open-circuit voltage, and short circuit current methods have simple algorithms and implementation but low efficiency [
3]. In the Perturbation and Observation (P&O) method, the current or voltage is perturbed continuously to find the direction of perturbation that results in increasing power [
7]. If radiation conditions change rapidly, the P&O method cannot track MPP accurately. To overcome this disadvantage, the Incremental Conductance (INC) Method was proposed. The INC method uses the power derivative with respect to voltage, namely, slope in the P–V curve. The MPP position is determined by the magnitude of the slope [
8]. The last two methods are the most commonly used methods for MPPT, because of their accurate MPP tracking despite the environmental change. Their high dependence on the changing value is one of their disadvantages. If the changing value is large, the tracking speed will be fast and the vibration will happen in the steady state, which leads to dramatically reduced efficiency. A small changing value results in decreasing the steady-state error and slowing the tracking speed. Gradually increasing or decreasing from the extremum value during initial operation is another disadvantage, because these methods are controlled in a predetermined order [
3]. Fuzzy control as an artificial intelligence approach is used to improve MPPT performance [
9]. Fuzzy control is able to handle nonlinear systems and needs no accurate model. Selecting the slope of the P–V curve and its changes as the fuzzy inputs and the control amount as the output is seen in the literature [
3]. The neural network is another artificial intelligence approach used for improving MPPT performance [
10]. The output of the neural network is power at MPP or voltage and current at MPP. The open-circuit voltage, short circuit current, and PV power or temperature and radiation are examples of neural network inputs. The shortcomings of the last two methods are their dependence on the human experience and requirement of a high-performance controller. A PV system traditionally operates in normal MPPT mode, while a derated PV system follows a flexible power point tracking algorithm to create virtual energy storage. In [
11], a neural network is used to determine the point of operation for the flexible power point tracking of the PV. The simulation results showed an improved system inertia response. The combination of P&O and open-circuit voltage [
12] and the combination of proportional-integral (PI) controller and fuzzy control method [
13] were proposed to improve MPPT performance. The different MPPT methods can be combined to overcome their disadvantages and benefiting from their strengths [
3]. In [
14], the authors proposed a drift-free P&O boost converter- based MPPT controller. The algorithm compares the sign of change in power, current, and voltage to detect a change in irradiance. In [
15], the authors proposed a weighted set point similarity method in which four consecutive duty cycles are stored to determine the direction of tracking. The algorithm uses an upper and lower power boundary limit which are iteratively reduced to converge towards the MPP. In [
16], an adaptive incremental resistance method has been proposed that shifts the operating point to RHS of the I–V curve under sudden variations in irradiance and load resistance. Although the algorithm promises a high tracking speed, it can deviate from the MPP tracking path under continuously changing irradiance.
In recent years, researchers have focused on developing improved MPPT algorithms to maintain a balance between the speed of tracking and the steady-state power oscillations [
17]. Moreover, every time a change in irradiance occurs, it shifts the MPP to a new set of I–V curves. Hence, the MPPT algorithm needs to constantly track this MPP regardless of fast or slow changes in atmospheric conditions [
18]. Solar PV inverters developed by manufacturers require the MPPT algorithm to be simple, cost-effective, and able to track the GMPP under partial shading conditions [
19,
20].
In [
21], two conventional and advanced categories are mentioned for Sliding Mode Control (SMC)-based MPPT algorithms. In all these methods, the main objective is to provide the maximum power continuously regardless of environmental conditions or load variations.
The first category includes conventional SMC approaches. First-order SMCs are one of the approaches of this category with a low calculation volume, easy practical implementation, and fast response, and without the reference signal to track MPP. The major drawback of this category is that the control signal across the sliding surface is discontinuous because of using the sign function. For the sake of duty cycle control, switching frequency should be high enough that it results in the phenomenon of chattering. In [
22], a first-order SMC was proposed and implemented with a fast and robust response under varying conditions, but it suffered from oscillation around MPP. P&O-based SMC is another conventional SMC-based MPPT technique with two loops. The internal loop is responsible for tracking the reference signal that the P&Q algorithm as an external loop generates for MPPT. As reported in [
23], the main problem of P&O-based SMC is chattering. INC-based SMC is another conventional SMC-based MPPT technique that benefits from the fast response in changing irradiation of the INC algorithm. In [
24], it was shown that the amount of chattering with the proposed variable-step INC-based SMC was less than that of the P&O algorithm. Another category approach is a linear-expression-based SMC that uses a curve-fitting technique to derive a reference signal. In [
25], reference voltage was generated using regression and the backstepping algorithm for minimizing the error between PV voltage and the reference voltage. In [
26], the reference current was defined as a linear expression, and the reference voltage was calculated based on that. A backstepping SMC (BSMC) was designed using Lyapunov stability criteria to ensure system stability and using a linear expression for the reference current. Its performance was better than that of Terminal SMC (TSMC).
The second category of SMC-based MPPT algorithms, namely, advanced SMC, includes TSMC. In [
27], an improved fast TSMC was employed for MPPT in the PV system, including a single-ended primary-inductor converter for fast convergence with higher output power and lower oscillations.
Artificial Intelligence (AI)-based SMC is another advanced SMC-based MPPT technique that uses the advantages of AI algorithms to generate the reference signal or optimize the controller parameters. In [
28], the P&O-fuzzy-logic controller generates the reference voltage, and BSMC tracks the reference voltage. The sliding surface is defined as the weighted combination of the error between current, its reference, and its integral. The control input contains a sign function. In [
29], PSO optimizes P&O-based SMC gains. The sliding surface is defined as the weighted combination of the error between current and its reference, voltage and reference, and derivative. The SALP Swarm Algorithm is used in [
30] to determine the value of the duty cycle optimally to track MPP by maximizing the output power of the photovoltaic system. The authors took into account the effect of partial shade and different atmospheric conditions in their optimization. Their results showed the superiority of their proposed method compared to P&O.
Super twisting algorithm-based SMC is another advanced SMC-based MPPT technique that uses the higher-order SMC to eliminate the chattering of the lower-order ones without compromising its performance. Tracking is accurate, but implementation is more complex using this method.
Despite the desirable performance, the second category suffers from implementation complexity. In [
26], it is shown that the performance of BSMC is better than that of TSMC. Hence, in this paper, we consider BSMC, presented in [
26], as the base for our proposed algorithm. As a significant contribution of this paper, we develop an FBSMC approach in which the control parameters are optimized using the PSO algorithm to solve the chattering problem of BSMC in MPPT of PV panels. The performance of the proposed FBSMC is compared with the BSMC [
26], ABC-based PI control [
31].
In [
3], various methods of MPPT for PV systems were reviewed. Among these methods, BSMC was one the methods with more advantages and fewer drawbacks. Therefore, it is used as the base for our proposed controller. It is obvious that chattering is the main drawback of SMC methods, including BSMC. Here, to eliminate this effect, a fuzzy system, inspired by [
32], is suggested to provide soft switching based on the ratio between power variation and the voltage variation of the PV panel in two consecutive sampling times, as well as its variations in two consecutive sampling times. Combining the structures of [
26,
32], we benefited from the advantages of both. However, their drawbacks are weakened dramatically. On the other hand, this combination increased the design parameters and complicated the selection of them by trial and error. In this paper, PSO as an old, well-known optimization method is used to optimize the design variables. Combining these three high-performance methods (BSMC, fuzzy, and PSO) to use the advantages of each one and attenuate their drawbacks is the main contribution and novelty of this paper.
The structure and performance of the present controller is compared to some of the reviewed controllers in
Table 1.
To accomplish this work, it was assumed that the model of the system is known and available and no identification process is considered.
The paper structure follows:
Section 2 includes the structure of the solar cell, its mathematical model, and the MPP search method for calculating the MPP voltage. In
Section 3, a fuzzy backstepping sliding mode controller is introduced. In
Section 4, the simulation results of the different temperature and solar radiation scenarios are given. They are analyzed quantitatively and qualitatively. Finally, in
Section 5, the conclusion is presented.
4. Case Study
The test system of this paper includes a PV module, a DC-DC boost converter, and a load resistance. Similar to [
35], the PV module considered in this paper is KC200GH-2P, consisting of 54 series-connected polycrystalline cells. The simulation is implemented in a MATLAB environment. The electrical characteristics of this PV module are shown in
Table 3 under standard conditions, that is,
and solar radiation (
) of
. The parameters of the DC-DC boost converter and resistance load are given in
Table 4.
The MPP voltage is the reference input, which is calculated online.
4.1. Simulation Results
Simulation was performed for several scenarios:
Scenario 1 (Normal operating conditions): In this scenario, the temperature is
and solar radiation is
.
Figure 4 compares the voltage, current, and power of the PV panel controlled with the BSMC [
26], ABC-based PI control [
31], and FBSMC. For rational comparison, the required parameters of all three controllers are optimized using PSO.
The simulation results show approximately equal rise time and better transient response in terms of less settling time, less error between the reference values of the current and voltage, and more extracted power by applying the FBSMC approach compared to the other controllers. In other words, faster convergence to MPP is obtained using this controller. Applying the ABC-based PI control [
16], the settling time becomes less than the BSMC [
26], while the extracted power is less. The results are presented for 0.2 s. As the simulation results show, the transient state is less than 0.2 s, which demonstrates how fast the proposed controller can adapt itself and reach the steady state.
It is obvious that in the control problem transient state response is as important as steady state response. In this scenario all the controllers have almost the same steady state response and superiority of the controller proposed in this paper is shown in terms of the transient response and how fast and smoothly it reaches the steady state.
Scenario 2 (Constant Temperature and variable solar radiation): The temperature is equal to
and the solar radiation increases linearly from
at
to
at
.
Figure 5 compares the voltage, current, and power of the PV panel of the controlled system with the BSMC [
26] and FBSMC.
The simulation results show that by applying the FBSMC approach in comparison to the other controllers, the MPP is tracked faster and more accurately. In other words, the best adaptation to the rapid variation in solar irradiation is obtained by this controller. The chattering phenomenon is eliminated using the fuzzy system in comparison to BSMC. Also, because of the nonlinear nature of the system, BSMC [
26] as a nonlinear controller outperforms ABC-based PI control [
31].
Simulations show that FBSMC has smoother behavior than BSMC while more power is gained. It demonstrates successful performance of the proposed controller using fuzzy.
Scenario 3 (Constant solar radiation and variable temperature): The solar radiation is equal to
. The cell temperature increases from
at
linearly to
at
.
Figure 6 compares the voltage, current, and power of the PV panel of the controlled system with the BSMC [
26] and FBSMC.
The simulation results illustrate that a faster and more precise convergence to MPP and faster adaptation to temperature variation is obtained by applying the FBSMC compared to the BSMC.
Simulations show that FBSMC has faster and smoother behavior than others while more power is gained. It demonstrates the superiority of the proposed controller in terms of transient response.
Scenario 4 (Partial shading): Natural conditions or obstructions in the passage of irradiation result in partial shading. Under partial shading conditions, the sunlight reaches only part of a PV array. To simulate partial shading, two KC200GH-2P PV modules are connected in parallel, the temperature is equal to
and the solar radiation for one of them is
while for the other it is
.
Figure 7 compares the voltage, current, and power of the PV panel of the controlled system with the BSMC [
26] and FBSMC.
Simulations show that FBSMC has the fastest adaption to environmental changes with a smooth trend while more power is gained.
Scenario 5 (Constant Temperature and variable solar radiation): The temperature is equal to
and the solar radiation changes according to
Figure 8.
Figure 9 compares the voltage, current, and power of the PV panel of the controlled system with the BSMC [
26] and FBSMC. Simulations demonstrate the superiority of FBSMC in terms of transient response while more power is gained.
The simulation results show that by applying the FBSMC approach in comparison to the other controllers, the MPP is tracked faster and more accurately. In other words, the best adaptation to the rapid variation in solar irradiation is obtained using this controller. The chattering phenomenon is eliminated, smoother behavior is achieved and transient response is improved using the fuzzy system in comparison to BSMC. Also, because of the nonlinear nature of the system, BSMC [
26] as a nonlinear controller outperforms ABC-based PI control [
31]. The superiority of FBSMC in particularly under partial shading conditions is demonstrated.
In all cases, in simulation time, the system passes its transient state and reaches its steady state. In other words, the settling time is less than the simulation time. This demonstrates how fast the proposed controller can respond to different changing situations.
4.2. Quantitative Analysis
To quantitatively evaluate the performance of the proposed control design, quantitative indices of the root mean square error between PV voltage and the reference voltage,
, and the root mean square error between PV power and maximum power,
are used as follows, respectively:
The dynamic efficiency (
) is calculated using the following equation as in [
36]:
The above-mentioned quantitative indices, mean power,
, and settling time are compared for the controlled system with the FBSMC, BSMC [
26], and ABC-based PI control [
31] in
Table 5.
In all scenarios, FBSMC outperforms the others in terms of the extracted power and dynamic efficiency. The power extracted using BSMC is more than that of ABC-based PI control. In the first scenario, the settling time decreases using BSMC, and the fuzzy system, respectively. In the second and third scenarios, where the climate condition changes rapidly, the ABC-based PI control cannot track MPP, as well as the others, and errors increase dramatically.
4.3. Robustness Evaluation
To evaluate the robustness as the performance of the controller under normal operating conditions, variations in the resistance load,
, and parameters of DC-DC converter components including capacitor capacitance,
, and inductance of the inductor,
, are considered.
Figure 10 and
Figure 11 show the PV power for a 20% increase and a 20% decrease in the above-mentioned parameters, respectively.
Table 6 and
Table 7 compare dynamic efficiency applying the BSMC [
26] and the FBSMC for a 20% increase and a 20% decrease in the parameters mentioned above. The results demonstrate the robust performance of BSMC and FBSMC while ABC-based PI control is not robust. Despite the substantial changes in the parameters, the FBSMC maintains its superiority over the others and has a lower variation percent in power extracted, which verifies its robust performance.
It is vital that combining three methods, in spite of improving its performance, increases the complexity of the proposed method which in turn makes its real implementation challenging, although tracking MPP in a smooth manner in variable environmental conditions is motivating. In other words, for real implementation the tradeoff between better performance and adaption with complexity have to be considered.