An intelligent hybrid GMPPT integrating with accurate PSC detection scheme for PV system using ESSA optimized AWFOPI controller

Abstract

This study proposes a hybrid global maximum power point tracking (GMPPT) scheme integrating an extreme learning machine with 0.8V_oc technique for PV system. An attempt is made to employ an anti-windup fractional-order proportional-integral controller for the MPPT. The controller parameters were tuned using an enhanced salp swarm algorithm. The algorithm integrates via an accurate detection scheme that distinguishes partial shading conditions (PSCs) from an irradiance uniform change. Furthermore, the computed irradiance is used to update PV array open-circuit voltage (V_{oc_Array}), preventing temperature and irradiance sensors from being used. Its performance was studied compared with MPPT controllers, i.e., deterministic particle swarm optimization, hybrid PSO, and Lagrange interpolation PSO. The proposed MPPT technique proved its ability to track GMPP with an average tracking efficiency of 99.20% and 99.10% for uniform and PSCs, respectively. The proposed scheme has significant speed and accuracy in tracking GMPP for complex PSCs and uncertain weather conditions. Irrespective of the environmental uncertainties, it has an average voltage tracking percentage error within ± 1% for ten hours test profile. The proposed technique is explored on OPAL-RT 4510 platform. The results depict its ability in GMPP tracking with an average tracking efficiency and tracking time of 99.15% and 0.12 s, respectively.

Introduction

With the rapid increase in demand and energy dependence in modern society, much interest has been paid towards utilizing renewable energy sources (RESs). Solar photovoltaic (SPV) energy is widespread among all RESs because of its various advantages. There are various types of SPV systems based on applications [1]. They are (a) low power stand-alone systems, (b) remote small photovoltaic (PV) systems, (c) large-scale grid-connected systems, and (d) hybrid PV systems. In all types, maximum power point tracking (MPPT) is equipped with a power converter for reliable operation by ensuring optimal use of large PV arrays. The power-voltage (P-V) characteristics portray an erratic maximum power point (MPP) for different environment conditions (i.e., temperature and solar irradiance change) and create a challenge to the tracking process. When the array is experiencing partial shading conditions (PSCs), the process becomes more complicated. The PSC occurred when a part of the PV array experienced non-uniform irradiance while all other modules are fully illuminated. The P-V curve is shown many local peaks (LPs) and one global pick (GP) during PSCs. If the GP is not tracked precisely, the MPPT algorithm is confined at one of the LPs, leading to significant power losses [2].

The authors have proposed various MPPT techniques in the literature. These are widely categorized into conventional and soft computing (SC) techniques. A review of various MPPT techniques under both categories has been presented in [3], [4]. Some widely used methods under conventional methods are perturb and observe (P&O) [5], incremental conductance (IC) [6], fractional-short circuit current (FSCC) [7], fractional-open circuit voltage (FOCV) [7], ripple correlation control (RCC) [8], sliding mode control (SMC) [9], and mathematical graphical approach [10]. They are quite efficient in tracking MPP for uniform irradiance conditions and showing an excellent speed of convergence. However, every tracking technique has its advantages, where the on-going oscillation around the MPP is a severe disadvantage. The oscillation action in steady-state contributes to a significant power loss. Many works are being done to overcome this issue but achieved at the cost of reduced tracking speed. Furthermore, no technique of this type can handle PSCs [11].

The SC MPPT techniques overcame the above limitation. Some approaches to utilize SC-based MPPT controllers reported in details [12] are: artificial neural network (ANN) [13], fuzzy logic controller (FLC) [14], differential evolution (DE) [15], deterministic particle swarm optimization (DPSO) [16], ant colony optimization (ACO) [17], grey wolf optimization (GWO) [18], cuckoo search (CS) [19], flashing fireflies (FF) [20], jaya algorithm (JA) [21], genetic algorithm (GA) [22], artificial bee colony (ABC) [23], radial movement optimization (RMO) [24], P&O assisted flower pollination algorithm (FPA) [25], spline model guided MPPT [26], novel grasshopper optimization (GHO) for complex PSCs [27]. SC algorithms are usually more complicated than traditional methods, despite their versatility. For example, ANN needed proper, prolonged training and a costly microprocessor to implement to achieve accurate performance because of its computationally intensive nature. FLC has an outstanding convergence speed. Its performance depends on the programmer’s knowledge of the PV module and the environmental conditions under which the system is connected. A distinct advantage of the SC MPPTs is the entire PV curve can be examined efficiently and its suitability for handling PSCs.

However, they are known much slower than their traditional counterparts in tracking speed [16]. Modified SC algorithms are implemented to boost the MPP convergence. An MPPT based on the Lagrange interpolation PSO (LIPSO) technique [28]. This technique eliminates the conventional method's problems by using a numerical calculation to initialize the particles around the GP. This is generally referred to as hybrid MPPT. The hybrid MPPT, which involves traditional and metaheuristic techniques, has recently grown in popularity [29], [30]. The authors implemented a hybrid PSO (HPSO) MPPT algorithm [30] by combined P&O with the PSO algorithm. Since the entire searching space is scrutinized, the global maximum power point (GMPP) is not missed for a number of middle peaks. The HPSO is sluggish due to the random numbers that allow the particles to travel randomly despite its effectiveness [16]. Therefore, the time of convergence is significantly increased without proper guidance.

The ANN has to be trained from time to time to assure true MPPT for the PV array with which it will be used. The iterative-based ANN MPPT technique reported in [13] has low computational speed, requires more hidden nodes for more accurate results, and requires more training data to train the network, affecting the tracking performance. So, an effective learning method is essential to track the MPP precisely. For training the single hidden layer feed-forward neural network (SLFN), an extreme learning machine (ELM) technique is proposed in [31]. It has better generalization performance compared to the iterative ANN. It has no learning rates, iteration process since weights and biases are allocated arbitrarily, and the pseudo-inverse determines the output weights. Hence the process was getting simple and less complicated, and easy to implement than iterative ANN.

It can be observed that the MPPs' position for uniform shading modules is located in the vicinity of an integer multiple of 80% of Voc (n × 0.8 × V_{oc_Module}) of a PV module, where 'n' is an integer. This voltage-based MPPT technique's tracking speed is high, as there is no derivative computation required, fewer sensors requirements, more efficient, and has fewer losses [32]. This is a cost-effective and straightforward approach with just one feedback loop. Although the literature suggests numerous intelligent MPPTs [33], it is unforeseen to note that most techniques not reliably and consistently recognize the occurrence of PSCs [18], [29], [30]. While an incorrect decision is made but PSC does not occur, an undesirable search for a GP starts. This confusion leads to a decrease in efficiency as the operating voltage is forced to move around the P-V curve to find the GP. Despite the extensive work being done to resolve this problem, an effective way of differentiating PSCs and uniform radiation changes is still to be found. Without the scheme, MPPT transient efficiency is lesser by 30–35% [34]. Because of the above discussion, a hybrid combination of noniterative ANN with a classical voltage-based MPPT technique with an accurate PSCs detection scheme may lead to better tracking results. This needs further investigation.

Several controls are mentioned in the MPPT literature [35], [36], [37] for effectively operates the PVS on MPP. Some are integral (I), proportional-integral (PI), and proportional integral derivative (PID). An MPPT based on incremental resistance (INR) for the fuel cell system was proposed [35]. In this case, the integrator gain is used to adjust the error signal to an appropriate range. The downside of the proposed technique is the optimal values for the controller parameter are unclear. The PID and FLC controller were reported in [37] to trace the MPP under PSC and shown that the PID controller outperforms FLC. Nevertheless, there is no precise method for tuning those PID controller gains. Researchers recently reported [38], [39], [53] that controllers using fractional-order derivatives and integrals could achieve robust performance results superior to those obtained with conventional integer-order controllers [35], [36], [37]. A variable fractional order incremental conductance (VFOINC) algorithm is proposed [38] in combination with extenics variable step size (EVSS) tracking. Once entering the MPP, VFO is picked. The results reveal that the FO method has a better MPPT effect. A fractional-order integral (FOI) incremental conductance MPPT is proposed [39]. The outcomes indicate high accuracy of monitoring for substantial environmental changes. The integration of fractional-order control (FOC) with the traditional INC algorithm raises the tracking speed by 41.67% [39]. Recognizing the potential of FOC to enhance system performance, the authors have found the use of FOC in MPPT one of the emerging areas of research. However, when there is distortion in the control signal, system output degradation occurs due to the controller’s integral action, a windup phenomenon. This phenomenon increases the output signal settling times, as well as system instability [40]. Therefore, the integrator windup problem must be corrected to prevent degradation of control performance by tuning the controller ignoring the actuator saturation by adding an anti-windup (AW) compensator [41]. In the literature mentioned above survey, studies are limited to a conventional and FOC for the MPPT controller. There may be better flexibility for improving results if the AW scheme is combined with the FOC. Thus, it requires further study.

For optimal control operation controller needs to find its best order and optimum gain. Various optimization techniques are reported in the literature [36], [42] to serve the purpose. For the stand-alone PVS with the vector-based swarm optimization (VBSO) [43], the authors have designed an optimal MPPT controller to adjust the PID-controller parameters, and the results are compared. The results demonstrate that VBSO can adjust the controller gains better. But the VBSO algorithm uses an entirely random search space distribution coefficient with a uniform distribution. It increases the time the particles need to converge, therefore increases computation time. A novel heuristic salp swarm algorithm (SSA) [38] inspired by slaps swarming behavior when exploring and searching the oceans is proposed. SSA is effective in finding the optimal solution for the system than other swarm-based algorithms [44]. However, classic SSA results in low convergence rates and unsatisfactory results for higher-dimensional problems. The SSA search process has identified a lack of exploration and exploitation in convergence efficiency, which improved further for better convergence and optimal solutions. An enhanced salp swarm algorithm (ESSA) is proposed in [45] to improve the SSA poor performance compared to other algorithms.

Taking into account the above fiction and facts, the aims of this work are the follows:

• Develop a model of hybrid GMPPT technique by integrating ELM with 0.8V_oc (HELM) techniques for a 12S PVS.

• An accurate PSCs detection scheme is incorporated with the HELM technique for MPPT in complex PSCs using ESSA optimized anti-windup fractional-order proportional-integral (AWFOPI) controller.

• HELM based on optimized AWFOPI MPPT is validated in different scenarios with uniform and PSCs compared with MPPT methods (i.e., DPSO, HPSO, and LIPSO).

• Validation of the proposed technique under complex PSCs (i.e., constant, static, and transient/dynamic) shading patterns and typical tropic country profile is tested.

• The proposed technique is tested on a real-time digital simulator platform for different irradiance levels and loading conditions.

The paper is organized as follows: Section “System configuration” depicts the overall system depiction, PV modeling, and DC-DC converter. The detection scheme of PSC is described in Section “Detection scheme of PSC”. The developed MPPT technique is described in Section “Proposed MPPT technique”. Simulation results are given in Section “Simulation results”. The conclusion of this study is given in Section “Conclusion”.