Abstract
This paper presents a mathematical model of 255 kW grid-connected solar photovoltaic (SPV) system. To study the performance characteristics of the grid-connected SPV system, a new hybrid adaptive grasshopper optimization algorithm with the recurrent neural network (AGO-RNN) control technique was implemented. Furthermore, the power quality at the point of common coupling (PCC) has been studied using the conventional (PSO) and proposed AGO-RNN controllers. The characteristics of the PV system were analyzed under varying environmental (variable irradiance and temperature) conditions considering 3 different cases such as (i) standard test conditions (STC), (ii) variable radiation with constant temperature, and (iii) variable radiation with variable temperature. For each case, the total harmonic distortion (THD) has been calculated using the proposed AGO-RNN control technique, and the results were compared with particle swarm optimization (PSO) technique. The 255 kW PV model is initially developed and connected to a three-level NPC inverter, an MPPT-based perturbation and observation algorithm. Later, the PV model is controlled by an AGO-RNN pulse width modulation (PWM) controller and is then integrated to the main grid at PCC. The main advantage of this technique is exploiting the separate DC-DC converter between the SPV module and the inverter. Finally, the proposed grid-connected SPV system was simulated on MATLAB for analyzing the performance of the system based on its I-V and P-V characteristics, inverter voltage, grid power, gird voltage, grid current, power factor, and THD under different environmental conditions. The simulation results demonstrate that the current magnitude and THD of the SPVGC system are improved with the cutting-edge AGO-RNN controller compared to PSO in all three different scenarios, and this value is less than 1.6%, which is within the permitted limits of IEC 61727 standards.
1. Introduction
1.1. Background and Motivation
In recent years, there has been a great interest in selling electricity generated from photovoltaic power plants. The installation of the PV plant attempts to maximize the value of the solar energy that is captured [1]. Various modeling and control strategies for grid-connected PV systems have been developed in different studies to aid in the intensive penetration of PV production into the grid. Existing methods for designing the various components of a PV plant are ineffective. As a result, a great deal of study is necessary for the general architecture of the grid-connected photovoltaic system, MPP tracking algorithm, inverter synchronization, and grid connection [2, 3].
1.2. Literature Review
To reach targets in the field of power generation, the Indian government and various government agencies encourage the implementation of grid-connected solar power generation systems or ground-mounted power generation systems [4]. Grid-connected solar PV systems operate in two ways, the first is the entire power generation fed to the main grid in regulated feed-in tariffs (FiT), and the second method is the net metering approach. The net metering system allows a two-way flow of electricity, i.e., self-consumption or feed-in to a DISCOM line; the net electricity bill will be paid by the consumer [5]. In recent years, the grid-connected photovoltaic system without energy storage has become more and more popular due to the drawbacks of the energy storage system. Implementation of distributed generation (DG) will reduce the aggregate technical and commercial (AT&C) losses in transmission and distribution systems and is one of the key points for implementing wind/solar power generation in any country.
Subudhi and Pradhan [6] investigated a comparative study of basic MPPT strategies, such as the hill climb equation, incremental conductance, short circuit current, open voltage, and ripple correlation algorithm. They concluded that these methods are ineffective for irregular radiation problems [6]. This is even though these strategies are quick and easy to apply. They cannot be used when there is a PSC problem because they take longer to track and use less energy. The method of implementing low and medium (from 1 kW to 100 kW) solar photovoltaic power plants, especially rooftop installations with grid connection, was presented in [7]. Perturbation and observation (P&O) tracking algorithms are better than traditional technologies. The trace starts in the wrong direction during the rapid change of radiation [8]. For the low-cost implementation of MPPT technology, the P&O algorithm is more convenient and easy to develop. But the operating point oscillates around the MPP, thus losing energy. Grid-connected single-phase inverter topologies such as with and without a transformer and decoupling capacitor and their efficiency with optimal cost are discussed in [9]. MPPs have two types, local MPP and global MPP. The designed MPPT needs to distinguish between these two types under scheduled and partial shade conditions (PSC). Classical techniques are slightly narrower in distinguishing between these two MPPs and locating the oscillations at peak energy. Much literature is available on the timing of operations and hybrid control techniques to give effective results [8–10], and detailed runs are tabulated. Murtaza et al. [11] describe a new procedure for reducing hot spots using a plate-bound MOSFET which is a thermal shaft equipped with an infrared (FLIR) camera that was used to perceive these hot spots. The FLC approach is a good MPPT strategy for PV modules because of its shorter run time and faster tracking speed. However, it still suffers from drift phenomenon due to the rapid variation in operating temperature and radiation level. Meticulously reengineered from MF, if the number increases, so will the complexity, as discussed by the authors [12]. In recent research, the traditional FLC-MPPT problem has been solved by using MPPT design with the ANN approach. It generates a heuristic output function based on digital data, and a comprehensive understanding of the PV parameters is not necessary when designing the MPPT [13]. However, appropriate training is required to develop an ANN-based MPPT. In all control applications based on artificial neural networks, the authors discussed the significant weakness in prediction [14]. To address these problems, several hybrid strategies have been put forward. The authors proposed in [15] a genetic algorithm that automatically selects the most effective data among all individual datasets. The results proved that the proposed controller works efficiently with photovoltaic systems under different conditions. Ref [16] calculated the characteristics of silicon-plastic photovoltaics using the Lambert W function and the ANN feed-forward approach. This method enhances the prediction of the PV curve of the ANN model. Ref [17] used the hybrid algorithm to reduce the error, and GA was mainly used for the optimal scaling of the hidden layer of the ANN. In [18–21], the authors proposed a hybrid controller with a PSO-gravity search algorithm (GSA) to calculate the appropriate activation function of ANN layers for peak power prediction. However, this method is complex. New optimization algorithms based on bee colony and grey wolf algorithms have been developed. These algorithms are aimed at improving the performance of different AI approaches discussed by the authors [22–24].
1.3. Contribution and Novelty
From the above literature, it was noticed that the DC link voltage must be more than the inverter output voltage. In a two-stage conversion process, a DC-DC converter and inverter are used to meet the required voltage. The novelty of the proposed work is to model a grid-connected SPV system without the use of a separate DC-DC converter; i.e., the PV power is injected into the grid with a single-stage converter (DC/AC) system by the use of an adaptive control technique. This will reduce investment costs and losses compared to the two-stage conversion process. A novel control technique is implemented with (1) crossover and (2) mutation in the converter. The proposed control scheme is the organized execution of both the adaptive grasshopper optimization algorithm and the recurrent neural network (AGO-RNN). Here, the searching behavior of the grasshopper is modified by using the neighborhood functions like crossover and mutation. In the proposed technique, the adaptive grasshopper optimization algorithm (AGOA) technique generates the optimal dataset of the control signal for the offline way in light of the power variety between the source side and the load side. The accomplished dataset is used by the RNN phase for the online way, and it leads the control procedure in less execution time and gives improved power quality.
This paper presents a mathematical model of a 255 kW solar PV grid-connected system, MPPT control technology, and inverter control using PSO and AGO-RNN in different cases. The proposed model has been simulated using MATLAB/Simulink, and the results were clearly explained with 3 different cases. This article has been divided into five sections. Section 2 presents the mathematical model formulation. Section 3 describes with MPPT and inverter control. Section 4 discusses the solar PV power plant simulation model and results under 3 different cases. Finally, Section 5 discusses the conclusion and future work.
2. Development of a Mathematical SPV Model
According to Figure 1, the SPV power-generating system is made up of solar panels, an MPPT controller, an inverter controller, and a utility grid. Transformers, cables, wires, switches, enclosures, fuses, earth fault detectors, surge arresters, etc. are also included in the model. The inverter is a key component in all solar power generation systems, including agricultural, commercial, residential, industrial, and solar gardens, as it transforms photovoltaic output from DC to AC to meet grid voltage and frequency. Since batteries are not used in the suggested concept because it is not a stand-alone system, there are no battery losses [5, 6]. The rate of the solar photovoltaic power plant is in kilowatts only, which means that the expected peak kilowatts of electrical energy from the system when the sun is indirect costs.
2.1. Photovoltaic Array
The basic component for converting solar radiation into electrical energy is the solar photovoltaic cell, which is made of silicon and forms a PN junction. The number of units used in a plant depends on the generation/capacity of the plant. PV cells are connected in series and parallel to make a module. This can be connected to an inverter connected to a grid-tied inverter through a parallel capacitor. Modules are connected in series to make a string, and each string is connected in parallel to form an array. To stabilize the power supply coming from the PV panel, a capacitor is required. Depending on the application and power plant capacity, the PV module/string/array interface with the inverter varies [4, 5]. The battery is not required by the grid system due to the lower start-up and maintenance costs of the plant. Figure 2 displays the connection model for a 255 kW PV array. Table 1 presents the ratings and specifications of the PV parameters.
Solar PV arrays are made up of various parallel strings (88) connected with several series-connected modules (7); 616 modules are taken into account in the simulation mode. Solar cells (128) are the fundamental parts of every module and are made of silicon. A solar cell is often a polycrystalline silicon p-n junction; for an equivalent circuit of a solar cell in a single diode form, see Figure 3 [7].
The efficiency of a PV array depends on the number of PV modules, the area of each one, average solar irradiation () (it is changed from country to country), and performance ratio (it depends on panel inclination and losses, default consider value is 0.75, and generally, its range varies between 0.5 and 0.9). Module efficiency can be defined as the ratio of PV panel output power () to input power of the PV panel (). calculation is very important; it varies from place to place and time to time [7, 8]. The calculated value below is under standard test conditions (STC), i.e., sunlight incident intensity and spectrum (1 kW/m2), the temperature of the cell (25°C), and cell area mm2) at AM1.5 spectrum conditions without partial shedding [7–9].
The efficiency calculation’s fill factor (FF), which can be defined as the ratio of the solar cell’s maximum voltage and current to its open circuit voltage and short circuit current, is crucial.
Finding the maximum voltage and current equivalent to the maximum power output from PV panels has been the subject of numerous research articles. Identifying the appropriate MPP is a difficult operation, and tracking MPP is vital under various climatic circumstances for improving PV power generation efficiency. Only when the solar panels are being used at the MPP does solar PV power generation utilize itself to its fullest potential. This study compared fluctuations in PV power with and without MPPT installation. MPP is monitored using the traditional perturbation and observation (P&O) method. MPP needs to be aware of the solar PV cell’s mathematical model to analyze. The output current and voltage of the module were calculated using the formulae below. Apply KVL for the similar circuit shown in Figure 3 to get the panel output current.
The current created by cell light is affected by temperature, irradiation (), diffusion length in P-N type materials, and area.
From the equivalent circuit, we can write .
At short circuit condition of the PV cell, output voltage () and shunt resistance voltage () are zero from equation (4)
In the above equation (6), the negative sign signifies the change in direction of the current in the cell. Now, total is equal to light generated current ().
At the open circuit condition of the PV cell, output current () is zero, with no current flowing in series resistance, so open circuit voltage is taken across the shunt resistance, from equation (4).
616 sun power SPR-415E modules, each of which produces 414.801 W of power, are taken into account to produce the 255 kW power from the PV array, which are connected in a combination of series and parallel connections, i.e., a string with seven modules connected in series and an array with eighty-four strings connected in parallel.
2.2. Inverter Control System
The control system contains two major Simulink-based subsystems.
2.2.1. MPPT Controller
The output power is affected by the P-V and I-V characteristics of the PV array/cell, which are nonlinear and regularly change with external climatic variables such as cell temperature (Oc), solar irradiation (), and load (Figure 4). Therefore, one algorithm is required to extract the maximum amount of power in any form of the solar cell, and that method is known as the maximum power point tracking (MPPT) methodology. Either a programming procedure or a mathematical block connection in Simulink can be used to achieve this MPPT technique.
(a)
(b)
Due to its simplicity and easier implementation, the standard perturbation and observation (P&O) algorithm was used for this paper. The simulation results use the P&O technique described in Section 3 to examine PV output power and voltage in various situations. There are other methods; however, P&O just needs a single voltage comparator or sensor to continually sense the PV voltage, making it more practical to build and less expensive.
The conventional P&O system is an instance of a hill-climbing algorithm. Differentiate the power equation for dv and set it to zero () to obtain the maximum power. The power equation in dc is well known. When the difference between the instant power and the real power is positive (), the direction of an operational point is located toward the MPP by increasing the reference voltage in the opposite direction. A PV voltage and current must be perturbed in both directions to find the MPP [13]. As a result, the temporal complexity of this algorithm is lowered; but, as it approaches the MPP, it continues to perturb in both directions rather than stopping there. When this happens, the value that the computation has generated is very close to the MPP, and we can then either set a reasonable error limit or use a wait function, both of which will end the unpredictability of the calculation duration. For a modal overview of the P&O algorithm, please see Figure 5.
The MPPT controller will generate the optimal reference value in the simulation mode to track the maximum power point and that will be converted into in terms of reference; this value is given to the voltage regulator in the inverter control circuit or DC-DC converter used before the inverter; upper and lower limits are required in P&O MPPT for the effective operation of converter/dc voltage regulator. The upper and lower bounds of MPPT in this paper’s simulation modal are 358 to 597 V. The upper limit is based on , while the lower limit is 60% in .
2.2.2. Current Controller
The proposed controller (AGO-RNN) has been implemented as a subsystem in the current controller.
(1) Adaptive Grasshopper Optimization Algorithm (AGOA). The necessity for optimization in any system is to operate the system as efficiently or effectively as possible, and the purpose of optimization techniques is to obtain the optimal result in the existing circumstances for a desired benefit. A correct balance of analysis and manipulation in any given optimization problem can result in the global optimum solution.
Nature-inspired intelligent optimization techniques (NIIOS) are faster and simpler than mathematical optimization strategies for determining the appropriate gain value formulation in a PI controller. These strategies replicate the normal biological habit of effectively seeking the best solution. The traditional NIIOS includes particle swarm optimization (PSO) [16], gravitational search algorithm (GSA) rule, ant lion optimizer (ALO), pigeon-inspired optimization (PIO), grey wolf optimizer (GWO), and grasshopper optimization algorithm (GOA) rule.
The traditional PSO algorithmic rule is modeled after the foraging behavior of schools of fish or flocks of flying organisms. The optimum answer in this algorithmic method is the solution obtained by the particle and that is best among the swarm of particles. The gravitational search algorithm (GSA) is based on Newton’s laws of gravity and motion. It is classified as a population-based strategy and is considered more natural; yet, its fundamental disadvantage is that it is a memoryless computation technique. Because of the collective performance of ants in finding the shortest path from the house to the source of nutrition, the ant colony optimization (ACO) algorithmic rule provides the simplest solution [14, 15]. The grasshopper optimization algorithm (GOA) is a leading advanced optimization method [18].
GOA replicates grasshopper convergence behavior and social association for benchmark capacity and real-world engineering difficulties. This technique outperforms certain other prominent NII optimization strategies in terms of convergence speed and precision [18, 20].
The main disadvantage of GOA is that it is more sensitive when perceiving the best values in the immediate vicinity. This means that when a grasshopper is formed, it is given a random start at the initial location. The grasshoppers have a proclivity to move in the direction of the target, given by the notation , toward the social communication network. Even though grasshoppers have been introduced in places concentrated close to the best of the nearby local community and far from the best of the global community, grasshoppers may be efficiently absorbed by the best of the neighborhood community at this time. This suggests that GOA is sensitive to the initial positions chosen. As a result, the augmentation strategies should be able to help search agents avoid the “local best” trap. Furthermore, the adaptive strategy on in the old GOA is just related to iteration (equation (12)) and does not take into account the dynamic feedback included in the search technique. It is, in particular, unable to dynamically react to the performance of the most recent iteration, making it somewhat equivalent to an open-loop system in the context of the control hypothesis. As a result, must be prepared as a variable parameter that is associated with both iterations and current population execution and is tied to both iteration and current population performance. That is, based on equation (12), the population’s present performance should be used as feedback. As the reference and control rule, the following population performance and the flexible adaptive strategy must be examined separately. The technique described above can be viewed as a closed-loop system in the context of the control hypothesis.
The adaptive grasshopper optimization algorithm, or AGOA, is one of the novel intelligence algorithms proposed for use in AGO-RNN to mimic grasshopper behavior. The recommended method outperforms intelligent algorithms such as the natural selection strategy, the democratic decision-making mechanism, and the dynamic feedback mechanism in terms of search capabilities. The proposed technique outperforms the traditional PSO as well. This section alters the grasshopper’s searching behavior by utilizing very effective neighborhood search processes like crossover and mutation. The phases of the method that are used to find the optimum gain. The mathematical expressions help to simulate the swarming behavior of grasshoppers and are done by initialization of proportional () and integral () gains of the PI controller and randomly generating the position of the grasshopper ().
The fitness function is applied and assessed for each movement of the grasshopper, and the appropriate function is evaluated using the following relation:
To update the fitness function, the following mathematical model is presented to simulate the predatory process: where and defines the position of the th and th grasshopper; is the social interaction between and , i.e., repulsion or attraction force; is the gravity force on the th grasshopper; is the number of grasshoppers; is the current iteration; shows the wind advection; is the upper bound in the th dimension; is the lower bound in the th dimension; is the value of the th dimension in the target (best solution found so far); and is a decreasing coefficient to shrink the comfort zone, repulsion zone, and attraction zone.
The attractive length scale(s) can be represented as follows:
The comfort zone is reduced by where is the maximum value, is the minimum value, indicates the current iteration, and is the maximum number of iterations.
(2) Recurrent Neural Network (RNN). RNNs use sequential information, which means they process data sequentially and store it in memory, allowing them to anticipate the next variable/data and call back when needed. In a typical neural network, we presume that all inputs and outputs are independent of one another, and it will fail at some tasks. If you want to predict the next word in a sentence, you must first understand what words occurred before it. RNNs are referred to as recurrent because they do the same task for each element of a sequence, with the output dependent on past calculations. In this proposed technique, AGOA is applied and combined with RNN to improve calculation speed and parameter quantity, which are important performance indicators for any neural network-based system. Designing an efficient neural network with fewer parameters and faster computation speed, on the other hand, is a difficult research problem. (1)Crossover and mutation
The crossover process involves the crossing of two chromosomes, which results in the formation of a new set of chromosomes. The process is then carried out between the chromosome of fitness value and the newly created chromosome. The genes are mutated at random during the mutation process dependent on the mutation rate. The crossover and mutation rate formula is slanted as follows: where represents the number of gene crossovers, represents the length of the chromosome, and represents the mutation point. Using the updated movement, find the fitness and check the objective function. (2)Termination
Determine iteration bounds; if the number of iterations reaches an extreme value, update the fitness; otherwise, increase the number of iterations by one. When the fitness function is changed, the AGOA returns optimal gain parameters to the PI current controller. The PI current controller outputs the quadrature axis current (), which is applied to the RNN input. RNN will forecast the reference three-phase current () based on the reference current and measured quadrature axis current inputs. The RNN process’s prediction procedure is discussed in the next section.
This section discusses the RNN training process, and the supervised learning methodology is used as the RNN training method in this section. It is a method of artificial training and testing that is based on the machine learning approach and consists of many artificial neurons [19]. An RNN will typically have three layers: an input layer, a hidden layer, and an output layer. The input layer in this situation is made up of both the reference quadrature axis current and the actual quadrature axis current. The output target layer is the current that is referenced by all three phases. The RNN was trained using the output target as well as the input corresponding to that output. The context layer gives the hidden layer its weight at the set time delay throughout the training process. This occurs throughout the training period. The weights of the RNN are modified using the backpropagation method. This is accomplished using the training dataset as its resource. With the help of this strategy, the error function will be decreased as much as possible. The error function is built using the actual quadrature current and the reference quadrature current and is conveyed as where and represent the actual quadrature current and the reference quadrature current and are the error function. The learning process is then used to construct the error function, and any necessary modifications are made. After all of the steps have been completed successfully, the learning algorithm produces the error function with the least value. Now that the RNN has been trained to produce the perfect current in all three phases, its output may be converted into useful control pulses. These optimal control pulses are transmitted to the MPPT controllers of the solar and wind-generating systems, as well as the proposed converter. Figure 6 depicts the flowchart for the suggested control method. The MATLAB/Simulink platform is used to assess the usefulness of the suggested strategy, and the proposed technique’s effectiveness is compared to the effectiveness of other strategies already in use. And Figure 7 shows the converter control circuit. The next section provides a succinct account of the implementation outcomes and the subsequent debate.
2.3. Inverter
A 3-level IGBT-based neutral point clamping (NPC) inverter with a PWM-controlled mode was used to convert the solar PV (SPV) DC power into distribution line power (AC). The NPC inverter shown in Figure 8 is a three-level DC-AC power converter with three arms and two neutral clamped diodes. Each arm consists of 4 switches with antiparallel diodes. Each capacitor should have an identical DC voltage, and voltage stress should be restricted to one capacitor level through clamping diodes ( and ). If the total DC link voltage is considered and the midpoint voltage is controlled to half of the total DC link voltage, the voltage across each capacitor is . The DC bus voltage is divided into three levels by connecting two series DC capacitors, and .
The NPC inverter can produce three voltage levels on the output: the DC bus plus voltage, zero voltage, and DC bus negative voltage. The two-level inverter can only connect the output to either the plus bus or the negative bus (refer to Figure 9 for the following example.) For a one-phase operation, when IGBTs Q1 and Q2 are turned on, the output is connected to ; when Q2 and Q3 are on, the output is connected to ; and when Q3 and Q4 are on, the output is connected to .
Switching states for the four IGBTs are listed in Table 2. Clamp diodes D4 and D5 provide the connection to the neutral point. From the switching states, it can be deduced that IGBTs Q2 and Q3 are on for most of the cycle, resulting in greater conduction loss than Q1 and Q4 but far less switching loss. In addition, the free wheel diodes for Q2 and Q3 are for most cases, soft switched as the IGBT parallel to the diode is on, thus holding the recovery voltage across the diode to that of the IGBT Vce.
The DC bus capacitors are connected in series and establish , the midpoint voltage. Due to available capacitor voltage rating, series-connected capacitors are generally required in inverters rated for 480 V and 600 V services. In NPC inverters, maintaining the voltage balance between the capacitors is important to the proper operation of the NPC topology. When SPV DC current 845 A and voltage 480 V are fed into an inverter, its power rating is 250 kVA, and it produces AC RMS voltage of 250 V and 1000 A.
The total capacitance required for connecting the PV array to the inverter is , at 3/4th of the power cycle; the equivalent capacitance is
Two capacitors are required for an NPC PV system, one for phase and one for neutral , which was represented as in the simulation model.
Table 3 shows the simulation’s inverter configuration as well as the needed input capacitor and output inductor values.
2.4. Filter
To decrease harmonics induced by the 3-level inverter, its output is connected to a 250 V/25 kV, 3-Ø, 250 kVA transformer via a choke RL and a filter before being interconnected to the utility grid. The choke coil specifications and are computed per unit system, and capacitance is determined using capacitive reactive power (), which is 10% of the nominal power, and minimal active power () of 50% of . values of choke coil R and XL are the ratio of actual values to base power ().
2.5. Inverter Efficiency
Because the PV array output power () is sent to a 3-level bridge converter for conversion into three-phase AC power (), inverter efficiency () is determined by the to ratio.
The product of total PV efficiency and inverter efficiency determines overall system efficiency; hence, PV modules and inverters play a significant role in plant efficiency enhancement. PV efficiency can be increased by increasing output power generation using various optimization/MPPT techniques, and inverter/converter efficiency can be increased by using optimal switching strategies [4].
3. Results and Discussions
Figure 10 depicts the simulation diagram for the 255 kW grid-connected photovoltaic system. The system’s performance was evaluated under a variety of environmental conditions, as illustrated in Table 4. The performance of the system in terms of total harmonic distraction (THD) is also investigated in each scenario using conventional (P&O) and proposed controllers (AGO-RNN). Irradiation and temperature are shown in one graph in simulation results, with the real temperature value (given in Table 4) multiplied by a constant 15 and then shown in all cases of the graph for clear sight of the temperature line.
3.1. Performance Analysis at Different Cases
3.1.1. Case 1: Standard Test Condition (STC)
In this scenario, the performance of a 255 kW solar photovoltaic (SPV) grid-connected system was investigated with constant irradiance and temperature, which were kept at 1000 W/m2 and 25°C, respectively. The grid-side converter voltage control approach is built with a conventional P&O MPPT algorithm, the converter current control architecture is designed with PSO, and THD was compared with the proposed AGO-RNN technique. The simulation results at various points are given in the figures below.
Figure 11(a) depicts the simulation time with constant irradiation at 1000 W/m2 and temperature at 25°C. Figure 11(b) depicts the mean, indicating that the photovoltaic system generates an output power of 255 kW. Figure 11(c) depicts the mean and reference created by MPPT; the mean initially oscillates before becoming constant at 510 V with minor fluctuations. Because of the constant irradiation and temperature, the mean and reference are almost identical.
(a)
(b)
(c)
3.1.2. Case 2: Constant Temperature and Variable Irradiation
In this case, constant temperature and variable are considered, which means that in the designed 255 kW SPV grid-connected system, irradiance is varied as shown in Table 3, starting at 1000 W/m2, then ramping down to 200 W/m2 at 0.4 seconds, then ramping up to 500 W/m2 at 0.6 seconds, and finally returning to 1000 W/m2 at 1 s onwards. The temperature is kept constant at 25 degrees Celsius until the simulation is completed. PSO and AGO-RNN techniques are used to create the grid converter current control structure. Finally, THD analysis is used to analyze the performance of the planned 255 kW SPV grid-linked system with PSO and AGO-RNN techniques. The simulation findings are discussed further below.
Figure 12(a) depicts the constant irradiance starting at 1000 W/m2, gradually decreasing to 200 W/m2 at 0.4 seconds, increasing to 500 W/m2 at 0.6 seconds, and eventually returning to 1000 W/m2 at 1 second. The temperature remains constant at 25 degrees Celsius at all times. Figure 12(b) depicts the mean, which is not fixed as in case 1 due to irradiation variance. SPV creates an initial output power of 255 kW, then gradually declines to 50 kW in 0.4 seconds, slowly increases to 135 kW, and ultimately returns to the starting value of 255 kW. mean and reference are shown in Figure 12(c), which is created using MPPT. The mean oscillates and becomes constant at 510 V. The voltage is distorted here due to disturbances on the source side.
(a)
(b)
(c)
3.1.3. Case 3: Variable Irradiation and Temperature
In this situation, 3 variable and are considered; i.e., in the proposed 255 kW SPV grid-linked system, irradiance starts at 1000 W/m2, drops to 200 W/m2 in 0.4 seconds, slowly climbs to 500 W/m2, and ultimately returns to 1000 W/m2. Temperature varies over the same time intervals, with interims of 25°C, 30°C, 35°C, and finally 45°C. PSO and AGO-RNN techniques are used to create the grid converter current control structure. Finally, THD analysis is used to analyze the performance of the planned 255 kW SPV grid-linked system with PSO and AGO-RNN techniques. The results of the simulation are displayed below.
Figure 13(a) depicts the irradiation and temperature variations mentioned in Table 4 (starting at 1000 W/m2, reducing to 200 W/m2, then incrementing to 500 W/m2, and again raising to 1000 W/m2 and temperature initially at 25°C, then incrementing to 30°C at 0.4 seconds, then incrementing to 35°C, and finally reaching 45°C). Figure 13(b) depicts the mean; i.e., the photovoltaic system generates an output power of 255 kW at first, then drops to 50 kW after 0.4 seconds, then increases to 135 kW, and ultimately returns to 255 kW. This is owing to the frequent interruptions on the source side. Figure 13(c) depicts the mean and reference. The mean oscillates and becomes constant at 510 V. From the Figure 13, we can see that the PV output voltage varies correspondingly with slight distortions with temperature variation.
(a)
(b)
(c)
Figure 14 depicts the 3-level grid-connected inverter output voltage, which is unaffected by changes in PV input energy.
The grid voltage is represented in Figure 15(a) and grid current in Figure 15(b); the grid voltage is constant at 20 kV even while the irradiance is changing. Figure 16 is due to the suggested control AGO-RNN technique. As the irradiance changes, the variations in grid current were observed during 0.4 s to 1.1 s in Figure 16.
(a)
(b)
(a)
(b)
The grid voltage is represented in Figure 17(a), and Figure 17(b) represents the grid current. Due to the application of AGO-RNN control technique, the grid voltage remains constant at 20 kV despite the fluctuating irradiance. As the irradiance varies, hence the grid current variations are further observed during 0.4 s to 1.1 s.
(a)
(b)
Figure 18 represents the grid power output waveform, which is almost constant due to fixed voltage and fixed grid currents depicted in Figure 15 for case 1. Since, in case 1, the temperature and irradiance are assumed to be constant. Figure 19 shows the grid power waveform for case 2, where the power output variations are observed during 0.4 s to 1.1 s. This is because of the change in current for case 2, depicted in Figure 16.
Figure 20 represents the grid power waveform for case 3, where the power output variations are observed during 0.4 s to 1.1 s. This is because of the change in current for case 3, which is depicted in Figure 17.
Case 1 in the preceding figures depicts the grid power of 255 kW at each phase, and the power factor in case 1 is continuous without any variation as shown in Figure 21. Even if the inputs of the SPV array are different in case 2 and case 3, the power and P.F. are roughly similar as depicted in Figures 22 and 23. The figures show that power drops to 50 kW after 0.4 seconds, then climbs to 135 kW, and then raises again to 255 kW. Due to the impact of irradiation, the power reduces to 50 kW after 0.4 seconds, then grows to approximately 135 kW after 0.6 seconds, and finally decreases to below 255 kW after 1 second.
The THD of the grid current with the PSO technique is 1.67% and the magnitude is 8.195 A as shown in Figure 24(a), whereas the THD of the grid current with the AGO-RNN technique is 1.54% and the magnitude is 8.205 A given in Figure 24(b) for case 1 study. Also, the THD of the grid current with the PSO technique is 1.74% and the magnitude is 8.199A as given in Figure 25(a), while the THD of the grid current with AGO-RNN technique is 1.55% and the magnitude is 8.189A which is depicted in Figure 25(b) for case 2 study. Similarly, the THD of the grid current with the PSO technique is 1.61% and the magnitude is 7.745 A as shown in Figure 26(a), while the THD of the grid current with AGO-RNN technique is 1.59% and the magnitude is 7.761 A as given in Figure 26(b).
(a)
(b)
(a)
(b)
(a)
(b)
From the above results, it can be concluded that the THD values (Figures 24(b), 25(b), and 26(b)) are lower in all three scenarios while applying AGO-RNN control than the THDs obtained using conventional PSO method of THD (which are given in (Figures 24(a), 25(a), and 26(a)). Hence, it concludes that the AGO-RNN technique is the optimal control topology for the given system with the lowest THD. The comparative analysis of the proposed system with existing work in the literature is given Table 5.
4. Conclusions
This manuscript focuses on two criteria: a mathematical model of the SPV grid-connected system and an analysis of its performance using a cutting-edge control approach under three radically distinct environmental situations. In MATLAB Simulink, a 255 kW solar-based PV grid-connected system was constructed. The numerical model of the SPVGC system is given in simple terms, and the MPP is tracked using an easy, smooth, and least complex P&O technique. It can be seen that when the control is comparatively changed due to irradiation variations, the grid power is proportionately adjusted, yet voltage is held up closer to consistency without being affected by temperature. This can be accomplished simply by employing an inverted voltage controller, which is a simple technique. Two current controlled approaches, PSO and suggested AGO-RNN, were utilized before PWM in the inverter control system to enhance THD at PCC and effective grid synchronization. The simulation results demonstrate that the current magnitude and THD of the SPVGC system are improved with the cutting-edge AGO-RNN controller compared to PSO in all three different scenarios, and this value is less than 1.6%, which is within the permitted limits of IEC 61727 standards. Due to the effective control technique, the inverter output voltage is not distorted even when G&T is adjusted. The fundamental issue with standard PSO/GOA is that it is easier to track local best values rather than global best values; however, this is not the case with the adoptive AGO-RNN approach.
5. Future Scope
(1)The installation of high-performance MPPT rather than the conventional P&O approach can improve tracking efficiency in this system. This article does not consider the partial shading condition (PSC), which should be investigated using the new MPP techniques(2)The second is that if the central inverter fails, the entire system fails; this can be avoided by incorporating microinverters with distributed MPPT (DMPPT), which produces the best results(3)In the event of a hybrid PV and wind energy grid-connected system, performance can also be monitored using the same controller
Data Availability
Data is available on request.
Conflicts of Interest
The authors declare no conflicts of interest.
Authors’ Contributions
The authors confirm the final authorship for this manuscript. All the authors have equally contributed to this manuscript.