Performance enhancement of maximum power point tracking for grid-connected photovoltaic system under various gradient of irradiance changes

This paper presents a new variant of smart adaptive algorithm of Maximum Power Point Tracking (MPPT) in the photovoltaic (PV) system. The algorithm was adopted from Modified Perturb and Observe (MP&O). The smart adaptive MPPT is used to search Maximum Power Point (MPP) of the PV system under various irradiance changes. This algorithm incorporates information of current change (ΔI), maximum operating point margin and dynamic perturbation step to prevent MPPT diverging away from the MPP and minimize the steady state oscillation. The smart adaptive MPPT algorithm performance is compared with the dI-P&O and conventional P&O to prove its effectiveness. The comparison is performed under the various gradient of irradiance change. It was found that, for all the tests, the smart adaptive algorithm scheme improve the tracking efficiency under various gradients of irradiance changes and increase the efficiency of extraction power from PV system.


Introduction
Renewable energy generation growth has been increasing quickly due to the continuous reduction of fossil fuel energy generation [1].The photovoltaic (PV) generation increasing 267.1 GW in the past 4 years [2].Solar is a viable alternative energy resource.Solar energy is becoming important as a renewable energy resource.Consequently, It gives great merits such as simple of installation, less maintenance, and no pollutions [3].PV generation is possible to substitute fossil energy generation.The grid-connected PV system is more advantageous in terms of cost and investment than the standalone PV system [4].
PV module conversion efficiency leans on the operating point on the characteristic curve of the PV module that is not linear and influenced by solar irradiations and ambient temperature [5].There is maximum power point (MPP) at each characteristic curve of the PV module.In order to make the PV generation system operate at its maximum power, MPPT is employed [6].MPPT is the main function that should include in every PV system to ensure that PV system generates possible maximum power [7].Since the MPPT is ordinarily areas to improve the effectiveness of PV conversion because it consists of software codes that can embed into the microcontroller and is the most economical method for increasing the maximum energy of a PV system [7,8].The existing algorithms of MPPT include the voltage feedback method [9,10].the current feedback method [11,12], the incremental conduction (INC) method, perturb and observe (P&O) method [13,14], artificial neural network (ANN) method [15,16], fuzzy logic methods [17,18].
The most used MPPT techniques are P&O and INC due to their simplified algorithm control and low-cost implementation.INC method is better than P&O because it based on the derivative of PV module output power with its voltage.Moreover, Compared INC method and P&O method show that INC has many advantages in the tracking speed, tracking accuracy and efficiency [19].Besides that, P&O method is simpler than INC method in practical applications.The algorithm operation is adversely affected by noise and errors on the measured control values [20].The P&O has two serious drawbacks.The first is oscillation around the MPP when steady state.Second is the false perturbation step when rapid change of irradiance.
The both drawbacks reduced the efficiency of output power and tracking efficiency.Many researchers have developed the MPPT techniques to minimize these two drawbacks.Adding PV module current (I pv ) as controlled variable is to resolve the false perturbation issue [21].However, the developed techniques are just for specific conditions.This technique only works when the algorithm deviated to the right of the P-V curve.There are other developed adaptive techniques to solve these drawbacks.The developed techniques exhibits high performance and solve both drawbacks.However, it needs the complex computational and make it difficult to implement.Another issue is the MPPT algortims performance benchmarking is tested on simple G profiles [22].It does not reflect the real environmental conditions.Authors in [23] have developed A Modified P&O to find the MPP and improved tracking efficiency.But, the performance of the MPPT PV system under the various gradient of irradiance changes conditions need to be developed and evaluated.
In this work, a new variant of smart adaptive algorithm for MPPT PV system based on Modified P&O [23] was developed and evaluated in the various conditions.The algorithm has additional structures that incorporates information of current change, maximum operating point margin and dynamic perturbation step to avoid MPPT from losing of tracking direction and to reduce steady-state oscillation under various conditions.The smart adaptive MPPT are simulated using PSIM and comprehensively benchmarked using MPPT efficiency test.This test order the algorithm to track various gradient G profile.Results show the compared performances of the developed MPPT algorithm and conventional MPPT algorithms.

Developed System Model 2.1. Photovoltaic Module Modeling
The fundamental component in PV module is a PV cell.The output of a PV cell is very low.In order to meet practical demands, they are connected in series (N S ) or in parallel (N P ) or in series-parallel [23] as shown in Figure 1.V pv is the PV cell voltage and N=N S /N P , the current I can be written as: (2) where I is the current and V is the voltage.R s and R p is the series resistance, and R p is the parallel resistance, while V T1 = V T2 is the thermal voltage of the diodes and a 1 = a 2 is the ideality factor of the diode.I PV is the light generated current which given by where G is the irradiance and T is temperature.Note that IPV_STC , GSTC , and TSTC are that measured in the standard test condition (STC).K I is short circuit current coefficient.I d1 = I d2 is the diode saturation current.It is written as: According to (5), I sc is short circuit current in STC and V oc is the open circuit voltage in STC.K V is the temperature coefficient of the voltage.In this study, we use JA Solar Cell JAP6 60-250 3BB panel.The JA Solar Cell JAP6 60-250 3BB panel is modeled using PSIM physical model of PV module by entering basic parameter from its datasheet as shown in Table 1.The JA Solar Cell JAP6 60-250 3BB is modeled using PSIM software.

Design of Grid-Connected PV System
In The grid-connected PV system is simulated to test the MPPT method.The system consist of PV module array, DC/DC Boost Converter, inverter, L filter, grid, and load as shows in Figure 2.

MPPT Algorithm
Many MPPT methods have been implemented in PV power plants.The P&O is the most popular to solve the efficiency problems of PV power generation.These methods are still developed to increase its performance.

Conventional Perturb and Observe (P&O)
This algorithm is the most popular in practice because of its simple control algorithm.However, it is not powerful enough to track the right MPP when rapid irradiance change [15,22,23].This algorithm searches MPP by changing the operating point with the perturbation step (ΔC) and then the change in P (ΔP) is measured.The formulation for changing operation point can be expressed as follows: The algorithm use controlled variable (C) and it can be either voltage (V), current (I) or duty cycle (d).C new and C old is the controlled variable at k and k-1 iteration, respectively.The perturbation step (ΔC) can be Δd, ΔV or ΔI according to the controlled variable that algorithm used.The direction of perturbation is indicated by multiplier slope: +1 or -1.They are increasing and decreasing controlled value (C).If ΔP>0, then the multiplier slope keep on its value until the algorithm reaches MPP.Once the ΔP<0, the multiplier slope goes to the opposite direction and the algorithm climbs the P-V curve until reaches the MPP again.It causes the operation point moves back and forth around the MPP.The perturbation step value (ΔC) is vital.If ΔC is small, it results in small oscillation.Consequently, the algorithm response is slow and vice versa [23].The flow diagram of P&O shown in Figure 3.During the increasing or decreasing of the voltage, slope value could be either positive or negative.These values is the sign multiplication of ΔP, ΔV, and ΔI.They will be used to determine whether the algorithm has converged to MPP and hence variations in the step value of the perturbation.This procedure is possible to detect the occurrence of oscillation precisely.Once it detected, the initial perturbation step value (2% of V oc_array ) is reduced 0.5% of V oc_array in every iteration until its reaches minimum value (0.5% of V oc_array ) to resolve the problem of oscillation.

Figure 5. Flowchart of dI-P&O algorithm
This algorithm avoid possible false tracking direction that conventional P&O is still suffering.Ordinarily, in the real environment the conventional P&O treat equally the variation of G with fixed perturbation step.When G changes slowly, the conventional P&O can cope with these changes.However, when fast increasing of G, the conventional P&O likely suffer from loss of tracking direction.The following procedure needs the value of irradiance change, but the method does not require irradiance sensor to measure the change of G.The MPPT takes consecutive sampling when irradiance starts changing.It is assumed that T will remain almost the same value.If T remains constant in two consecutive samples, the relationship of the samples as given by [23].
From [23] can be concluded that the normalized power change is equivalent to the normalized irradiance change.In this simulation, the minimum change in power (ΔP/P) is set to be 0.0002 based on trial and error observation from simulation as do as in [24].This value selected as the as the threshold ΔTr.If the ΔP/P ≥ ΔTr, the perturbation step value is returned to the initial value to track the MPP.Elseways, it kept to the minimum value.In some case, the MPP voltage will decrease when level of irradiance is high due to the panel series  [25] and it causing PV power is decrease and high change in power is raise.This threshold can solve the problem caused by that situation.
The mechanism how the algorithm minimizes the possible false of MPPT direction shown in Figure 6.In order to eliminate this problem, the voltage margin is imposed once steady-state oscillation is detected.Initial voltage margin is set to 0.5 % of Open circuit voltage (V oc ) to 0.95 % V oc_array at the beginning, V oc_array = V oc x N s .When the steady-state oscillation detected, the voltage margin change to V mpp -0.05 V oc_array to V mpp + 0.05 V oc_array .Note that V mpp is the target Voltage of MPP.When G increases, the V mpp actually shifts 5 % of V oc_array to the right.The operating point is restricted to 5 % of V oc_array margin to force the operating point always close the MPP.This procedure can avoid the loss of tracking.The flow diagram of smart adaptive MPPT algorithm is presented in Figure 7.

. Simulation Setup
The performance and efficiency of the system are evaluated with simulation cases by using PSIM.Figure 8 shows the PSIM simulation to simulate the MPPT.The dc/dc converter of boost converter is used to implement the MPPT with following specifications: inductor (L) = 0.188 mH, switching frequency (f) = 15 kHz, capacitor C1 and C2 are 2.35 mF and 0.173 mF, respectively.The voltage and current of the PV array are designed to be input of the MPPT.The both values are measured by using the voltage sensor and current sensor.The output of MPPT is V ref .It is subtracted from V pv to find the error voltage as an input of PI controller.Then PI controller output compare by saw tooth waveform to produce Pulse Width Modulation (PWM) signal for controlling the IGBT switch.This controlling signal keep the converter to operate at the V mpp .The simulation model of the grid connected PV system is shown in Figure 9.The 15 kW PV array capacity consist of 60 PV modules connected to the grid through inverter.PV system and grid generate power to the load.The power flow of the PV-grid system is also presented due to different load conditions.The smart adaptive MPPT algorithm is programmed with C Block as shown in Figure 10.The system test parameters are listed in Table 2.

Result and Discussion
The PV system consists of 10 modules in series and 6 modules in parallel.Different level of irradiance is applied to modules.The study is to verify response of the MPPT method under the various gradient of irradiance change, as shown in Figure 11.Moreover, the power flow of the PV-grid system is also presented due to different load conditions.
from (7), the average efficiency calculate using From ( 7), ( 8), P max is the theoretical maximum power which PV array possible to generate.P out is the extracted power from PV array by the MPPT algorithm.From the simulation, the smart adaptive MPPT can handle the drawbacks of the two MPPT algorithm, i.e.P&O and dI-P&O.The dynamic perturbation step can minimize the oscillation as show in Figure 12, image 1 and image 2.Moreover, image 3 shown the smart adaptive improved the dynamic responses under various change of irradiance.When rapid irradiance changes occur, the conventional P&O and dI-P&O loss their tracking direction as shown in Figure 12, image 4. Their efficiency drops due to this G change.The smart adaptive MPPT is capable to track MPP under various conditions.Consequently, it can track the irradiance gradient locus.
Figure 12.MPPT simulation result Figure 13 shows the efficiency of MPPT method under various conditions.The graph shows the effect of oscilation and diverging of MPPT tracking.In the rapid change of irradiance at time 1-1.7 s, the conventional P&O and dI-P&O loss their tracking direction as shown in Figure 12.Their effeciency drops due to this irradiance change.It can track follow the irradiance gradient locus.
The deviating can occur towards to the left or to the right of the P-V curve.Figure 14 (a) shows the smart adaptive MPPT performance to track the MPP. Figure 14 (b) shown the V mpp of smart adaptive MPPT is restricted by voltage margins to prevent the voltage diverging away from the V mpp , while the P&O and dI-P&O voltage diverging away from the V mpp .As a result, the efficiency increase and there is not a negligible deviation of power.In some period, power is not  Figure 15 shows the initial load is 10 kW.Consequently, the PV system is capable of supplying power to the load.The remaining power 2 kW flows into the grid.When the load is 5 kW, PV system can supply 5 kW of power to the load and 7 kW to the grid.When the 12 kW PV power supply the 15 kW load, grid supply the remaining 3 kW of power.It shows that the PV system still supplies 12 kW power to the load.The simulation shows that the PV system is capable during grid-connected condition and transient load condition.

Conclusion
In this work, the new smart adaptive MPPT algorithm is implemented to simulate the grid-connected PV system.This algorithm is adapted from modified P&O with adding information of current change (ΔI), maximum operating point margin and dynamic perturbation step.The dI-P&O and Conventional P&O MPPT method are implemented in the system to do performance comparison with new smart adaptive MPPT algorithm and they are tested under various gradients of irradiance changes.As a result, the new smart adaptive MPPT algoritm provides highest efficiency tracking and power conversion compared to the dI-P&O and conventional P&O in the rapid changes of irradiances.In the different load conditions, PV system is capable supplying the power to the load.The power flow from the grid depends on the supply power of the PV system and capacity load.

Figure 1 .
Figure 1.The equivalent circuit of twodiode PV cell

Figure 2 .
Figure 2. Block diagram of grid-connected photovoltaic system


ISSN: 1693-6930 TELKOMNIKA Vol.17, No. 4, August 2019: 1973-1984 1982 at its P mpp .Ordinarily, this is because of the voltage not exactly on V mpp , and it stays around to the V mpp .

Figure 14 .
Figure 14.(a) The modified P&O tracking performance (b) Voltage operation of the MPPT