10
Modified incremental conductance MPPT algorithm to mitigate inaccurate responses under fast-changing solar irradiation level Kok Soon Tey, Saad Mekhilef Power Electronics And Renewable Energy Research Laboratory (PEARL), Department of Electrical Engineering, University of Malaya, Malaysia Received 9 February 2013; received in revised form 30 December 2013; accepted 2 January 2014 Available online 29 January 2014 Communicated by: Associate Editor Jan Kleissl Abstract During the increment of solar irradiation, the conventional incremental conductance algorithm responds inaccurately at the first step change in the converter duty cycle. This paper presents the conventional algorithm confusion and proposes a modified incremental con- ductance algorithm that responds accurately when the solar irradiation level increases. Moreover, the proposed algorithm shows zero oscillation in the power of the solar module after the maximum power point (MPP) is tracked. MATLAB simulation is carried out with the modified incremental conductance algorithm under a fast-changing solar irradiation level. Results of the modified, conventional and variable step size incremental conductance algorithms are compared. Finally, the hardware implementation, consisting of a single-ended primary-inductor converter (SEPIC) and a PIC controller, is applied as the maximum power point tracking (MPPT) controller. The sim- ulation and experimental works showed that the proposed algorithm performs accurately and faster during the increment of solar irra- diation level. Ó 2014 Elsevier Ltd. All rights reserved. Keywords: MPPT; Incremental conductance; Photovoltaic (PV) system; SEPIC converter; Fast changing solar irradiation 1. Introduction Electricity generation using solar energy has gained increasing popularity in recent years (Mekhilef et al., 2011, 2012). However, electricity generated by photovoltaic (PV) panels is an unstable energy source because of its firm dependence on factors such as solar irradiation level and surrounding temperature. Therefore, maximum power point tracking (MPPT) controller is needed to improve the efficiency of the PV system by ensuring that the PV module continuously supplies maximum power despite changes in weather conditions (de Cesare et al., 2006; Houssamo et al., 2010). To ensure the high efficiency of the PV system, numer- ous MPPT algorithms have been developed, such as Fractional Open-Circuit Voltage (FOCV), Fractional Short-Circuit Current (FSCC), Fuzzy Logic, Neural Net- work, Perturbation and Observation (P&O), and Incre- mental Conductance (Reza Reisi et al., 2013; Gounden et al., 2009; Chaouachi et al., 2010; Altas and Sharaf, 2008; Faranda et al., 2008; de Brito et al., 2011; Ishaque et al., 2012). The simplest algorithms are FOCV and FSCC, which use the linearity of open-circuit voltage or short-circuit current to the maximum power point (MPP) voltage or current. However, these algorithms require intermittent disconnection of the PV module to obtain the open-circuit voltage or short-circuit current. Thus, overall efficiency of the PV system is lower because of the power losses during the disconnection (Qiang et al., 0038-092X/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.solener.2014.01.003 Corresponding author. E-mail address: [email protected] (S. Mekhilef). www.elsevier.com/locate/solener Available online at www.sciencedirect.com ScienceDirect Solar Energy 101 (2014) 333–342

Modified incremental conductance MPPT algorithm to ... · Modified incremental conductance MPPT algorithm to mitigate inaccurate responses ... maximum power point tracking ... SEPIC

Embed Size (px)

Citation preview

Available online at www.sciencedirect.com

www.elsevier.com/locate/solener

ScienceDirect

Solar Energy 101 (2014) 333–342

Modified incremental conductance MPPT algorithmto mitigate inaccurate responses under fast-changing

solar irradiation level

Kok Soon Tey, Saad Mekhilef ⇑

Power Electronics And Renewable Energy Research Laboratory (PEARL), Department of Electrical Engineering, University of Malaya, Malaysia

Received 9 February 2013; received in revised form 30 December 2013; accepted 2 January 2014Available online 29 January 2014

Communicated by: Associate Editor Jan Kleissl

Abstract

During the increment of solar irradiation, the conventional incremental conductance algorithm responds inaccurately at the first stepchange in the converter duty cycle. This paper presents the conventional algorithm confusion and proposes a modified incremental con-ductance algorithm that responds accurately when the solar irradiation level increases. Moreover, the proposed algorithm shows zerooscillation in the power of the solar module after the maximum power point (MPP) is tracked. MATLAB simulation is carried out withthe modified incremental conductance algorithm under a fast-changing solar irradiation level. Results of the modified, conventional andvariable step size incremental conductance algorithms are compared. Finally, the hardware implementation, consisting of a single-endedprimary-inductor converter (SEPIC) and a PIC controller, is applied as the maximum power point tracking (MPPT) controller. The sim-ulation and experimental works showed that the proposed algorithm performs accurately and faster during the increment of solar irra-diation level.� 2014 Elsevier Ltd. All rights reserved.

Keywords: MPPT; Incremental conductance; Photovoltaic (PV) system; SEPIC converter; Fast changing solar irradiation

1. Introduction

Electricity generation using solar energy has gainedincreasing popularity in recent years (Mekhilef et al.,2011, 2012). However, electricity generated by photovoltaic(PV) panels is an unstable energy source because of its firmdependence on factors such as solar irradiation level andsurrounding temperature. Therefore, maximum powerpoint tracking (MPPT) controller is needed to improvethe efficiency of the PV system by ensuring that the PVmodule continuously supplies maximum power despitechanges in weather conditions (de Cesare et al., 2006;Houssamo et al., 2010).

0038-092X/$ - see front matter � 2014 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.solener.2014.01.003

⇑ Corresponding author.E-mail address: [email protected] (S. Mekhilef).

To ensure the high efficiency of the PV system, numer-ous MPPT algorithms have been developed, such asFractional Open-Circuit Voltage (FOCV), FractionalShort-Circuit Current (FSCC), Fuzzy Logic, Neural Net-work, Perturbation and Observation (P&O), and Incre-mental Conductance (Reza Reisi et al., 2013; Goundenet al., 2009; Chaouachi et al., 2010; Altas and Sharaf,2008; Faranda et al., 2008; de Brito et al., 2011; Ishaqueet al., 2012). The simplest algorithms are FOCV andFSCC, which use the linearity of open-circuit voltage orshort-circuit current to the maximum power point (MPP)voltage or current. However, these algorithms requireintermittent disconnection of the PV module to obtainthe open-circuit voltage or short-circuit current. Thus,overall efficiency of the PV system is lower because of thepower losses during the disconnection (Qiang et al.,

Fig. 1. Block diagram of the PV system.

334 K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342

2011). Alternatively, Fuzzy Logic or Neural Network ischosen to obtain an accurate MPPT algorithm because oftheir ability to handle the non-linearity and dispensabilityof an accurate mathematical model, but these algorithmshighly dependent on user experience on the PV modulecharacteristics (Gounden et al., 2009; Chaouachi et al.,2010; Altas and Sharaf, 2008; Liu et al., 2013; Whei-Minet al., 2011). The PV module’s characteristics change alongwith time because of degradation. Therefore, the algorithmparameters need to be updated.

Among the aforementioned algorithms, P&O and incre-mental conductance are frequently used. These two algo-rithms work in accordance with the power againstvoltage (P–V) curve of the PV module. Both algorithmstune the DC–DC converter duty cycle in the PV systemto ensure that the latter operates at the MPP. For P&O,steady-state oscillation occurs after the MPP is reachedbecause the perturbation continuously changes in bothdirections to maintain the MPP (Weidong and Dunford,2004; Weidong et al., 2007; Fermia et al., 2007; Femiaet al., 2005; Abdelsalam et al., 2011). Steady-state oscilla-tion causes system power losses. For incremental conduc-tance, the slope of the P–V curve is used for MPPT. Theconventional incremental conductance algorithm deter-mines the gradient of the P–V curve. In addition, the dutycycle of the converter is tuned in fixed step size until thepeak (gradient is equal to zero) of the P–V curve is reached.However, the algorithm speed is slow when fixed step size isused. Therefore, Rahman et al. (2013) and Fangrui et al.(2008) introduced the variable step size in MPPT. In thealgorithm, the fixed step size is multiplied with the slopeof the P–V curve. Thus, the duty cycle step size becomessmaller when the PV system operates near to the MPP(peak of the P–V curve). Meanwhile, the step size is largerand possesses faster tracking speed when the PV systemoperates far from the MPP. Theoretically, when the peakof the P–V curve is found, no further perturbation of dutycycle or no oscillation in the power of the PV module(Qiang et al., 2011; Rahman et al., 2013; Fangrui et al.,2008; Kakosimos and Kladas, 2011) occurs. However, dur-ing implementation, the zero value is rarely obtained on theslope of the P–V curve because of the truncation error indigital processing. Conventional and variable step sizeincremental conductance algorithms are also unable torespond accurately at the first step change in the duty cycleof the converter after the increase in solar irradiation level(Zbeeb et al., 2009). Therefore, the present paper intro-duces new tracking steps to identify the changes in solarirradiation level. The variations in current (dI) and voltage(dV) of the PV module are used to track the increase insolar irradiation instead of the slope (dP/dV) of the P–V

curve. The proposed algorithm can respond accurately dur-ing the changes in solar irradiation level. Moreover, a smallpermitted error is used in the proposed algorithm to ensurethat no steady-state oscillation occurs. The results of bothsimulation and hardware implementation are provided.The results of the proposed algorithm are compared with

those of the conventional algorithm. The comparisonreveals that the proposed algorithm performs better byeliminating the steady-state oscillation and by respondingfaster and accurately to the variation in solar irradiationlevel. Fig. 1 shows the block diagram of the PV system.

1.1. PV module characteristic

PV module is a current source that comprises a numberof solar cells connected in series or parallel to generate elec-trical energy when there is sunlight. Solar cell is a type ofsemiconductor that produces electricity when sunlight isemitted onto it (Bennett et al., 2012). To understand solarcell characteristics, mathematical models have been devel-oped. Few types of models have been created, such as sin-gle-diode and two-diode models (Bennett et al., 2012;Ishaque and Salam, 2011; Ishaque et al., 2011; Gonzalez-Longatt, 2005). For simplicity, some models ignore theshunt resistance Rsh, which is normally very large andcan be considered as an open circuit. In this paper, sin-gle-diode model from Gonzalez-Longatt (2005) is chosento model the solar cell. The mathematical expression forthe equivalent electric circuit used in the model is shownbelow:

I ¼ npIsc � npIo�fexp½qðV =ns þ RsIÞ=nkT k� � 1g ð1Þ

where V is the output voltage of PV module, I the outputcurrent of PV module, Rs the series resistance of cell, q

the electron charge, Isc the light-generated current, k theBoltzman constant, Tk the temperature (K), n the diodeideality factor, ns the number of PV cells connected in ser-ies, np the number of PV cells connected in parallel, and Io

the reverse saturation current.As shown in Gonzalez-Longatt (2005), Isc is affected by

temperature and solar irradiation while Io is only affectedby temperature. The PV module model is generated usingEq. (1) and the Newton–Raphson method in MATLAB.

The maximum power available from the PV moduledepends on the surrounding temperature and solar irradia-tion level. Fig. 2 shows the two main characteristic curves,namely, current against voltage curve (I–V curve) and P–V

curve. These curves are used to investigate the effect of tem-perature and solar irradiation level on the PV module.When a load is connected to the PV module, a load lineis imposed on the I–V curve. The voltage and current of

Fig. 2. PV current against voltage curve (I–V Curve) and PV poweragainst voltage curve (P–V curve).

K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342 335

the PV module are at the point where the load line inter-sects with the I–V curve. Then, the power is simply the mul-tiplication of the current and voltage at the intersectionpoint. The load line position on the I–V curve dependson the impedance of the load (Coelho et al., 2009; Ying-Tung and China-Hong, 2002). Therefore, a DC–DC con-verter is needed between the PV module and load.

The duty cycle of the DC–DC converter is tuned toensure that the load line always intersects with the I–V

curve at the MPP (Fig. 2). MPPT algorithm is used toensure that the PV module constantly operates at theMPP, regardless of the intensity of sunlight, surroundingtemperature, and load resistance.

2. Conventional and variable step size incremental

conductance algorithm

Incremental conductance algorithm detects the slope ofP–V curve. The MPP is tracked by searching the peak ofthe P–V curve. This algorithm uses the instantaneous con-ductance I

V and the incremental conductance dIdV for MPPT.

Using these two values, the algorithm determines the loca-tion of the operating point of the PV module in the P–V

curve. Eq. (2) shows that the PV module operates at theMPP. Meanwhile, Eq. (3) shows that the PV module oper-ates at the left side of the MPP, whereas Eq. (4) shows thatthe PV module operates at the right side of the MPP in theP–V curve (Qiang et al., 2011; Fangrui et al., 2008; Safariand Mekhilef, 2011; Zhou et al., 2008; Bo-Chiau andChun-Liang, 2011) .

dIdV¼ � I

Vð2Þ

dIdV

> � IV

ð3Þ

dIdV

< � IV

ð4Þ

The equations above are obtained from the concept wherethe slope of the P–V curve at MPP is equal to zero, asshown in Eq. (5):

dPdV¼ 0 ð5Þ

By rewriting Eq. (5), the following equations are obtained:

dPdV¼ I

dVdVþ V

dIdV

ð6Þ

dPdV¼ I þ V

dIdV

ð7Þ

I þ VdIdV¼ 0 ð8Þ

In the conventional incremental conductance algorithm,Eq. (8) is used to detect the MPP and flowchart of the algo-rithm is as shown in Fig. 3(a). The voltage and current ofthe PV module are sensed by the MPPT controller. If Eq.(3) is satisfied, the duty cycle of the converter needs to bedecreased, and vice versa if Eq. (4) is satisfied. No changeon the duty cycle occurs if Eq. (8) is satisfied.

The variable step size incremental conductance algo-rithm proposed in Fangrui et al. (2008) is able to increasethe tracking speed of the MPPT controller. The flow chartof the algorithm is shown in Fig. 3(b). The sequences in thealgorithm are mostly similar to the conventional incremen-tal conductance and the only difference is the step size cal-culation. Eq. (9) is used in the variable step size algorithmto vary the duty cycle step size where N is the scalingfactor.

Step ¼ N �absðdP=dV Þ ð9Þ

2.1. Weakness of conventional and variable step size

incremental conductance

Incremental conductance algorithm uses the slope of theP–V curve to detect the MPP. If the algorithm finds thatthe operating point is at the peak of the P–V curve (slopeequals to zero) and if Eq. (5) is satisfied, then the duty cycleof the DC–DC converter is fixed and no oscillation duringthis stage occurs until changes in the slope arise. However,in real life, the zero slope condition is rarely achieved, asmentioned in Weidong et al. (2007), because of the trunca-tion error of the numerical differentiation.

Apart from steady-state oscillation, the conventionalalgorithm is confused when the solar irradiation increases.As shown in Fig. 4, when the irradiation is at 0.4 kW/m2,the MPPT algorithm adjusts the duty cycle to ensure thatthe PV system operates at load line 2 and the MPP (pointB) is tracked. After some time, the solar irradiationincreases, but the duty cycle is maintained at load line 2.Therefore, point G will be found by load line 2 in theI–V curve of 1.0 kW/m2, corresponding to the power atpoint C in the P–V curve. The conventional incrementalconductance algorithm calculates the gradient betweenpoints B and C. The result is a positive gradient. In fact,the power found by load line 2 is at point C. The gradientbetween point C and the MPP (point A) of 1.0 kW/m2 hasa negative value. Without noticing this error, the conven-tional incremental conductance algorithm increases thevoltage of the PV module. Therefore, an inaccurate firststep change is implemented by the conventional algorithmwhen solar irradiation level changes from low to high.

Fig. 3. Flow chart for (a) conventional incremental conductance algorithm and (b) variable step size incremental conductance algorithm.

Fig. 4. I–V and P–V curves of irradiation 1000 W/m2 and 400 W/m2.

Table 1Changes in PV voltage and current during the change in solar irradiationand load resistance.

Voltage changes (dV) Current changes (dI)

Solar Increase IncreaseIrradiation Decrease DecreaseLoad Increase Increase DecreaseResistance Decrease Decrease Increase

336 K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342

However, this problem does not occur when the solar irra-diation level decreases from high to low. The reason isbecause from points E to H, or in the P–V curve frompoints A to D, the gradient is positive. The gradientbetween points B and D is also positive.

3. Proposed incremental conductance algorithm

The incremental conductance algorithm depends on theslope of the P–V curve, which is affected by the solar irra-diation level and load resistance. The algorithm uses thecurrent and voltage of the PV module in the calculation.Therefore, the effect of solar irradiation and load changeson the current and voltage of the PV module must be wellconsidered in the algorithm.

Table 1 shows the summary of changes in the voltageand current of the PV module against the changes in solarirradiation level and load resistance. As shown in Fig. 4,when the PV system operates at load line 2 (point F) andsolar irradiation suddenly increases, the operating pointof the PV system moves to point G. Therefore, both volt-age and current increase. Conversely, when the PV systemoperates at load line 1 (point E) and solar irradiation sud-denly decreases, the operating point of the PV system

moves to point H. Thus, both voltage and current decrease.In the conventional incremental conductance algorithm,these two types of changes are not well considered. Mean-while, if the PV system operates at load line 1 and loadresistance increases, the PV system will operate at load line2. Therefore, the PV module voltage increases and the PVmodule current decreases. Alternatively, the voltagedecreases and the current increases when the load resis-tance decreases.

A permitted error is applied to eliminate the steady-stateoscillation. Eq. (8) is rewritten as

I þ VdIdV

����

����< 0:06 ð10Þ

Using the permitted error of 0.06, the steady-state error isapproximately 0.7% with a duty cycle step size of 0.005 inthe proposed algorithm. In order to select the suitable dutycycle step size for the proposed system, simulation data iscollected and the PV module used in the simulation isKyocera KC85T (the parameters of the PV module are inTable 3). Fig. 5 shows the power of the PV module forthree different step size values of 0.001, 0.005, and 0.01when the duty cycle is varied around the MPP. Table 2summarizes the characteristics for the three different typesof step sizes. All step sizes can track the MPP at approxi-mately 86.37 W. Although the 0.01 step size responds veryfast, the power value differs at 1.16 W between two consec-utive step changes, which is recognized as the largestamong the three step sizes. Thus, the MPPT controller is

Table 2Comparison between three different step size values.

Characteristics Step size

0.001 0.005 0.01

Maximum power 86.375 W 86.375 W 86.374 WMinimum slope value 0.0165 0.0527 0.0967Response time Slow Medium FastPower different between two

consecutive step changes0.00349 W 0.28115 W 1.1642 W

Table 3Parameters of KC85T PV Module at STC: temperature = 25 �C, insola-tion = 1000 W/m2.

Maximum power (Pmax) 87 WVoltage at MPP (Vmpp) 17.4 VCurrent at MPP (Impp) 5.02 AOpen-circuit voltage (Voc) 21.7 VShort-circuit current (Isc) 5.34 A

Fig. 5. PV power for three different step size values at solar irradiationlevel 1000 W/m2 when the duty cycle varies near the MPP.

K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342 337

not precise enough. For the 0.001 step size, the power val-ues are close to one another (shown in Fig. 5) when theduty cycle is varied. Therefore, the MPPT controller re-sponse is slow. The 0.005 step size shows that the differencein the power of the PV module is acceptable among thethree different step sizes. The controller can respond pre-cisely and rapidly with this duty cycle step size. Therefore,0.005 is chosen as the step size of the duty cycle in the sys-tem. In Table 2, the minimum slope value is used to set thepermitted error in Eq. (10). As shown in the table, the min-imum slope value for the 0.005 step size is 0.0527. Thus, thepermitted error is set to 0.06 to stop the varying of step sizewith the slope value below this permitted error.

Finally, the proposed algorithm is seen in the flow chartshown in Fig. 6. Initially, a flag value is set to zero. Thisflag value is used to indicate that the MPP is reached whenit is set to 1. If the flag value is zero, the conventional incre-mental conductance algorithm is implemented with the useof Eq. (10). When the condition in Eq. (10) is reached, thesystem operates at the MPP. Therefore, the algorithm setsthe flag to 1, and then goes into the improved algorithm. Inthe improved algorithm, the program continues checkingthe condition of Eq. (10). If the solar irradiation and loadresistance remain constant, no variation is made on the

duty cycle. When changes occur in either solar irradiationor load, the algorithm sets the flag value to zero and thendetermines the changes in the voltage and current of thePV module. If the algorithm finds that both the currentand voltage increased, then the duty cycle is also increased.Consequently, the incremental conductance algorithm ismodified to overcome the inaccurate response during theincrease in solar irradiation.

4. Results and discussion

4.1. Simulation results

Fig. 7 shows the MATLAB Simulink model of the entireMPPT system, which consists of the PV module model, asingle-ended primary-inductor converter (SEPIC), and anMPPT controller. The PV module model is designed basedon a Kyocera PV module (KC85T module). Table 3 showsthe parameters of the module. The values of the compo-nents in the converter are as follows: Cin and Cout = 3900lF, L1 and L2 = 125 lH, Cs = 1000 lF, and load = 10 Xresistance. The switching frequency of the converter is setto 20 kHz.

The simulation is conducted to investigate the differ-ences between the three algorithms. In the simulation, thesolar irradiation is changed from low to high and then tolow again. The duty cycle step size for the converter in con-ventional and proposed algorithm is 0.005, the scaling fac-tor, N in the variable step size is 0.03 and the sampling timefor the MPPT controller is 0.05 s. Fig. 8(a) shows the sim-ulation results of the conventional incremental conduc-tance algorithm. At the beginning of the simulation, thesolar irradiation is set to 0.8 kW/m2, and the MPP isreached at t = 0.2 s. The duty cycle fluctuates between0.61 and 0.62. Simultaneously, oscillation occurs in thepower (69.40–69.78 W) of the PV module. At t = 0.716 s,the solar irradiation is increased to 1.0 kW/m2, but theduty cycle remains at 0.62. Consequently, the power ofthe PV module increases. At t = 0.75 s, the MPPT control-ler is sampled with the conventional incremental conduc-tance algorithm. The duty cycle first step change isconfusing and obtains inaccurate responses. Therefore,the power of the PV module is decreased, as shown inFig. 8(a), point A. Afterward, the algorithm reverts direc-tion and the power of the PV module is increased. Att = 1.0 s, the MPP is reached for solar irradiation of1.0 kW/m2. Oscillation in the power of the PV module(86.04–86.37 W) occurs again. At t = 1.44 s, the solar irra-diation level decreases to 0.8 kW/m2, and the conventionalincremental conductance algorithm works accurately.Fig. 8(b) shows the results of the variable step size algo-rithm. When the solar irradiation is set to 0.8 kW/m2, theMPP is reached at t = 0.2 s. The duty cycle fluctuatesbetween 0.61 and 0.62. Due to the use of Eq. (9) in thealgorithm, the duty cycle changes become smaller as theoperating point near to the peak. Therefore, the oscillationin the power (69.35–69.78 W, point D) of PV module is

Fig. 6. Flow chart of the proposed incremental conductance algorithm.

Fig. 7. MATLAB simulation of an MPPT controller with SEPICconverter.

338 K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342

also found to become smaller along with time. Att = 0.716 s, the solar irradiation is increased to 1.0 kW/m2. The same problem faced by the variable step size algo-rithm as the inaccurate duty cycle changes is applied by theMPPT controller as shown in Fig. 8(b), point C. This is dueto the algorithm sequence which similar to the conven-tional algorithm is used during the variation in solar irradi-ation. During the steady state condition, the power of PVmodule is also oscillating (86.24–86.37 W). When the solarirradiation level decreases, the algorithm is able to responseaccurately and in fact it has faster response due to the useof variable step size.

Fig. 8(c) shows the simulation results of the proposedincremental conductance algorithm. At t = 0.2 s, theMPP for 0.8 W/m2 is reached and the duty cycle is main-tained at 0.615. The power of the PV module is maintainedat 69.78 W. Thus, the power losses are lower due to thereduction in steady-state oscillation. At t = 0.716 s, thesolar irradiation is increased to 1.0 kW/m2. Then, att = 0.75 s, the proposed algorithm detects the increase insolar irradiation and performs an accurate variation in

the duty cycle, as shown in Fig. 8(c), point E. Therefore,the power is increased from the first step until the MPPreached t = 0.9 s and the power of the PV module is main-tained at 86.37 W. The proposed algorithm only needs foursteps to reach the MPP. The response of the proposed algo-rithm on the changes in solar irradiation is faster comparedwith that of the conventional and variable step size algo-rithm, which needs six steps and five steps respectively.The proposed algorithm is 0.1 s faster compared with theconventional algorithm during the increase in solarirradiation level. Moreover, the steady-state oscillation isreduced when the permitted error reaches 0.06. Finally,the computational time needed by the proposed algorithmis only 4 instructions more than the conventional algorithmwhich are (i) initialize the flag value to zero, (ii) check theflag value is equal to one or zero, (iii) clear the flag valueif the permitted error is not meet, (iv) check the variationin both the current and voltage of PV modules. Theproposed algorithm determines the variation direction inthe voltage and current before increasing or decreasingthe duty cycle in the next sampling time immediately afterthe variations in solar irradiation. Meanwhile, the conven-tional algorithm responds in the next sampling timeimmediately after the variation in solar irradiation. Theuse of a microcontroller in the hardware implementationcauses the processing time of the steps introduced in theproposed algorithm to be less than the sampling time.Therefore, the proposed algorithm does not have anyimplementation issue with the microcontroller. Table 4summarizes the comparison between both algorithms.

Fig. 8. Simulation results: (a) Conventional incremental conductance algorithm. (b) Variable step size algorithm. (c) Proposed incremental conductancealgorithm.

K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342 339

4.2. Experimental results

For hardware implementation, both conventional andproposed algorithms are implemented using the PIC con-troller. The PIC used in the controller is MicrochipPIC18F4520 with 40 MHz processing speed. A 10-bitADC conversion is used to convert the voltage and current

of the PV module. Current sensor (LEM LA25-NP) andvoltage sensor (LEM LV25-P) are used to sense the voltageand current of the PV module. In the experiment, thecomponents of the SEPIC have the same value as in thesimulation. The solar array simulator (SAS) from Agilent(E4360A) is used to generate the output characteristics ofa PV array. The Agilent SAS is a DC power source with

Table 4Comparison between the conventional and proposed incremental conductance algorithms.

Evaluated parameters Conventional algorithm Variable step size algorithm Proposed algorithm

Tracking steps during increase in solar irradiation level 6 steps 5 steps 4 stepsSteady-state oscillation Large Small NoWrong response during increase in solar irradiation Yes Yes NoPractical implementation Yes Yes Yes

Fig. 9. Agilent solar array simulator and experimental setup of MPPT system.

Fig. 10. I–V curve and P–V curve from the SAS. (a) Conventional incremental conductance algorithm. (b) Proposed incremental conductance algorithm.

Fig. 11. Experimental results: waveform for PV power, voltage, and current. (a) Conventional incremental conductance algorithm. (b) Proposedincremental conductance algorithm.

340 K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342

Fig. 12. Experimental results for variable step size algorithm: (a) waveform for PV power, voltage, and current (b) P–V curve.

Fig. 13. Waveform for PV power, voltage, and current when the loadresistances change.

K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342 341

600 W output and it is primarily a current source with verylow output capacitance and is capable of simulating theI–V curve of different arrays under various conditions.The desired I–V curve is programmable over the Ethernetand is conveniently generated within the SAS. Fig. 9 showsthe SAS and the experimental setup.

The solar irradiation level in the hardware implementa-tion is varied as in the simulation which is from 0.8 kW/m2

increased to 1.0 kW/m2 and then decreased back to0.8 kW/m2. The experiment results are similar with the sim-ulation results. For conventional and variable step sizealgorithm, confusion occurs with the increase in solar irra-diation, as shown in Figs. 11(a) and 12(a). Oscillation alsooccurs in the power of the PV module when the MPP isreached. From the P–V curve given by the SAS, the oscil-lation in the power of the PV module by using the conven-tional algorithm is from 85.08 W to 86.3 W, the lowestoperating point in the steady state is as shown inFig. 10(a). Meanwhile, for the variable step size, the oscil-lation of power is from 84.64 W to 86.3 W and the lowest

operating point in the steady state is as shown inFig. 12(b). The larger oscillation in the variable step sizealgorithm is due to the operating point does not reach nearto the MPP yet. When the MPP is reached, the oscillationbecome smaller and even without oscillation as shown inFig. 12(a). The proposed algorithm responds accuratelyagainst the increase in solar irradiation level, as shown inFig. 11(b). In addition, no steady-state oscillation occurswhen the MPP is reached. The operating point of the PVmodule during the steady state is shown in Fig. 10(b),86.3 W.

To ensure that the algorithm is functioning well, theresistive load is varied. Fig. 13 shows the results. Initially,the resistive load is at 10 O, and the MPP is tracked. Next,the resistive load changes to 6.67 O at point D. The currentof the PV module is increased while the voltage isdecreased. The PV module operates at the left side of theP–V curve (Fig. 13). The algorithm decreases the duty cycleof the converter to reach the MPP again. Afterward, theresistive load is switched to 10 O at point E, the currentof the PV module is decreased, and the voltage of the PVmodule is increased. The PV module operates at the rightside of the P–V curve. The algorithm increases the dutycycle and reaches MPP again.

5. Conclusion

In this paper, the proposed incremental conductancealgorithm was used to track the MPP for PV module undera fast-changing solar irradiation level. The confusion facedby the conventional algorithm was discussed. Modifica-tions were applied on the conventional incremental con-ductance algorithm to mitigate the inaccurate response.The simulation and experimental results showed that theproposed algorithm responds and tracks the MPP accu-rately. Subsequently, the proposed algorithm does notshow steady-state oscillation, thereby reducing the powerlosses. In conclusion, the proposed algorithm performsaccurately and better than the conventional algorithm.

342 K.S. Tey, S. Mekhilef / Solar Energy 101 (2014) 333–342

Acknowledgement

The financial support of this research by High ImpactResearch (HIR-MOHE), under the Project code UM.C/HIR/MOHE/ENG/24 and University of Malaya ResearchGrant (UMRG), under the Project code RP015A-13AETare greatly appreciated.

References

Abdelsalam, A.K., Massoud, A.M., Ahmed, S., Enjeti, P.N., 2011. High-performance adaptive perturb and observe MPPT technique forphotovoltaic-based microgrids. IEEE Trans. Power Electron. 26,1010–1021.

Altas, I.H., Sharaf, A.M., 2008. A novel maximum power fuzzy logiccontroller for photovoltaic solar energy systems. Renew. Energy 33,388–399.

Bennett, T., Zilouchian, A., Messenger, R., 2012. Photovoltaic model andconverter topology considerations for MPPT purposes. Sol. Energy 86,2029–2040.

Bo-Chiau, C., Chun-Liang, L., 2011. Implementation of maximum-power-point-tracker for photovoltaic arrays, in: 6th IEEE Conference onIndustrial Electronics and Applications (ICIEA), pp. 1621–1626.

Chaouachi, A., Kamel, R.M., Nagasaka, K., 2010. A novel multi-modelneuro-fuzzy-based MPPT for three-phase grid-connected photovoltaicsystem. Sol. Energy 84, 2219–2229.

Coelho, R.F., Concer, F., Martins, D.C., 2009. A study of the basic DC–DC converters applied in maximum power point tracking, in: PowerElectronics Conference, pp. 673–678.

de Brito, M.A.G., Sampaio, L.P., Luigi, G., e Melo, G.A., Canesin, C.A.,2011. Comparative analysis of MPPT techniques for PV applications,in: International Conference on Clean Electrical Power (ICCEP), pp.99–104.

de Cesare, G., Caputo, D., Nascetti, A., 2006. Maximum power pointtracker for portable photovoltaic systems with resistive-like load. Sol.Energy 80, 982–988.

Fangrui, L., Shanxu, D., Fei, L., Bangyin, L., Yong, K., 2008. A variablestep size INC MPPT method for PV systems. IEEE Trans. Industr.Electron. 55, 2622–2628.

Faranda, R., Leva, S., Maugeri, V., 2008. MPPT techniques for PVsystems: energetic and cost comparison, in: Power and Energy SocietyGeneral Meeting – Conversion and Delivery of Electrical Energy in the21st Century, pp. 1–6.

Femia, N., Petrone, G., Spagnuolo, G., Vitelli, M., 2005. Optimization ofperturb and observe maximum power point tracking method. IEEETrans. Power Electron. 20, 963–973.

Fermia, N., Granozio, D., Petrone, G., Vitelli, M., 2007. Predictive &adaptive MPPT perturb and observe method. IEEE Trans. Aerosp.Electron. Syst. 43, 934–950.

Gonzalez-Longatt, F.M., 2005. Model of Photovoltaic Module inMatiabe. Int. J. Comput. Technol. Elect. Eng. (IJCTEE) 2.

Gounden, N.A., Ann Peter, S., Nallandula, H., Krithiga, S., 2009. Fuzzylogic controller with MPPT using line-commutated inverter for three-phase grid-connected photovoltaic systems. Renew. Energy 34, 909–915.

Houssamo, I., Locment, F., Sechilariu, M., 2010. Maximum powertracking for photovoltaic power system: development and experimen-tal comparison of two algorithms. Renew. Energy 35, 2381–2387.

Ishaque, K., Salam, Z., Syafaruddin, 2011. A comprehensive MATLABSimulink PV system simulator with partial shading capability based ontwo-diode model. Sol. Energy 85, 2217–2227.

Ishaque, K., Salam, Z., Taheri, H., 2011. Simple, fast and accurate two-diode model for photovoltaic modules. Sol. Energy Mater. Sol. Cells95, 586–594.

Ishaque, K., Salam, Z., Amjad, M., Mekhilef, S., 2012. An improvedParticle Swarm Optimization (PSO)-based MPPT for PV with reducedsteady-state oscillation. IEEE Trans. Power Electron. 27, 3627–3638.

Kakosimos, P.E., Kladas, A.G., 2011. Implementation of photovoltaicarray MPPT through fixed step predictive control technique. Renew.Energy 36, 2508–2514.

Liu, Y.-H., Liu, C.-L., Huang, J.-W., Chen, J.-H., 2013. Neural-network-based maximum power point tracking methods for photovoltaicsystems operating under fast changing environments. Sol. Energy 89,42–53.

Mekhilef, S., Saidur, R., Safari, A., 2011. A review on solar energy use inindustries. Renew. Sustain. Energy Rev. 15, 1777–1790.

Mekhilef, S., Safari, A., Mustaffa, W.E.S., Saidur, R., Omar, R., Younis,M.A.A., 2012. Solar energy in Malaysia: current state and prospects.Renew. Sustain. Energy Rev. 16, 386–396.

Qiang, M., Mingwei, S., Liying, L., Guerrero, J.M., 2011. A novelimproved variable step-size incremental-resistance MPPT method forPV systems. IEEE Trans. Industr. Electron. 58, 2427–2434.

Rahman, N.H.A., Omar, A.M., Saat, E.H.M., 2013. A modification ofvariable step size INC MPPT in PV system, in: Power Engineering andOptimization Conference (PEOCO), 2013 IEEE 7th International, pp.340–345.

Reza Reisi, A., Hassan Moradi, M., Jamasb, S., 2013. Classification andcomparison of maximum power point tracking techniques for photo-voltaic system: a review. Renew. Sustain. Energy Rev. 19, 433–443.

Safari, A., Mekhilef, S., 2011. Simulation and hardware implementationof incremental conductance MPPT with direct control method usingCuk converter. IEEE Trans. Industr. Electron. 58, 1154–1161.

Weidong, X., Dunford, W.G., 2004. A modified adaptive hill climbingMPPT method for photovoltaic power systems, in: IEEE 35th AnnualPower Electronics Specialists Conference, pp. 1957–1963.

Weidong, X., Dunford, W.G., Palmer, P.R., Capel, A., 2007. Applicationof centered differentiation and steepest descent to maximum powerpoint tracking. IEEE Trans. Industr. Electron. 54, 2539–2549.

Whei-Min, L., Chih-Ming, H., Chiung-Hsing, C., 2011. Neural-network-based MPPT control of a stand-alone hybrid power generation system.IEEE Trans. Power Electron. 26, 3571–3581.

Ying-Tung, H., China-Hong, C., 2002. Maximum power tracking forphotovoltaic power system, in: Industry Applications Conference, pp.1035–1040.

Zbeeb, A., Devabhaktuni, V., Sebak, A., 2009. Improved photovoltaicMPPT algorithm adapted for unstable atmospheric conditions andpartial shading, in: International Conference on Clean ElectricalPower, pp. 320–323.

Zhou, Y., Liu, F., Yin, J., Duan, S., 2008. Study on realizing MPPT byimproved Incremental Conductance method with variable step-size, in:3rd IEEE Conference on Industrial Electronics and Applications,pp. 547–550.