Development of adaptive perturb and observe-fuzzy control maximum power point tracking for photovoltaic boost dc–dc converter

www.ietdl.org

IE

d

Published in IET Renewable Power GenerationReceived on 6th December 2012Revised on 7th May 2013Accepted on 17th May 2013doi: 10.1049/iet-rpg.2012.0362

T Renew. Power Gener., 2014, Vol. 8, Iss. 2, pp. 183–194oi: 10.1049/iet-rpg.2012.0362

ISSN 1752-1416

Development of adaptive perturb and observe-fuzzycontrol maximum power point tracking forphotovoltaic boost dc–dc converterMuhammad Ammirrul Atiqi Mohd Zainuri1, Mohd Amran Mohd Radzi1, Azura Che Soh1,

Nasrudin Abd Rahim2

1Department of Electrical and Electronic Engineering, Faculty of Engineering, Universiti Putra Malaysia, UPM Serdang,

Selangor, Malaysia2University of Malaya Power Energy Dedicated Advanced Centre (UMPEDAC), Kuala Lumpur, Malaysia

E-mail: [email protected]

Abstract: This study presents an adaptive perturb and observe (P&O)-fuzzy control maximum power point tracking (MPPT) forphotovoltaic (PV) boost dc–dc converter. P&O is known as a very simple MPPT algorithm and used widely. Fuzzy logic is alsosimple to be developed and provides fast response. The proposed technique combines both of their advantages. It should improveMPPT performance especially with existing of noise. For evaluation and comparison analysis, conventional P&O and fuzzy logiccontrol algorithms have been developed too. All the algorithms were simulated in MATLAB-Simulink, respectively, togetherwith PV module of Kyocera KD210GH-2PU connected to PV boost dc–dc converter. For hardware implementation, theproposed adaptive P&O-fuzzy control MPPT was programmed in TMS320F28335 digital signal processing board. The othertwo conventional MPPT methods were also programmed for comparison purpose. Performance assessment covers overshoot,time response, maximum power ratio, oscillation and stability as described further in this study. From the results and analysis,the adaptive P&O-fuzzy control MPPT shows the best performance with fast time response, less overshoot and more stableoperation. It has high maximum power ratio as compared to the other two conventional MPPT algorithms especially withexisting of noise in the system at low irradiance.

1 Introduction

In recent years, various research works have been done on theuse of photovoltaic (PV) energy as alternative resource. PVenergy is one of the most promising renewable energyresources and it is much clean, inexhaustible and free toharvest [1]. Several applications employing this technologyhave been developed such as satellite power systems, solarpower generations, solar battery charging stations and solarvehicles [2–5]. The main disadvantage of PV is the lowefficiency of energy conversion as compared with otheralternative resources. PV is a non-linear source that dependson irradiation and temperature in its operation. Maximumpower point tracking (MPPT) is introduced to extract themaximum power from the PV array.Currently, perturb and observe (P&O) algorithm is the

most popular and used widely since it is the simplestalgorithm and easy to be implemented as compared withother methods [6–9]. However, it still has commondrawbacks as follows:

† Poor tracking, not intelligent enough and less efficientduring rapid change of the irradiance because it movesaway from the real maximum power point (MPP) [10, 11].

† Inability to verify whether the higher new output powervalue is because of the new irradiation amount or the newduty cycle value [10].† Continuous oscillations around the optimal operating pointmake the average power level is deviated from the MPPespecially at low irradiance [12, 13].† It goes back and forth around the MPP and unable to stickexactly [14].† Slow time response [15].

Another possible significant drawback of this algorithm isits inability to perform well with existing of noise. In suchPV system, noise factor exists and must seriously beconsidered not only because of non-linearity of PV source,but may come from the embedded controller, voltage andcurrent measurements [16]. The noises from bothmeasurements have significant effect on the decisions madeby the MPPT algorithm [16–18]. In specific case, forexample, at the MPP, the algorithm is expected to make anequal number of decisions for incrementing anddecrementing the reference, leaving the operating pointconstant on average [16]. With noisy voltage and currentmeasurements, some of the decisions made by thealgorithm become incorrect. The noise from voltagemeasurement will affect the right hand side of the PV

183& The Institution of Engineering and Technology 2014

www.ietdl.org
curve, and meanwhile, the noise from current measurementwill affect both sides of the PV curve [16]. Oscillationsmay occur which again contributes to power losses.A latest progress in development of MPPT is to use the
artificial intelligence control techniques such as fuzzy logic,neural network, neuro-fuzzy [19–27] and genetic algorithm[28]. The advantages of using these techniques are fast timeresponse and more stable as compared to the classical ortraditional algorithms. Fuzzy logic control (FLC) algorithmis the most popular among them since it is simple andpractical to be implemented. It has been implemented withnotable improvement in efficiency or maximum power ratio[29–31].Even though FLC algorithm is better as compared with

P&O algorithm, there is still drawback occurs where it doesnot able to locate accurate enough the MPP. Its inputs, errorand change of error, need to be calculated first. Thecalculation process may contribute drawbacks in terms oftime response and accuracy of the MPPT to track the MPPwhen the algorithm is implemented in digital controller.Furthermore, most works on this algorithm did not considernoise factor and it would not work appropriately with lowirradiation which resulted high oscillations. Fig. 1 showsthe block diagram of proposed PV system to evaluate theMPPT algorithm. PV’s voltage and current are measuredand used by MPPT algorithm to produce appropriate dutycycle to the boost dc–dc converter which is connectedbetween the PV panel and the load.This paper presents work on improving MPPT algorithm

which should be able to increase performance of the PVsystem. This work introduces adaptive combination offeatures in both conventional P&O and FLC to form asingle algorithm for MPPT of PV dc–dc boost converterwhich should be simple and fast, and the most significancecontribution and achievement is its ability to response wellto existing of noise. The proposed MPPT algorithm wasevaluated and compared with both conventional P&O andFLC algorithms under steady state and dynamicperformances with various irradiance conditions. Anotheradditional focus in this work is to cover evaluation ofMPPT performance at low irradiance which may contributesignificant different among these three algorithms.To further explain about this work, Section 2 in this paper

covers theory of boost dc–dc converter of the PV system, andfollowed by discussion on MPPT algorithms in term of theirstructures and improvements that have been made for each

Fig. 1 Block diagram of proposed PV system


algorithm in Sections 3. Simulation work and hardwareimplementation including the results are discussed inSections 4 and 5, respectively. Finally, Section 6 concludesfindings from this work.

2 dc–dc boost converter

PV systems typically employ a dc–dc or dc–ac converter, withor without battery charging capabilities, which make use ofMPPT [32]. dc–dc converter can be used asswitching-mode regulator to convert a dc voltage, normallyunregulated, to a regulated dc output voltage. Theregulation is normally achieved by pulse-width modulation(PWM) technique and the switching device is normallyMOSFET or IGBT. Boost dc–dc converter’s function is tostep up dc voltage. Maximum power is reached when theMPPT algorithm changes and adjusts the PWM’s dutycycle of the boost dc–dc converter. The value of inductor is120 μH and the capacitor is about 470 μF. For the load,resistor is used with value of 50 Ω.

3 Maximum power point tracking

3.1 Perturb and observe

The conventional and popular MPPT method is P&Oalgorithm [33–35]. This algorithm works by using themethod of perturbation on the desired maximum point. Twotypes of P&O are by using direct duty cycle to produceoutput (also known as hill climbing), or by using voltagereference. The advantage of using direct duty cycle is it canavoid using proportional-integral algorithm to controlovershoot and steady-state error of the output. However, itsdisadvantage is hard to tune to obtain the desired dutycycle. Fig. 2 shows flowchart of classic P&O algorithmwith duty cycle method [33–35]. When PV power andvoltage are increasing, a perturbation will increase a stepsize ΔD to be added with the duty cycle D, in order togenerate next cycle of perturbation and to force theoperating point moving towards the MPP. If the PV powerdecreases and PV voltage increases, the P&O will workvice versa [33]. The inputs of the P&O algorithm arecurrent and voltage of the PV.Previous works evaluated this algorithm in steady-state

operation. It was noticed some oscillations occurred aroundthe MPP which caused power losses [36]. Lower MPPT

IET Renew. Power Gener., 2014, Vol. 8, Iss. 2, pp. 183–194doi: 10.1049/iet-rpg.2012.0362

Fig. 2 P&O algorithm with direct duty cycle

www.ietdl.org

performance including achievement of MPP ratio for thisalgorithm at the steady-state operation has significantlybeen reported and discussed. Among reported MPP ratiowere around 80% [37], 85.99% [15] and up to 90% [38].However, the worst MPP ratio was reported in [39], whichwas about 68% only [39]. In [40], range of MPP ratio wasreported rather than specific value, which was around 81–85% only.Besides assessment in the steady-state operation, MPPT

performance can be evaluated under dynamic approach withcontinuous varying irradiance intensities. The dynamicapproach has been regarded as a good performanceindicator for MPPT, as highlighted in European StandardEN 50530, which was released on April 2010 [41]. In thisstandard, the irradiance profile for evaluating dynamicMPPT performance is presented. Further previous works onP&O algorithm in dynamic operation reported the MPPratio or efficiency were above 98 [42] and 99.5% [11],respectively, with important highlight that both of themwere operated at high irradiation condition.Meanwhile, there were some research works that tried to

improve the P&O algorithm. In [43], both peak currentcontrol and instantaneous value of power were used tocalculate the direction of next cycle. These workscontributed to increasing of the speed response of thesystem and reducing the amplitude of power oscillationsaround the MPP. Another work in [44] proposed a novelMPPT control algorithm for a half bridge inverter with acontrol unit consisting of a multistage buck–boostconverter. It allowed each of the PV modules to generate its


maximum power by simply detecting only the total outputpower of the PV system. In [45], a three-point weightcomparison for P&O method was proposed to track theMPP under rapid changes of irradiance. Besides that,sampling rate may contribute to better performance ofMPPT, and thus, optimisation of sampling rate was done toimprove time response and efficiency of the algorithm tomake appropriate decision [45, 46].

3.2 Fuzzy logic control

Application of artificial intelligence is the new trend ofresearch works in MPPT. Typical or conventional FLC forMPPT uses two inputs such as error E and change in errorCE at sample time k, which are defined by (1) and (2),while the output of FLC is the duty cycle D [46].

E(k) = P(k)− P(k − 1)

V (k)− V (k − 1)(1)

CE(k) = E(k)− E(k − 1) (2)

FLC is divided into four categories, which includefuzzification, fuzzy inference, rule-base and defuzzificationas shown in Fig. 3. During fuzzification, the numericalinput variables are converted into linguistic variables basedon the membership functions. Various fuzzy levels could beused for input and output variables.In operation of MPPT control, after E and CE are

calculated, these inputs are converted into linguistic


Fig. 3 Block diagram of conventional FLC MPPT algorithm

www.ietdl.org

variables and then the output D is generated by looking up ina rule-base table. The FLC tracks the MPP based on masterrule of ‘If X and Y, Then Z’ [47]. To determine the outputof the fuzzy logic, the fuzzy inference is used. There aremany methods for inference but the popular one isMamdani [23]. Other methods include compositional ruleof inference, generalised Modus Ponens and Sugenoinference method. Usually, weights are added to the rules toimprove reasoning accuracy and to reduce undesirableconsequent. The fuzzy output is converted back tonumerical variable from linguistic variable duringdefuzzification. The most common method used for thisdefuzzification is the centroid of area (COA) since it hasgood averaging properties and it produces more accurateresults. Other defuzzification methods include bisector andmiddle of maxima [19].Evaluation of MPP ratio was also carried in some works

related to this MPPT algorithm. In [29], the MPP ratio forthis work was reported about 90–96% of the maximumpower. In another work [30], the MPP ratio of FLC washigher around 2.53% from classical P&O MPPT algorithm.FLC MPPT algorithm also caused power losses around6.28% from the overall maximum power [31]. Besides that,there were some previous research works on improving theconventional FLC for MPPT. Among reported works wererelated to additional of more variables in the membershipfunction [48], use of three or more inputs to the fuzzy logic

Fig. 4 Comparison flowchart between P&O and adaptive P&O-fuzzy M


[49], employing of fuzzy cognitive networks to enhance theMPPT performance [50] and use of a single input of FLCby applying the ‘signed distance method’ to changeconventional inputs such as error and change of error [51].

3.3 Adaptive P&O–FLC

This method uses existing inputs of P&O algorithm such asdifferential power and differential voltage. The proposedchange in this algorithm is to replace comparing andswitching methods with fuzzy logic approach, as shown inFig. 4. Fig. 5 shows implementation of the algorithm inMATLAB-Simulink and furthered can be directlyprogrammed, compiled and loaded to the targeted board forhardware implementation. Same four categories of FLCwith new inputs are used. Five proposed fuzzy sets forinput/output variables are N (negative), ZE (zero), PS(positive small), P (positive) and PB (positive big). Fig. 6shows its membership functions for input of differencebetween the current power and the previous power ΔP,input of difference between the current voltage and theprevious voltage ΔV and output of difference between thecurrent duty cycle and the previous duty cycle ΔD.After ΔP and ΔV are calculated, they are converted into

linguistic variables and then the output ΔD is generated bylooking up in a rule-base table as shown in Table 1, whichconsists of 25 rules. This algorithm uses Mamdani fuzzy

PPT


Fig. 5 Implementation of adaptive P&O-fuzzy MPPT

Fig. 6 Membership functions for

a Input of ΔP and ΔVb output of ΔD

Table 2 Parameters of Kyocera KD210GH-2PU

www.ietdl.org

inference system to determine the output. For thedefuzzification, the common COA is used.In its operation, when ΔP and ΔV are either positive or

negative values, the FLC will decide the best and accuratevalue of ΔD. Therefore ΔD will be changed according tothe condition of ΔP and ΔV at certain time wheremeasurement process is done. It will obey the membershipfunction designed in the proposed FLC.

4 Simulation results

The proposed PV module was connected to boost dc–dcconverter to form a unit of PV system. The PV module ofKyocera KD210GH-2PU was selected which would beused in the hardware implementation later. It is apolycrystalline silicon type that produces 210 W at 1000 W/m2 and its parameters are shown in Table 2. It wasaccurately modelled in MATLAB-Simulink and its I/V andP/V curves are shown in Fig. 7. By using this PV module,simulation works were carried out under steady state anddynamic conditions with proposed adaptive P&O-fuzzyalgorithm, and further with conventional P&O and FLCalgorithms, respectively, for evaluation and comparisonanalysis. The input of dc–dc converter was 26 V, the output

Table 1 Rule-base for adaptive P&O-fuzzy MPPT

ΔP\ΔV N ZE PS PS PB

N ZE PS P PB PBZE ZE ZE PS P PBPS N ZE ZE PS PP N N ZE ZE PSPB N N N ZE ZE


was 34 V and the duty cycle of PWM was about 24%. Theinductor value was 120 µH, the input capacitor was 250 µF,the output capacitor was 470 µF and the load was 50 Ω.The main importance factors used to analyse performance

of each MPPT algorithm are maximum power ratio, timeresponse, oscillation, overshoot and stability. The maximumpower ratio is determined from ratio of the output obtainedfrom simulation work to expected output obtained in thedatasheet of the selected PV panel which may differ fromvarious irradiances. The simulation work also consideredthe noise from the current and voltage sensors.

4.1 Steady-state simulation

Steady-state test was done for various irradiances such as150, 200, 400 and 600 W/m2. Fig. 8 shows effect of eachMPPT algorithm towards the MPP, followed by detailedanalysis results as described in Table 3. Fig. 9a showseffect of each algorithm towards the boost voltage,followed by detailed analysis results as described inTable 4. Fig. 9b shows the time response of each MPPT

Item Value

maximum power (Pmpp) 210 Wmaximum system voltage (Vmpp) 26.6 Vmaximum system current (Impp) 7.9 Ashort-circuit current (Isc) 8.58 Aopen-circuit voltage (Voc) 33.2 Vtemperature coefficient of open-circuit voltage − 0.36%/Ktemperature coefficient of short-circuit current 0.06%/Ktemperature coefficient of maximum power − 0.46%/Knormal operating cell temperature 25°C


Fig. 7 PV module of Kyocera KD210GH-2PU as modelled in MATLAB-Simulink

a I/V curveb P/V curve

Fig. 8 Simulation result of maximum power point for each MPPT algorithm

www.ietdl.org



Table 3 Comparison results for maximum power point

MPPT algorithm MPP ratio,%

Oscillation,ms

Stability

P&O 85–90 120 not thatstable

FLC 90–95 45 stableadaptiveP&O-fuzzy

98–100 20 more stable

Table 4 Comparison results for converter’s voltages

MPPT method Overshoot,V

Oscillation,ms

Stability

P&O 15 120 not thatstable

FLC 0.05 45 stableadaptiveP&O-fuzzy

0.05 20 more stable

www.ietdl.org

algorithm. All algorithms could extract MPP of the PVmodule. The critical situation was when the irradiation wasabout 0–400 W/m2. In specific situation, the conventionalP&O did not work well with low irradiation and causedpower losses as shown in Fig. 8. From Table 3, it alsocontributes to the slowest time response, high oscillationand not that stable. By comparing it with FLC as shown inFig. 8 and listed in Table 3, the FLC shows that it is muchbetter as compared with P&O. The FLC also shows agood time response, good efficiency, low oscillation andstable. However, the proposed algorithm, the adaptiveP&O-fuzzy really shows the best performance as comparedwith both conventional MPPT algorithms. The efficiency ishigh, time response is fast, really low oscillation exists andoperation is more stable.Despite effect towards MPP, the algorithms should also

affect the boost dc–dc converter. From Fig. 8 and Table 4,the P&O algorithm produces high overshoot and oscillation.The conventional FLC performs much better with lowovershoot and oscillation. The proposed adaptiveP&O-fuzzy algorithm shows the best performance again in

Fig. 9 Simulation results of

a dc voltage outputb Time response for each MPPT algorithm


the existing of noise, its overshoot is same as theconventional FLC and oscillation of the voltage is low.Fig. 9a clearly shows difference between adaptiveP&O-fuzzy and both conventional MPPT algorithmstowards oscillation, overshoot, time response and stability.The proposed adaptive P&O-fuzzy successfully performswell with existing of noise as compared to the other twoMPPT algorithms with low overshoot, fast response, lowoscillation and more stable operation.Fig. 9b shows time responses, overshoots and oscillations

when the all MPPT algorithms start to work and boost upthe voltage output from 26 to 34 V. The main differencesbetween these three algorithms are effects on time responseand overshoot. For P&O algorithm, the time response isabout 30 ms before reaching to stable state, overshoot isabout 8 V and high oscillation exists. For FLC algorithm,the time response is about 25 ms before reaching stablestate, overshoot is about 3 V and high oscillation isobserved. The best result is from adaptive P&O-fuzzyalgorithm that achieves time response at only 15 ms beforereaching stable state, overshoot is about 1 V and only low


www.ietdl.org
oscillation exists. Therefore, from the simulation results,performance of adaptive P&O-fuzzy algorithm is muchbetter as compared to both conventional MPPT algorithmsin terms of time response, overshoot, maximum powerratio, oscillation and stability.
4.2 Dynamic simulation

Simulation work under dynamic operation was carried outaccording to the test conditions addressed in EuropeanStandard EN 50530 [41]. In this part, the MPP ratio undervarying irradiance conditions could be characterised byusing a ramp sequence (slope) consisting of irradiance rampwith different gradients as well as irradiance levels [51–55].Three tests were carried out which covered low-mediumirradiance, medium-high irradiance and startup and

Fig. 10 Simulation results of dynamic MPPT performances for

a Low-medium irradianceb Medium-high irradiancec Startup and shutdown test


shutdown for low irradiance. The low-medium irradiance isbetween 100 and 500 W/m2 with slope range of 0.5–50 W/m2/s, the medium-high irradiance is between 300 and 1000W/m2 with slope range of 10–100 W/m2/s, the startup andshutdown for low irradiance is between 10 and 100 W/m2

with slope of 0.1 W/m2/s.Fig. 10 shows dynamic MPPT performance under all

three tests and detailed measurement results are shown inTables 5–7. At low-medium irradiance, Fig. 10a showsclear result that confirms adaptive P&O-fuzzy producesbetter performance in terms of dynamic MPP ratio,stability and time response. P&O shows the worstperformance especially at the low irradiance with lowdynamic MPP ratio, high overdamped behaviour and notthat stable. FLC is much better as compared with P&O.For medium-high irradiance as shown in Fig 10b, at high


Table 7 Dynamic MPP ratio for startup and shutdown test forlow irradiance

Slope, W/m2/s 0.1 W/m2/s, %

P&O 86.9FLC 92.3adaptive P&O-fuzzy 99.8

Table 5 Dynamic MPP ratio for low-medium irradiance

Slope, W/m2/s 1 W/m2/s, % 7 W/m2/s, % 50 W/m2/s, %

P&O 98.5 98.8 99.4FLC 98.8 99.3 99.8sdaptive P&O-fuzzy 99.5 99.8 100

Table 6 Dynamic MPP ratio for medium-high irradiance

Slope, W/m2/s 20 W/m2/s,%

50 W/m2/s,%

100 W/m2/s,%

P&O 99.4 99.5 99.8FLC 99.6 99.7 99.94adaptiveP&O-fuzzy

99.98 100 100

Fig. 11 MPPT of each algorithm for

a 200 W/m2

b 1000 W/m2

www.ietdl.org

irradiance condition, generally all three MPPT algorithmsperform well and produce high dynamic MPP ratio, goodstability and fast time response. In more specific, theadaptive P&O-fuzzy still shows the best performanceespecially in term of dynamic MPP ratio. In the startupand shutdown test for low irradiance as shown inFig 10c, adaptive P&O-fuzzy works well at low irradianceas compared to the other two MPPT algorithms. P&Oshows the worst dynamic MPP ratio, slow time responseand not stable at low irradiance condition. Highoverdamped behaviour is observed for P&O which maycause power losses. FLC also produces low dynamic MPPratio, slow time response, not that stable and lowoverdamped behaviour as compared with adaptiveP&O-fuzzy but still is much better than P&O algorithm.From all tests in dynamic conditions, adaptive P&O-fuzzyshows the best performance in term of dynamic MPPratio, stability and time response including at lowirradiance condition. P&O has high dynamic MPP ratio atmedium to high irradiance, but at low irradiance, lowratio is observed. Meanwhile, FLC is much better ascompared with P&O, but not as good as adaptiveP&O-fuzzy MPPT algorithm. The big difference at lowirradiance confirms the proposed MPPT algorithm workswell at low irradiance.

Table 8 Maximum power ratios for difference irradiances

MPPT method 200 W/m2,%

600 W/m2,%

1000 W/m2,%

P&O 80 75 86FLC 85 92 90adaptiveP&O-fuzzy

91 94 95

5 Hardware implementation

Laboratory prototype was developed to evaluate the proposedalgorithm. The dc–dc boost converter was constructed assame as already modelled in MATLAB/Simulink. Digitalsignal processing TMS320F28335 board was used toperform the MPPT algorithm. The prototype was testedwith real PV panel of Kyocera KD210GH-2PU. In the realPV system, there is noise occurs from the sensors used.Noise factor cannot be neglected because it can affect


accuracy of certain MPPT algorithms, especially when thesealgorithms are implemented in real system.Fig. 11 shows comparison results of all three MPPT

algorithms to achieve MPPs at low irradiance and highirradiance. Performance in low irradiance is critical where atthis level the power is very low. Both conventional MPPTalgorithms show low ability to extract MPPs as comparedto the adaptive P&O-fuzzy algorithm. Table 8 showsmaximum power ratios with difference irradiances.At 1000 W/m2, the maximum power ratio for each MPPTalgorithm is higher because of the power is much higher atthat time and all three MPPT algorithms are observed towork well with higher irradiance.The adaptive P&O-fuzzy and FLC algorithms have stable

output waveforms as compared with P&O algorithm. Highoscillation is observed from P&O algorithm and thisoscillation may contribute to power losses. Fig. 12 shows


Fig. 12 Time responses of boost voltage for

a Adaptive P&O-fuzzyb P&Oc FLC

www.ietdl.org

time responses when all three MPPT algorithms start tooperate dc–dc converter to boost up the voltage output from26 to 34 V. The fastest response is achieved by the adaptiveP&O-fuzzy algorithm in about 20 ms, followed by theconventional FLC algorithm which is around 35 ms, andthe P&O algorithm is the worst performance with about 40ms. The obtained experimental results are almost same asthe results obtained from simulation works.However, the experimental results show lower

performance as compared to the simulation results. Thesimulation work was carried out with programmableapproaches where the PV module was modelled in theMATLAB-Simulink which included the test conditionsmentioned in EN 50350, whereas the hardwareimplementation involved directly field test which could notbe programmed. Furthermore, the laboratory prototypemay contribute to small losses especially at boost dc–dcconverter whereas in the simulation work, it was an idealsystem. However, as shown consistently with thesimulation results, from all three MPPT algorithms, theadaptive P&O-fuzzy managed to show the bestperformance in the actual field operation.


6 Conclusions

This paper has presented a new MPPT algorithm withcombination of P&O and FLC algorithms. This algorithmuses simple features available in both P&O and FLCwhich contribute to reduction of complexity in itsoperation without compromising high performance target.Wide range of irradiance level has been considered whichcontribute to uniqueness of this work especially duringoperation at low irradiance. Analysis in steady-stateoperation has widely been used before, and throughadditional analysis with dynamic operation, morecomprehensive results and findings could be obtained forfurther assessment.The proposed algorithm has been demonstrated, and

comparative evaluation has been carried out withconventional P&O and FLC in order to obtain results whichverify its better performance. Steady state and dynamicsimulation works confirm the best performance of adaptiveP&O-fuzzy MPPT algorithm to achieve high MPP ratiowith low oscillation and overshoot, which contributes tohigh stability operation. Significant difference has been


www.ietdl.org
observed at low irradiance level which can appropriately behandled by this algorithm.Hardware implementation should confirm effectiveness of
this algorithm by using real PV module with possibleexisting of noise especially in disturbing measurementprocess. Although limitation of this work is unavailabilityof real PV simulator, the real PV module should be able totest the proposed algorithm in real field operation. Forfuture work, the PV simulator which follows standard ofEN 50530 can be used to further confirm the bestperformance of the proposed algorithm where directcomparison could later be analysed between simulation andexperimental works. In this existing work on hardwareimplementation, high different range between low and highirradiance levels provided good justification to confirm thebest performance of the proposed algorithm. Achievementin fast time response of boost voltage gives someindications of significant reduction in calculation processdone by the proposed algorithm at its inputs and accuratedecision made to achieve its MPP. Therefore the proposedalgorithm has future prospect with further evaluation couldbe carried at higher level rating of PV system.

7 Acknowledgment

This work was supported by the Research University GrantScheme of Universiti Putra Malaysia under05-02-12-1919RU, and Ministry of Higher Education ofMalaysia Higher Institution Centre of Excellence and HighImpact Research Grant under H-16001-D0032.

8 References

1 Faranda, R., Leva, S.: ‘Energy comparison of MPPT techniques for PVsystems’, WSEAS Trans. Power Syst., 2008, 3, pp. 446–455

2 Ko, S.-H., Chao, R.-M.: ‘Photovoltaic dynamic MPPT on a movingvehicle’, Sol. Energy, 2012, 86, pp. 1750–1760

3 Chen, W., Shen, H., Shu, B., Qin, H., Deng, T.: ‘Evaluation ofperformance of MPPT devices in PV systems with storage batteries’,Renew. Energy, 2007, 32, pp. 1611–1622

4 Andrejasic, M., Jankovec, M.: ‘Topic. Comparison of direct maximumpower point tracking algorithms using EN 50530 dynamic testprocedure’, IET Renew. Power Gener., 2011, 5, (4), pp. 281–286

5 Long, X., Liao, R., Zhou, J.: ‘Low-cost charge collector of photovoltaicpower conditioning system based dynamic DC/DC topology’, IETRenew. Power Gener., 2011, 5, (2), pp. 167–174

6 Houssamo, I., Locment, F., Sechilariu, M.: ‘Maximum power trackingfor photovoltaic power system: development and experimentalcomparison of two algorithms’, Renew. Energy, 2010, 35,pp. 2381–2387

7 Jiang, J., Huang, T., Hsiao, Y., Chen, C.: ‘Maximum power tracking forphotovoltaic power systems’, Tamkang J. Sci. Eng., 2005, 8,pp. 147–153

8 Mellit, A., Rezzouk, H., Messai, A., Medjahed, B.: ‘FPGA-based realtime implementation of MPPT-controller for photovoltaic systems’,Renew. Energy, 2011, 36, pp. 1652–1661

9 Tafticht, T., Agbossou, K., Doumbia, M.L., Cheriti, A.: ‘An improvedmaximum power point tracking method for photovoltaic systems’,Renew. Energy, 2008, 33, pp. 1508–1516

10 Alqarni, M., Darwish, M.K.: ‘Maximum power point tracking forphotovoltaic system: modified perturb and observe algorithm’. Proc.47th Int. Universities Power Engineering Conf. (UPEC), 2012, pp. 1–4

11 Femia, N., Petrone, G., Spagnuolo, G., Vitelli, M.: ‘Increasing theefficiency of P&O MPPT by converter dynamic matching’. Int. Symp.on Industrial Electronics, 2004, vol. 2, pp. 1017–1021

12 Xiao, W., Elnosh, A., Khadkikar, V., Zeineldin, H.: ‘Overview ofmaximum power point tracking technologies for photovoltaic powersystems’. Proc. 37th Annual Conf. on IEEE Industrial ElectronicsSociety (IECON), 2011, pp. 3900–3905

13 Yu a, G.J., Jung, Y.S., Choi, J.Y., Kim, G.S.: ‘A novel two-mode MPPTcontrol algorithm based on comparative study of existing algorithms’,Sol. Energy, 2004, 76, pp. 455–463


14 Ishaque, K., Salam, Z.: ‘A review of maximum power point trackingtechniques of PV system for uniform insulation and partial shadingcondition’, Renew. Sustain. Energy Rev., 2013, 19, pp. 475–488

15 Younis, M.A., Khatib, T., Najeeb, M., Mohd Ariffin, A.: ‘An improvedmaximum power point tracking controller for PV systems using artificialneural network’, Prz. Elektrotech. (Electr. Rev.), 2012, 3, pp. 116–121

16 Atrash, H., Rustom, K.: ‘Statistical modeling of DSP based hill-climbingmppt algorithms in noisy environments’. Proc. 20th Annual IEEEApplied Power Electronics Conf. and Exposition, 2005, vol. 3,pp. 1773–1777

17 Wenkai, W., Pongratananukul, N., Weihong, Q., Rustom, K., Kasparis,T., Batarseh, I.: ‘DSP-based multiple peak power tracking forexpandable power system’. Proc. 18th Annual IEEE Applied PowerElectronics Conf. and Exposition (APEC), 2003, vol. 1, pp. 525–530

18 Hussein, K.H., Muta, I., Hoshino, T., Osakada, M.: ‘Maximumphotovoltaic power tracking: an algorithm for rapidly changingatmospheric conditions’, IEE Proc., Gener. Transm. Distrib., 1995,142, (1), pp. 59–64

19 Wu, Y., Zhang, B., Lu, J., Du, K.L.: ‘Fuzzy logic and neuro-fuzzysystem: a systematic introduction’, Int. J. Artif. Intell. Expert Syst.(IJAE), 2011, 2, pp. 47–80

20 Ammasai Gounden, N., Peter, S.A., Nallandula, H., Krithiga, S.: ‘Fuzzylogic controller with MPPT using line-commutated inverter forthree-phase grid-connected photovoltaic systems’, Renew. Energy,2009, 34, pp. 909–915

21 Khaehintung, N., Sirisuk, P., Kurutach, W.: ‘A novel ANFIS controllerfor maximum power point tracking in photovoltaic systems’. Proc. FifthInt. Conf. on Power Electronics and Drive Systems (PEDS 2003), 2003,vol. 2, pp. 833–836

22 Iqbal, A., Abu-Rub, H., Ahmed, Sk.M.: ‘Adaptive neuro-fuzzyinference system based maximum power point tracking of a solar PVmodule’. IEEE Int. Energy Conf., 2010, pp. 51–56

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

24 Kassem, A.M.: ‘MPPT control design and performance improvementsof a PV generator powered DC motor-pump system based on artificialneural networks’, Electr. Power Energy Syst., 2012, 43, pp. 90–98

25 Algazar, M.M., AL-monier, H., Abd EL-halim, H., El Kotb Salem, M.E.: ‘Maximum power point tracking using fuzzy logic control’, Electr.Power Energy Syst., 2012, 39, pp. 21–28

26 Ramaprabha, R., Balaji, M., Mathur, B.L.: ‘Maximum power pointtracking of partially shaded solar PV system using modified Fibonaccisearch method with fuzzy controller’, Electr. Power Energy Syst.,2012, 43, pp. 754–765

27 Syafaruddin, Karatepe, E., Hiyama, T.: ‘Artificial neural network-polarcoordinated fuzzy controller based maximum power point trackingcontrol under partially shaded conditions’, IET Renew. Power Gener.,2009, 3, (2), pp. 239–253

28 Messai, A., Mellit, A., Guessoum, A., Kalogirou, S.A.: ‘Maximumpower point tracking using a GA optimized fuzzy logic controller andits FPGA implementation’, Sol. Energy, 2011, 85, pp. 265–277

29 Chao, P.C.P., Chen, W.D., Chang, C.K.: ‘Maximum power tracking of ageneric photovoltaic system via a fuzzy controller and a two-stageDC–DC converter’, Microsyst. Technol., 2012, 18, pp. 1267–1281

30 Al Nabulsi, A., Dhaouadi, R.: ‘Fuzzy logic controller based perturb andobserve maximum power point tracking’. Int. Conf. on RenewableEnergies and Power Quality, 2012, pp. 1–6

31 Kottas, T.L., Boutalis, Y.S., Karlis, A.D.: ‘New maximum power pointtracker for PV arrays using fuzzy controller in close cooperation withfuzzy cognitive networks’, IEEE Trans. Energy Convers., 2006, 21,(3), pp. 793–803

32 Bennett, T., Zilouchian, A., Messenger, R.: ‘Photovoltaic model andconverter topology considerations for MPPT purposes’, Sol. Energy,2012, 86, pp. 2029–2040

33 Bouchafaa, F., Beriber, D., Boucherit, M.S.: ‘Modeling and simulationof a gird connected PV generation system with MPPT fuzzy logiccontrol’. Proc. Seventh Int. Multi-Conf. on Systems, Signals andDevices, 2010, pp. 1–7

34 Harjai, A., Bhardwaj, A., Sandhibigraha, M.: ‘Study of maximum powerpoint tracking (MPPT) techniques in a solar photovoltaic array’. Thesis,Department of Electrical Engineering National Institute of Technology,Rourkela, Orissa, 2011, pp. 43–49

35 Ngan, M.S., Tan, C.W.: ‘A study of maximum power point trackingalgorithms for stand-alone photovoltaic systems’. IEEE Applied PowerElectronics Colloquium (IAPEC), 2011, pp. 22–27

36 Houssamo, I., Locment, F., Sechilariu, M.: ‘Experimental analysis ofimpact of MPPT methods on energy efficiency for photovoltaic powersystems’, Int. J. Electr. Power Energy Syst., 2013, 46, pp. 98–107


www.ietdl.org
37 Hua, C., Shen, C.: ‘Study of maximum power tracking techniques and
control off DC/DC converters for photovoltaic power system’. Proc.29th Annual IEEE Power Electronics Specialists Conf. (PESC), 1998,vol. 1, pp. 86–93

38 Surya Kumari, J., Sai Babu, C.h., Kamalakar Babu, A.: ‘Design andanalysis of P&O and IP&O MPPT techniques for photovoltaicsystem’, Int. J. Modern Eng. Res. (IJMER), 2012, 2, (4), pp. 2174–2180

39 Tafticht, T., Agbossou, K., Doumbia, M.L., Chériti, A.: ‘An improvedmaximum power point tracking method for photovoltaic systems’,Renew. Energy, 2008, 33, pp. 1508–1516

40 Hohm, D.P., Ropp, M.: ‘Comparative study of maximum power pointtracking algorithms’, Prog. Photovolt., Res. Appl., 2003, 11, pp. 47–62

41 European Standard: ‘Overall efficiency of grid connected photovoltaicinverters EN 50530’. European Committee for Electro technicalStandardization, 2010, pp. 1–36

42 Valentini, M., Raducu, A., Sera, D., Teodorescu, R.: ‘PV inverter testsetup for european efficiency, static and dynamic MPPT efficiencyevaluation’. Proc. 11th Int. Conf. on Optimization of Electrical andElectronic Equipment (OPTIM), 2008, pp. 433–438

43 Villalva, M.G., Gazoli, J.R., Filho, E.R.: ‘Analysis and simulation of theP&O algorithm using a linearized PV Array Model’. IndustrialElectronics Conf., 2009, pp. 231–236

44 Hashimoto, O., Shimizu, T., Kimura, G.: ‘A novel high performanceutility interactive photovoltaic inverter system’. Conf. Record of the2000 IEEE on Industry Applications Conf., 2000, vol. 2254,pp. 2255–2260

45 Ying-Tung, H., China-Hong, C.: ‘Maximum power tracking forphotovoltaic power system’. Proc. 37th IAS Annual Meeting onIndustry Applications Conf., 2002, vol. 1032, pp. 1035–1040

46 Femia, N., Petrone, G., Spagnuolo, G., Vitelli, M.: ‘Optimization ofperturb and observe maximum power point tracking method’, IEEETrans. Power Electron., 2005, 20, pp. 963–73


47 Aït Cheikh, M.S., Larbes, C., Tchoketch Kebir, G.F., Zerguerras, A.:‘Maximum power point tracking using a fuzzy logic control scheme’,Revue des Energies Renouvelables, 2007, 10, pp. 387–395

48 Mahmoud, A.M.A., Mashaly, H.M., Kandil, S.A., El Khashab, H.,Nashed, M.N.F.: ‘Fuzzy logic implementation for photovoltaicmaximum power tracking’. Proc. 26th Annual Conf. IEEE onIndustrial Electronics Society (IECON), 2000, vol. 731, pp. 735–740

49 Masoum, M.A.S., Design, S.M.: ‘Simulation and implementation of afuzzy-based maximum power point tracker under variable insulationand temperature conditions’, Iranian J. Sci. Technol., 2005, 29, (6),pp. 1–6

50 Kottas, T.L., Boutalis, Y.S., Karlis, A.D.: ‘New maximum power pointtracker for PV arrays using fuzzy controller in close cooperation withfuzzy cognitive networks’, IEEE Trans. Energy Convers., 2006, 21,pp. 793–803

51 Byung-Jae, C., Seong-Woo, K., Byung Kook, K.: ‘Design and stabilityanalysis of single-input fuzzy logic controller, systems, man, andcybernetics, Part B: cybernetics’, IEEE Trans., 2000, 30, pp. 303–309

52 Bründlinger, R., Henze, N., Häberlin, H., Burger, B., Bergmann, A.,Baumgartner, F.: ‘prEn 50530-The new European standard forperformance characterisation of PV Inverters’. Proc. 24th EuropeanPhotovoltaic Solar Energy Conf., 2009, pp. 3105–3109

53 Yu, B., Yu, G., Kim, Y.: ‘Design and experimental results of improveddynamic MPPT performance by EN5053’. IEEE 33rd Int.Telecommunications Energy Conf. (INTELEC), 2011, pp. 1–4

54 Andrejasic, T., Jankovec, M., Topic, M.: ‘Comparison of directmaximum power point tracking algorithms using EN 50530 dynamictest procedure’, IET Renew. Power Gener., 2011, 5, (4), pp. 281–286

55 D’Souza, N.S., Lopes, L.A.C., Xuejun, L.: ‘An intelligent maximumpower point tracker using peak current control’. IEEE 36th PowerElectronics Specialists Conf. (PESC), 2005, pp. 172


Documents

Development of adaptive perturb and observe-fuzzy control maximum power point tracking for photovoltaic boost dc–dc converter