Maximum Power Point Tracking Control for Photovoltaic System Using Adaptive Neuro- Fuzzy ANFIS

Transcription

1 Author manuscrpt, publshed n " Eghth Internatonal Conference and Exhbton on Ecologcal Vehcles and Renewable Energes (EVER), Monaco : Monaco ()" Maxmum Power Pont Trackng Control for Photovoltac System Usng Adaptve Neuro- Fuzzy Boutabba Tarek, Drd sad and M.E.H. Benbouzd Unversty of Khanchela Laboratore LSPIE, L.S.P.IE Research Laboratory; Unversty of Batna, Algera. Unversty of Brest, EA 45 LBMS Rue de Kergoat - CS 987, 98 Brest, France Emal: boutabba_tarek@yahoo.fr hal-874, verson - 7 Oct Abstract Due to scarcty of fossl fuel and ncreasng demand of power supply, we are forced to utlze the renewable energy resources. Consderng easy avalablty and vast potental, world has turned to solar photovoltac energy to meet out ts ever ncreasng energy demand. The mathematcal modelng and smulaton of the photovoltac system s mplemented n the MATLAB/Smulnk envronment and the same thng s tested and valdated usng Artfcal Intellgent (AI) lke. Ths paper presents Maxmum Power Pont Trackng Control for Photovoltac System Usng Adaptve Neuro- Fuzzy. The PV array has an optmum operatng pont to generate maxmum power at some partcular pont called maxmum power pont (MPP). To track ths maxmum power pont and to draw maxmum power from PV arrays, MPPT controller s requred n a stand-alone PV system. Due to the nonlnearty n the output characterstcs of PV array, t s very much essental to track the MPPT of the PV array for varyng maxmum power pont due to the nsolaton varaton. In order to track the MPPT conventonal controller lke Adaptve Neuro-Fuzzy and fuzzy logc controller s proposed and smulated. The output of the controller, pulse generated from PWM can swtch MOSFET to change the duty cycle of boost DC-DC converter. The result reveals that the maxmum power pont s tracked satsfactorly for varyng nsolaton condton. Keywords Photovoltac; Pulse Wdth Modulaton; Proportonal Integral Controller; boost DC-DC;;fuzzy logc;mppt. I. INTRODUCTION Today photovoltac (PV) systems are becomng more and more popular wth ncrease of energy demand and there s also a great envronmental polluton around the world due to fossls and oxdes. Solar energy whch s free and abundant n most parts of world has proven to be economcal source of energy n many applcatons []. The energy that the earth receves from the sun s so enormous and so lastng that the total energy consumed annually by the entre world s suppled n as short a tme as half an hour. The sun s a clean and renewable energy source, whch produces nether green house effect gas nor toxc waste through ts utlzaton. It can wthstand severe weather condtons, ncludng cloudy weather. The watt peak prce s decreased snce the seventes, ths leads to large scale promsng areas. It does not have any movng parts and no materals consumed or emtted. Unfortunately,[] ths system has two major dsadvantages, whch the low converson effcency of electrc power generaton (9 to 6%), especally under low rradaton condtons and the amount of electrc power generated by solar array changes contnuously wth the weather condtons lke rradaton and temperature. To overcome ths problem, maxmum power pont trackng (MPPT) technque wll be used. Fuzzy logc control and adaptve neuro-fuzzy control. The trackng algorthm ntegrated wth a solar PV system has been smulated wth boost DC-DC converter n stand - alone PV system. The proposed PV system wth boost DC-DC converter s shown n Fg.. The gven model operates very fast n comparson on wth avalable methods and has proper accuracy n maxmum power pont trackng (MPPT). Fg.. Photovoltac module wth DC-DC boost converter. II. MATHEMATICAL MODELING OF PHOTOVOLTAIC CELL A PV cell can be represented by an equvalent crcut [9] as shown n Fg.. The characterstcs of ths PV cell can be obtaned usng standard equaton (). VRs I [ Va t ] V Rs I I Ipv I e () R p Fg. Equvalent crcut of PV cell

2 hal-874, verson - 7 Oct I PV = photovoltac current I O = saturaton current V t = N S k T/q, thermal voltage of array N s = cell connected n seres T = s the temperature of the p-n juncton k = Boltzmann constant q = electron charge R S = equvalent seres resstance of the array R P = equvalent parallel resstance of the array a = dode dealty constant Fg. shows the sngle dode model. A sngle solar cell wll produce only a lmted power. Therefore t s usual practce n order to get desred power ratng the solar cells are connected n parallel and seres crcuts whch form a module. Such modules are agan connected n parallel and seres to form a solar array or panel to get requred voltage and current. The equvalent seres and parallel resstance of the array are denoted by the symbol R S and R P respectvely n the equvalent crcut. From the general I-V characterstc of the practcal photovoltac devce one can observe that the seres resstance R S value wll domnate n the voltage source regon and the parallel resstance R P value wll domnate n the current source regon of operaton. The general equaton of a PV cell descrbes the relatonshp between current and voltage of the cell. Snce the value of shunt resstance R P s hgh compared to value of seres resstance R S the current through the parallel resstance can be neglected. The lght generated current of the photovoltac cell depends lnearly on the solar rradaton and s also nfluenced by the temperature [] gven by the equaton () G Ipv [ I pv, n KI T ] () Gn I PV = s the lght generated current at nomnal condton (5 C and W/ m) ΔT = T Tn T = actual temperature [K] T n = nomnal temperature [K] K I = current coeffcents G = rradaton on the devce surface [W/m] G n = nomnal rradaton The current and voltage coeffcents K V and K I are ncluded as shown n equaton () n order to take the saturaton current I O whch s strongly dependent on the temperature. Isc, n KIT I () o Voc, n KV T av t e K V = voltage coeffcents K I = current coeffcents Fg. (a) and Fg. (b) gve the power voltage (P-V) characterstcs of a PV module for dfferent values of solar radaton and temperature. The short crcut current s clearly nverse dependence to the temperature; an ncrease n temperature causes a reducton of the open-crcut voltage (when suffcently hgh) and also more maxmum output power. Pussance Ppv (V ) (a) (b) Fg.. Smulated waveforms showng the effect of (a) Irradaton and (b) temperature on P-V characterstcs. The manufacturer s data at standard condtons are gven as Pmax = 8W, Imax = 4.55 A and Vmax =.6V. The smulaton results obtaned were: Pmax = 78.5W, Imax = 4.5 A and Vmax = 8. V. [4-8] Smulated I-V, P-V characterstcs for the maxmum power pont trackng (MPPT) s shown n Fg.4. At ths Maxmum Power Pont (MPP), the solar array s matched to ts load and when operated at ths pont the array wll yeld the maxmum power output. From Fg. 4 (a) & (b), t s observed that the power output has an almost lnear relatonshp wth array voltage unt, hence the MPP s attaned. Any further ncrease n voltage results n power reducton. [5] Courant Ipv (A) MPP Tenson Pussance Ppv (V) Tenson (a) (b) Fg.4. PV array smulated curves I-V curve (5 C) and (b) P-V curve (w/m). III. DC DC BOOST CONVERTER A dual stage power electronc system comprsng a boost type dc-dc converter and an nverter s used to feed the power generated by the PV array to the load. To mantan the load voltage constant a DC-DC step up converter s ntroduced between the PV array and the nverter. The block schematc of the proposed scheme s shown n Fg.5. Fg. 5 A MPPT Controllers n a PV System MPP

3 hal-874, verson - 7 Oct In ths scheme a PV array feeds DC-DC converter used n stepup confguraton.. For a dc-dc boost converter, by usng the averagng concept, the nput output voltage relatonshp for contnuous conducton mode s gven by Vo (4) Vn D Where, D = duty cycle. Snce the duty rato D s between and the output voltage must be hgher than the nput voltage n magntude. [-] IV. A. Adaptve neuro-fuzzy prncple ADAPTIVE NEURO-FUZZY MPPT CONTROLLER A typcal archtecture of an s shown n Fg. 6, n whch a crcle ndcates a fxed node, whereas a square ndcates an adaptve node. For smplcty, we consder two nputs x, y and one output z. Among many FIS models, the Sugeno fuzzy model s the most wdely appled one for ts hgh nterpretablty and computatonal effcency, and bult-n optmal and adaptve technques. For a frst order Sugeno fuzzy model, a common rule set wth two fuzzy f then rules can be expressed as: Rule : f x s A and y s B ; then z px qy r Rule : f x s A and y s B ; then z px q y r Where A and B are the fuzzy sets n the antecedent, and p, q and r are the desgn parameters that are determned durng the tranng process. As n Fg. 6, the conssts of fve layers: Layer Layer Layer Layer 4 Layer 5 Fg. 6 Archtecture of. Layer : Every node n the frst layer employ a node functon gven by: O A ( x), =, (5) O B ( x), =,4 Where A and B can adopt any fuzzy membershp functon (MF) [5]. Layer : Every node n ths layer calculates the frng strength of a rule va multplcaton O ( x) ( x), =, (6) A B Layer : The -th node n ths layer calculates the rato of the -th rule s frng strength to the sum of all rules frng strengths: z O Where, =, s referred to as the normalzed frng strengths. (7) Layer 4: In ths layer, every node has the followng functon: 4 O z p x q y r, =, (8) Where s the output of layer, and p, q, rs the parameter set. The parameters n ths layer are referred to as the consequent parameters[4-6]. Layer 5: The sngle node n ths layer computes the overall output as the summaton of all ncomng sgnals, whch s expressed as: 5 z z O z, =, (9) The output z n Fg. can be rewrtten as [7-9]: z x p y q r x p y q r (). B. Adaptve neuro-fuzzy controller The controller generates change n the duty cycle D, based on speed error ( e ) and dervate n the speed error ( de ) defned as: e P dp de dt Where ΔP s changes n power In ths study frst order Sugeno type fuzzy nference was used for and the typcal fuzzy rule s: f e s A and de s B ; then z f( e, de) Where A and B are fuzzy sets n the antecedent and z = f (e,de) s a crsp functon n the consequent[]. The sgnfcances of structure are: Layer : Each adaptve node n ths layer generates the membershp grades for the nput vectors A, =,,5. In ths paper, the node functon s a trangular membershp functon:, e a e a, a e b b a c e O A ( e),, b e c c b, c e Layer : The total number of rule s 5 n ths layer. Each node output represents the actvaton level of a rule: O mn ( e) ( de), =,...,5. A B Layer : Fxed node n ths layer calculate the rato of the - th rule's actvaton level to the total of all actvaton level:

4 O n j j errors obtaned from the are 4*7. -6 and 5*. -6 respectvely. Input Input mf Rule Output mf Output Layer 4: Adaptve node n ths layer calculate the contrbuton of -th rule towards the overall output, wth the followng node functon: 4 O z p e q de r. Layer 5: The sngle fxed node n ths layer computes the overall output as the summaton of contrbuton from each rule: 5 z z O z, The parameters to be traned are (a, b, and c ) of the premse parameters and (p, q, and r )of the consequent parameters. Tranng algorthm requres a tranng set defned between nputs and output [ - ]. Although, the nput and output pattern set have 5 rows. Fg.7 shows optmzed membershp functon for e and de. Input Layer Layer Layer Layer Layer 4 Layer 5 Fg. 8 The graphcal representaton of system two nput and one output varable RESULTS AND DISCUSSION hal-874, verson - 7 Oct (a) (b) Membershp Functons of the st Input Varable st (E) Smulaton of neural fuzzy algorthm as MPPT on PV system s done by provdng nterference n the form of solar rradaton and temperature changes because these two varables affect PV output voltage and current. A. Solar rradaton and temperature Fxe Les fgures a, b, c, d shows the smulaton results of the PV system wth a PV Power of W wthout dsturbance at rradaton of W/m and temperature of 5 o C. The responce of the system when we use control s better than control but wth an overshootng n the dynamc response. In the both system of controls produce a maxmum power pont voltage of.7v as shown n Fg. a and Fg. b correspondng to characterstc P-V and I-V. The outputs of and regulators are connected to the boost converter, so they produce a duty cycle as shown n Fg. d. At steady state condtons output power reached the value of W. Membershp Functons of the nd Input Varable nd ( E) Fg. 7 Optmzed membershp E ΔE The number of epochs was 5 for tranng. The number of MFs for the nput varables e and de s 7 and 7, respectvely. The number of rules s then 49 (7*7 = 49).The trangular MF s used for two nput varables. It s clear from (5) that the trangular MF s specfed by two parameters. Therefore, the used here contans a total of 95 fttng parameters, of whch 8 (7* + 7* = 8) are the premse parameters and 47 (* 49 = 47) are the consequent parameters. The tranng and testng root mean square (RMS)

5 Temperature (Celcus) Solar Irradaton (W/m) 8 T (C e lc u s ) 6 4 E (W/m) Tme (s) a) Fg.9 Solar rradaton and temperature change Tme (s) e) 7 6 MPPT P V Ppv (V) 4 Ppv (V) 4 hal-874, verson - 7 Oct Ipv (V) b) c) Fg. MPPT characterstc P-V Fg. MPPT characterstc I-V Ip v (A ) f) g) Pout (W ) Pout (w) Tme (s) d) Fg. Power (W) output of the boost Tme (s) h) Duty cycle D..5 d Tme (s) Fg. Duty cyle (d) Tme (s)

6 hal-874, verson - 7 Oct B. Solar rradaton and temperature change In ths condton the output voltage transent response of buck converter shows the smulaton results by gvng solar rradaton dsturbance. At the ntal condtons for solar rradaton of W/m and changes untl 8W/m n t=5s and n t=5 ncreased to ts ntal condtons. At the ntal condtons, the PV temperature of 5 o C then decreased at a temperature 5 o C wth dfferent value of solar rradaton. From t=5s to t=s the fxed temperature 5 o C and solar rradaton decreased from W/m to 8 W/m, PV output power of W and wll decreases to low value whch causes by the decrease of the current n the output of the boost. Ths suggests that changes n solar rradaton wll lead to changes n the PV output current. The two MPPT s controller produces changes n d, as shown n Fg h. From t=s to t=5s the fxed solar rradaton 8 W/m and temperature decreased from 5 o C to 5 o C, PV output power of 88 W and wll ncreases to hgh value whch causes by the decrease of the current n the output of the boost as shown n Fg e,caused by the changng of the value of the duty as shown n Fg h. Ths suggests that a change n PV temperature s nversely proportonal to changes n the PV output current. From t=s to t=5s the fxed temperature 5 o C and solar rradaton ncreased from 8 W/m to W/m, PV output power of 98 W and wll ncreases to hgh value as shown n Fg f,,whch causes by the ncrease of the current n the output of the boost as shown n Fg g. The change the value of D s very small as well as the envronmental parameters change. Ths change of D value wll affect the converter output voltage that reaches steady at.5 V. Because the D value changes only very lttle then the converter output voltage change s also small as shown n Fg a. V. CONCLUSION Ths paper has presented neural fuzzy for controllng PV system output voltage to operate at maxmum power pont although happened temperature and rradaton changes. Applcatons of neural fuzzy controller on MPPT of PV showed a good performance. The system was analyzed and desgned, and performance was studed by smulaton wth Smulnk/Matlab. PV system can operate at maxmum power pont although occur temperature and sun rradaton change that can shft maxmum power pont. REFERENCES [] Bmal K. Bose, Global Warmng Energy: Envronmental Polluton and the Impact of Power Electroncs, IEEE Industral Electroncs Magazne, pp.-7, March. [] M. Masoum, Desgn, Constructon and Testng of a Voltage-based Maxmum Power Pont Tracker (VMPPT) for Small Satellte Power Supply, th Annual AIAA/USU Conference on Small Satellte. [] Cong L. Hou, Applcaton of Adaptve of Solar Cell Battery Charger, IEEE DRPT4. [4] Adedamola Omole, Analyss, Modellng and Smulaton of Optmal Power Trackng of Multple-Modules of Paralleled Solar Cell Systems, Master of Scence Thess, The Florda State Unversty College of Engneerng, 6. [5] V. D Do, D. La Casca, R. Mcel, A Mathematcal Model to Determne the Electrcal Energy Producton n Photovoltac Felds under Msmatch Effect, Proceedngs of the 978() IEEE, pp.46-5, August 9. [6] Geoff Walker, Evaluatng of MPPT Converter Topologes Usng a MATLAB PV Model, the Unversty of Queensland. [7] A. Saad and A. Mouss, Optmaton of Buck Boost Converter By MPPT Technque Wth A Varable Reference Voltage Appled to Photovoltac Water Pumpng System Under Varable Weather Condtons, Asan Journal of Informaton Technology 6, 7. [8] A. Mellt, An -Based Predcton for Monthly Clearness Index and Daly Solar Radaton: Applcaton for szng of a Stand-Alone Photovoltac System, Journal of Physcal Scence, Vol. 8(), pp.5 5, 7. [9] N. Patcharaprakt and S. Premrudeepreechacharn, Maxmum PowerPont Trackng Usng Adaptve Fuzzy Logc Control for Grd connected Photovoltac System, PESW, volume, PP:7-77,. [] Marcelo Gradella Vllalva, Jonas Rafael Gazol, and Ernesto Ruppert Flho, Comprehensve Approach to Modelng and Smulaton of Photovoltac Arrays, IEEE Transactons on Power Electroncs, vol 4, no. 5, 9, pp [] N. Fema, Optmzng Duty-cycle Perturbaton of P&O MPPT Technque, 4 5th Annual IEEE Power Electroncs Specalsts Conference. [] Jan, S.K.; Agrawal, P.; Gupta, H.O.; Fuzzy logc controlled shunt actve power flter for power qualty mprovement IEE Proceedngs- Electrc Power Applcatons, Volume 49, Issue5, Sept. Page(s):7-8. [] Fabo Lma, Walter Kaser, Ivan Nunes da Slvay, Azaur Albano de Olvera Junory; speed neuro-fuzzy estmator for ndrect flux orented nducton motor control XVIII Congresso Braslero de Automátca / a 6-setembro-, Bonto-MS. [4] Kalka S., L. Rajaj, Subhash Gupta Gupta, H.O.; Intellgent Technque Based Modelng for PVPS Internatonal Journal of Engneerng and Innovatve Technology (IJEIT) Volume, Issue 6, December. [5] T.R. Sumthra, A. Nrmal Kumar, R. Ramesh Kumar An adaptve neuro-fuzzy nference system () based Predcton of Solar Radaton:A Case study Journal of Appled Scences Research, 8(): 46-5, ISSN X. [6] Samra DIB, Brahm FERDI, Chellal BENACHAIBA Adaptve Neuro-Fuzzy Inference System based DVR Controller Desgn Leonardo Electronc Journal of Practces and Technologes ISSN 58-78, Issue 8, January-June.