A Hybrid (ACO-PSO) algorithm Based on Maximum Power Point Tracking and its Performance Improvement within Shadow Conditions

Transcription

1 A Hybrd (ACO-PSO) algorthm Based on Maxmum Power Pont Trackng and ts Performance Improvement wthn Shadow Condtons Ashty M. Aaref 1, Sameen F. Mohammed 1, Abbas B. NOORI 1, Al M. Humada 2 Abstract- In ths study, a hybrd Ant Colony Optmzaton-Partcle Swarm Optmzaton (ACO-PSO) algorthm was proposed to optmally determne the maxmum power pont trackng (MPPT) parameters. The man goal of current study s to mprove the overall performance of the MPPT system. The efforts of oscllaton flterng and nose suppresson were taken n ths desgn, as well as the tme response and the settlng tme. The proposed method s employed to track the global MPP under dfferent shadow condtons, based on three dfferent rradaton levels to test the ablty and accuracy of the proposed method. The results of trackng MPP by the proposed MPPT technque showed that the mproved method tracked the MPP for all the tested cases wth a reasonable accuracy and n a very short convergence tme as compared to the P&O method. 1. Introducton Photovoltac (PV) module only converts a relatvely low percentage of ts ncdent solar rradaton nto electrcal energy, whch means ts effcency s modest [1]. Man step n ncreasng the PV cell effcency s by employng one of the MPPT technques to obtan the maxmum probable power from a varyng PV sources. Normally, MPP analytcally determned from the power-volt (P-V) characterstc curve and voltage-current (I-V) characterstc curve. Each curve has a unque pont named MPP, whereby the module s operatng at ts maxmum effcency to yeld the maxmum output power, dependng on ts operatng condtons (radaton and temperature). Therefore, can be defned MPPT method as a technque that constantly tracks the MPP on the characterstc curve of P V or I V and to ensure that the PV module can operate at the optmal pont where most power s obtaned. Dfferent MPPT algorthms have been used prevously to acheve that task alongsde a DC-DC converter whose duty cycle s vared and s beng used on the load sde powered by a solar panel [2-5]. For ths purpose, there are dfferent methods used to track the MPP. These methods were delberated n prevous studes. Perturbaton and observaton (P&O) technque was used as one of the most well-known algorthms employed to track the MPP [6-8]. The smplcty and ease of mplementaton render t an attractve method to use [6]. Nonetheless, one of the major drawbacks of ths algorthm s the oscllaton around the MPP. Once the algorthm reaches a maxmum pont (local or global), t wll keep oscllatng back and forth around t, by a value called ΔV. Therefore, perturbaton

2 sze corresponds to the power oscllaton around the MPP. The Incremental Conductance (INC) method can overcome some of the drawbacks of P&O method [9, 10]. It s stll a popular method, although t s more complex than the P&O. However, ncremental conductance s slghtly harder to mplement than P&O [11].n addton, these methods stll conventonal methods and cannot dstngush between the local and global MPP. Therefore, t has become necessary to mplement MPPT technques dealng wth such cases. Partcle Swarm Optmzaton (PSO) algorthm was one of the most wdely used, snce t s a method utlzed towards predctng the PV system s MPP accurately [12, 13]. Meanwhle partally shaded I V curve exhbts multmodal characterstcs, PSO methods are envsaged and very effectve to track global MPP and dstngush t from the local ones under ths condton. Fast changng n the weather condton needs to do teraton fast, but long convergence tme s requred for PSO based MPPT, whch make t less effectve n case of rapd weather changes. Ant Colony Optmzaton (ACO) method s supposed to be one of the most effectve methods n the operaton of trackng MPP from a PV. Ths s because t combnes a postve feedback applance, dstrbuted calculatng, as well as greedy search method. In the process of searchng for the optmal soluton, t has a sturdy level of ablty lke, the postve feedback mechansm whch guarantees that the ACO s almost capable n percevng fast and optmal soluton n the early stages. By combnng the advantages of both methods, PSO and ACO technques, a desrous search, adequate results are quckly obtaned and effcent MPP results are acheved. Therefore In ths paper, a new technque based on hybrd ACO-PSO algorthm to track the global MPP for a PV system operates wth dfferent partal shadow condtons and rapd weather change has been presented. It combnes both, the smplcty n the search process and accuracy of the acqured results. To verfy ts valdty, the presented technque s then compared wth PSO based on P&O MPPT algorthm [8,19]. The mplemented method s smulated and tested usng Matlab Smulnk software. 2. Proposed method As mentoned earler, the mportance of MPPT technques dedcated n ts ablty of achevement the best value of extracted power pont from PV panels through any operaton condton. Most mplemented methods succeed n trackng and extractng the maxmum power pont. However, each method has ts obstacles and mplemented n a certan weather condton

3 for a specfc system. Also, there s no method consdered all avalable shadow condtons; lke shadow condtons wth rapd change n the weather condtons. Therefore, a new method developed n ths study to track the global MPP under both rapd and partal shadow condtons. In addton, the presented method combnes the smplcty and mprove effcency of the extracted power va dfferent weather condtons. Consequently, the proposed MPPT based on ACO-PSO and varable step sze P&O appled and tested through three dfferent shadow condtons. In all these shadow condtons, the MPPT method and optmal PID controller presented a clear enhancement for all oscllaton round MPP, the settlng tme and the response tme. The ACO s a combnaton of postve feedback mechansm, dstrbuted computng, and a greedy search algorthm. It has strong abltes to look for the optmal soluton. For nstance, the postve feedback mechansm ensures that the ant colony algorthm s able to dentfy the optmal soluton n earler stage. By usng the greedy search, the acceptable soluton s quckly found and effcency of the system s mproved. Therefore, the MPPT convergence speed problem n a PV system s solved through modfed ACO-based optmzaton. ACO-based MPPT s appled n ths study to fnd the shortest path, to ncrease convergence speed. The equaton below descrbes the ant s behavour towards fndng the shortest path to the global MPP pont, denotng the amount of pheromone toward the shortest path. Ph Ph ( 1 ) Ph xy xy Where Phxy s the amount of pheromone deposted for a state transton xy, s the pheromone evaporaton coeffcent, ΔPhxy s the amount of pheromone deposted by th ant. xy (1) Ph xy K / L f ant uses curve xy n ts tour 0 otherwse Where L s the cost of the th ant's tour (typcally length) and K s a constant. Due to aforementoned merts of usng ACO, the problem of slow MPP trackng for PV arrays could be solved va adoptng ACO based MPPT [14]. The PSO algorthm reles on populaton. It uses several possble solutons n order to establsh the best one or a combnaton of solutons to a partcular problem. Its objectve s to provde the global optmum of the ftness functon (the real valued functon) ascrbed to a partcular search space. The Pbest and Gb must be kept up-to-date at each teraton all through

4 the optmzaton process. The mathematcal form of acqurng the velocty and updatng the locaton of each partcle s shown n Eqs. (2) and (3), ndvdually. The equatons below are the bass for computng the poston and velocty of ndvduals. V c r ( n n n wvn ) Pbest X t ) c2r ( Gbest X t (2) X 1 n X n V t Where the subscrpton n represents the number of used partcles; refers to the number of teratons; V n stands for the velocty vector of the nth partcle and X n for locaton vector, at th teraton. The search operaton of MPPT wthn shadow condtons s more dffcult, due to mult peaks appearng on the P-V characterstc curve. Dfferent local ponts and only one global pont wll be generated on the P-V characterstc curve. A technque of usng Perturbaton and Observaton method (P&O) combned wth a Partcle Swarm Optmzaton method (PSO) has shown an effcent result for trackng global MPP and dfferentate t from the local ones [15, 16]. However, t has the dsadvantage of long convergence tme and falure to track global MPP rapdly, especally when the array s exposed to fast changng weather condtons. Therefore, to overcome these drawbacks, ths study proposes a new method of combnng the P&O method wth a hybrd method of ACO-PSO. Ths s because the above mentoned merts of PSO n dfferentatng the global pont and trackng t effcently, the ablty of ACO algorthms n adaptng to the rapd changes n weather condton and ts fast convergence speed [17, 18]. Ths wll lead to producng an accurate MPPT method wth fast convergence speed. In ths method, DC/DC converter (boost converter) s used to control the maxmum global pont va employng the proposed MPPT algorthm, whch adjusts the duty cycle of boost converter. The schematc dagram of nterleaved proposed MPPT controller s shown n Fgure 1. In the proposed system, to control the step sze, frstly the perturb and observe algorthm are actvated. They are used because of the ablty to track the frst local pont and changng the step sze of the search n a proper way. In the next step, the calculaton process of a hybrd ACO-PSO algorthm s carred out as ths can track the global MPP n a very fast convergence speed and effcent mode. n (3)

5 Fgure 1. Block dagram of proposed ACO-PSO based P&O MPPT and PID controller. As a matter of fact, the performance of any optmzaton technque depends on the selected objectve functon or ftness functon. The ntegral tme absolute error (ITAE) helps n shrnkng the peak overshoot. Also, the ntegral tme square error (ITSE) decreases the settlng tme, whch cannot be acheved wth the ntegral absolute error (IAE) or the ntegral square error (ISE) based-tunng [19]. Consequently, n ths research, ITAE and ITSE are used as objectve functons or cost functons to mprove the PID desgn process, as shown n Eqs. (4) and (5). Whle the whole combnaton system shown n Fgure 2. Fgure 2. PV system wth proposed MPPT dagram & optmal PID controller. ITAE ITSE T T 0 0 t e( t). dt te( t) 2. dt (4) (5) A combnaton of ACO-PSO algorthm have been used n ths study and tunng PID controller for the purpose of attanng steady state condton of extracted power n order to acheve the system stablty and relablty. The study algorthm appled on PV system from

6 four PV modules operated n two dfferent show condtons. These processes are represented n Fgure 3, whch llustrates the block dagram of PV modules system. Fgure 4. Block dagram of a PV array wth four PV modules. 3. MPPT mplementaton under Partal Shadows Under partal shadowng condtons, there are many MPPT technques ntroduced and proposed n the lterature. However, each method had ts restrctons and lmtatons. Some methods could track the MPP but t so complcated whereas other methods have an oscllaton around the MPP, but faled to track MPP wthn fast changng weather condtons. Ths s because when the PV system operates under shadow condtons the I-V and P-V characterstc curves wll generate many local maxmum ponts and only one global pont. Trackng ths MPP global pont and dstngushng t among the local ones s a bg constrant for most MPPT methods. Also, transent ssues of extractng power are added up to the lmtatons facng trackng operaton. Therefore, the proposed method serves to resolve the most mportant ssue; whch s to track the MPP. To nsure the capablty of proposed method to treat the most mportant ssue for any change n weather condton, ths study was appled n three dfferent shadow condtons to examne the relablty of method used.

7 4.1. Frst Case Steady The frst appled shadow condton was as n the followng sequence; 400 W/m 2 ; 400 W/m 2 ; 800 W/m 2 ; 800 W/m 2, for each PV module and connected n seres. The P V and I V characterstc curves of the array of the PV system s array are as presented n Fgure 5, the output whch represents the general characterstcs of the PV model operatng n the shadow condton. The results of trackng the MPP usng the proposed method and the P&O comparson method are presented and plotted n Fgures 6 and 7, whch were recorded under two dfferent levels of shadow condtons. The frst level s 400 W/m2 and appled on the 1 st and the 2 nd PV modules, and the second level s 800 W/m2 appled on the 3rd and 4th PV modules, when these four PV modules are connected n a seres. The second mportant ssue s the attanment of steady state condton for the extracted power. a Fgure 5. a: P V characterstcs of the PV array, b: I V characterstcs of the PV array operatng under shadow condtons, (400 W/m 2 ; 400 W/m 2 ; 800 W/m 2 ; 800 W/m 2 ). b

8 0.82s 3.5s Fgure 6. Extracted power usng the proposed method (ACO-PSO), under shadow condtons (400 W/m 2 ; 400 W/m 2 ; 800 W/m 2 ; 800 W/m 2 ). 1.65s 4.05s Tme (s) Fgure 7. Extracted power usng the P&O method, under shadow condtons (400 W/m 2 ; 400 W/m 2 ; 800 W/m 2 ; 800 W/m 2 ). The effectveness of the proposed MPPT method was tested and nvestgated through three dfferent shadow condtons, whch represented the most common shadow condtons that could happen on the PV system. Fgure 6 dsplays the extracted power s smulaton results

9 through the proposed MPPT technque. Whereas, Fgure 7 s dsplayng the extracted power s smulaton results by the P&O comparson method. By usng the proposed method, t can be seen that at the thrteenth selecton nterval, the algorthm tracked the local MPP at a value of 231W and at a tme of 0.82 s. The algorthm contnued to search untl fndng the second pont whch represented the global pont. It was tracked accurately at 3.5 s and 447 W. However, n the comparson method (P&O method), the frst local pont tracked at the eghteenth selecton nterval wth power value rangng from 225 W and 235 W. The global pont was tracked at the tme of 4.05 s wth power value rangng between 443 W and 453 W. From the two Fgures (6 and 7), t can be concluded that n addton to the fast trackng tme of the proposed method, t also recorded other advantages n terms of effcency and tme response besdes acqurng the steady state condton around MPP value. In the trackng operaton of the frst shadow condton, settlng tme of MPP n the proposed method was shorter than the P&O method, whch occurred at 3.5 s for global pont. However, t took 4.05 s n the P&O method. Furthermore, the other mportant ssue consdered n the proposed method s the oscllaton pattern around MPP, whch was oscllaton-free n comparson wth the P&O method; rangng from 443 W to 453 W along the tme after the global MPP trackng was acheved Second Case Steady The second case selected shadow condton as per the followng sequence: 600 W/m 2 ; 1000 W/m 2 ; 1000 W/m 2 ; 800 W/m 2 and the shape of P V and I V characterstc curves of the PV system s array are presented n Fgure 8. The output represents the general characterstcs of the PV model operatng n shadow condton. a Fgure 8. a:p V characterstcs of the PV array, b: :I V characterstcs of the PV array, for a system operatng under shadow condtons, (600 W/m 2 ; 1000 W/m 2 ; 1000 W/m 2 ; 800 W/m 2 ). b

10 0.82s 5.3s Fgure 9. Extracted power usng the proposed method (ACO-PSO), under shadow condtons (600 W/m 2 ; 1000 W/m 2 ; 1000 W/m 2 ; 800 W/m 2 ). 1.4s 6s 7.4s Fgure 10. Extracted power usng P&O method, under shadow condtons (600 W/m 2 ; 1000 W/m 2 ; 1000 W/m 2 ; 800 W/m 2 ). The result of trackng MPP n second operatng condton, usng the proposed method s presented n Fgure 9 and by usng the comparatve P&O method presented n Fgure 10. In ths shadow condton, the same way of trackng the frst shadow condton was used. However, n ths method, two local ponts and one global pont were taken. In the trackng process, t ntally tracked the frst local pont at 0.82 s and 445 W n the proposed method. Afterward,

11 the searchng process contnued and t tracked the global pont at 5.3 s and 648 W. In the second local pont, the tracker mantaned the same power value of global pont, although the change s observed n the power characterstc curve, as shown n Fgure 9. Ths change s due to the fact that the proposed method tred to mantan the maxmum value of extracted power. However, as shown n Fgure 10, n the second local MPP, the power extracted n P&O method contnued to track the MPP and the value of power plunges very fast. Under ths P&O method, the maxmum tracked power was at 245 W. However, n the proposed method, t mantaned at 648 W. Another most mportant comparatve pont, wthn the trackng process, s the proposed algorthm, whch tracked the frst local MPP at the twelfth selecton nterval of 445 W and 0.82 s. However, n the P&O, t was tracked at power value n the range of 437 W and 447 W and 1.4 s. The trackng process contnued searchng untl t fndng the global pont at 648 W and 5.3 s. Smlarly, n the P&O method, the global pont was tracked at 5.88 s wth a power value rangng between 640 W and 650 W. In the comparatve method, the search contnued, resultng n trackng the thrd MPP pont (.e. the second local pont) at 7.4 s wth a maxmum power value of 245 W. Therefore, the mportant dfference between the proposed method and the P&O method; s that the proposed method was successful n trackng the global pont, whle P&O faled n trackng t. Contradctorly, the P&O method contnued trackng each local pont on the power curve, whch fnally resulted n a decrease n power value at all tme values. But, n the proposed method, the focus was on the global pont. As a general comparson between the proposed and P&O method durng ths trackng process, the proposed algorthm tracked the frst local MPP at the tenth selecton nterval at 0.7 s and 147 W, but the compared P&O method MPP tracked t at 1.35 s and power value rangng from 139 W to 149 W. The second local MPP was tracked at the twelfth selecton nterval, but n the compared P&O method, t was tracked at the sxteenth selecton nterval at 3.9 s and power value between 373 W and 383 W. The thrd MPP (global MPP) was tracked at the fourteenth selecton nterval, but n the compared P&O method, t was tracked at the nneteenth selecton nterval of 6.2 s and power value rangng from 451 W to 461 W. Therefore, for all the case studes tested n ths study, the MPPT proposed method recorded fast tracks tme as compared to the P&O method whch was much slower than the proposed method. In addton to that, the proposed method also depcts other benefts n terms of power effcency and tme response, as well as the steady state occurrng around the MPP

12 values. Addtonally, n the proposed method, the oscllaton pattern around MPP was not found, ndcatng that t had attaned the steady state condton. In comparson wth P&O method where the oscllaton around MPP was present, the thrd shadow condton was rangng from 451 W to 461 W, along wth the tme after the global MPP trackng was attaned. Concluson In ths study, a new MPPT algorthm based on hybrd ACO-PSO algorthm has been developed. The proposed method mplemented by referrng to the output of the conventonal P&O algorthm and nonlnear characterstcs of photovoltac as an nput to the PID controller. The strategy s to create varable and adaptve perturbaton ntenstes for acceleratng the MPPT velocty and mprovng the MPPT effcency. To mprove the overall performance of the PV system, an optmzed PID controller has been mplemented for the purpose of oscllaton flterng and nose suppresson as taken n ths desgn, as well as the tme response and settlng tme. In addton, the proposed method has been used to track the maxmum global power pont under shadow condtons. For ths purpose, two dfferent shadow condtons were tested as a case studes. The results of trackng MPP by the proposed MPPT technque showed that the mproved method tracked the MPP for all the tested cases wth a reasonable accuracy and n a very short settlng tme as compared to the P&O method. It was found that the extracted power had no oscllaton around the MPP, when comparng wth the P&O method wth a notceable oscllaton around the MPP, whch resulted n decreasng the effcency of the extracted power from the PV system. References [1] A. M. Humada, A. M. Aaref, H. M. Hamada, M. H. Sulaman, N. Amn, and S. Mekhlef, "Modelng and characterzaton of a grd-connected photovoltac system under tropcal clmate condtons," Renewable and Sustanable Energy Revews, [2] R. F. Coelho, F. Concer, and D. C. Martns, "A study of the basc DC-DC converters appled n maxmum power pont trackng," n Power Electroncs Conference, COBEP'09. Brazlan, 2009, pp [3] A. I. Bratcu, I. Munteanu, S. Bacha, D. Pcault, and B. Rason, "Cascaded dc dc converter photovoltac systems: Power optmzaton ssues," Industral Electroncs, IEEE Transactons on, vol. 58, pp , [4] Y.-H. J, D.-Y. Jung, J.-G. Km, J.-H. Km, T.-W. Lee, and C.-Y. Won, "A real maxmum power pont trackng method for msmatchng compensaton n PV array under partally shaded condtons," Power Electroncs, IEEE Transactons on, vol. 26, pp , [5] K. S. Lee, "Maxmum power pont trackng method," ed: Google Patents, 2014.

13 [6] N. Fema, G. Petrone, G. Spagnuolo, and M. Vtell, "A technque for mprovng P&O MPPT performances of double-stage grd-connected photovoltac systems," IEEE transactons on ndustral electroncs, vol. 56, pp , [7] N. Fema, G. Petrone, G. Spagnuolo, and M. Vtell, "Optmzng samplng rate of P&O MPPT technque," n Power Electroncs Specalsts Conference, PESC IEEE 35th Annual, 2004, pp [8] M. G. Vllalva, J. R. Gazol, and E. Ruppert Flho, "Analyss and smulatonof the P&O MPPT algorthm usng alnearzed PV array model," n Power Electroncs Conference, COBEP'09. Brazlan, 2009, pp [9] B. Lu, S. Duan, F. Lu, and P. Xu, "Analyss and mprovement of maxmum power pont trackng algorthm based on ncremental conductance method for photovoltac array," n Power Electroncs and Drve Systems, PEDS'07. 7th Internatonal Conference on, 2007, pp [10] M. A. Elgendy, B. Zahaw, and D. J. Atknson, "Assessment of the ncremental conductance maxmum power pont trackng algorthm," IEEE Transactons on sustanable energy, vol. 4, pp , [11] S. Qn, M. Wang, T. Chen, and X. Yao, "Comparatve analyss of ncremental conductance and perturb-and-observaton methods to mplement MPPT n photovoltac system," n Electrcal and Control Engneerng (ICECE), 2011 Internatonal Conference on, 2011, pp [12] K. Ishaque, Z. Salam, M. Amjad, and S. Mekhlef, "An mproved Partcle Swarm Optmzaton (PSO) based MPPT for PV wth reduced steady-state oscllaton," Power Electroncs, IEEE Transactons on, vol. 27, pp , [13] F. M. de Olvera, S. A. O. da Slva, F. R. Durand, L. P. Sampao, V. D. Bacon, and L. B. Campanhol, "Grd-ted photovoltac system based on PSO MPPT technque wth actve power lne condtonng," IET Power Electroncs, vol. 9, pp , [14] L. L. Jang, D. L. Maskell, and J. C. Patra, "A novel ant colony optmzaton-based maxmum power pont trackng for photovoltac systems under partally shaded condtons," Energy and Buldngs, vol. 58, pp , [15] R. Suryavansh, D. R. Josh, and S. H. Jangamshett, "PSO and P&O based MPPT technque for SPV panel under varyng atmospherc condtons," n Power, Sgnals, Controls and Computaton (EPSCICON), 2012 Internatonal Conference on, 2012, pp [16] A. R. Res, M. H. Morad, and S. Jamasb, "Classfcaton and comparson of maxmum power pont trackng technques for photovoltac system: A revew," Renewable and Sustanable Energy Revews, vol. 19, pp , [17] A. Besheer and M. Adly, "Ant colony system based PI maxmum power pont trackng for stand alone photovoltac system," n Industral Technology (ICIT), 2012 IEEE Internatonal Conference on, 2012, pp [18] Z. Salam, J. Ahmed, and B. S. Merugu, "The applcaton of soft computng methods for MPPT of PV system: A technologcal and status revew," Appled Energy, vol. 107, pp , [19] H. Fang, L. Chen, and Z. Shen, "Applcaton of an mproved PSO algorthm to optmal tunng of PID gans for water turbne governor," Energy Converson and Management, vol. 52, pp , 2011.