Modeling and Control of a Low Power Wind Turbine

Transcription

1 4 th International Conference on DEVELOPMEN AND APPLICAION SYSEMS, Suceava, Roania, May 4-6, 08 Modeling and Control of a Low Power Wind urbine Dana-Alexandra Ciupăgeanu, Student Meber, IEEE, Gheorghe Lăzăroiu, Meber, IEEE Power Engineering Faculty University Politehnica of Bucharest Bucharest, Roania dana_ciupageanu@yahoo.co, glazaroiu@yahoo.co Viorel Berbece Mechanical Engineering and Mechatronics Faculty University Politehnica of Bucharest Bucharest, Roania vberbece@caz.ecen.pub.ro Mihai îrșu, Victor Galbură he Institute of Power Engineering Acadey of Sciences of Moldova Kishinev, Republic of Moldova tirsu@ie.as.d Abstract he unpredictability characterizing renewable energy sources severely ipacts power systes as their share continuously enlarges. Choosing the right technology to exploit the available potential of the intended resource and ipleenting appropriate plant control represents a solution that iproves syste's operating conditions. his paper presents siulations results for a Peranent Magnet Synchronous Generator (PMSG) driven by a Wind urbine (W) considering a control strategy eploying a Neural Network Predictive Controller (NNPC), as one of the ost viable aong suitable options for rising wind energy share while itigating its highly stochastic characteristics. he odel syste was represented in Matlab prograing environent, using Siulink/Siscape/Power Systes blocks. he behavior of the syste is analyzed both in steady and dynaic conditions, in the hypothesis that the controller provides the torque reference for the generator. Keywords low-power wind turbine; Matlab/Siulink; neural network predictive control; peranent agnet synchronous generator; wind energy I. INRODUCION Given the increasingly restrictive conditions regarding environental conservation and fossil fuels depletion, concurrent with the evolutionary trends of electricity deand and actually changing status of the consuers, power systes are continuously facing change [-]. Only last decade, renewable energy generating capacities installed worldwide doubled, reaching W at the end of 06; Wind Energy Conversion Systes (WECS) account near 4% of it. A great decreasing trend with respect to wind power costs encouraged by governent policies and arket dynaics led to halving its price [4-5]. Under the circustances of progressively copetitive renewable energy technologies and still very active research interest in the area, their variable nature involves any issues related to power network integration, especially concerning aintaining an adequate reliability level of the supply. Even with the constant growth of WECS rated power, considering the large dispersion of exploitable potential and the strong variability of the resource, sall scale systes could represent the solution that answers not only this proble, but also facilitates integration into the power syste. Furtherore, in low-power applications the power output fluctuations do not deterine frequency variation [-5]. A proising technology in order to achieve variable speed wind turbine operation without using a gearbox outstands the Peranent Magnet Synchronous Generator (PMSG), especially considering the lately rearkable advances in peranent agnet aterials that increased their affordability [], [6]. A sall scale WECS odel consisting of a W directly driving a PMSG which further supplies a three-phase load is described in this paper. Its purpose is to establish an appropriate structure of a Neural Network Predictive Controller (NNPC), including choosing suitable input in order to achieve iniu deviation of the output. According to steady and dynaic siulations results, the behavior of the controller shows iproved perforance. II. PMSG DRIVEN BY A WIND URBINE A. Advantages Although ost of the WECS are based on Doubly Fed Induction Generators (DFIG), the use of PMSG becae very popular lately. hat is due to nuerous advantages, such as higher energy densities, which allows saller diensions of turbine's rotor for the sae output as DFIG, self excitation syste and the possibility of direct coupling to the wind turbine, avoiding therefore the use of gearboxes representing an eleent of echanical vulnerability in the configuration, increasing its reliability, eaning lower aintenance and operation costs [7-8]. Moreover, PMSG allows a full control towards axiu wind power extraction within a large operating range and it is characterized by very good efficiency, precision and enhanced fault ride through capability. For these reasons, several anufacturers produce and develop such equipents, arket deand highlighting new opportunities especially in lowpower applications [9-]. Regarding the power electronic interface between the WECS and the electrical grid, the fully scaled ac/dc/ac converter configuration ensures both optial operating point, in order to extract axiu power fro the air strea, and quality of electricity output. Moreover, this highly reliable his work was supported in part by a grant of the Roanian Ministry of Research and Innovation, CCCDI UEFISCDI, project nuber PN-III- P-.-PCCDI /PCCD/08, within PNCDI and 7BMPNIII- P-99/06-I05.6.0, within PNCDI III /8/$ IEEE

2 architecture enables total decoupling fro the grid, protecting the generator in fault conditions [], [7], []. hough, it has to be entioned that increased installation costs and higher converter losses represent still a drawback of this technology [7], [9]. Finding a balance between investents and perforances is, as always, a challenge that can be et by ipleenting adequate control in order to satisfy often eergent objectives. B. Matheatical odel he atheatical odel of the PMSG is given as it follows [], [6], [9]: d vds = Rsids + Ld ids ωs Lqiqs () d vqs = Rsiqs + Lq iqs + ω s Ld ids + ωsλ () where v ds, v qs are direct and quadrature axes stator voltages; Rs is stator's resistance; L d, Lq are direct and quadrature axes stator inductances; λ is rotor's flux; ωs is the steady state electrical speed. he electroagnetic torque is given as: e [ i + ( L L ) i i ] = p λ qs d q ds qs () where p represents the nuber of pair of poles of the achine. he dynaic equation of the rotor: dω J = Fω (4) e where J is the oent of inertia; F is the viscous friction coefficient; is the echanical oentu developed by the driving turbine. he echanical power of the prie over coupled to the PMSG, here represented by the wind turbine, results according to (5) [-]: P = cp ρ A w (5) where cp is turbine's power coefficient, ρ is the density of air flow, A is the surface swept by the rotor and w is the wind speed. he power coefficient depends on the pitch angle of rotor's blades and the tip speed ratio as in (6) [4-5]: c ( λ, β ) = k P k k λ i β k 4 e k5 λ i + k 6 λ (6) where k i,i =, 6 are real coefficients, λ is the tip speed ratio, β is the blade's pitch angle. he value of deterined according to (7): = λ λ β 0.05 i β + λ i can be he echanical torque at turbine's shaft can be calculated based on (8): III. (7) = P P ω = ωe (8) p PLAN CONROL he extensive developent of WECS highlights the necessity of establishing adequate plant control to enhance power syste reliability and iprove operating conditions [8], []. Regarding control strategies of the ain subsystes of a wind energy based power plant, aong the ost widely used are: the Field Oriented Control (FOC) and Direct orque Control (DC) for the achine side converter; the Voltage Oriented Control (VOC) and Direct Power Control (DPC) for the grid side converter; as for the pitch angle control, this becoes iportant at higher values of the wind speed, at lower values being fixed [], [9], [7]. Modern power plants based on WECS need to overcoe the uncertainty related not only to the renewable source itself but also to other interfering factors, such as load dynaics or power grid state. Conventional controllers encounter several probles in perforing these functions, precisely because of the power output unpredictability, beside the nonlinearity and coplexity of the syste [], [5-6]. o iprove plants and, forwards, networks perforances, new control strategies should be designed and optiized considering all involved subsystes and the quality of the output. here is a considerable probability that classical PI controller shall encounter difficulties relatively to indefinite derivatives in conditions of highly variable wind speed [7], [0]. he Neural Network Predictive Controller (NNPC) represents a very suitable option in stochastic input conditions. Based on biological evolution considerations and taking into account huan expertise, it brings the great benefits of robustness and easy understanding without jeopardizing control perforances [6]. Previous approaches to ipleenting NNPC based strategies eployed as input either turbine's echanical power, generator's speed or blade pitch angle aiing to provide an estiate of variables of interest such as optial pitch angle, power factor, wind speed, torque at the shaft or axiu power output. Depending on both the coplexity of the syste and the considered objective, training algoriths and network structure vary [8]. 7

3 IV. CASE SUDY A. Syste paraeters he plant architecture discussed previously was ipleented in Matlab/Siulink software, using Siscape/Power Systes blocks. he overall concept on which NNPC relies is a closed loop control, feeding the easured output of the plant to the input of the controller, together with the set reference, as shown in Fig.. alternatives the pitch angle and the torque reference, and considering the power supplied to the load as the output, the best results in ters of load following were obtained if the NNPC provides the torque reference, as presented in Fig.. his allows avoiding the use of integral coponent, siplifying and stabilizing torque control behavior in a wide range of conditions. he values of the coefficients involved in calculating the variable power coefficient of the wind turbine, according to (6), are entioned in able II [4]. ABLE II. WIND URBINE COEFFICIENS Coefficient Value k Fig.. General structure of the siulation odel he odel considered for the siulation of the wind power plant ipleents a variable pitch wind turbine, directly driving a 500 kw PMSG which in turn feeds a three phase load. First, it was required to establish the input and output of the plant, in ters of the control strategy. able I presents the paraeters of the three-phase PMSG used in the siulation [7], [9-0]. ABLE I. PMSG PARAMEERS Characteristic quantity Sybol Measureent unit Value Stator phase resistance RS Ω 0.05 k 6 k 0.4 k 4 5 k 5 k he axiu power coefficient is obtained for a pitch angle of β = 0, corresponding to a tip speed ratio of λ = 8. : P = he characteristics cp = cp( λ ) for the wind turbine eployed in the odel are illustrated in Fig.. c ax d-axis inductance Ld H q-axis inductance Lq H Flux linkage established by agnets ψ V s Inertia J kg 0.0 Viscous daping F N s Pole pairs p - 0 Static friction f N 0 Fig.. Wind turbine charateristics Fig.. Wind power plant odel he neural network, trained as it shall be described forward, elaborates the control signal, representing plant input. After analyzing several options for the input, including as B. NNPC design he general structure of the NNPC consists of a double hidden layer aidst the input and the output of the controller. Further, its design aied finding the paraeters that will ensure iproved perforance. Following several preliinary tests, the cost and control horizons, cost weighting factor and search paraeter resulted as in Fig. 4. 8

4 Fig. 4. NNPC block paraeters In order to generate the training saples, the reference provided to the plant was given as a rando nuber, following a noral distribution having the ean equal to the rated capacity of the load and the distribution equal to the allowable percentage of variation. Plant input was considered to vary between zero and noinal value of the torque, deterined according to (8): Fig. 7. raining data of the NNPC C. Siulation results After establishing the NNPC structure and paraeters, the response of the odel in steady state conditions and subsequent two sudden changes of wind speed, one increasing fro 8 /s to /s and one decreasing between sae values, was analyzed. he variation was considered to occur at t=5 s..5 0 N he response of the plant to the variable reference is presented in Fig. 5, for a 000 saples training set. During siulation tie (t s=0 s), wind speed was considered variable, with a axiu value of 5 /s. Fig. 8. Plant response under constant wind speed As pictured in Fig. 8 and Fig. 9, plant output under constant reference and unifor or variable wind speed has very siilar evolution trend, highlighting controller's capacity to adapt and ensure load supply. Fig. 5. raining set Further, perforing neural network training eploying different algoriths and considering various nubers of neurons on the hidden layer, best perforances led to a first hidden layer with a log-sigoid activation function and 0 neurons, and a second one with a single neuron and linear transfer function. Fig. 6. Neural network architecture he training algorith which perfored best and behaved the fastest convergence was Levenberg-Marquar backpropagation. Fig. 9. Plant response in dynaic conditions It can be noticed that the behavior of the syste in dynaic conditions stabilizes. V. CONCLUSIONS NNPC represents a control strategy that can overcoe the issues related to highly variable renewable sources, such as wind energy. Siulation results presented for a NNPC providing the torque reference for the PMSG in WECS show good perforances in different conditions, steady and transient deterined by wind speed variation. he proposed configuration allows achieving control objectives without using an integral coponent for regulation, which akes the syste ore stable. 9

5 Control strategies relying on NNPC show a proising perspective regarding ipleentation in renewable energy systes, and further in hybrid power configurations. Selecting convenient input-output pair is particular to each application and strongly depends on the objective pursued. Siulations on this subject are therefore of ajor interest and shall be developed in following research. ACKNOWLEDGMEN his work was supported by a grant of the Roanian Ministry of Research and Innovation, CCCDI UEFISCDI, project nuber PN-III-P-.-PCCDI / PCCDI/08 and 7BMPNIII-P-99/06-I05.6.0, within PNCDI III. REFERENCES [] C. Zhao, U. opcu, N. Li, "Design and stability of load-side priary frequency control in power systes", IEEE ransactions on Autoatic Control, vol. 59, no. 5, pp , 04. [] S. M. ripathi, A. N. iwari, D. Singh, "Grid-integrated peranent agnet synchronous generator energy conversion systes: A technology review", Renewable and Sustainable Energy Reviews, vol. 5, pp , 05. [] D.A. Ciupăgeanu, G. Lăzăroiu, M. îrșu, "Carbon dioxide eissions reduction by renewable energy eployent in Roania", Proceedings of 07 International Conference on Electroechanical and Power Systes (SIELMEN 07), - October, Iași, Roania, pp. 8-85, 07. [4] IRENA (08), "Renewable power generation costs in 07", International Renewable Energy Agency, Abu Dhabi [5] D.A. Ciupăgeanu, G. Lăzăroiu, O. Zachia, "he influence of wind turbine generators on power systes dynaic behavior", Proceedings of 8th Syposiu on heral Science and Engineering of Serbia, 7-0 October, Sokobanja, Serbia, pp , 07. [6] Y. Xia, K. H. Ahed, B. W. Willias, "A new axiu power point traking technique for a peranent agnet synchronous genrator based wind energy conversion syste", IEEE ransactions on Power Electronics, vol. 6, no., pp , 0. [7] A. Rolán, A. Luna, G. Vásquez, D. Aguilar, G. Azevedo, "Modeling of a variable speed wind turbine with a peranent agnet synchronous generator", IEEE Iternational Syposiu on Industrial Electronics ISIE 009, 5-8 July, Seoul, Korea, pp , 009. [8] X. Liu, X. Kong, "Nonlinear odel predictive control for DFIG-based wind power generation", IEEE ransactions on Autoation Science and Engineering, vol., no. 4, pp , 04. [9] M. Chinchilla, S. Arnaltes, J. C. Burgos, "Control of peranent-agnet generators applied to variable-speed wind-energy systes connected to the grid", IEEE ransactions on Energy Conversion, vol., no., pp. 0-5, 006. [0] M. E. Haque, M. Negnevitsky, K. M. Muttaqi, "A novel control strategy for a variable speed wind turbine with a peranent agnet synchronous generator", IEEE ransactions on Industry Applications, vol. 46, no., pp. -9, 00. [] M. A. Rahan, D. M. Vilathgauwa, M. N. Uddin, K.-J. seng, "Nonlinear control of interior peranent-agnet synchronous otor", IEEE ransactions on Industry Applications, vol. 9, no., pp , 00. [] K. Aei, Y. akayasu,. Ohji, M. Sakui, "A axiu power control of wind turbine generator syste using a peranent agnet synchronous generator and a boost chopper circuit", Proceedings of the Power Conversion Conference, -5 April, Osaka, Japan, vol., pp , 00. []. Senjyu, R. Sakaoto, N. Urasaki,. Funabashi, H. Fujita, H. Sekine, "Output power leveling of wind turbine generator for all operating regions by pitch angle control", IEEE ransactions on Energy Conversion, vol., no., pp , 006. [4] [5] D. Kuar, K. Chatterjee, "A review of conventional and advanced MPP algoriths for wind energy systes", Renewable and sustainable energy reviews, no. 55, pp , 06. [6] A. Medjber, A. Guessou, H. Belili, A. Mellit, "New neural network and fuzzy logic controllers to onitor axiu power for wind energy conversion systes", Energy, no. 06, pp. 7-46, 06. [7] S. J. Underwood, "On-line paraeter estiation and adaptive control of peranent agnet synchronous achines", PhD hesis, University of Akron, Akron, Ohio, USA, 006. [8] R. Ata, "Artificial neural network applications in wind energy systes: a review", Renewable and sustainable energy reviews, no. 49, pp , 05. [9] M. J. Corley, R. D. Lorentz, "Rotor position and velocity estiation for a salient-pole peranent agnet synchronous achine at standstill and high speeds", IEEE ransactions on Industry Applications, vol. 4, no. 4, pp , 998. [0] M. Linke, R. Kennel, J. Holtz, "Sensorless position control of peranent agnet synchronous achines without liitation at zero speed", IEEE 8th Annual Conference on the Industrial Electronics Society IECON0, 5-8 Noveber, Sevilla, Spain, pp , 00. 0