Available online at ScienceDirect. International Conference on Electrical Engineering and Informatics 2013 (ICEEI 2013)

Transcription

1 Avalable onlne at ScenceDrect Proceda Technology 11 ( 13 ) Internatonal Conference on Electrcal Engneerng and Inforatcs 13 (ICEEI 13) Speed Trackng of Indrect Feld Orented Control Inducton Motor usng Neural Network Azuwen Ada Bohar a, Wahyu Mulyo Utoo a, Zanal Ala Haron a,, Nooradzane Muhd. Zn a, Sy Y S a, and Roslna Mat Arff a a Faculty of Electrcal and Electronc Engneerng, Unverst Tun Hussen Onn Malaysa, 860 Part Raja, Batu Pahat, Johor, Malaysa Abstract Neural network based ndrect feld orented control of nducton otor drve s presented n ths paper. The purpose of ths otor s to nvestgate the perforance of neural network controller whch s appled on three phase nducton otor drve syste. The ntroducton of neural network usng ndrect feld orented control prncples for decouplng control of nducton otor are dscusses. A decouplng controller and flux estator were pleented upon ths neural network usng MATLAB/SIMULINK software. The tranng procedure of the neural network, descrpton of the control syste s gven n ths paper. The coparatve results show that the neural network based ndrect feld orented control schees of nducton otor drve s ore effectve and, hence, found to be a sutable replaceent of the conventonal proportonal ntegral dervatve controller. Keywords:Inducton otor, feld orented control, neural network controller. 1. Introducton An nducton otor (IM) s desgned to operate fro a three phase source of alternatng voltage and t s a type of synchronous AC otor. The advantages of the nducton otor ncludes hgh relablty, relatvely sple, has rugged structure, low cost, has a wde speed range, hgh effcences and robustness. These advantages are proven to be a crucal developent of the electrcal drves, takng the nducton achne as the executon eleent, for all related aspects lke startng, brakng, speed reversal and speed change. The dynac operaton of the nducton achne drve syste s a part of the overall perforance of the syste. The advanced control ethods have ade the nducton otor for hgh perforance applcatons possble due to the rapd proveents n power electroncs and croprocessor. The nducton otor control n the feld of electrc otor drves has attracted a lot of researchers over te[1-14]. To eet the hgh standards of perforance set by dc drves, varable speed ac drves eployng nducton otors and the feld-orented control ethod have been developed n recent years. Decouplng ters between the flux and the torque s hghly advocates to provde fast dynac response as well as good steady-state perforance. Hgh dynac perforance n nducton otor can be acheved by eans of feld-orented control where t provdes a sutable atheatcal descrpton of three phase nducton otor. The translaton of coordnate fro the fxed references stator frae to the frae of rotatng synchronous s pled by the vector control [1]. Independent control of the torque and flux of the coplex dynac of the otor were pertted but ths decouplng technque, ntroduced n early 1970s [2].Thus, a good perforance wth paraeter nsenstve property cannot be acheved wthout applyng ore sophstcated control technques such as adaptve control [3]. Artfcal neural networks attracted a great attenton n the feld of power electroncs due to ther nherent parallels whch allows for hgh speed processng and perts pleentaton of real te control applcaton. The ablty to perfor n nose envronents and the tolerance to faults and to ss data are possessed by the NN s[4]. a E-al address: he110247@sswa.uth.edu.y The Authors. Publshed by Elsever Ltd. Open access under CC BY-NC-ND lcense. Selecton and peer-revew under responsblty of the Faculty of Inforaton Scence & Technology, Unverst Kebangsaan Malaysa. do: /j.protcy

2 142 Azuwen Ada Bohar et al. / Proceda Technology 11 ( 13 ) To dentfy and control nonlnear dynac systes and nonlnear paraeter estaton, the use of Neural Networks (NN s) has been proposed [5][6]. An estaton of the stator flux and traned to ap the nonlnear behavor of a rotor flux s perfored on the proposed NN s [7]. Hgh perforance control has been the objectves of [8], whch presents robust neural controller and aganst varaton of paraeter. Wthout knowng the exact atheatcal odel of the syste, the desred appng between the nput and outputs sgnals of the syste can be learned by the ANN s. The NN s are excellent estators n non-lnear syste because t does not use the atheatcal odel of the syste [9-10]. Ths paper wll present a neural network (NN) based ndrect feld orented control (IFOC) of nducton otor (IM) drve. The perforance of NN controller whch s appled n three phase IM drve syste wll be nvestgated n ths paper. The theoretcal prncple, nuercal sulaton procedures and the results of these ethods are dscussed and copared wth conventonal PI controller. 2. Dynac Modelng of Inducton Motor The nducton otor odel has been derved n a nuber of dfferent reference fraes. Ths akes t easer to fx the reference frae to a partcular otor quantty and adjust the odel accordngly. Most of nducton otors are the rotary type wth bascally a statonary stator and a rotatng rotor. The dynac odel of the nducton otor s derved by transforng the three phase quanttes nto two phase drect and quadrature axes quanttes. The atheatcal odel n copact for can be gven n the statonary reference frae as follows [11]. Vds Rs Lsp 0 Lp 0 ds V qs 0 Rs Lsp 0 Lp qs (1) V dr L p rl Rr Lrp rl r dr Vqr rl Lp rlr Rr Lrp qr The nducton odel n the stator-fxed d-q reference frae s descrbed as: Whereas the echancal equaton s gven as follow: The nducton otor stator flux and torque are calculated n the flux and torque calculator as follows: The electroagnetc torque of the achne can be presented as follow:

3 Azuwen Ada Bohar et al. / Proceda Technology 11 ( 13 ) IFOC Control Syste Descrpton IFOC s a control procedure for operatng the otor that results n fast dynac response and energy effcent operaton at all speeds. It coutates the otor by calculatng voltage and current vectors based on otor current feedback. It antans hgh effcency over a wde operatng range and allows for precse dynac control of speed and torque. The IFOC controls the stator currents represented by a space vector. It transfors three-phase stator currents (A, B, C) nto a two-phase te varant syste (α,β). A two-coordnate te nvarant syste (dq) s obtaned fro that te varant syste. In ths structure, the otor flux generatng part s d (drect) and a toque generatng part s q (quadrature). 4. Neural Network Inspred by the successful functon of the huan brans, the neural network (NN) was developed for solvng any large scale and coplex probles. Based on ablty to process soe nforaton and also to analyze the nput and output sultaneously, t akes NN sutable for dynac and nonlnear syste. The developent of the structure and learnng algorth of the NN s explaned as follows [12] Proposed NN speed controller Ths paper proposed a NN control ethod of IFOC based on hysteress to reduce the overshoot and torque rpple. The NN control s added to the speed controller to produce the torque reference. The block dagra of the proposed NNIFOC of IM drve s shown n Fg 1. Fg. 1. Coplete block dagra of proposed NNIFOC 4.2. Structure of NNIFOC To desgn the neural network control soe nforaton about the plant s requred. Bascally, the nubers of nput and output neuron at each layer are equal to the nuber of nput and output sgnals of the syste respectvely. Further the nuber of hdden layers and the total neurons s depended on the coplexty of the syste and the requred tranng accuracy. To pleent search effcency optal control of an nducton otor drve, a ultlayer perceptrons neural network control s developed. Based on the type of the task to be perfored, the structure of the proposed NNIFOC s as shown n Fg 2. Fg.2. Block dagra of neural network speed control for nducton otor drve

4 144 Azuwen Ada Bohar et al. / Proceda Technology 11 ( 13 ) The controller conssts of nput layer, hdden layer and output layer. Based on nuber of the neuron n the layers, the NNFOC s defned as a network structure. The frst neuron of the output layer s used as a torque reference sgnal (a 2 1= f ). The connectons weght paraeter between j th and th neuron at th layer s gven by w j, whle bas paraeter of ths layer at th neuron s gven by b. Transfer functon of the network at th neuron n th layer s defned by: n 1 S j 1 w j a 1 j b The output functon of neuron at th layer s gven by: a f n ) Where f s actvaton functon of the neuron. In ths desgn the actvaton functon of the output layer s unty and for the hdden layer s a tangent hyperbolc functon gven by: ( 2 f ( n ) 1 2n 1 e Updatng of the connecton weght and bas paraeters are gven by: w b j F( k) ( k 1) wj ( k) w j F( k) ( k 1) b ( k) b where k s saplng te, α s learnng rate, and F perforance ndex functon of the network. 5. Sulaton Results A sulaton was constructed for the entre syste n Fg 1 wth the pleentaton of NN based IFOC schee for the three-phase nducton otor usng Borland C++, and then ebedded as S-functon n Sulnk-Matlab software. In ths paper, the pattern of the reference speed that wll be use to verfy the effectveness of the proposed controller s shown n Fg 3 the paraeters for the otor are 2V, 50Hz, Poles=4, Rs=0.3Ω, Rr=0.25Ω, Ls=0.0415H, Lr=0.0412H, L=0.03H, J=0.1kg-2. 1 REFERENCE SPEED Te (s) Fg. 3. Pattern of the reference speed

5 Azuwen Ada Bohar et al. / Proceda Technology 11 ( 13 ) The output response of the syste based on the speed reference patern s shown n Fg 4 and Fg Te (s) Te (s) Fg.4. Acceleraton rap response of NN and PID speed controller Te (s) Te (s) Fg.5. Deceleraton rap response of NN and PID speed controller Referrng to the Fg 4, the acceleraton rap of the syste by usng both controllers n between te 0 to 2 seconds, shows that NN speed controller follow the shape of speed reference pattern and have a response proved fro the conventonal PID controller. Whle, n Fg 5 between te 2 to 3 seconds, the speed contnue at 130 rad/s. Then, NN speed controller follow the shape of deceleraton rap at the te of 3 to 5 seconds. The results for the both fgure shows the great reduce n rpple of the transent respose and proved the perforance at the sae te rather than PID controller. 6. Concluson The NN based IFOC of nducton otor (IM) drve syste has been presented n ths paper. The purpose of ths otor s to nvestgate the perforance of NN copared to the PID controller. The coparatve results show that NN based IFOC schees of IM drve s ore effectve and, hence, found to be a sutable replaceent of the conventonal PI controller. It s concluded that feld orented control by usng neural network are useful and able to prove the perforance. The result shows that the perforance of the syste s proved by reducng the rpple. Acknowledgeent The author would lke to grattude Unversty Tun Hussen Onn Malaysa for any valuable supports durng conductng ths research and n preparng ths anuscrpt.

6 146 Azuwen Ada Bohar et al. / Proceda Technology 11 ( 13 ) References [1] Goyat, S. and Ahuja, R. K.., 12. Speed Control of Inducton Motor usng Vector or Feld Orented Control, IJAET. [2] Blaschke, F., The prncple of feld orentaton as appled to the new transvector closed-loop control syste for rotatng-fled achnes, Seens. Vol.34, pp [3] Von Zuben, F.J., Netto, M.L.A., B, E., Szajner, J., "Adaptve vector control of a three-phase nducton otor usng neural networks," Neural Networks, IEEE World Congress on Coputatonal Intellgence., 1994 IEEE Internatonal Conference on, vol.6, no., pp vol.6. [4] Ba-Razzouk, A., Chert, A., Olver, G.,1997."A neural networks based feld orented control schee for nducton otors," Industry Applcatons Conference, Thrty-Second IAS Annual Meetng, IAS '97., Conference Record of the 1997 IEEE, vol.2, no., pp vol.2. [5] Wshart, M.T., Harley, R.G., "Identfcaton and control of nducton achnes usng artfcal neural networks," Industry Applcatons, IEEE Transactons on, vol.31, no.3, pp [6] Narendra, K.S., Parthasarathy, K., "Identfcaton and control of dynacal systes usng neural networks," Neural Networks, IEEE Transactons on, vol.1, no.1, pp [7] Ba-Razzouk, A., Chert, A., Olver, G., Scard, P., 1995, "Feld orented control of nducton otors usng neural networks decouplers," Industral Electroncs, Control, and Instruentaton, 1995., Proceedngs of the 1995 IEEE IECON 21st Internatonal Conference on, vol.2, no., pp vol.2. [8] Mloud, A., Draou, A., 01. "Neural controller desgn for speed control of an ndrect feld orented nducton achne drve," Industral Electroncs Socety, 01. IECON '01. The 27th Annual Conference of the IEEE, vol.2, no., pp vol.2. [9] Ohnsh T., Myazak H., and Oktsu H., Hgh Effcency Drve of an Inducton Motor by Means of V/F Rato Control. IECON Proceedngs 14 annual Conference, vol. 3, pp [10] Lorenz R.D., Yang S.M., Effcency-optzed flux trajectores for closed-cycle operaton of feld-orentaton nducton achne drves IEEE Transactons on Industry Applcatons, Vol. 28, pp [11] Kuar, Rajesh, Gupta, R.A., Bhangale, S.V., Gothwal, Hanshu; 07. "Artfcal neural network based drect Torque Control of Inducton Motor drves," Inforaton and Councaton Technology n Electrcal Scences (ICTES 07), 07. ICTES. IET-UK Internatonal Conference on, vol., no., pp , -22. [12] Yat A.H.M., and Utoo W.M., 04. On-Lne Optal Control of Varable Speed Copressor Motor Drve Usng Neural Control Model. PECon Proceedngs Conference, pp