Switching-off angle control for switched reluctance motor using adaptive neural fuzzy inference system

Size: px

Start display at page:

Download "Switching-off angle control for switched reluctance motor using adaptive neural fuzzy inference system"

Merilyn Bryan
5 years ago
Views:

1 Intenational Jounal of Enegy and Powe Engineeing 0; (): 9- Published online Febuay, 0 ( doi: 0.68/.iepe ISSN: 6-97X (Pint); ISSN: 6-960X (Online) Switching-off angle contol fo switched eluctance moto using adaptive neual fuzzy infeence system Liu Zhi Jian, *, Nguyen Le Minh Ti,, Nguyen Le Thai, Phan Xuan Le, Faculty of Electic Powe Engineeing, Kunming Univesity of Science and technology, Kunming City, Yunnan Povince, China Faculty of Electic and Electonic Engineeing, Tuy Hoa Industial College, Tuy Hoa City, Phu Yen Povince, Vietnam Faculty of Electical Engineeing, Kien Giang Technology and Economics College, Kien Giang Povince, Vietnam addess: alz067@sina.com (Liu Zhi Jian), nlmti@kiengiangtec.edu.vn (N. L. M. Ti), thai.nguyenle.tt@gmail.com (N. L. Thai), phanxuanle.ts@gmail.com (P. X. Le) To cite this aticle: Liu Zhi Jian, Nguyen Le Minh Ti, Nguyen Le Thai, Phan Xuan Le. Switching-off Angle Contol fo Switched Reluctance Moto Using Adaptive Neual Fuzzy Infeence System. Intenational Jounal of Enegy and Powe Engineeing. Vol., No., 0, pp. 9-. doi: 0.68/.iepe Abstact: Switched eluctance motos (SRM) have a wide ange of applications in industies due to the special popeties of this moto. Howeve, because of its dynamical nonlineaities, so the poblems contol of SRM is complex. This pape poposed an adaptive intelligent contolle fo SRM with the aim to impove the ipple of toque. Fist, we use a fuzzy logic contolle to contol switch-off angle, and then poposes a new contolle by means of Adaptive Neual Fuzzy Infeence (ANFIS). Simulation esults ae given to show the efficacy of the poposed method. Keywods: SRM, Fuzzy logic, ANFIS. Intoduction The switched eluctance moto (SRM) has simple stuctue, low-cost, and obust moto that makes it suitable fo vaiable speed and sevo-type applications. Futhe, it has a simple convete and contol equiements. Thus, SRM has eceived geat attention in the dive industy. [-] The disadvantages of SRM ae the stepping natue and inheent nonlinea chaacteistics, which causes an undesied effect on beaing. If the poblems can be solved, SRM can be an altenative to the othe motos []. These nonlinea chaacteistics include the nonlinea toque function of cuent, oto position and the magnetic satuation at cetain opeation egions. Thee ae two choices fo educing the toque ipple minimization that is impoved the magnetic stuctue of the moto o impoved moto contolle [6]. In ecent yeas, fuzzy logic combination with neual netwok to contol nonlinea obects was developed stongly. Specially, that is the used of ANFIS to contol switched eluctance moto [7, 8, 9, 0]. In [7] an adaptive neuo-fuzzy infeence system (ANFIS) is used to estimate phase inductance. An Adaptive Neual Fuzzy Infeence System also is used to develop a new model fo SRM [8]. In [9], the authos studied the poblems of speed contol fo switched eluctance moto by means of an adaptive neuo-fuzzy contolle. An adaptive neuo-fuzzy contolle of switched eluctance moto is pesented in [0]. Fom the easons mentioned above, the ANFIS algoithm can be used to contol of nonlinea systems is effective. In this pape, we popose an ANFIS contolle to contol switching-off fo SRM with the pimay pupose is to educe the toque ipple minimization. The emainde of the pape is oganized as follows. Section descibes SRM model. The Switch-Off Contolle fo SRM is pesented in section. Simulations ae given in Section and conclusions ae summaized in Section.. SRM Model In this Section, we pesent the pinciple of contol switch-off and the theoetical basis of the SRM to povide a basis fo pocess design calculations... The Pinciple of Contol Switch-Off The schematic diagam of the switch-off contolle is shown in Figue. The contolle inputs ae obtained by manipulating the speed efeence and feedback, while the contolle output is integated to contol switch-off paamete of SRM, which the main pupose is to educe toque ipple.

2 0 Liu Zhi Jian et al.: Switching-off Angle Contol fo Switched Reluctance Moto Using Adaptive Neual Fuzzy Infeence System conside the following thid hamonic (,, ) = 0 cos( ( ) ) λ N θ i a ak kn θ θ s i (9) k = T ( N, θ, i ) = i N ka sin ( kn θ ( i) θ ) (0) e k s k = Figue. Schematic of Switch-off contolle.. Electomagnetic Equations The switched eluctance moto has a simple stuctue, but the solution of its mathematical models is elatively difficult due to its dominant nonlinea behaviou. The mathematical equations of switched eluctance moto ae descibed as following The voltage equation is: di V = Ri s ( θ, i ) + E ( ω, θ, i ) () dt The motion equation is: dω = ω () dt J Te Bm TL dl( i, θ ) T = i = i N ka sin( kn θ ) () e k dθ k = The linea inductance model is given = ( θ ) () L( N, θ ) a a cos kn 0 k k = Fouie expansion equation () and get to Level (coesponding to the thid hamonic) Phase Phase Phase Phase -th ( ) ( ) ( N θ ) L ( N, θ ) = a a cos N θ a cos N θ 0 a cos ( ) ( ) ( θ θ ) L ( N, θ) = a a cosn θ θ a cos N θ θ 0 s s a cosn s ( ) ( ) ( θ θ ) L ( N, θ ) = a a cosn θ θ a cos N θ θ 0 s s a cosn s () (6) (7) ( ( ) ) (8) L ( N, θ ) = a a cos kn θ θ 0 k s k = Wite the geneal fom equation inductance, the flux linkage, the moment of SRM moto at phase -th only Whee V is the phase voltage -th i is the Phase cuent -th R s is the phase winding esistance θ is the shaft speed J is the moment of inetia B m is the fiction λ is the flux linked by the winding -th T L is the toque load N s is the numbe of stato poles N is the numbe of oto poles θ is the position of the oto θ is the position of the oto -th θ s is the phase cuent displacement angle T e is the total toque -th In SRM dive, it is impotant to synchonize the stato phase excitation with the oto position. Theefoe, the infomation of oto position is an essential fo the pope switching opeation. By synchonizing the appopiate oto position with the exiting cuent in one phase the effectiveness of SRM can be achieved.[] Contol method of switch-off angle is intoduced fo vaiable load, which is based on two command signals fo switching-on and switching-off angle independently. Accoding to the moto speed and load condition, a pope switching-on angle θ on is set at the coss point of negative slope of the senso signal and the switching-on command signal Von as follows as V on θon = ( θo θa ) + θ a () Vmax The maximum switching-on angle is in the minimum inductance egion. Thus, a fast build up of cuent is possible at the ated load. The minimum switching-on angle is in the inceasing egion of inductance. Theefoe a smooth build up of cuent is possible at a light load with a smooth toque poduction. Similaly, the delay angle is set at the coss point of positive slope of the signal and the switching off command signal V off as V θ θ θ θ () off off = ( d o ) + o Vmax In addition, the dwell angle is the inteval of switching-on and switching-off angles, which takes the fom Whee θdwell = θoff θ on ()

Intenational Jounal of Enegy and Powe Engineeing 0; (): 9- θ on is the switching-on angles θ off is the switching-off angles θ o is the Ovelap of phase flux linkages o inductances θ a is the Advance

3 Intenational Jounal of Enegy and Powe Engineeing 0; (): 9- θ on is the switching-on angles θ off is the switching-off angles θ o is the Ovelap of phase flux linkages o inductances θ a is the Advance angle θ d, θ dwell is the Cuent dwell angle In SRM moto, though expeiments showed that, with each diffeent speed and toque of diffeent loads, thee will be a coesponding switch-off, at which the ipple of toque is the smallest.. Switch-Off Contolle fo SRM This Section will pesent the design of fuzzy logic contolle and then develop a new contolle by means of Adaptive Neual Fuzzy Infeence Systems (ANFIS).. Fuzzy Logic Contolle Fuzzy logic contolles ae the one which is mainly used in system contol fo industies. The fuzzy logic contolle has the following main functions as follows Figue. Block diagam of Fuzzy Logic Contolle system () Fuzzification is the pocess of conveting eal scala value in to membeship fuzzy set values. In this pape Gaussian membeship function is used fo inputs and tiangula membeship function fo output. () Infeence: The fuzzy infeence decides how to pocess ules using fuzzy input. The inputs fo the fuzzy contolle will be eo and change in eo. The contol signals will vay accoding to eo and change in eo. Once the fuzzy contolle eceives input the ule base is evaluated []. () Rule base Design: Basically a ule base is a linguistic contolle, which is designed using IF THEN statements. Hee we have two input conditions and one output esponse. Based on that, ules ae designed fo pope contol of the system. The Figue shows the ule base design fo the speed contol of Switched Reluctance Moto []. () Defuzzification: Defuzzification is the pocess of conveting the degees of membeship of output linguistic vaiables with thei linguistic tems in to cisp values. The defuzzification method used in this pape is cente of aea which will change the switch-off angle of the contolle accodingly. Figue. Rule data base fo Fuzzy Logic Contolles.. ANFIS Contolle The Adaptive Neual Fuzzy Infeence Systems is a kind of neual netwok which is based on Takagi Sugeno fuzzy infeence system. Since it integates both neual netwoks and fuzzy logic pinciples, it has ability to captue the advantages of both in a single famewok. Its infeence system coesponds to a set of fuzzy IF THEN ules that it is capable of leaning to appoximate nonlinea functions []. Hence, ANFIS is egaded to be a univesal estimato []. Fuzzy if-then ules and how the adial basis function netwok elate to this kind of simplified ANFIS Fo simplicity, we assume the fuzzy infeence system unde consideation has two inputs e ω and de / dt and one ω output f. Suppose that the ule base includes two fuzzy if-then ules of Takagi and Sugeno s type [6, ]. In this study using set of fuzzy ule, such as: Rule : if x is A and yis B, then f = px + q y +, Rule : if x is A and yis B, then f = p x + q y + Rule : if x is A and yis B, then f = p x + q y + Rule : if x is A and yis B, then f = p x + q y + Rule : if x is A and yis B, then f = p x + q y +. Fo the taining of the netwok, thee is a fowad pass and a backwad pass. We now look at each laye in tun fo the fowad pass. The fowad pass popagates the input vecto though the netwok laye by laye. In the backwad pass, the eo is sent back though the netwok in a simila manne to back-popagation. The coesponding equivalent ANFIS achitectue is in as Figue. Figue. ANFIS achitectue

4 Liu Zhi Jian et al.: Switching-off Angle Contol fo Switched Reluctance Moto Using Adaptive Neual Fuzzy Infeence System The node functions in the same laye ae of the same function family as descibed below: - Laye : Evey node in this laye is a squae node with a node function. O i A ( x) = µ () Whee i=, = ; x is the input to node i and A is the linguistic label (small, lage, etc.); associated with this node function. In othe wods O is the membeship function of i A, and it specifies the degee to which the given x satisfies the quantifie A. In this study we choose () to be bell-shaped with maximum equal to and minimum equal to 0, such as () and (6). ( x) = ( eω ) = µ µ A A bi i de µ µ eω + c + a ω ( y) = = B B bi + i deω + ci dt a i () (6) With {a i, b i, c i } is the paamete set. As the values of these paametes change, the Bell-shaped functions vay accodingly, thus exhibiting vaious foms of membeship functions on linguistic label A, B. Paametes in this laye ae efeed to as pemise paametes. - Laye : Evey node in this laye is a cicle node labeled II which multiplies the incoming signals and sends the poduct out, as in (7 ). w = µ A ( x). µ ( ) = ( ). B y µ A e ω µ B (7) w = µ A ( e ). ω µ B (8) w = µ A ( e ). ω µ B (9) w = µ A ( e ). ω µ B (0) w = µ ( e ) µ () A. ω B w = µ A ( e ). ω µ B () Each node output epesents the fiing stength of a ule, (In fact, othe T-nom opeatos that pefom genealized AND can be used as the node function in this laye). - Laye : Evey node in this laye is a cicle node labeled N. The i-th node calculates the atio of the i-th ule s fiing stength to the sum of all ules fiing stengths, as in (): w w = w + w + w + w + w () Fo convenience, outputs of this laye will be called nomalized fiing stengths - Laye : Evey node in this laye is a squae node with a node function: ( ) O = w f = w p x + q y + = w peω + q + () Whee w is the output of laye, and {p, q, } is the paamete set. Paametes in this laye will be efeed to as consequent paametes. O = w f = w pe ω + q + O = wf = w peω + q + O = wf = w peω + q + O = wf = w peω + q + O = wf = w peω + q + () - Laye The single node in this laye is a cicle node labeled E that computes the oveall output as the summation of all incoming signals, i.e, as in (6): O = f = UANF IS = w f (6) = Whee output signal of in laye is Figue... Hybid Leaning Algoithm f = θ, shown in off Fom the poposed ANFIS achitectue in Figue, it is obseved that given the values of pemise paametes, the oveall output can be expessed as a linea combinations of the consequent paametes. Moe pecisely, the output f in Figue can be ewitten as in (7). f = w f + wf + wf + wf + wf (7). Simulation Results In this section, compaative simulations ae given to veify the poposed methods and show the impoved pefomance. In the following simulations, the paametes of the moto ae

5 Intenational Jounal of Enegy and Powe Engineeing 0; (): 9- used fo ou studies ae given in Table. Table. Switched Reluctance Moto Paametes Symbol Paametes Value m Mumbe phase N s Numbe of stato poles 6 N Numbe of oto poles I max Phase cuent 0(A) U đm Phase teminal voltage 00(V) R s Phase winding esistance.(ω) L max Maximum phase inductance 0.0(H) L min Minimum phase inductance 0.0(H) s Stato pole ac 0 0 Roto pole ac 0 0 J Moment of inetia 0.00(kgm ) Bm Fiction (Nm)/(ad/s) Tload Moment load.0(nm) Then two diffeent cases ae conducted: Case : We select the mode of SRM woks at speed 00pm, the chaacteistic speed of SRM is shown in Figue. The figue 6 gives the simulation esults of the toque ipple of thee diffeent appoaches. The black line shows the toque ipple when uses PI contolle. The ed line povides the toque ipple of SRM with fuzzy logic, and the blue line is the toque ipple of SRM unde the ANFIS contolle. The esults in Figue 6 show that, the fuzzy logic contolle achieves slightly bette pefomance compaed to PI contolle. Nevetheless, the poposed method with ANFIS contolle can obtain best pefomance. 600 Speed n [pm] Figue. The chaacteistic speed of SRM woks at 00pm. Toque With PI With FUZZY Figue 6. The toque ipple of SRM woks at 00pm

6 Liu Zhi Jian et al.: Switching-off Angle Contol fo Switched Reluctance Moto Using Adaptive Neual Fuzzy Infeence System Case : We select the mode of SRM is woking at speed 000 pm, and we also use the paametes of the moto in Table. The chaacteistic speed of SRM is given as Figue 7. Figue 8 shows that the simulation esults of the toque ipple of thee diffeent appoaches. The black line shows the toque ipple when use PI contolle. The ed line povides the toque ipple of SRM with fuzzy logic, and the blue line is the toque ipple of SRM unde ANFIS contolle. Figue 8 also shows that the fuzzy logic contolle achieves slightly bette pefomance compaed to PI contolle, and the poposed method by using ANFIS contolle can obtain best pefomance. Hee, we can see that the poposed methods in this pape can povide oveall impoved pefomance compaed to othe methods. 00 Speed n [pm] Figue 7. The chaacteistic speed of SRM woks at 000 pm. Toque With PI With FUZZY Figue 8. The toque ipple of SRM woks at 000pm. Conclusion This aticle poposes hybid ANFIS algoithm, its pupose to minimize the toque ipple of the switched eluctance motos. To see the effectiveness of ANFIS contolle cleae, we also analyzed the classic contolle, fuzzy contolle and the finally develop an intelligent contolle by using Adaptive Neual Fuzzy Infeence Systems (ANFIS). Based on Matlab/Simulink softwae to simulate the esults of the study. The simulation esult showed that ANFIS can adapt with the nonlinea plant vey well, its pefomance to educe toque ipple is bette than the classic contolle and fuzzy contolle. This has poven to be obust ANFIS algoithm to minimize toque ipple of the switched eluctance motos.

7 Intenational Jounal of Enegy and Powe Engineeing 0; (): 9- Refeences [] Yasuhau Ohdachi, Yoshihio Kawase, Yutaka Miua and Yoi Hayashi, Optimum design of switched eluctance motos using dynamic finite element analysis, IEEE Tans Magnet, Vol., No., pp.0 06, 997. [] Peez GE, Otiz PM, Ramiez HS, Passivity-based contol of switched eluctancemotos with nonlinea magnetic cicuits, IEEE Tans Contol System Technol, Vol., No., pp.9 8, May 00. [] McCann R, Islam M, Application of a sliding-mode obseve foposition and speed estimation in switched eluctance moto dives, IEEE Tans Ind Appl, Vol. 7, No., pp. 8, 00. [] Yamai H, Kaneda M, Optimal switched eluctance moto dive fohydaulic pump unit, Poc IEEE Int Conf Ind Appl, Italy, Rome, pp. 6, 000. [] R. Kishnan, Switched Reluctance Moto Dives, Boca Raton, FL:CRC Pess, 00. [6] T. S. Chuang, and C. Pollock, Robust speed contol of a switched eluctance vecto dive using vaiable stuctue appoach, IEEE Tans. on Industial Electonics, vol., no. 6, pp , Dec [7] Fehat Daldaban, Nuettin Ustkoyuncu, Keim Guney, Phase inductance estimation fo switched eluctance moto using adaptive neuo-fuzzy infeence system, Enegy Convesion and Management, pp.8 9, 006. [8] Wen Ding, Deliang Liang, Modeling of a 6/ Switched Reluctance Moto Using AdaptiveNeual Fuzzy Infeence System, IEEE Tansactions on Magnetics, Vol., No. 7, pp , July 008. [9] Hany M. Hasanien, Speed Contol of Switched Reluctance Moto Using an Adaptive Neuo-fuzzy Contolle, Poceedings of the Wold Congess on Engineeing 0 Vol II, pp , July 0. [0] Ahmed Tahou, Hamza Abid, Abdel Ghani Aissaoui, Adaptive Neuo-Fuzzy Contolle of Switched Reluctance Moto, Sebian Jounal Of Electical Engineeing Vol., No., pp, June 007. [] Jin-Woo Ahn Toque Contol", book edited by Moulay Taha Lamchich, ISBN , pp0-, Febuay, 0. [] S. Viayan, S. Paamasivam, R. Aumugam, S. S. Dash, K. J. Poonaselvan, "A Pactical appoach to the Design and Implementation of Speed Contolle fo Switched Reluctance Moto Dive using Fuzzy Logic Contolle, Jounal of Electical Engineeing, vol.8, No., pp. 9-6, 007. [] Gamal M. Hashem, Hany M. Hasanien, Speed Contol of Switched Reluctance Moto Based on Fuzzy Logic Contolle, oceedings of the th Intenational Middle East Powe Systems Confeence (MEPCON 0), Caio Univesity, Egypt, pp.88-9, 00. [] J-S-R.Jang, C-T.Sun, E.Mizutani, Neuo-fuzzy anh soft computing, Pentice Hall Uppe Saddle ive, NJ 078, 997. [] A. Abaham, "Adaptation of Fuzzy Infeence System Using Neual Leaning, Spinge-Velag Belin Heidelbeg, 00. [6] T. Takagi and M. Sugeno, Fuzzy identification of systems and its applications to modeling and contol, IEEE tansactions on systems, man, and cybenetics, Vol. SMC-, No., 98, pp. 6, Januay/Febuay. [7] Chee-mun Ong, Dynamic Simulation of Electic Machiney using Matlab Simulink, Pentice Hall PTR, 998.

Design of an LLC Resonant Converter Using Genetic Algorithm

Design of an LLC Resonant Convete Using Genetic Algoithm H. H. Nien, C. K. Huang, S. K. Changchien, C. H Chan Dept. of Electical Engineeing, Chienkuo Technology Univesity E-mail: nien@ctu.edu.tw Dept,