Indonesian Journal of Electrical Engineering and Computer Science V ol. 2, No . 3, J une 2016, pp . 486 500 DOI: 10.11591/ijeecs .v2.i3.pp486-500 486 Routh Appr o ximation: An Appr oac h of Model Or der Reduction in SISO and MIMO Systems D . K. Sambariy a* and Omveer Sharma Depar tment of Electr ical Engg., Rajasthan T echnical Univ ersity Ra w atbhata Road, K ota, 324010, India *Corresponding author , e-mail: dsambar iy a 2003@y ahoo .com Abstract In this paper the Routh Appro ximation method is e xplored f o r getting the reduced order model of a higher order model. The reduced order modeling of a large system is necessar y to ease the analysis of the system. The approach is e xamined and compared to single-input single-output (SISO) and m ulti-input m ulti- output (MIMO) systems . The response compar ison is considered in ter ms of step response par ameters and g r aphical compar isons . It is repor ted that the reduced order model using proposed Routh Appro ximation (RA) method is almost similar in beha vior to that of with or iginal systems . K e yw or ds: Model order reduction (MOR), Single-input single-output (SISO), Multi-input m ulti-outp ut (MIMO), Routh Appro ximation method. Cop yright c 2016 Institute of Ad v anced Engineering and Science . All rights reser v ed. 1. Intr oduction The analysis of high order systems (HOS) is gener ally v er y m uch complicated and costly . On other hand it became easy to analysis of lo w er order system [1, 2]. The reduced models f or the or iginal high order system is achie v ed b y using mathematical optimization procedures or simplification procedures based on ph ysical consider ations [3]. Thus analysis , synthesis and sim ulation of reduced lo w order systems is easier and pr acticab le as compared to it’ s high order systems [4]. An approach to get reduced order model of a higher order system using time- moments method is presented b y [5]. The reduced model ha v e a prob lem of stability because of mathematical appro ximation in model reduction technique . The reduced model m a y be unstab le e v en though the high order system is stab le [6]. The instability prob lem of reduced models w as studied b y Hutton [7], Shamash [8], Gut- man et. al. [9] and W an [10]. Some method based on stability cr iter ion and other not based on stability cr iter ion b ut the reduced model f or a stab le high order system (HOS) is alw a ys stab le [11, 12]. Diff erent methods giv e diff erent approach some giv es batter result in r ise time , some giv es batter results in settling time [13]. The combin ation of these methods giv es batter results . The reduced model using combination of methods is near ly wit h its higher order system. The com- bination of Routh appro ximation and par ticle s w ar m optimization (PSO) is presented in [14]. The concept of preser v ation of stability is presented in [15]. The diff erentia tion method f or reduction of systems is prese nted in [16]. The diff erentiation method is used to der iv e reduced order model of single machine infinite b us po w er system in [17]. The application of Routh stability algor ithm is presented in [18, 19]. The application of soft computing techniques ha v e been presented in liter ature in the field of model order reduction [20]. The concept used is minimization of integ r al squared err or using bat algor ithm [20]. The application of fire fly algor ithm in model order reduction is presented in [21]. The application of par ticle s w ar m optimization (PSO) is pr esented in [22]. The application of Routh appro ximation with Cuc k oo search algor ithm f or model order reduction is presented in [23]. The h ybr id application of stability equation method with self-adaptiv e bat algor ithm to reduce po w er system to a reduced model is presented in [24]. Receiv ed J an uar y 1, 2016; Re vised Apr il 24, 2016; Accepted Ma y 10, 2016 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 487 In this paper the application of Routh appro ximation method is presented f or der iving reduced order model of the higher order L TI systems which include s bechmar k prob lems . The statement of prob lem is presented in section 2.. The detailed procedur al steps on Routh appro x- imation method are included in section 3.. The systems under consid er ation and their reduced order models are presented in section 4.. The results of or iginal system and reduced models are subjected to step input and compared in this section. Finally the man uscr ipt is concluded in section 5. and f ollo w ed b y ref erences . 2. Pr ob lem Form ulation Consider a high order tr ansf er function of a system represented as in Eq. 1. G ( s ) = n 1 P i =0 b i s i n P i =0 a i s i (1) where , the G ( s ) represents a high order sy st em with the order of n . The p ur pose of man uscr ipt is to reduce the order of such high order system to r . The reduced order model ma y be represented as in Eq. 2. R ( s ) = r 1 P j =0 d j s j r P j =0 c j s j (2) where , a i , b i , c j and d j are the scalar constants of or iginal high order system and the reduced order system. The obje ctiv e is to find a reduced r th order system model R ( s ) such that it retains the impor tant proper ties of G ( s ) f or the same types of inputs . 3. Re vie w on Routh appr o ximation This method n umber of useful proper ties lik e if or iginal system is stab le then reduce model will be stab le ,con v erge monotonically of or iginal system in ter ms of step and impulse response . By increase order of appro ximation poles and z eros of the appro ximants mo v e to w ards the poles and z eros of the or igina l. In this method Routh T ab le f or or iginal system is use to constr uct the appro ximate in a manner that it will stab le f or stab le or iginal system [22]. 3.1. Description of Method G ( s ) = b n s ( n 1) + b n s ( n 2) + : : : + b 1 a n s n + a ( n 1) s ( n 1) + : : : + a 0 (3) By taking reciprocal of Eq. 3 and sho wn in Eq. 4 ^ G ( s ) = 1 s G 1 s = b 1 s ( n 1) + : : : + b n a 0 s n + a 1 s ( n 1) + : : : + a n (4) If s i , represents the i th pole/z eros of the or iginal system then 1 =s i , the i th poles/z eros of the reciprocal system. Routh Appro ximation: An Approach of Model Order Reduction ... (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
488 ISSN: 2502-4752 3.2. Alpha-Beta e xpansion The tr ansf er function of Eq. 4 can be e xpanded in the canonical f or m as presented in Eq. 5. ^ G(s) = 1 F 1 (s)+ 2 F 1 (s)F 2 (s)+ 3 F 1 (s)F 2 (s)F 3 (s) + : : : + n F 1 (s)F 2 (s)F 3 (s) : : ::: F n (s) = n P i=1 i Q i i =1 F j ( s ) (5) The F i ( s ) can be defined b y the contin ued fr action e xpansions as sho wn in Eq. 6. F i ( s ) = 1 i s + 1 i +1 s + 1 i +2 s +  . . . n 1 s + 1 n s (6) In Routh T ab le 1, the first tw o ro ws of tab le are f or med b y coefficients of the denominator of function ^ G ( s ) and taking assumption that the entr ies of a 0 J = a I ( J 1) = 0 f or j > n . a i +1 0 = a i 1 2 i a i 2 a i +1 2 = a i 1 4 i a i 4 . . . a i +1 n i 1 = a i 1 n i i a i n i (7) where , Eq. 7 stands f or i = 1 ; 2 ; 3 ; : : : ; n 1 . If the v alue of n i as odd, the last ter m in Eq. 7 is replaced b y as sho wn in Eq. 8. a i +1 n i 1 = a i 1 n i 1 (8) F or i = 1 ; 2 ; 3 ; : : : ; n , the marginal entr ies f or i are calculated as in Eq. 9. i = a i 1 0 a i 0 (9) The i coefficients of the canonical f or m Routh tab le are deter mined using coefficients of the n umer ator of ^ G ( s ) and is sho wn in Eq. 10. i = b i 0 a i 0 (10) b i +2 j 2 = b i j i a i j (11) The Routh T ab le 1 is equiv alent to constr uction of f ollo wing finite contin ued fr action e xpansion as sho wn in Eq. 12. ^ D ( s ) = 1 s + 1 2 s + 1 3 s + . . . n 1 s + 1 n s (12) It could be easy to sa y that the system wit h all par ameters being positiv e ref ers to an asymptot- ically stab le system [1]. IJEECS V ol. 2, No . 3, J une 2016 : 486 500 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 489 T ab le 1. Alpha tab le I st ro w a 0 0 = a 0 a 0 2 = a 2 a 0 4 = a 4 2 nd ro w a 1 0 = a 0 a 1 2 = a 3 a 1 4 = a 5 1 = a 0 0 a 1 0 a 2 0 = a 0 2 1 a 1 2 a 2 2 = a 0 4 1 a 1 4 a 2 4 = a 0 2 1 a 1 6 2 = a 1 0 a 2 0 a 3 0 = a 1 2 2 a 2 2 a 3 2 = a 1 4 2 a 2 4 : : : 3 = a 2 0 a 3 0 a 4 0 = a 2 2 3 a 3 2 a 4 2 = a 2 4 3 a 3 4 : : : 4 = a 3 0 a 4 0 a 5 0 = a 3 2 4 a 4 2 : : : : : : 5 = a 4 0 a 5 0 a 6 0 = a 4 2 5 a 5 2 : : : : : : 6 = a 5 0 a 6 0 : : : : : : : : : T ab le 2. Beta tab le I st ro w b 1 0 = b 1 b 1 2 = b 3 b 1 4 = b 5 2 nd ro w b 2 0 = b 2 b 2 2 = b 4 b 2 4 = b 6 1 = b 1 0 a 1 0 b 3 0 = b 1 2 1 a 1 2 b 3 2 = b 1 4 1 a 1 4 : : : 2 = b 2 0 a 2 0 b 4 0 = b 2 2 2 a 2 2 b 4 2 = b 2 4 1 a 2 4 : : : 3 = b 3 0 a 3 0 b 5 0 = b 3 2 3 a 3 2 : : : : : : 4 = b 4 0 a 4 0 b 6 0 = b 4 2 4 a 4 2 : : : : : : 5 = b 5 0 a 5 0 : : : : : : : : : 3.3. Routh Con ver g ent The reduced k th order tr ansf er function as ^ R k ( s ) f or an or iginal tr ansf er function G ( s ) is der iv ed b y tr uncating the e xpansion and r ational arr angement of the results . The ter ms appear ing k +1 ; : : : ; n and k +1 ; : : : ; n are eliminated using e xpansion. In this w a y the the resultant is dependent on the first k-ter ms [7, 25]. Assuming a set of k-functions , which are defined b y G i;k f or i = 2 ; 3 ; : : : ; k and is repre- sented as in f ollo wing Eq. 13 [25]. G i;k ( s ) = 1 i s + 1 i +1 s + 1 i +2 s + . . . k 1 s + 1 k s (13) The abo v e method possess slight modification f or i = 1 . The I st ter m in the contin ued fr action e xpansion is 1 + 1 s instead of 1 s . In this w a y , the k th con v ergent ma y be giv en b y as in Eq. 14 [1, 25]. ^ R k (s) = 1 G 1 ; k (s)+ 2 G 1 ; k (s)G 2 ; k (s)+ + k G 1 ; k (s)G 2 ; k (s) G k ; k (s) = k P i =1 i I Q i =1 G i;k ( s ) (14) The A k ( s ) is the denominator of the k th con v ergent while B k ( s ) represents the n umer ator of it. In Routh Appro ximation: An Approach of Model Order Reduction ... (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
490 ISSN: 2502-4752 this w a y , the k th con v ergent ma y be represented as in f ollo wing Eq. 15 [26]. A 1 ( s ) = 1 s + 1 B 1 ( s ) = 1 A 2 ( s ) = 1 2 s 2 + 2 s + 1 B 2 ( s ) = 2 1 s + 2 A 3 ( s ) = 1 2 3 s 3 + 2 3 s 2 + ( 1 + 3 ) s + 1 B 3 ( s ) = 2 3 1 s 2 + 3 2 s + ( 1 + 3 ) A k ( s ) = k sA k 1 ( s ) + A k 2 ( s ) B k ( s ) = k sB k 1 ( s ) + B k 2 ( s ) + k A 1 ( s ) = 1 ; B 1 ( s ) = 0 A 0 ( s ) = 1 ; B 0 ( s ) = 0 (15) The ^ R k ( s ) represents the appro ximation of ^ G ( s ) with preser ving the frequency beha viour . The k th appro ximate can be der iv ed b y consider ing the reciprocal of ^ R k ( s ) as sho wn in Eq. 16 [25]. R k ( s ) = 1 s ^ R k 1 s (16) 3.4. Algorithm of Routh appr o ximation The f ollo wing steps can be f ollo w ed f or deter mining the reduced order of a high order system. (i) Initially deter mine the reciprocal ( ^ G ( s ) ) of the full order system G ( s ) (ii) Der iv e the elements (iii) Deter mine k th con v ergent using ^ R k ( s ) = B k ( s ) A k ( s ) (iv) Reciprocate ^ R k ( s ) f or k th order Routh appro ximation R k ( s ) . 4. Results and Discussions 4.1. Example-1: SISO Consider ing the 8 th order system presented in Shamash, 1975 [8] and presented in Eq. 17. G ( s ) = 18 s 7 + 514 s 6 + 5982 s 5 + 36380 s 4 + 122664 s 3 + 222088 s 2 + 185760 s + 40320 s 8 + 36 s 7 + 546 s 6 + 4536 s 5 + 22449 s 4 +67284 s 3 + 118124 s 2 + 109584 s + 40320 (17) The reduced 2 nd order and 3 r d order models are presented in Eq. 18 and Eq. 19, respectiv ely using Routh Appro ximation method. R 2 ( s ) = 1 : 990 s + 0 : 432 s 2 + 1 : 174 s + 0 : 432 (18) R 3 ( s ) = 4 : 968 s 2 + 4 : 331 s + 0 : 940 s 3 + 2 : 545 s 2 + 2 : 555 s + 0 : 940 (19) The step response compar ison of the or iginal system [ 8] and it’ s reduced 2 nd and 3 r d order models are g r aphically compared in Fig. 1. It can be obser v ed that the stability of the system that of with reduced models are retain ed e xcept slight v ar iation in r ise time , settling time , peak v alue and peak time as included in T ab le 3. Since , the impor tant proper ties of the higher order system are preser v ed in it’ s reduced ( 2 nd order) system, consequently the mathematical ease is increased g reatly . IJEECS V ol. 2, No . 3, J une 2016 : 486 500 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 491 0 5 10 15 20 0 0.5 1 1.5 2 2.5 Time (s) Amplitude  Step response     Original (Shamash, 1975) RA−MOR: 2 nd  order (Proposed) RA−MOR: 3 rd  order (Proposed) Figure 1 . Step response of the or iginal 8 th order system [8] and its reduced model R 2 ( s ) (Eq. 18) and R 3 ( s ) (Eq. 19) using the Routh appro ximation method T ab le 3. Step response compar ision of or iginal system in Example-1, with reduced models using Routh Appro ximation T r ansf er Rise Settling P eak P eak Function Time Time V alue Time Or iginal: G 8 ( s ) [8] 0.0569 4.8201 2.2035 0.4493 MOR-RA: R 2 ( s ) 0.5514 8.7327 1.5717 2.3235 MOR-RA: R 3 ( s ) 0.1973 7.0765 2.1128 1.1637 4.2. Example-2: SISO Consider ing the 4 th order system presented in Hw ang,1996 [27] and presented in Eq. 20. G ( s ) = 10 s 4 + 82 :s 3 + 264 s 2 + 396 s + 156 2 s 5 + 21 s 4 + 84 :s 3 + 173 s 2 + 148 s + 40 (20) The reduced 2 nd order and 3 r d order models are presented in Eq. 21 and Eq. 22, respectiv ely using Routh Appro ximation method. R 2 ( s ) = 1 : 990 s + 0 : 432 s 2 + 1 : 174 s + 0 : 432 (21) R 3 ( s ) = 4 : 968 s 2 + 4 : 331 s + 0 : 940 s 3 + 2 : 545 s 2 + 2 : 555 s + 0 : 940 (22) The step response compar ison of the or iginal system [27] and it s reduced 2 nd and 3 r d order models are g r aphically compared in Fig. 2. It can be obser v ed that the stability of the system that of with reduced models are retain ed e xcept slight v ar iation in r ise time , settling time , peak v alue and peak time as included in T ab le 4. In this case the r ise-time of the or iginal, 2 nd and 3 r d order reduced models are 2.7456, 2.6830 and 2.5549 seconds , respectiv ely . The diff erence in the r ise times is minimal and is enough to pro v e similar ity of the or iginal and reduced models . The other step response data are enlisted in T ab le 4. 4.3. Example-3: SISO Consider ing the 7 th order system presented in J amshidi, 1983 [28] and presented in state-space f or m b y Eq. 23 - 24 and in tr ansf er function b y Eq. 25. It represents the SMIB po w er Routh Appro ximation: An Approach of Model Order Reduction ... (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
492 ISSN: 2502-4752 0 5 10 15 20 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Time (s) Amplitude  Step response     Original (Hwang, 1996) RA−MOR: 2 nd  order (Proposed) RA−MOR: 3 rd  order (Proposed) Figure 2. Step response of the or iginal 4 th order system [27] and its reduced model R 2 ( s ) (Eq. 21) and R 3 ( s ) (Eq. 22) using the Routh appro ximation method T ab le 4. Step response compar ision of or iginal system in Example-2, with reduced models using Routh Appro ximation T r ansf er Rise Settling P eak P eak Function Time Time V alue Time Or iginal: G 8 ( s ) [27] 2.7456 5.4346 3.8944 10.4392 MOR-RA: R 2 ( s ) 2.6830 3.9639 3. 9748 6.4974 MOR-RA: R 3 ( s ) 2.5549 5.5932 3. 8992 15.6589 system and the details are giv en in [29]. _ x ( t ) = 2 6 6 6 6 6 6 6 6 4 0 : 58 0 0 0 : 269 0 0 : 2 0 0 1 0 0 0 1 0 0 0 5 2 : 12 0 0 0 0 0 0 0 377 0 0 0 : 141 0 0 : 141 0 : 2 0 : 28 0 0 0 0 0 0 0 0 : 0838 2 173 66 : 7 116 40 : 9 0 66 : 7 16 : 7 3 7 7 7 7 7 7 7 7 5 x ( t ) + 2 6 6 6 6 6 6 6 6 4 1 0 1 0 1 0 1 3 7 7 7 7 7 7 7 7 5 u ( t ) (23) y ( t ) = 1 1 1 1 0 1 0 x ( t ) (24) G ( s ) = 2 s 6 + 420 : 4 s 5 + 9435 s 4 + 1 : 39 10 5 s 3 +4 : 663 10 5 s 2 + 4 : 342 10 5 + 1 : 877 10 5 s 7 + 23 : 48 s 6 + 331 : 7 s 5 + 2640 s 4 + 1 : 757 10 4 s 3 +5 : 165 10 4 s 2 + 3 : 534 10 4 s + 1 : 729 10 4 (25) The reduced 2 nd order and 3 r d order models are presented in Eq. 26 and Eq. 27, respectiv ely using Routh Appro ximation method. R 2 ( s ) = 10 : 085 s + 4 : 360 s 2 + 0 : 821 s + 0 : 402 (26) IJEECS V ol. 2, No . 3, J une 2016 : 486 500 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 493 T ab le 5. Step response compar ision of or iginal system in Example-3, with reduced models using Routh Appro ximation T r ansf er Rise Settling P eak P eak Function Time Time V alue Time Or iginal: G 8 ( s ) [28, 29] 0.1126 5.9294 16.0310 0.4307 MOR-RA: R 2 ( s ) 1.1998 7.4456 13.92 03 3.3225 MOR-RA: R 3 ( s ) 0.5740 9.0915 13.22 69 2.1099 R 3 ( s ) = 29 : 318 s 2 + 27 : 948 s + 12 : 081 s 3 + 3 : 26 s 2 + 2 : 275 s + 1 : 113 (27) 0 5 10 15 20 0 2 4 6 8 10 12 14 16 18 Time (s) Amplitude  Step response     Original (Jamshidi, 1983) RA−MOR: 2 nd  order (Proposed) RA−MOR: 3 rd  order (Proposed) Figure 3. Step response of the or iginal 7 th order system [28, 29] and its reduced model R 2 ( s ) (Eq. 26) and R 3 ( s ) (Eq. 27) using the Routh appro ximation method In this e xample , the considered system is from the po w er system engineer ing. The or ig- inal syste m and it’ s reduced 2 nd and 3 r d order models are subjected to step signal and super- imposed to compare the responses in Fig. 3. It can be seen that the response due to or iginal system is ha ving more oscillations as compared to that of with the reduced order models . The step response inf or mation of these responses are enlisted in T ab le 5. 4.4. Example-4: SISO Consider ing the 9 th order boiler system represented in tr ansf er function f or m in Eq. 28 as presented in [26, 30]. The reduced 2 nd order and 3 r d order models are presented in Eq. 29 and G ( s ) = 146 : 4 s 8 + 9 : 81 10 4 s 7 + 5 : 999 10 7 s 6 + 3 : 206 10 10 s 5 + 3 : 582 10 12 s 4 +1 : 113 10 14 s 3 + 1 : 154 10 15 s 2 + 3 : 971 10 15 s + 3 : 063 10 15 s 9 + 659 : 8 s 8 + 4 : 136 10 5 s 7 + 2 : 13 10 8 s 6 + 2 : 422 10 10 s 5 + 8 : 737 10 11 s 4 +1 : 523 10 13 s 3 + 1 : 221 10 14 s 2 + 3 : 636 10 14 s + 2 : 406 10 14 (28) Routh Appro ximation: An Approach of Model Order Reduction ... (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
494 ISSN: 2502-4752 T ab le 6. Step response compar ision of or iginal system in Example-4, with reduced models using Routh Appro ximation T r ansf er Rise Settling P eak P eak Function Time Time V alue Time Or iginal: G 9 ( s ) [26, 30] 0.5432 2.2753 12.6986 4.5555 MOR-RA: R 2 ( s ) 0.6375 2.9668 13.28 09 1.6504 MOR-RA: R 3 ( s ) 0.2577 2.4749 12.69 20 4.5431 Eq. 30, respectiv ely using Routh Appro ximation method. R 2 ( s ) = 35 : 448 s + 27 : 343 s 2 + 3 : 246 s + 2 : 148 (29) R 3 ( s ) = 90 : 835 s 2 + 319 : 054 s + 246 : 1 s 3 + 9 : 662 s 2 + 29 : 214 s + 19 : 331 (30) The considered 9 th order system is a pr actical boiler system as presented in [26, 30]. The 0 1 2 3 4 5 6 7 8 0 5 10 15 Time (s) Amplitude  Step response     Original (Salim, 2009) RA−MOR: 2 nd  order (Proposed) RA−MOR: 3 rd  order (Proposed) Figure 4. Step response of the or iginal 9 th order system [26, 30] and its reduced model R 2 ( s ) (Eq. 29) and R 3 ( s ) (Eq. 30) using the Routh appro ximation method system is reduced to 2 nd and 3 r d order models using Routh appro ximation method. The or iginal and reduced models are subjected to step input and the g r aphical compar ison is presented in Fig. 4. It can be seen that the or iginal and main proper ties of the or iginal higher order system are retained in it’ s reduced model responses with compar ativ ely reduced o v ershoots . The step response inf or mation are included in T ab le 6. 4.5. Example-5: MIMO A po w er p lant system can be classified as a m ultiv ar iab le large-scale system. Numerous methods of analysis and synthesis f or such processes ha v e been de v eloped, b ut the remar kab le dimensions of the model str ucture mak es their implemen tation v er y difficult. Consider ab le atten- tion has theref ore been de v oted to the prob lem of der iving reduced-order models f or such sys- tems . The siz e and comple xity of current electr ic po w er netw or ks in v olv es methods f or studying appro ximated models to in v estigate the dynamic beha viour of such system types in a more suit- ab le w a y; the methods currently used f or deter mining reduced-order dynamic models f or po w er systems in m ulti-b us , m ulti-machine fr ames are gener ally ref erred to as ”dynamic equiv alents”. IJEECS V ol. 2, No . 3, J une 2016 : 486 500 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 495 An electr ic po w er system consisting of a salient-pole synchronous gener ator connected to an infinite b us-bar is considered. T aking into account the w ell kno wn perf or mance equations of both the machine and the tr ansmission line , a v er y accu r ate non-linear mathematical model in the state-space f or m has been der iv ed. As state v ar iab les of the electr ical par t of the synchronous machine , the set of winding currents of the q d equiv alent circuit has been chosen. The se v enth- order state v ector of the or iginal sys t em consists of the stator currents i d , i q , the field circuit current i f d , tw o damping circuit currents i k q , i k d and the mechanical quantities and ! . The input v ector , in the chosen representation, consists of tw o quantities , the mechanical torque T m and the v oltage V f . As output v ar iab les , the machine v oltage VI and the mechanical state v ar iab les and ! ha v e been chosen [3]. By consider ing small v ar iations ( ) around a steady-state oper ating point, a linear model has been der iv ed. The v alues of the par ameters , steady-state w or king conditions and fur ther details on the adopted model are repor ted b y Ramamoor ty and Ar um ugan [31]. w e indicate with: Change in mechanical torque ( T m ) as input 1 Change in field v oltage ( V f ) as input 2 Change in ter minal v oltage ( V t ) as output 1 Change in po w er angle ( ) as output 2 Change in speed ( ! ) as output 3 The tr ansf er function of m ulti-input m ulti-output (MIMO) single-machine infinite-b us (SM IB) po w er system can be represented as in Eqn. 31. The tr ansf er function of the system with output V t to input T m can be represented b y G 11 ( s ) = g 11 ( s ) =d ( s ) and si,ilar ly f or others . The considered MIMO SMIB consists of six diff erent tr ansf er function with diff erent sets of input and output sig- nals . The denominator of these systems is common and represen ted b y d ( s ) in Eqn. 32. The polynomials presented in Eqn. 33 - Eqn. 38, are the n umer ators of diff erent tr ansf er functions due to diff erent sets of input and output signals . G ( s ) = 2 4 g 11 ( s ) g 21 ( s ) g 12 ( s ) g 22 ( s ) g 13 ( s ) g 23 ( s ) 3 5 d ( s ) (31) d ( s ) = 8 > > < > > : s 7 + 258 : 7 s 6 + 4 : 31 10 5 s 5 +4 : 835 10 7 s 4 + 1 : 853 10 9 s 3 +2 : 54 10 10 s 2 + 5 : 973 10 10 s +1 : 886 10 10 (32) g 11 ( s ) = 8 < : 12 : 41 s 4 + 1 : 213 10 4 s 3 2 : 866 10 6 s 2 3 : 325 10 8 s 6 : 404 10 9 (33) g 12 ( s ) = 8 < : 12 : 41 s 5 + 1 : 213 10 4 s 4 2 : 866 10 6 s 3 3 : 325 10 8 s 2 6 : 404 10 9 s + 0 : 0006087 (34) g 13 ( s ) = 8 > > < > > : 0 : 2005 s 6 + 47 : 88 s 5 +3 : 928 10 4 s 4 + 5 : 122 10 6 s 3 +2 : 288 10 8 s 2 + 3 : 434 10 9 s +5 : 492 10 9 (35) Routh Appro ximation: An Approach of Model Order Reduction ... (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.