Full Paper Int. J. on Recent Trends in Engineering and Technology, Vol. 7, No.

2, July 2012

Large Scale Linear MIMO System Model Order Reduction using Artificial Bee Colony Optimization Algorithm
1
1,2

G.Vasu and 2P.Murari

Assistant professor, Dept. of Electrical Engineering, S.V.P Engineering College Visakhapatnam, A.P , India E-mail: vasuganji_kvs@yahoo.com and murarinitt@gmail.com quency domain [5],[20-21]. Unfortunately, very few methods are available for the reduction of Linear MIMO systems. Many of the methods available in the international literature for SISO can be easily extended for the reduction of linear MIMO systems. Earlier a method for the reduction of MIMO system was given by Bendekas et.al.,[9], which is a mixed method of time moment matching and Pade approximates. Taiwo et.al.,[10] developed methods based upon Pade approximated generalized time moments. Chen [11] has given a method based on matrix continued fraction technique. The Routh approximation method given by Hutton [12], was extended to MIMO systems by Sinha [13]. Numerous methods of order reduction are available in the literature [3],[4],[6] which are based on the minimization of the Integral square error (ISE). However, a common feature in these methods is that the values of the denominator coefficients of the reduced order model (ROM) are chosen arbitrarily by some stability preserving method such as Dominant pole, Routh approximation, Stability equation methods etc., and then the numerator coefficients of the ROM are determined by minimizing the ISE. Hewitt and Luss [14] suggested a technique, in which both the numerator and denominator coefficients are considered to be free parameters and are chosen to minimize the ISE in Impulse or Step responses. In spite of several methods available, no single approach always gives the best results for all systems. Recently, The Artificial Bee Colony Algorithm (ABC) appeared as promising evolutionary technique for handling the optimization problems, which is based on the intelligent foraging behaviour of honey bee swarm, proposed by Karaboga in 2005[15-16]. This swarm algorithm is very simple and flexible when compared to the other existing swarm based algorithms. It can be used for solving uni-model and multimodel numerical optimization problems. This algorithm uses only common control parameters such as colony size and maximum cycle number. It is a population based search procedure in which individuals called food positions(reduced order model coefficients) are modified using the artificial bees with time and the aim of the bees is to discover the places of food sources(reduced order model) with high nectar amount and finally choose source with the highest nectar amount among the other resources. 13

Abstract: In this paper, a computationally simple approach is proposed for order reduction of large scale linear dynamic MIMO (Multi input-Multi output) system using Artificial Bee Colony (ABC) optimization algorithm. The method is based on minimization of the Integral square error (ISE) between the transient responses of original and reduced order models pertaining to unit step input. The reduction procedure is simple, efficient and computer oriented. Stability and accuracy of the reduced order model is always assured in the proposed method. The validity of algorithm is illustrated with the help of two numerical examples considered from the literature and the results are compared with other recently published reduction techniques to show its superiority. Keywords: Model Order Reduction, MIMO Systems, Artificial Bee Colony Optimization, Integral Square Error, Stability.

I. INTRODUCTION The modelling of complex dynamic systems is one of the most important subjects in Engineering and Science. The mathematical procedure of system modelling often leads to higher order differential equations which are too complicated to use either for analysis or controller synthesis. So approximation (reduction) procedures based on physical considerations (or) mathematical approaches are used to achieve simpler models for the original one. These reduction techniques are well-established part of the control system designers toolkit. At the forefront, these techniques have been those that deal with the linearized system models in both time and frequency domains[1-4]. Each of the more established order reduction methods has its relative merits and is selected according to the system characteristics being approximated. Among them the frequency domain techniques are developed from the overwhelming amount of design information gathered by classical methods (e.g. Nyquist, Bode and Root-Locus plots), leading to the transfer function formation. This approach is still favoured by many designers and consequently the frequency domain order reduction methods continue to be of major importance. The most desirable properties of any order reduction techniques is that preserving stability of the original model in the reduced model and matching of time responses. Several reduction methods are available in literature for reducing the order of large scale linear SISO systems in fre 2012 ACEEE DOI: 01.IJRTET.7.2.20

Full Paper Int. J. on Recent Trends in Engineering and Technology, Vol. 7, No. 2, July 2012 In the present work authors proposed a new method for order reduction of large scale linear MIMO systems using ABC optimization algorithm. In which common denominator and numerator coefficients of reduced MIMO system are determined by minimizing the integral square error (ISE) pertaining to unit step input. The reduced model preserves the stability and good matching of time response characteristics of original higher order model. The results of proposed method are compared with the recently published methods. II. ARTIFICIAL BEE COLONY(ABC) ALGORITHM ABC is a population based optimization algorithm based on intelligent behaviour of honey bee swarm [15]. In the ABC algorithm, the foraging bees are classified into three categories; Employed bees, Onlookers and Scout bees. A bee waiting on the hive for making decision to choose a food source is called an Onlooker and a bee going to the food source visited by it previously is named an Employed bee. A bee carrying out random search is called a Scout. The employed bees exploit the food source and they carry the information about the food source back to the hive and share information with onlookers. Onlooker bees are waiting in the hive at dance floor for the information to be shared by the employed bees about their discovered food sources and scouts bees will always be searching for new food sources near the hive. Employed bees share information about food sources by dancing in the designated dance area inside the hive. The nature of dance is proportional to the nectar content of food source just exploited by the dancing bee. Onlooker bees watch the dance and choose a food source according to the probability proportional to the quality of that food source. Therefore, good food sources attract more onlooker bees compared to bad ones. Whenever a food source is exploited fully, all the employed bees associated with it abandon the food source and become scout. Scout bees can be visualized as performing the job of exploration, where as employed and onlooker bees can be visualized as performing the job of exploitation. In the ABC algorithm, each food source is a possible solution(reduced order model) for the problem under consideration and the nectar amount of a food source represents the quality of the solution which further represents the fitness value(ISE). The number of food sources is same as the number of employed bees and there is exactly one employed bee for every food source. At the first step, the ABC generates a randomly distributed initial population P (C=0) of SN solutions (food sources position), where SN denotes the size of population. Each solution (food sources) is a D-dimension vector. Here D is number of optimization parameters. After initialization, the population of the position (solution) is subjected to repeated cycles, of the search process of the employed bees, onlookers and scouts. The production of new food source position is also based on comparison process of food sources position. However, in the model, the artificial bees 14 2012 ACEEE DOI: 01.IJRTET.7.2.20 do not use any information in comparison. They randomly select a food source position and produce a modification on the existing, in their memory as described in Eq.(5) provided that the nectar amount of the new source is higher than that of the previous one of the bee memorizes the new position and forgets the old position. Otherwise she keeps the position of the previous one. An onlookers bees evaluate the nectar information taken from all employed bees and choose a food source depending on the probability value associated with that food source , calculated by the following equation: (1) Where is the fitness value of the solution evaluated by its employed bee, which is proportional to the nectar amount of food source in the position and SN. In this way, the employed bees exchange their information with the onlookers. In order to produce a new food position from the old one, the ABC uses following expression (2): (2) Where and are randomly chosen indexes. Although k and l are determined randomly, they have to be different from . is a random number between [-1, 1]. It controls the production of neighbour food source position around and the modification represents the comparison of the neighbour food positions visualized by the bee. Equation (2) shows that as the difference between the parameters of the decreases, the perturbation on the position decreases too. Thus, as the search approaches to the optimum solution in the search space, the step length is adaptively reduced. If its new fitness value is better than the best fitness value achieved so far then the bee moves to this new food source abandoning the old one, otherwise it remains in its old food source. When all employed bees have finished this process, they share the fitness information with the onlookers, each of which selects a food source according to probability given in Eq. (1). With this scheme, good food sources will get more onlookers than the bad ones. Each bee will search for better food source around neighbourhood path for a certain number of cycles (limit), and if the fitness value has not improve then that bee becomes scout bee and discovers a new food source replaced by . This operation can be defined as (3)

Full Paper Int. J. on Recent Trends in Engineering and Technology, Vol. 7, No. 2, July 2012 functions, where each entry is the transfer function relationship between each individual input, and each individual output. Let the transfer function matrix of the higher order n having p inputs and m outputs be

(4)

Or transfer matrix. The general form of taken as: of

is a is

(5) Let the transfer function matrix of the reduced order r having same number of inputs and outputs to be synthesized be :

(6)

Or transfer matrix.The general form of is taken as of

is a

(7) For getting rth order reduced transfer matrix coefficients of reduced common denominator , the and the

numerator of the reduced transfer matrix are successively determined by minimizing the Integral square error (E) between the transient responses of original [ and reduced E is given as order models using ABC algorithm and

(8)
Fig.1. Flow chart of ABC algorithm

III. DESCRIPTION OF T HE PROPOSED ALGORITHM In recent years, it was realized that the MIMO systems seems to be inevitable in accelerated evolution of high data rates applications. This intends to provide highlights of the current research topics in the field of MIMO system, to offer a snapshot of the recent advances and major issues faced today by the researchers in the MIMO related areas. Systems with more than one input and/or more than one output are known as Multi-Input Multi-Output systems, or they are frequently known by the abbreviation MIMO. This is in contrast to systems that have only a single input and a single output (SISO). For MIMO systems, with p inputs and m outputs, the transfer function matrix will contain p m transfer 2012 ACEEE DOI: 01.IJRTET.7.2.20 15

and

are the unit step response of the original and reduced order model, respectively. IV. NUMERICAL EXAMPLES Two numerical examples are chosen from the literature to show the flexibility and effectiveness of the proposed reduction algorithm than other existing methods, and the response of the original and reduced models are compared. Example- 1: Consider a 6th order two input two output system [4] described by the transfer function matrix:

Full Paper Int. J. on Recent Trends in Engineering and Technology, Vol. 7, No. 2, July 2012
TABLE I. COMPARISON OF REDUCTION METHODS

(9) Where is the common denominator polynomial. As can be verified the MIMO system is a 6th order system and is given as

The coefficients of the reduced common denominator and the numerator of the reduced transfer matrix are successively determined by minimizing the , where between the transient responses of original

[ and reduced order models using ABC algorithm. In the present study, a population size of SN=50, and maximum number of cycles ( have been used. Finally the second order reduced model obtained as:
Fig. 4(a) Comparison of step responses of and

To show the accuracy of the proposed reduction method step responses of the reduced-order models are compared with the original system and are shown in Figure 4(a)-(d). The integral square error (ISE) in between the transient parts of the original and reduced order is calculated and compared with other well known order reduction techniques available in literature which are shown in Table I.

Fig. 4(b) Comparison of step responses of

and

2012 ACEEE DOI: 01.IJRTET.7.2.20

16

Full Paper Int. J. on Recent Trends in Engineering and Technology, Vol. 7, No. 2, July 2012 By Using the proposed method, Reduced 2nd order MISO System is obtained as:

The step responses of the reduced-order models compared with the original system shown in Figure 4(e)-(f).

are

of Example-2 are

Fig. 4(c) Comparison of step responses of

and

Fig. 4(e) Comparison of step responses of

&

Fig. 4(d) Comparison of step responses of

th

and

Example 2: Consider a 6 order two input one output system [7] described by the transfer function matrix:

Fig. 4(f) Comparison of step responses of

&

V. CONCLUSION In this paper, a new method for order reduction of large scale linear dynamic MIMO systems has been presented using ABC optimization algorithm. The reduction algorithm is based on minimization of the Integral Square Error by ABC 2012 ACEEE DOI: 01.IJRTET.7.2.20 17

Full Paper Int. J. on Recent Trends in Engineering and Technology, Vol. 7, No. 2, July 2012 Optimization Technique pertaining to unit step input. Based on the results obtained for the illustrative examples, it is concluded that proposed method generates better approximations for a original higher order linear MIMO system and preserves the stability. A comparison of ISE between the proposed method and other existing reduction techniques are also shown in TABLE I, from which it is clear that proposed method is superior over the other existing techniques. REFERENCES
[1] R. Genesio and M. Milanese, A note on the derivation and use of reduced order models, IEEE Trans. Automat. Control, Vol. AC-21,No. 1, pp. 118-122, February 1976. M. Jamshidi, Large Scale Systems Modelling and Control Series, NewYork, Amsterdam, Oxford, North Holland, Vol. 9, 1983. G.Parmar et al., World Academy of science engg. and tech. Vol26 ,2007. C.B. Vishwakarma et al., Hindawi publishing corp, Modelling and simulation in engg,vol-2009, Article id-540895, 5pages. C.B.Vishwakarma and R.Prasad, Indian journal of engg.and material sciences,vol-15, 2008 pp :447-451. G.Vasu, J.Nancy Namratha, V.Rambabu, Large Scale Linear Dynamic System Reduction Using Artificial Bee Colony Optimization Algorithm, Int Journal of Electrical Engineering and Technology(IJEET), Vol 3, Issue 1, June-2012, PP130140. P.M.Rao et al., A Method for reduction of MIMO system using Interlacing property and coefficient matching , Int Journal of Computer application(0975-8887), Vol-1-No.9, 2010 pp 14-17. [8] R.Prasad and J.Pal , Use of CFE for stable reduction of linear multi variable systems, journal of the institution of engineers, Vol-72, pp:43-47, 1991. [9] Bendekal DV et al.,The Moment and Pade approximate method Applied to the order reduction of MIMO Linear system, Journal of Franklin. Vol .329, No.3, 1992. [10]Taiwo O , The Simplification of Multivariable pade approximation, International conference on control and Modeling Tehron , Iran,1990. [11] Chen C.F, International journal of control, Vol.20, No.2, 1974. [12] Hutton.M.F and Friedland,B, IEEE Trans. On Automatic Control, Vol.20, 1975, PP.329-337. [13] Sinha .N.K , Electrical letter Vol.16,No.20,1980 [14] G.D. Howitt, and R. Luus, Journal of Franklin Inst., Vol. 327, pp. 343-357, 1990. [15] D.Karaboga, An idea based on Honey Bee Swarm for Numerical Optimization, Technical ReportTR06,Erciyes university, Computer Engineering Department, OCT-2005. [16] D.Karaboga, B.Basturk, A powerful and efficient algorithm for numerical function optimization:ABC algorithm,, Journal of Global Optimization, Vol: 39, Issue 3, pp: 459-471, Springer Netherlands, 2007. [17] S.K.Mishra, S.Panda, S.Padhy, C.Ardil, World Academy of science engg.and tech, Vol-76, 2011. [18] M.G.Safonov et al., International journal of adaptive control and signal processing, Vol-2, No.4, pp.259-272,1988. [19] R.Prasad and J Pal etal., Journal of institution of engineers, India ,IE(I) journal- EI Vol-76 PP186-188, Nov-1995. [20] G.Vasu et al., Reduction of Higher Order Linear Dynamic SISO and MIMO Systems Using the Advantages of Improved Pole Clustering and PSO, Proc. of ICACEEE-2012, P0151. [21] G.Vasu et al, Reduction of Large Scale Linear Dynamic SISO and MIMO Systems Using DE Optimization Algorithm, IEEE Student Conference (SCEECS),Mar-2012,PP180-185.

[2]

[3] [4] [5] [6]

[7]

2012 ACEEE DOI: 01.IJRTET.7.2.20

18