Indonesian Journal of Electrical Engineering and Computer Science V ol. 5, No . 1, J an uar y 2017, pp . 19 32 DOI: 10.11591/ijeecs .v5.i1.pp19-32 19 Optimal design of PID contr oller f or load frequenc y contr ol using Harmon y sear c h algorithm D . K. Sambariy a* and Seema Shrangi 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 This paper deals with soft computing technique , used f or tuning PID controller . Controller tuned with a har mon y search algor ithm is used f or controlling the frequency and Tie-line po w er responses of a non reheat tw o area po w er system. Step load per turbation has been giv en in both areas sim ultan eously . The dynamic results obta ined b y the proposed controller are compared with PID controller of recent pub lished paper . The per f or mance of the controllers is sim u lated using MA TLAB/Sim ulation softw are . The results of tuned PID are compared with con v entional controller on the basis of settling-time , peak o v er-shoot and peak under-shoot. Proposed PID giv es b etter results than the con v entional controller . The compar ativ e results also tab ulated as a compar ativ e perf or mance . K e yw or ds: Load frequency control (LFC), A utomatic control error (A CE), Propor tional Intergal Der iv ativ e controller (PID), Har mon y search algor ithm (HSA). Cop yright c 2016 Institute of Ad v anced Engineering and Science . All rights reser v ed. 1. Intr oduction P o w er system oper ator has a responsibility that adequate po w er m ust deliv er to con- sumer . Reliability and economically m ust maintain. F or this control, str ategy is needed [1]. It means that there should balance betw een real and reactiv e po w er . As reactiv e po w er depends on the v oltage and real po w er on the gener ation, it means frequency [2]. The analysis f or control- ling frequency and v oltage can be tak en separ ately . In this paper , automatic frequency control is tak en. As the load is changed ; the frequency is changed to meet the requi rement of the connected load [3]. The load is v ar ied with f ast r ate and slo w r ate w e are consider ing the f ast v ar iation; slo w v ar iation is not considered. In larger iner ter-connected po w er system tie line po w er should be maintained in a toler ab le limit with respect to load change [4]. It becomes necessar y to regulate the input to the turbine b y opening of v alv e the input can be steam o r h ydro f or h ydro alter nator . F or this , an efficient control str ategy is required. The controlling of electr ic po w er of the gener ator b y this method is automatic gener ation control (A GC) [5, 6]. In this paper , a step disturbance is considered, and it is seen that the frequency steady state error and tie line steady sta te error f ollo wing the step disturbance should be z ero [7]. In the ALFC , there are tw o loops pr imar y loops and secondar y loop , pr imar y loop is f ast, and it is called uncontrolled loop . The secondar y loop is slo w , which is called controlled loop . Sec- ondar y control loop is done to mak e the automatic control error to z ero . The con v entional control str ategy f or LFC prob lems is to be tak en the integ r al of the control error as the control signal, an integ r al controller pro vides z ero steady-state error . Ho w e v er , it e xhibits poor dynamic beha vior , to impro v e the dynamic beha vior man y control techniques are used such as linear f eedbac k, optimal control, and v ar iab le str ucture control ha v e been proposed [ 8]. Man y algor ithms f or controlling the con v entional controller gains are proposed. The algor ithms are PSO , genetic algor ithm, ZN algor ithm [9]. In this paper , the har mon y search algor ithm is used f or tuning the free coef ficients of PID controller . Har mon y search algor ithm uses a Meta heur istic approach to solv e man y optimiza- Receiv ed No v ember 18, 2016; Re vised December 15, 2016; Accepted December 28, 2016 Evaluation Warning : The document was created with Spire.PDF for Python.
20 ISSN: 2502-4752 tion prob lems this algor ithm w as inspired b y the method used to impro v e the tuning or pitches of the m usic to obtain better har mon y [10]. Har mon y search algor ith m w or ks on the mimic king the impro visation of m usic pla y ers . As the most str uctur al optimization methods are based on mathematical algor ithms that require strongly b uilt g r adient inf or mation [11]. The har mon y search algor ithm does not require initial v alues and uses a r andom search instead of a g r adient search; so that the der iv ativ e inf or mation is unnecessar y . This algor ithm is conceptual using the m usical impro visation process of searching f or a perf ect state of har mon y (i.e har mon y memor y) to auto- matically adjust par ameter v alues . This algor ithm does not require to fix e initial condition. This algor ithm is used b y engineers to solv e the optimization prob lem [12]. Boroujeni et al [13], proposed w or k f or tw o are po w er system. A uthor T uned the PI b y HSA f or quenching the de viation in frequency and Tie-line po w er due to diff erent loading conditions . Compare the result of tuned PI b y tuned IP . Sho w eff ectiv eness of both controllers in industr ies . Sambar iy a and Nath, 2015, ha v e proposed the application of par ticle s w ar m optimization (PSO) in optimal tuning of PID par ameters f or tw o-area load frequency po w er system in [14]. The ne w NARMA L2 controller is presented in [15]. The fuzzy logic based controller f or m ulti-area system is presented in [16]. The application of adaptiv e nero fuzzy logic controller is presented in [17]. Abedinia et al [18], proposed w or k to solv e the LFC prob lem. By implementing a ne w m ulti-stage fuzzy (MSF) controller based on m ulti-objectiv e Har mon y search algor ithm (MOHSA). Membership Function of fuzzy is designed automatically b y the proposed MOHSA method. Com- pare results with other controllers used f or LFC . Sanpala and V akula [19], in proposed w or k tw o interconnected are with identical par ameters are considered f or study . The presented w or k better utiliz ed the har mon y search algor ithm to design the optimal par ameters of Fuzzy-PID . Sambar iy a and Pr asad [20], ha v e designed a Fuzzy logic po w er system stabiliz er , whose par ameters are tuned b y Har mon y search algor ithm. The P ar ameters of the PD controller are considered as the nor maliz ed f actors of FPSS and T uned using har mon y search algor ithm. In this paper , con v entional PID controllers are compared with HSA-PID controller . It is f ound t hat tuned PID controller is better than the con v entional controller . The settling time , o v er- shoot time , steady-state error is reduced b y using a proposed controller as compared to the con- v ention controller . By MA TLAB softw are the frequency response , tie line po w er , go v er nor po w er disturbance response is compared of both the area. The paper is organiz ed in 5 sections . The prob lem f or m ulation with system descr iption, process and objectiv e function is presented in section 2.. The har mon y search algor ithm used to tune the par ameters of PID controller is presented in section 3.. The system responses with proposed PID controller in ter ms of frequency and tie-line po w er are compared to that of the controllers a v ailab le in liter a ture and mentioned in section 4.. The paper is concluded in section 5. and f ollo w ed with appendix, nomenclature and ref erences . 2. Pr ob lem f orm ulation 2.1. System description The system under consider ation is a tw o-area po w er system. It consists of go v er nor , non-reheat turbine and load-iner tia. The models of the components are solv ed f or obtaining state v ector matr ix b y state space modeling method. Defining the state v ar iab les in ter ms of diff erential v ar iab les of the system. The state v ar iab les as x 1 ; :::; x 9 are defined b y diff erential v ar iab les as in Eqn. 1. The b loc k diag r am representation of the system is sho wn in Fig. 1. The u 1 ; u 2 and IJEECS V ol. 5, No . 1, J an uar y 2017 : 19 32 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 21 d 1 = p d 1 ; d 2 = p d 2 are representing the control and disturbance v ar iab les . x 1 = f 1 x 2 = p t 1 x 3 = p g 1 x 4 = f 2 x 5 = p t 2 x 6 = p g 2 x 7 = p tie 12 x 8 = R AC E 1 dt x 9 = R AC E 2 dt (1)     R   ! R s T g         s T t             !   D H   s T g   ! !   s T t   ! !         ! D H     L P     m P     l P     m P     ! l P     x   x   x   u   x   x   x   x   x   u   x   f     f   s T   ! !     Figure 1. Bloc k Diag r am of tw o-area interconnected system State space equations f or tw o area are giv en as f ollo wing: _ x 1 = D 1 2 H 1 x 1 + 1 2 H 1 x 2 + 1 2 H 1 x 7 + ( d 1 ) 2 H 1 (2) _ x 2 = 1 T t 1 x 2 + 1 T t 1 x 3 (3) _ x 3 = 1 R 1 T g 1 x 1 1 T g 1 x 3 + 1 T g 1 u 1 (4) _ x 4 = D 2 2 H 2 x 4 + 1 2 H 2 x 5 + 1 2 H 2 x 7 d 2 2 H 2 (5) _ x 5 = 1 T t 2 x 2 + 1 T t 2 x 6 (6) Optimal design of PID controller f or LFC using HS algor ithm (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
22 ISSN: 2502-4752 2 6 6 6 6 6 6 6 6 6 6 6 6 4 _ x 1 _ x 2 _ x 3 _ x 4 _ x 5 _ x 6 _ x 7 _ x 8 _ x 9 3 7 7 7 7 7 7 7 7 7 7 7 7 5 = 2 6 6 6 6 6 6 6 6 6 6 6 6 6 4 D 1 2 H 1 1 2 H 1 0 0 0 0 0 0 0 0 1 T t 1 1 T t 1 0 0 0 0 0 0 1 R 1 T g 1 0 T g 1 0 0 0 0 0 0 0 0 0 D 2 2 H 2 1 2 H 2 0 1 2 H 2 0 0 0 1 T t 2 0 0 0 1 T t 2 0 0 0 0 0 0 1 R 2 T g 2 0 1 T g 2 0 0 0 2 T o 0 0 2 T o 0 0 0 0 0 B 1 0 0 0 0 0 1 0 0 0 0 0 B 2 0 0 1 0 0 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 2 6 6 6 6 6 6 6 6 6 6 6 6 4 x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 3 7 7 7 7 7 7 7 7 7 7 7 7 5 (12) _ x 6 = 1 R 2 T g 2 x 4 1 T g 2 x 6 + 1 T g 2 u 2 (7) _ x 7 = 2 T 0 x 1 2 T 0 x 4 (8) _ x 8 = ( B 1 ) x 1 x 7 (9) _ x 9 = ( B 4 ) x 4 + x 7 (10) The Eqns . 2 - 10 can be organiz ed as a v ector matr ix as f ollo wing in Eqn. 11. _ x = Ax + B u + C d (11) where x is state v ector , u is control v ector and d is the disturbance v ector . The complete matr ix representation is sho wn in f ollo wing Eqn. 12. The tw o systems interconnected are sho wn in Fig. 1. The Go v er nor , turbine and load iner tia b loc k are rep resented with their time constant. The both areas are interconnected using tie-line . The rele v ant data of the system is presented in the Appendix. 2.2. Pr ocess Study of load f requency control is considered f or tw o interconnected area po w er system. Step Disturbance of 0.01 p .u is giv en to both areas . The u 1 ; u 2 are tw o controlled signal giv en to go v er nor f or controlling dynamic response de viations . Har mon y Search algor ithm is used f or tuning t he par ameters of PID controller . The designed controller is used control the frequency and tie - line po w er de viations in the interconnected area are controlled. The main impact of HSA-PID is to control the tr ansient response of the system. The tuning scheme of PID par ameters is sho wn in Fig. 2. 2.3. Objective function The area control error of both of the areas is sensed and the diff erence (i.e . error of A CE) is considered as the signal to be minimiz ed under optimization. The par ameters of both PID controllers are optimiz ed under optimization using HS algor ithm subjected to the upper and lo w er bounds of the par ameters as sho wn in Eqn. 13. K min pj K pj K max pj K min ij K ij K max ij K min d j K d j K max d j (13) where , j stands f or the PID controller connected an area i.e j = 1 ; 2 . These par ameters of the PIDs are optimiz ed using the objectiv e function as sho wn in Eqn. 14. It sho ws the minimization of integ r al square error (ISE) of the area control error f or the sim ulation time T sim . I S E = T sim Z 0 AC E 1 ( t ) AC E 2 ( t ) dt (14) IJEECS V ol. 5, No . 1, J an uar y 2017 : 19 32 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 23 S Kp Kd Pow er Sys tem Model Obj . Fun . Harmony S ea rc h Algorithm + - PID Controller Kp S + + + 1 f D 1 A C E Figure 2. The scheme of optimization of PID par ameters using har mon y search algor ithm 3. Harmon y sear c h algorithm The har mon y search algor ithm (HSA) w as de v eloped b y Geem.et.al in 2001 [21]. The HSA is made efficient b y using impro visation technique used in m usic making. The m usician selects the pitches of the instr ument f or getting better state of har mon y . A m usician impro visation is same as the search process in optimization. Ev er y m usic pitch is equal to the appreciation of beauty of quantity [22]. The per-f or mation and efficiency of most of the Meta heur istic algor ithm depends on the e xtent of balance betw een div ersification and intensification dur ing the search process . The e xploitation is the ability of an algor ithm to e xploit the search space in the neighborhood of the current good solution using the inf or mation already collected [12]. The pitch adjustment technique helps in finding the best solution in the har mon y memor y [23]. The har mon y search algor ithm depends on the points such as Har mon y memor y siz e (HMS), Har mon y memor y consider ation r ate (HMCR), when a m usician is impro ving, he or she has three possib le choice:(1) pla ying an y f amous e xactly from his or her memor y ,(2) pla ying some- thing similar to the f or a mentioned tune(thus adjusting the pitch slig htly),(3) composition ne w or r andom notes [24, 25]. 3.1. Steps of HS algorithm Each ro w of har mon y memor y (HM), consists of N decision v ar iab les and the fitness score ! ([ x 1 ; x 2 ; :::; ! ]). The HM is initializ ed with HMS r andomly gener ated solution v ectors [24]. 3.1.1. Initialization Let in an optimization prob lem, the objectiv e function is represented b y Minimization of F ( x ) , which is subjected to x i 2 X i , i = 1 ; 2 ; 3 N . Where , x is the set of design v ar iab les ( x i ), ( x L i x i x U i ), and N is the count of design v ar iab les . Define the v ar iab le limits as lo w er ( x L i ) and upper ( x U i ) or x L i x i x U i . Deciding the v alue of har mon y memor y siz e (HMS), from the r ange 10 H M S 100 . Decide v alue of HMCR (har mon y memor y consider ation r ate) within the r ange 0 : 0 H M C R Optimal design of PID controller f or LFC using HS algor ithm (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
24 ISSN: 2502-4752 1 : 0 . x 0 i   x 0 i l f x 1 i ; x 2 i ; :::; x H M S i : pr ob: H M C R x 0 i 2 X i : pr ob: (1 H M C R ) (15) Decide the v alue of P AR (pitch adjustment r ate) from the r ange 0 : 0 P AR 1 : 0 . x i j = x L j + r and (0 ; 1) ( x U j x L j ) (16) x 0 i   y es w ith pr obabil ity P AR no w ith pr oba bil ity (1 P AR ) (17) Compute the step siz e ( b i ) as b i = x U i x L i N . Specify the maxim um limit of iter ation n umber . 3.1.2. HM Initiation The HM matr ix as in Eqn. 18 is filled with r andomly gener ated possib le solution v ectors f or HMS and is sor ted b y the v alues of the objectiv e function f ( x ) . H M = 2 6 6 6 6 6 4 x 1 1 x 1 2 ::: x 1 N 1 x 1 N x 2 1 x 2 2 ::: x 2 N 1 x 2 N . . . . . . . . . . . . . . . x H M S 1 1 x H M S 1 2 ::: x H M S 1 N 1 x H M S 1 N x H M S 1 x H M S 2 ::: x H M S N 1 x H M S N 3 7 7 7 7 7 5 ) 2 6 6 6 6 6 4 f ( x 1 ) f ( x 2 ) . . . f ( x H M S 1 ) f ( x H M S ) 3 7 7 7 7 7 5 (18) 3.1.3. Impr o visation A Ne w Har mon y v ector x 0 = ( x 0 1 ; x 0 2 ; :::; x 0 n ) is gener ated based on three cr iter ia: r andom selection, memor y consider ation, and pitch adjustment. Random Selection : T o decide the v alue of x 0 1 f or the Ne w Har mon y x 0 = ( x 0 1 ; x 0 2 ; :::; x 0 n ) , the HS algor ithm r andomly selects a v alue from r ange with a probability of 1 H M C R . Memor y Consider ation : T o decide the v alue of x 0 1 , the HS algor ithm r andomly selects a v alue from the HM with a probability of H M C R , where j = 1 ; 2 ; :::; H M S . It can be represented as in Eqn. 18. Pitch Adjustment : Each element of the Ne w HM v ector x 0 = ( x 0 1 ; x 0 2 ; :::; x 0 n ) is subjected to deter mine whether it should be pitch-adjusted or not. The selected x 0 i is fur ther adjusted b y adding an amount to the v alue with a probability of P AR. The P AR par ameter is called the probability of pitch adjustment and is represented b y Eqn. 17 [26]. The 0 no 0 with probability 1 P AR represents that the probability of not adding an y amount. On the other hand, if the pitch adjustment decision f or x 0 i is y es , then it is replaced b y x 0 i   x 0 i bw ; where , bw is distance ban dwidth in case a contin uous v ar iab le . The pitch adjustment is perf or med on e v er y v ar iab le of the Ne w HM v ector . IJEECS V ol. 5, No . 1, J an uar y 2017 : 19 32 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 25 3.1.4. Updating HM Let the HM v ector is x 0 = ( x 0 1 ; x 0 2 ; :::; x 0 n ) , which is resolv ed b y minimization of objectiv e function is better than the w orst har mon y present in the HM. Theref ore , the Ne w Har mon y is inser ted into the HM, while , the w orst har mon y is remo v ed from the HM. 3.1.5. Stopping criterion If the maxim um count of impro visations is reached and the stopping cr iter ion as maxim um n umber of iter ations is satisfied, then the process of computation is ter minated. Otherwise , go to steps impro visatio n and updating HM as abo v e to repeat the process . The HS algor ithm is sho wn in the Fig. 3. Start Initialization S e l e ct H M CR S e lect P A R Com p ute new H a rm on y Modify H M by new Harmon y Stop Cri te ri on U pda t ion of HM Yes No S t op Yes No Im prov isa ti on Figure 3. W or king of Har mon y Search Algor ithm 4. Result and discussion 4.1. Optimization of PID parameter s The system sho wn in Fig. 1 equipped with PID controllers with unkno wn par ameters as K p 1 ; 2 ; K i 1 ; 2 ; K d 1 ; 2 is sim ulated to tune these par ameters using har mon y search algor ithm. The prob lem of optimization is subjected to minimization of ISE based objectiv e function as sho wn Optimal design of PID controller f or LFC using HS algor ithm (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
26 ISSN: 2502-4752 in Eqn . 14 and the par ameter bounds as in Eqn. 13. The optimiz ed PID par ameters using HS algor ithm are sho wn in T ab le 1. The plot of the fitness function with iter ation counts is sho wn in Fig. 4. It can be seen that the fitness function gets reduced as the iter ation count is increased and around 190 it becomes constant. It is the instant, where the optimal par ameters are considered. 0 50 100 150 200 0.5 1 1.5 2 2.5 3 x 10 −6 Interations F min  Fitness value Figure 4. The v ar iation of fitness function with iter ations using har mon y search algor ithm T ab le 1. Compar ison of par ameters of PID controller in liter ature and proposed HSA-PID controller Controllers K p 1 K i 1 K d 1 K p 2 K i 2 K d 2 HSA-PID (Prop .) 1.8677 1.9934 1.9099 1.4324 1.9139 1.8247 PSO-PID [27] 3.0 0.8531 0.35 1.98 0.3919 0.1978 Con v-PID [28] 0.1109 0.2742 0.1110 0.0121 0.2019 0.0030 PSO-PID [5] 22.8070 2.0734 17.4628 22.8070 2.0734 17.4628 BFO A-PID [29] 0.1317 0.4873 0.2506 0.1317 0.4873 0.2506 PSO-PID [30] 0.790 1.4252 0.4652 0.790 1.4252 0.4652 4.2. P erf ormance comparison In this presented w or k, a tw o area interconnected system with non-reheat turbine is con- sidered f or load frequency control (LFC) prob lem. A step load of 0.01 p .u are giv en in both areas , which sho w rob ustness of the system. Gener ally , liter ature sur v e y on LFC sho w that authors con- sidered in gener al 1% step load per turbation in an area is giv en f or an optimal results . In proposed w or k, disturbance is giv en in both area. It has been used f or PID tuning. The HSA-PID results are compared with recent pub lished papers b y consider ing their PID par ameters . F requency re- sponses of area-1 and area-2 are sho wn in Fig. 5 - Fig. 6. The tie - line po w er response is sho wn in Fig. 7. It is seen that proposed controller results are f ar better than the con v entional con- troller . The tie-line po w er response of an interconnected area obtained b y the proposed controller is compared with PID controller . Compar ison of frequency responses on the basis of settling - time , o v er - shoot and under-shoot as giv en in the tab ular f or m in T ab le 2 - T ab le 3 and the tie IJEECS V ol. 5, No . 1, J an uar y 2017 : 19 32 Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS ISSN: 2502-4752 27 -line compar ison is giv en in T ab le 4. It is f ound that HSA tuned PID is better than controllers in [5, 27–30]. 0 5 10 15 20 25 30 35 −6 −5 −4 −3 −2 −1 0 1 x 10 −4 Time (s) Frequency deviation Frequency response of two area−1 system     Prop. (HSA−PID) PID (Kavya, 2015) PID (Ikhe, 2013 ) PID (Bassi, 2011) PID (Ali, 2013) PID (Dhanlakshmi, 2015) Figure 5. F req uency response of area-1 obtained b y HSA-PID and compared with controllers in [5, 27–30] 0 5 10 15 20 25 30 −10 −5 0 5 x 10 −4 Time (s) Frequency deviation Frequency response of two area−2 system     Prop. (HSA−PID) PID (Kavya, 2015) PID (Ikhe, 2013 ) PID (Bassi, 2011) PID (Ali, 2013) PID (Dhanlakshmi, 2015) Figure 6. F req uency response of area-2 obtained b y HSA-PID and compared with controllers in [5, 27–30] Optimal design of PID controller f or LFC using HS algor ithm (D . K. Sambar iy a) Evaluation Warning : The document was created with Spire.PDF for Python.
28 ISSN: 2502-4752 T ab le 2. Compar ison of frequency response of area-1 of HSA-PID compared with con v entional PID Controllers Settling time Ov er shoot Under shoot HSA-PID 12 0 : 085 10 6 2 : 03 10 4 PSO-PID [27] 14.1 7 : 34 10 4 4 : 71 10 4 Con v-PID [28] 28.35 - 6 : 51 10 4 PSO-PID [5] 33.16 2 : 2 10 4 6 : 3 10 4 BFO A-PID [29] 17.28 - 3 : 9 10 4 Con v-PID [30] 18.32 7 : 34 10 4 3 : 8 10 4 T ab le 3. Compar ison of frequency response of area-2 of HSA-PID compared with con v entional PID Controllers Settling time Ov er shoot Under shoot HSA-PID 12.2 0 : 31 10 4 3 : 24 10 4 PSO-PID [27] 13.0 1 : 65 10 4 6 : 32 10 4 Con v-PID [28] 27.50 1 : 05 10 4 9 : 70 10 4 PSO-PID [5] 27.10 4 : 4 10 4 1 : 7 10 4 BFO A-PID [29] 16.20 2 : 8 10 5 5 : 4 10 4 Con v-PID [30] 28.55 3 : 00 10 4 5 : 5 10 4 T ab le 4. Compar ision of Tie-line po w er response of area-12 of HSA-PID compared with con v en- tionl PID Controllers Settling time Ov er shoot Under shoot HSA-PID 26.0 1 : 70 10 4 - PSO-PID [27] 27.2 4 : 4 10 4 4 : 4 10 4 Con v-PID [28] 43.0 1 : 043 10 4 - PSO-PID [5] - 2 : 25 10 4 - BFO A-PID [29] 29 3 : 87 10 5 0 : 98 10 4 Con v-PID [30] 27.0 3 : 35 10 4 1 : 0 10 4 IJEECS V ol. 5, No . 1, J an uar y 2017 : 19 32 Evaluation Warning : The document was created with Spire.PDF for Python.