International Journal of Automation and Computing 8(4), November 2011, 429-436 DOI: 10.1007/s11633-011-0600-6 A Reliable and Accurate Calculation of Excitation Capacitance Value for an Induction Generator Based on Interval Computation Technique Rajesh Kumar Thakur1 1 Vivek Agarwal2 Paluri S. V. Nataraj1 Systems and Control Engineering, Interdisciplinary Programme (IDP),Mumbai, 400076, India 2 Department of Electrical Engineering, IIT–Bombay, Powai, Mumbai 400076, India Abstract: A squirrel cage induction generator (SCIG) offers many advantages for wind energy conversion systems but suffers from poor voltage regulation under varying operating conditions. The value of excitation capacitance (Cexct ) is very crucial for the selfexcitation and voltage build-up as well as voltage regulation in SCIG. Precise calculation of the value of Cexct is, therefore, of considerable practical importance. Most of the existing calculation methods make use of the steady-state model of the SCIG in conjunction with some numerical iterative method to determine the minimum value of Cexct . But this results in over estimation, leading to poor transient dynamics. This paper presents a novel method, which can precisely calculate the value of Cexct by taking into account the behavior of the magnetizing inductance during saturation. Interval analysis has been used to solve the equations. In the proposed method, a range of magnetizing inductance values in the saturation region are included in the calculation of Cexct , required for the self-excitation of a 3-φ induction generator. Mathematical analysis to derive the basic equation and application of interval method is presented. The method also yields the magnetizing inductance value in the saturation region which corresponds to an optimum Cexct(min) value. The proposed method is experimentally tested for a 1.1 kW induction generator and has shown improved results. Keywords: d-q model, hull consistency, induction generator, interval analysis, interval arithmetic, optimization, range of interval, realpaver, self-excitation, transients. 1 Introduction A wind power generation system generates electricity from wind energy and typically comprises an induction generator coupled to a wind turbine. In a wind power generation system, the mechanical energy of the wind turbine is converted into electrical energy by the induction generator. A squirrel cage induction generator (SCIG) is highly suitable to be driven by wind turbine because of its small size and weight, robust construction and reduced maintenance cost[1] . In order to initiate voltage generation by the induction generator (self-excitation), a leading reactive power is provided to the stator windings of the generator by connecting a capacitor bank to the stator windings. The induced e.m.f. and current in the stator winding starts rising and attains its steady-state value with frequency dependent on rotor speed and machine parameters. The generated voltage is sustained at this operating point till reactive power balance is maintained[2] . The voltage so generated is unstable in the sense that its value changes drastically under various loading conditions. This, in turn, changes the generated torque and the rotor speed varies causing further changes in the generated voltage. This leads either to a collapse of the terminal voltage or building up to an excessively high value depending upon the values of the magnetizing inductance and the terminal (excitation) capacitance[2−5] . The magnetizing inductance value is a function of rotor speed and is nonlinear[3] . Hence, it is necessary to determine the range of capacitance value to keep the machine in excitation mode and regulate the generator terminal voltManuscript received March 30, 2010; revised September 26, 2010 age within the limits. Tuning of capacitance is required with the variation of rotor speed and loading conditions. This is an impractical task due to the inter-dependence of the system variables, changing rotor speed and the system0 s nonlinearity. It is essential that the capacitance value of the capacitor bank is such that at a given rotor speed, the generated voltage in the stator windings does not undergo large transients[5, 6] . If the voltage generated in the stator windings is not sustained and smooth, there will be vibrations in the wind turbine leading to wear and tear of the wind turbine thereby reducing its life. In order to convert the wind energy into electrical energy efficiently, it is important that the induction generator operates smoothly, i.e., voltage generated in the stator windings is sustained and is without transients. Therefore, calculation of the capacitance value of the capacitor bank is very critical for the desired operation of the induction generator. It is known that the induction generator operates in the saturated region during the self-excitation phase. It is also known that the maximum value of magnetizing inductance (Lmax ) in the saturated region leads to sustained voltage generation in the stator windings. Therefore, in order to accurately calculate the capacitance value, it is essential to know the maximum value of the magnetizing inductance in the saturated region. But this value is not known and unavailable. However, it is known that the magnetizing inductance value in the saturated region lies between zero and the Lmax value in the unsaturated region. For accurate calculation of the capacitance value, it is necessary to take into consideration the magnetizing inductance value for the saturated region both in the steady state and dynamic state operating conditions of the induction generator. 430 International Journal of Automation and Computing 8(4), November 2011 Two models are reported in [7–12] for the analysis of selfexcited induction machine. One is based on the per-phase equivalent approach and the other, the d-q axis model, is based on the generalized machine theory. Most of the work is based on per-phase equivalent circuit approach, which includes loop-impedance method[7−9] and the nodaladmittance method[10, 11] . The d-q method has been used by Elder et al. and others[2, 4, 6, 12] . In the per-phase equivalent circuit approach, highly nonlinear simultaneous equations in per unit frequency or magnetizing reactance are solved using iterative methods to obtain minimum and maximum values of the excitation capacitance. Apart from being time consuming, this method uses the steady-state model and unsaturated magnetizing inductance value in the calculations. Hence, the minimum value of capacitance so obtained is an over estimation. This capacitance value is able to initiate self-excitation, but yields satisfactory results only under steady-state conditions. Higher value of capacitance is not technically viable as there may be a possibility that the current flowing in the capacitor might exceed the rated current of the stator, due to the fact that capacitive reactance reduces as the capacitance value increases[2, 4, 6] . This will cause higher transient currents during voltage build-up and hence more fluctuation in the mechanical load on the prime mover which may excite oscillatory mode of the prime mover. The d-q model based approach yields better results as it also captures transient phenomenon of the machine. But this is also an iterative approach and makes use of unsaturated value of magnetizing inductance. Therefore, this method also yields an over estimation of capacitance value. In view of the ongoing discussion, it is clear that there is a need to calculate the minimum value of capacitance taking the magnetizing inductance value in the range of saturated region which ensures self-excitation for a given induction generator. This paper presents an “interval technique” based novel method for determining the minimum capacitance (optimum value) required for self-excitation of a 3-φ induction generator using d-q model of induction machine and using a range of magnetizing inductance values in saturated region. This method offers the following advantages: 1) It yields all possible values of excitation capacitance. 2) It guarantees that the entire range of excitation capacitance values is found to a user-specified accuracy. 3) The range helps in selection of a suitable excitation capacitance for wider operating speed variations. The remaining paper is organized as follows. Section 2 describes the system modeling. Section 3 presents the evolution of the equations needed to be solved to determine the capacitance value. This section also describes the application procedure of the interval analysis method. Section 4 presents the simulation results. Experimental results and observations are included in Section 5. Discussion of results and major conclusions of this work are included in Section 6. Notations. Rs , Rr : Per phase stator, rotor (referred to stator) Ls , Lr : Per phase stator and rotor (referred to stator) inductance (H) Lm : Inductance (referred to stator) (H) C: Per phase terminal excitation capacitive (F) Iqs , Ids : Stator quadrature and direct axis currents (A) Iqr , Idr : Rotor quadrature and direct axis currents (A) Vqs , Vds : Stator quadrature and direct axis voltage (V) Vqr , Vdr : Rotor quadrature and direct axis voltage (V) ωr : Angular speed of rotor (rad/s) ω: Angular speeds of synchronous, reference frame (rad/s) p : d/dt, derivative with time. 2 d -qq based system modelling The parameters of induction machine are obtained by operating the machine as a motor. It has been reported[2, 4] that per phase resistance and inductance of the stator and rotor (referred to the stator side) remain nearly constant over the frequency range 2 Hz–50 Hz. Even if there is a small variation in the values of these parameters, it has been observed that there is no significant impact on the system behavior. Fig. 1 shows the laboratory based schematic diagram of a 3-φ, induction generator connected with 3-φ capacitor bank and driven by the prime mover (DC motor). The d-q axis model of induction generator based on the generalized machine theory is shown in Fig. 2. Fig. 1 3-φ induction generator with prime mover DC motor and capacitor bank Fig. 2 d-q model of SCIG at no load (a) d-axis (b) q-axis The terminal voltage equation is obtained from the equivalent circuits (see Fig. 2) for a stationary reference frame by setting ω = 0. The terminal voltage equation obtained is R. K. Thakur et al. / A Reliable and Accurate Calculation of Excitation Capacitance Value for · · · as follows: "Vqs# Rs + pLs + Vds 0 = Vqr Vdr 3 1 pC pLm ωr Lm 0 1 Rs + pLs + pC ωr Lm pLm pLm " # iqs 0 0 pLm Rr + pLr ωr Lr ωr Lm Rr + pLr ids iqr idr (1) where the subscripts “s” and “r” are used to represent stator and rotor quantities and “d” and “q” represent direct and quadrature axis quantities, respectively. For a stationary reference frame (ω = 0) and before excitation, all terminal voltages are considered to be zero. Laplace transform of (1) in complex variable form “s” results in: V (s) = A (s) I (s) (2) where Vqs (s) V (s) ds V (s) = Vqr (s) Vdr (s) Iqs (s) I (s) ds , I (s) = Iqr (s) Idr (s) (3) 431 Evolution of equations and interval analysis For an induction generator to be self-excited, the machine has to operate in saturation region and one pair of poles of the machine must lie on the imaginary axis or in the right half of s-plane as shown in Fig. 3. The minimum value of capacitance corresponds to location of this pair of poles on imaginary axis with maximum value of magnetizing inductance. Expression 3 is complex polynomial and if a pair of poles has to lie on the imaginary axis for a given capacitance value, then this capacitance value will be minimum for the given set of machine parameters. The real and imaginary parts of the 6-th order polynomial are generated by replacing the complex variable “s” in the 6-th order polynomial by jω and substituting the value of Lm as [0 −Lmax value in the unsaturated region] and value of C as [0 −∞]. and 1 0 sLm 0 sC 1 0 Rs + sLs + 0 sLm A (s)= . sC sLm ωr Lm Rr + sLr ωr Lr ωr Lm sLm ωr Lm Rr + sLr (4) Solution of (2) describes the behaviour of the system. The dynamic behaviour is determined by the location of the poles of the polynomial or eigen-values of matrix “A”. The characteristic polynomial of the terminal voltage equation (2) is obtained by taking the determinant of the matrix “A” in (4). Rs + sLs + detA(s) = C 2 X12 s6 + 2C 2 X1 X2 s5 + Fig. 3 s-plane pole location for Cexct > Cexct(min) and Cexct = Cexct(min) (C 2 (X22 + X1 (2Rs Rr + ωr2 X1 ))+ 2CL2 X1 )s4 + 2(Rs (Rr X2 + L2 ωr2 X1 )C 2 + 3 (L2 X2 + Rr X1 )C)s + (C 2 The real and imaginary parts are generated as follows: Rs2 X3 )+ det A(s = jω) = −C 2 X12 ω 6 + 2C 2 X1 X2 jω 5 + (2C(Rs Rr L2 + Rr X2 + ωr2 L2 X1 )+ L22 ))s2 + 2(Rs CX3 + Rr L2 )s + X3 . (C 2 (X22 + X1 (2R1 R2 + ωr2 X1 )) + 2CL2 X1 )ω 4 − (5) (C 2 R12 X3 + 2C(R1 R2 L2 + R2 X2 + ωr2 L2 X1 ) + L22 )ω 2 + The 6-th order polynomial obtained is as follows: 2(R1 CX3 + R2 L2 )jω + X3 . C 2 X12 s6+2C 2 X1 X2 s5 + (C 2 (X22 + X1 (2Rs Rr + ωr2 X1 ))+ 2CL2 X1 )s4 + 2(Rs (Rr X2 + L2 ωr2 X1 )C 2 + (L2 X2 + Rr X1 )C)s3 + (C 2 Rs2 X3 + The real part on the right half side of s-plane (RHS) in (8) is 2CL2 X1 )ω 4 − (C 2 R12 X3 + 2C(R1 R2 L2 + R2 X2 + (6) ωr2 L2 X1 ) + L22 )ω 2 + X3 . (9) The imaginary part on the RHS in (8) is where L1 = Ls + Lm ; L2 = Lr + Lm ; X1 = L1 L2 − L2m X2 = Rs L2 + Rr L1 ; X3 = Rr2 + ωr2 L22 . (8) − C 2 X12 ω 6 + (C 2 (X22 + X1 (2R1 R2 + ωr2 X1 ))+ (2C(Rs Rr L2 + Rr X2 + ωr2 L2 X1 ) + L22 ))s2 + 2(Rs CX3 + Rr L2 )s + X3 = 0 (2(R1 (R2 X2 + L2 ωr2 X1 )C 2 + (L2 X2 + R2 X1 )C)jω 3 − C 2 X1 X2 ω 5 − (R1 (R2 X2 + L2 ωr2 X1 )C 2 + (7) (L2 X2 + R2 X1 )C)ω 3 + (R1 CX3 + R2 L2 )ω. (10) 432 International Journal of Automation and Computing 8(4), November 2011 Both (9) and (10) are set to zero and a pair of equations is obtained in terms of C and Lm as follows: Fig. 4 (a) and (b), respectively. It is observed that as speed increases, the capacitance value decreases. Table 1 − C 2 X12 ω 6 + (C 2 (X22 + X1 (2R1 R2 + ωr2 X1 ))+ 2CL2 X1 )ω 4 − (C 2 R12 X3 + 2C(R1 R2 L2 + R2 X2 + ωr2 L2 X1 ) + L22 )ω 2 + X3 )ω = 0. (11) Sr No. Rotor speed Synch speed Range of C C(min) (µF) (rad/s) (rad/s) (µF) 11 377 376 0 – 23 20.1 2 377 370 0 – 25 23.92 25 3 377 360 0 – 34 25.5 34* 4 314 313 0 – 33 31.56 33 5 314 310 0 – 35 33.3 35 6 314 300 0 – 146 42.07 135 7 250 249 0 – 52 49.96 52 8 250 245 0 – 57 54.76 57 9 200 199 0 – 81 78.37 81 10 150 149 0 – 150 140 150 11 100 99 0 – 330 317 320 12 50 49 0 – 1400 1246 1400 13 45 44 0–+∞ 1791 2002 14 40 39 0–+∞ No solution C 2 X1 X2 ω 5 − (R1 (R2 X2 + L2 ωr2 X1 )C 2 + 3 (L2 X2 + R2 X1 )C)ω + (R1 CX3 + R2 L2 )ω = 0. (12) Equations (11) and (12) are nonlinear equations of the order six. It is highly cumbersome to solve such equations using numerical techniques. In many practical situations in engineering, data are only known to lie within intervals and only ranges of values are computed experimentally[13] . In such cases interval computation yield desired results[14, 15] . Interval analysis method provides a direct means of solving the two highly nonlinear equations (11) and (12) to find the value of capacitance and predicting the behaviour of a self-excited induction generator under balanced operating conditions. The coefficients of both the equations vary as the excitation builds up because of the change in the value of magnetizing inductance from unsaturated to saturated region. Hence, these coefficients can only be represented by interval of values calculated based on interval arithmetic operation. It is reported that the magnetizing inductance first increases from its value of unsaturated region and then starts decreasing while operating in saturated region[2, 4] . The above equations are solved simultaneously using interval analysis method to obtain all the solutions for C and Lm . The capacitance value and corresponding magnetizing inductance (Lmax ) value for the saturated region are extracted from the solutions obtained for C and Lm for smooth sustained voltage build-up in an induction generator. 4 Minimum values of excitation capacitance for different rotor speeds 23 Simulation and results RealPaver[16] is a modeling language for numerical constraint solving software. Numerical constraint is a relation between the unknowns (variables) of a problem which is defined by an analytic expression: f (x1 , x2 , · · · , xn )♦0 where ♦ = {6, >, =} and f : Rn → R. Each domain is represented by a closed interval: [r1 , r2 ] = {r ∈ R|r1 6 r 6 r2 } It uses: A set of real-valued variables, e.g., (x1 , x2 , · · · , xn ); A set of interval domain, e.g., (x1 , x2 , · · · , xn ); A set of numerical constraints, e.g., (c1 , · · · , cm ) over the given set of variables; Solution is obtained for all the consistent values with respect to all constraints. Realpaver is used for interval analysis on a 3-φ, Y connected induction generator with machine rating of 220 V, 4.8 A and 60 Hz. Other important machine parameters (in per unit) are, Rs = 0.0946, Rr = 0.0439, Xs = Xr =0.0865, Xm(max) = 2.12[6] . Results are obtained for different sets of constants and are summarized in Table 1. The variations required in the excitation capacitance w.r.t rotor speed and synchronous speed are shown in Fig. 4 Graph showing decrease in capacitance as (a) rotor speed increases and (b) synchronous speed increases 433 R. K. Thakur et al. / A Reliable and Accurate Calculation of Excitation Capacitance Value for · · · The result obtained by this method and that mentioned in [6] is summarized in Table 2. Data at Sr. No. 1 is from [6], Data at Sr. No. 2 is from the presented work. Table 2 Sr. No. Comparison of minimum values of capacitance Rotor speed Synchronous speed Cexct(min) (rad/s) (rad/s) (µF) 1 377 360 84.7 2 377 360 34 It is observed that actual value of minimum capacitance for the given set of parameters lies in between 25.5 µF and 34 µF which is much smaller than the value obtained by eigen-value sensitivity method[6] . The simulation result for Cexct(min) = 84.7 µF is shown in Fig. 5 (a) and the transient interval therein is marked as “A” whereas that for Cexct(min) = 34 µF is shown in Fig. 5 (b) with the transient interval marked as “B”. A comparison of Fig. 5 (a) and (b) reveals that when operated with Cexct(min) = 34 µF (determined by the proposed method), the system experiences a smooth transient interval, which, in turn, reduces the over voltage stress during the transient excitation of the induction generator and minimizes the adverse effects on the turbine and associated equipment. It is also observed that with Cexct(min) = 34 µF, the system can withstand wider variations in operating speeds without pronounced transients as compared with Cexct(min) = 84.7 µF. In fact, the proposed method yields a range of all possible values of excitation capacitance (see Table 1) for the entire operating range. The range permits trade-off between the specifications on speed of response and oscillatory nature of voltage build-up. These aspects are further highlighted in Section 5. Fig. 5 Simulation results showing self-excitation process of voltage buildup for: (a) C = 84.7 µF, calculated using eigen-value sensitivity method; (b) C = 34 µF, calculated using interval method (proposed method) 5 Experimental results and observations A 3-φ, star connected induction generator coupled with DC motor is used in this experiment. The DC motor is used as a prime mover to provide mechanical power at the induction generator rotor shaft in the form of kinetic energy which, in turn, is converted into electrical energy. Residual flux in the induction generator initiates terminal voltage generation. A 3-φ star connected capacitor bank is used to provide the requisite reactive power to the induction generator for building up the terminal voltage through selfexcitation process. The capacitor bank, with value determined using the proposed calculation method, is connected to the generator. The voltage build-up is captured using Agilent0 s Infiniium digital storage oscilloscope (DSO), with a multiplying factor of 20 on the vertical axis. The waveforms so obtained are shown in Figs. 6 and 7. The Cexct(min) values, for different (example) rotor speeds of the SCIG, are obtained with the proposed method. Experiments are conducted with each example operating speeds and the performance of the SCIG is tested with the various values of Cexct as summarized in Table 3. Key waveforms corresponding to each experiment are captured using the DSO and are shown in Fig. 6. These are discussed as follows. Table 3 Example Minimum capacitance values corresponding to the various example rotor speeds IM rotor speed Synchronous Minimum capacitance (rad/s) speed (rad/s) value (µF) 1 329 314 62 2 301 288 67 3 270 258 84 4 245 233 103 The capacitance value by proposed method is found to be 62 µF at a rotor speed of 329 rad/s and synchronous speed of 314 rad/s. On performing the experiment with a capacitance value of 60 µF at a rotor speed of 314 rad/s, no initiation of self-excitation was observed. Next, a capacitance value = 70 µF is used and self-excitation process is observed, as shown in Fig. 6 (a). Also, the minimum capacitance value for a rotor speed of 301 rad/s and synchronous speed of 288 rad/s is found to be 67 µF. When the experiment was performed at this operating point with capacitance value = 60 µF, no initiation of self-excitation was observed. The next capacitance value used was 70 µF and the self-excitation process was observed as shown in Fig. 6 (b). Similarly, at rotor speeds of 270 rad/s and 245 rad/s, Cexct(min) = 84 µF and 103 µF, respectively, were used and it was observed that the machine failed to excite with capacitance value of 80 µF at rotor speed 270 rad/s and capacitance value of 100 µF at rotor speed of 245 rad/s, whereas self-excitation took place with capacitance values of 90 µF and 110 µF, respectively, as shown in Fig. 6 (c) and (d). Thus, it is experimentally verified that the capacitance value calculated by using the proposed interval method is accurate and reliable. Further investigations were carried out by performing experiments with various capacitance 434 International Journal of Automation and Computing 8(4), November 2011 values at different operating conditions to analyze its effect on the transients of the induced voltage through cases 1–6 as summarized in Table 4. Table 4 Various case studies with different combinations of the capacitance value and rotor speed Sr. No. Case No. Cexct , (µF) Rotor speed (rad/s) 1 Case 1 90 301 2 Case 2 90 329 3 Case 3 110 251 4 Case 4 110 257.6 5 Case 5 110 270 6 Case 6 110 282 The following observations were made: 1) As the value of rotor speed was increased from 301 rad/s to 329 rad/s with a capacitance value of 70 µF, a significant increase in voltage fluctuation was observed (Fig. 6 (a) and (b)). But the speed of response increased. 2) As the value of rotor speed was increased through 270 rad/s, 301 rad/s and 329 rad/s with a capacitance value of 90 µF, a significant increase in voltage fluctuation (Figs. 6 (c), 7 (a), and 7 (b)) took place. The speed of response also increased considerably. 3) As the value of rotor speed is increased through 245 rad/s, 251 rad/s, 257.6 rad/s, 270 rad/s and 282.7 rad/s with a capacitance value of 110 µF, there was significant increase in voltage fluctuation (Figs. 6 (d), 7 (c)–7 (f)). Once again, the speed of response was found to have increased. Thus, it was experimentally verified that if the value of capacitance used is large, there are large transients during the self-excitation process, which can have adverse effects on the turbines. Thus, it may be concluded that the proposed method gives more suitable/accurate and optimum value of the excitation capacitance. 6 Fig. 6 Experimental results showing voltage building at the terminals of the induction generator: (a) Rotor speed = 329 rad/s; Cexct(min) = 62 µF; (b) Rotor speed = 301 rad/s; Cexct(min) = 67 µF; (c) Rotor speed = 270 rad/s; Cexct(min) = 84 µF; (d) Rotor speed = 245 rad/s; Cexct(min) = 103 µF. Discussion of results and conclusions A direct method based on interval computation technique has been proposed to calculate minimum capacitance value required for self-excitation of an induction generator. The proposed method is fast and can be directly used for transient analysis investigation of the generator voltage. A salient feature of the proposed calculation method is that it also yields a range of capacitance values which will not be able to induce self-excitation in the generator for the given operating conditions. The use of the interval (range) of magnetizing inductance values leads to an accurate prediction of whether or not self-excitation will occur for various capacitance values. The minimum capacitance value calculated by the interval analysis method, for different rotor speeds, is found to be very close to the capacitance value at which the induction machine initiates the self-excitation process, rendering an accurate value of the self-excitation capacitance. R. K. Thakur et al. / A Reliable and Accurate Calculation of Excitation Capacitance Value for · · · 435 Fig. 7 Experimental results showing voltage building at the terminals of the induction generator. (a) Case 1: Capacitance value = 90 µ, rotor speed =301 rad/s; (b) Case 2: Capacitance value = 90 µ, rotor speed = 329 rad/s; (c) Case 3: Capacitance value = 110 µF, rotor speed = 251 rad/s; (d) Case 4: Capacitance value = 110 µ, rotor speed = 257.6 rad/s; (e) Case 5: Capacitance value = 110 µF, rotor speed = 270 rad/s; (f) Case 6: Capacitance value = 110 µF, rotor speed = 282.7 rad/s References [1] R. C. Bansal. Three-phase self-excited induction generators: An overview. IEEE Transactions on Energy Conversion, vol. 20, no. 2, pp. 292–299, 2005. [2] C. Grantham, D. Sutanto, B. Mismail. Steady-state and transient analysis of self-excited induction generators. Electric Power Applications, vol. 136, no. 2, pp. 61–68, 1989. [3] S. Wekhande, V. Agarwal. Simple control for a wind-driven induction generator. IEEE Industry Applications Magazine, vol. 7, no. 2, pp. 44–53, 2001. [4] D. Seyoum, C. Grantham, F. Rahman. The dynamics of an isolated self-excited induction generator driven by a wind turbine. In Proceedings of the 27th Annual Conference of the IEEE Industrial Electronics Society, IEEE, Denver, USA, vol. 2, pp. 1364–1369, 2001. [5] R. K. Thakur, V. Agarwal. Effect of excitation capacitance value on the transient behaviour of induction generator in wind energy conversion system. In Proceedings of National Power Electronics Conference, Bangalore, India, 2007. [6] L. Wang, C. H. Lee. A novel analysis on the performance of an isolated self-excited induction generator. IEEE Transactions on Energy Conversion, vol. 12, no. 2, pp. 109–117, 1997. [7] N. H. Malik, S. E. Haque. Steady state analysis and performance of an isolated self-excited induction generator. IEEE 436 [8] [9] [10] [11] [12] [13] [14] [15] [16] International Journal of Automation and Computing 8(4), November 2011 Transactions on Energy Conversion, vol. 1, no. 3, pp. 134– 140, 1986. S. S. Murthy, O. P. Malik, A. K. Tandon. Analysis of selfexcited induction generators. Generation, Transmission and Distribution, vol. 129, no. 6, pp. 260–265, 1982. N. H. Malik, A. H. Al-Bahrani. Influence of the terminal capacitor on the performance characteristics of a self excited induction generator. Generation, Transmission and Distribution, vol. 137, no. 2, pp. 168–173, 1990. L. Ouazene, G. Mcpherson. Analysis of the isolated induction generator. IEEE Transactions on Power Apparatus and Systems, vol. 102, no. 8, pp. 2793–2798, 1983. T. F. Chan. Capacitance requirements of self-excited induction generators. IEEE Transactions on Energy Conversion, vol. 8, no. 2, pp. 304–311, 1993. J. M. Elder, J. T. Boys, J. L. Woodward. Self-excited induction machine as a small low-cost generator. Generation, Transmission and Distribution, vol. 131, no. 2, pp. 33–41, 1984. R. K. Thakur, V. Agarwal. Application of interval computation technique to fixed speed wind energy conversion system. In Proceedings of IEEE International Conference on Sustainable Energy Technologies, IEEE, Singapore, pp. 1093–1098, 2008. P. S. V. Nataraj, M. Arounasalamme. A new subdivision algorithm for the Bernstein polynomial approach to global optimization. International Journal of Automation and Computing, vol. 4, no. 4, pp. 342–352, 2007. R. E. Moore. Methods and Application of Interval Analysis, USA: SIAM Studies in Applied Mathematics, 1979. RealPaver User0 s Manual, Solving Nonlinear Constraints by Interval Computations, Edition 0.3, 2003. Rajesh Kumar Thakur received the Bachelor0 s degree in electrical engineering from Muzaffarpur Institute of Technology, Bihar University, Muzaffarpur, India, Master0 s degree in electrical engineering (control systems) from the Institute of Tecnology, Banaras Hindu University, Varanasi, India, and pursuing for Ph. D. degree in systems and control engineering from the Indian Institute of Technology Bombay, India. In 1992, he joined the Department of Instrumentation & Control Engineering, Regional Engineering College Jalandhar, India, where he served as lecturer till 2000. He is currently serving as associate professor at College of Military Engineering, Pune, India. He is a life member of the Institution of Engineers. His research interests include modeling and simulation of large systems, robust stability and control, nonlinear system analysis and control, and conditioning of energy from non-conventional sources. E-mail: rkthakur@sc.iitb.ac.in (Corresponding author) Vivek Agarwal received the Bachelor’s degree in physics from St. Stephen0 s College, Delhi University, Delhi, India, Master0 s degree in electrical engineering from the Indian Institute of Science, Bangalore, India, and the Ph. D. degree in electrical and computer engineering from the University of Victoria, Victoria, BC, Canada. Subsequently, he worked as a research engineer with Statpower Technologies, Burnaby, BC, Canada. In 1995, he joined the Department of Electrical Engineering, Indian Institute of Technology–Bombay, Mumbai, India, where he is currently a professor. He is a fellow of the Indian National Academy of Engineering, a fellow of the Institute of Electronics and Telecommunication Engineers (IETE), and a life member of the Indian Society for Technical Education. He is a senior member of IEEE and serves on the editorial boards of IEEE Transactions on Power Electronics and Smart Grid. He has received multiples awards and honors for his contributions towards research in various areas. His research interests include modeling and simulation of new power converter configurations, intelligent and hybrid control of power electronic systems, power quality issues, electromagnetic interference (EMI)/electromagnetic compatibility (EMC) issues, and conditioning of energy from nonconventional sources. E-mail: agarwal@ee.iitb.ac.in Paluri S. V. Nataraj is a professor of Systems and Control Engg Group at IIT Bombay. He obtained his Ph. D. from IIT Madras in process dynamics and control in 1987. He then worked in the CAD center at IIT Bombay, India for about one and half years before joining the faculty of the Systems and Control Engineering Group at IIT Bombay in 1988. He is an associate editor of International Journal of Systems Assurance Engineering and Management (Springer), and editor of two international journals– International Journal of Automation and Control (Inderscience) and Opsearch (Springer). His research interests include chemical process control, global optimization, robust stability and control, nonlinear system analysis and control, and reliable computing. E-mail: nataraj@sc.iitb.ac.in