584

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 20, NO. 3, SEPTEMBER 2005

A Synchronous Machine Model With Saturation and Arbitrary Rotor Network Representation
Dionysios C. Aliprantis, Member, IEEE, Scott D. Sudhoff, Senior Member, IEEE, and Brian T. Kuhn, Member, IEEE
AbstractThis paper addresses equivalent circuit and magnetic saturation issues associated with synchronous machine modeling. In the proposed synchronous machine model, the rotor equivalent circuits are replaced by arbitrary linear networks. This allows for elimination of the equivalent circuit parameter identication procedure since the measured frequency response may be directly embedded into the model. Magnetic saturation is also represented in both the - and -axis. The model is computationally efcient and suitable for dynamic time-domain power system studies. Index TermsElectric machines, modeling, power system stability, realization theory, rotating machine transient analysis, synchronous generator transient analysis, synchronous machines, transfer functions.

, , ,

, ,

Rotor network state equation matrices . Column vectors of the matrix . Row vectors of the matrix . Voltage-behind-reactance of an averagevalue exciter-rectier system representation (in volts). Field winding current (in amperes). -axes magnetizing branch current (in amperes). Current owing into the armature side of the rotor networks (in amperes). Stator windings current (in amperes). Moment of inertia ( ). Parks transformation matrix. Stator winding leakage inductance (in Henries). Transient inductance of an average-value exciter-rectier system representation (in Henries). Minimal polynomial of transfer function matrix . Field-to-armature turns ratio. Number of -axes rotor network states.

, , , , , ,

Manuscript received September 17, 2003; revised March 31, 2004. This work was supported by the Naval Combat Survivability Effort under Grant N00024-02-NR-60427. Paper no. TEC-00249-2003. D. C. Aliprantis is with the Greek Armed Forces (e-mail: aliprantis@ alumni.purdue.edu). S. D. Sudhoff is with the Department of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907-1285 USA (e-mail: sudhoff@ecn.purdue.edu). B. T. Kuhn is with SmartSpark Energy Systems, Inc., Champaign, IL 61820 USA (e-mail: b.kuhn@smartsparkenergy.com). Digital Object Identier 10.1109/TEC.2005.845455

Order of -axes rotor network minimal polynomials. Number of poles. Electric power supplied to coupling eld (in watts). Characteristic polynomial of transfer function matrix . Effective resistance of an average-value exciter-rectier system representation (in ohms). Field winding resistance (in ohms). Stator winding resistance (in ohms). Complex frequency (in radians per second). Electromagnetic torque ( m). Prime-mover torque ( m). Voltage across the eld side of the -axis rotor network (in volts). Field winding voltage (in volts). -axes magnetizing branch voltages (in volts). Armature voltage (in volts). Coupling eld energy (in Joules). -axes rotor network states. -axis rotor two-port network transfer function matrix ( ). -axis rotor network transfer function ). ( Constant related to the -axis rotor network admittance at dc ( ). Elements (transfer functions) of . Incremental inverse magnetizing inductance matrix. Inverse magnetizing inductances ( ). Saliency-dependent magnetizing path characteristic constants. Constants of the -axis rotor network transfer functions . Constants of the -axis rotor network , transfer functions . Electrical rotor position (in radians). Mechanical rotor position (in radians). Stator windings leakage ux linkage ). Effective magnetizing ux linkage . -axes magnetizing ux linkage .

0885-8969/$20.00 2005 IEEE

ALIPRANTIS et al.: SYNCHRONOUS MACHINE MODEL WITH SATURATION AND ARBITRARY ROTOR NETWORK REPRESENTATION

585

Stator windings ux linkage . Electrical rotor speed (in radians per second). Mechanical rotor speed (in radians per second). I. INTRODUCTION ATHEMATICAL models of physical devices are abstractions of reality. Their purpose is to portray, with sufcient accuracy and simplicity, the important characteristics of the systems under consideration. For synchronous machines, relatively low-order models may be employed to predict their steady-state or dynamic behavior, with the orthogonal -axes theory [1] used in the majority of cases. Over the years, a plethora of models has been proposed to represent phenomena related to the distributed circuit behavior of rotor windings, magnetic saturation, and saliency-related issues. However, the common thread among them is that they were focused onand, thus, limited bythe specic problem at hand. This work provides an integrated perspective on synchronous machine modeling, realizing the need for a generalized formulation that is compatible with modern computational tools. Synchronous machines are usually employed as generators connected to a power system. In the current IEEE Standard [2], it is acknowledged that synchronous machines may be accurately modeled by two lumped-parameter equivalent circuits representing the - and the -axis, as shown in Fig. 1. The number of rotor damper branches is selected in accordance with the rotor design. In low-order models, these branches correspond to the actual amortisseur windings; higher order models utilize these branches to represent the distributed effects in the rotor iron. The rotor-related part of the equivalent circuits proposed in the standard differs from other conventional equivalent circuits [1], [3] since it contains the differential leakage induc. Its use was proposed by Canay in his seminal paper tance of 1969 [4], to signify unequal coupling between the rotor windings and the stator, and between the rotor circuits themselves. A variety of other circuit structures has also been proposed in the literature [5][8]. equal to zero may be obtained with New circuits having an appropriate transformation, as was shown by Canay in his original paper, and again by Kirtley in [9]. This corresponds to forcing equal magnetic coupling between all rotor circuits and the stator windingoften called the equal mutuals base. However, the equal mutuals base is restricted to the case of three coupled circuits; that is, when the rotor has, at most, a eld winding and one damper [10]. The most unconstrained form of equivalent circuit is one with differential leakage inductances between all damper circuits. Important theoretical results about equivalent circuits were published in [11] and [12] where it was shown that ladder network representation of a given there is no unique rotor two-port network with a prescribed impedance matrix. A further disadvantage of using a mathematical transformation to ensure a specic network structure (one with no differential leakage) is that it causes the new values of the stator leakage and magnetizing inductance to lose their original physical meaning.

Fig. 1.

IEEE Standard model of type 3.3.

Fig. 2. Proposed synchronous machine model.

If magnetizing path saturation is to be consideredeither by adjusting the magnetizing branch inductance or by replacing it by a nonlinear elementthe physical signicance of these elements should be preserved [13]. Although some physical signicance may be attributed to the parameters of models with up to three damper circuits [14], it is gradually lost as the order is further increased, as is recommended for accurate transient stability or subsynchronous resonance studies [15]. The purpose of this work is to address these equivalent circuit issues. In the model that is set forth herein, the equivalent-circuit structure of the rotor is replaced by a completely arbitrary linear circuit, a two-port network for the -axis, and a single branch for the -axis, as shown in Fig. 2. This approach has been adopted from previous modeling work on induction machines [16], emphasizing the importance of the rotors actual inputoutput behavior rather than the debatable physical meaning of the equivalent circuit parameters. It offers the advantage that once the transfer functions of these linear circuits are determinedusing standstill frequency response tests, for example, they may be immediately incorporated in the model [17]. It does not require the tedious and time-consuming process of equivalent circuit parameter identication to which numerous studies have been devoted [18], [19]. Indeed, this problem is highly nonlinear, and the solution algorithms often face convergence issues, accentuated by the fact that there is no unique solution. Furthermore, for modern computer simulation software, such as Matlab/Simulink [20] or ACSL [21], the entry of a specic equivalent circuit

586

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 20, NO. 3, SEPTEMBER 2005

structure is not required. Rather, it is more efcient to directly provide a state-space representation of the system to be simulated. The proposed model possesses such a form. Apart from equivalent circuit-related problems, synchronous machine research has also focused on the accurate incorporation of magnetic saturation, which has been shown to considerably affect their operating characteristics [22]. Different methods to model this complicated phenomenon were analyzed in a variety of publications, such as [23][31]. An alternative approach is to employ articial neural networks to represent the magnetizing path saturation [32][34], or the variation of rotor parameters [35], [36]. The form of the proposed model readily lends itself to the modication of the magnetizing inductances for saturation (and crosssaturation) modeling in both axes. Recent evidence suggests that lumping saturation effects in the magnetizing branch is a reasonable assumption [37], [38] and this hypothesis is further corroborated by our results. Although the leakage ux paths may saturate due to excessive current or by high magnetizing ux levels, synchronous generators are usually operated close to their nominal values of ux, so the leakage inductance may be assumed to remain constant for a wide range of studies. A derivation of the models torque equation, and restrictions on the magnetizing inductances arising from the assumption of a lossless (conservative) coupling eld will be presented in later sections. II. PROPOSED SYNCHRONOUS MACHINE MODEL From a computer simulation point of view, the proposed model is of the voltage-in, current-out type; the stator and eld winding voltages are the inputs, while the stator and eld winding currents are the outputs. The model is computationally efcient in the sense that it is noniterative at each time step, and uses only a minimum number of states obtained from a minimal realization of the measured inputoutput behavior. It is especially suitable for time-domain dynamic simulations of power systems, as well as for the design and optimization of control schemes. A. Notation In order to assist the reader, the papers nomenclature is dened. Throughout this work, matrix and vector quantities appear in bold font. The primed rotor quantities denote referral to the stator through the turns ratio, which is dened as the ratio of eld-to-armature turns [39]. The analysis takes place in the rotor reference frame; the often used superscript [1] is omitted for convenience. The electrical rotor poand electrical rotor speed are times the mesition and mechanical speed, , where chanical rotor position is the number of poles. The transformation of stationary to variables in the rotor reference frame is dened by [1] (1) where

B. Voltage Equations The stator voltage equations may be expressed in ables as vari-

(3) where , , and denote stator winding (phase-toneutral) voltages, currents owing into the machine terminals, and ux linkages, respectively, and is the stator winding resistance. Transforming (3) to the rotor reference frame yields (4) where . It will be assumed hereafter that the zero-sequence variables can be neglected. The state equations of the -axis two-port network (Fig. 2) as may be expressed by a linear system of order

(5) (6) , , , and . The contents of and were written as the column , , and the row vectors , . The magnevectors tizing branch voltages are equal to the derivatives of the corresponding magnetizing ux linkages, which will be computed in the ensuing analysis. may be eliminated from the equations, since The voltage it is related to the eld voltage and current by . Using this, the following state equation is obtained: where (7) Similarly, the -axis state equations are (8) (9) where , , , , and is the order of the -axis system. It will be useful to derive expressions for the current derivatives. From (6)(9)

(10) (2) (11)

ALIPRANTIS et al.: SYNCHRONOUS MACHINE MODEL WITH SATURATION AND ARBITRARY ROTOR NETWORK REPRESENTATION

587

C. Rotor Transfer Functions In the frequency domain, the rotor currents and voltages are related by the transfer functions

The formulation of the -axis transfer function is more straightforward due to the absence of a eld winding. In particular, the -axis transfer function may be expressed as (22) In this case, poles and zeros at the origin are not allowed.

(12) (13) and the tilde are used to denote phasor quantiwhere ( , , ), and ties. The elements of are strictly proper rational polynomial functions of the complex frequency of the form (14) and may not be simultaneously equal to The coefcients zero since that would result in a pole-zero cancellation at the origin. This is a representation of a most general form, but a simplied version may be obtained for the -axis if the physics of the rotor are taken into account. To this end, by applying Faradays law to the eld winding ), it can be seen that there is no (i.e., dc voltage drop besides the ohmic drop of the windings resistance (which is external to the two-port network). This implies that there exists a direct path for dc current between the primary and secondary sides of the two-port network. of the polynomials Under this assumption, the constants will all have the same absolute value, and the system will possess a pole at the origin. To see this, assume that the eld side so that of the admittance block is short circuited (15) (16) As the frequency approaches zero , the hypothesis that at low frequency the two-port network behaves as an ideal series , so inductor implies that (17) Since the denominator of (14) corresponds to the least common denominator of all elements of , the element of is equal to of . Similarly, setting and letting the frequency approach zero yields (18) Thus, the element of is equal to of . This observation, coupled with the fact that these admittances be(again because of the assumption that come innite as the network acts as an ideal series inductance at sufciently low frequency), implies that the -axis transfer functions must have the following forms: (19) (20) (21)

D. Realization Theory The rotor transfer functions are the starting point for the computation of the time-domain state matrices, as used in (7) and (8). The problem is to determine an internal, state-space description of a linear system, given its external, inputoutput description. This is the subject of system realization theory [40]. Numerous algorithms exist for deriving a realization. However, for increased computational efciency, it is desirable to obtain a system of the least possible order, a so-called minimal realization. In contrast to the single-input/single-output case, where it is rather straightforward to obtain a realization, the multiple-input/multiple-output case (like the two-port -axis rotor system) is more complicated. The difculty arises when determining the minimal realization order. Consider a transfer func, where is a tion written in the form matrix of polynomials and is the minimal polynomial of .1 The roots of constitute a subset of the eigenvalues of the minimal realization; hence, a minimal polynomial with two roots may correspond to a system with a second-, third-, or fourth-order minimal realization. In the case of the proposed model, the transfer function matrix entries are the functions (19)(21), the degree of the minimal polynomial is , and the order of the minimal realization is . This is proved using the following theorem: The order of any minimal realization is equal to the degree of the characteristic polynomial of the transfer function matrix [40, p. 397].2 , given the functional forms It remains to compute (19)(21). Let us write the -axis transfer function as (23) where (24) , It is assumed that no common factors exist between , , and . To obtain , it is necessary to . The rst-order minors compute all nonzero minors of , and their monic least common denomare the entries of . The inator is (by denition) the minimal polynomial second-order minor is equal to the determinant (25) Taking into account the specied forms of the transfer functions has a (19)(21), it is readily shown that the numerator of
1The minimal polynomial is dened herein as the monic least common denominator of the transfer function entries. A polynomial is monic when the highest order term has a coefcient of one. 2The characteristic polynomial p (s) is dened as the monic least common denominator of all nonzero minors of (s).

588

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 20, NO. 3, SEPTEMBER 2005

root at the origin, which cancels out with one of the two zero . Hence roots of (26) The degree of the characteristic polynomialthe order of the . An algorithm to minimal realizationis thus obtain a minimal state-space realization from the transfer function matrix is described in the Appendix. The order of the -axis realization, which is a single. input/single-output system, is E. Leakage and Magnetizing Path Magnetics The stator ux linkage is separated into leakage and magnetizing ux terms as (27) is the stator leakage ux, and is the magnewhere is used tizing ux. The same value of leakage inductance for both the - and the -axis. The functional forms of main path saturation in the proposed model are similar to [30]. Since in the general case of a salient rotor machine, the magnetizing magnetomotive force (mmf) and ux vectors are not aligned, the following relations may be assumed for the magnetizing path: (28) (29) The magnitude of the effective magnetizing ux vector is dened by (30) where is a saliency-dependent parameter. Differentiating (28), (29) with respect to time yields (31) where

the electric power supplied to the coupling eld may be expressed after manipulation as

(33) The electrical energy provided to the coupling eld is partly stored, and the rest is transferred to the mechanical system. If denotes the coupling eld energy and the EM torque, then the power balance may be written as (34) where the product is positive when energy is supplied to the rotor. Equating (33) and (34), and solving for the rate of change of the eld energy yields

(35) Hence, the differential change of the coupling eld energy may be written as (36) where (37) (38) (39) The change of coupling eld energy from an initial state to an arbitrary nal state is obtained by integrating (36)

(32) The subscript denotes incremental value. F. Torque Equation The proposed models electromagnetic (EM) torque equation may be derived by examining the energy balance of the machines coupling eld [1]. The coupling eld is created by the magnetizing ux, which links both stator and rotor circuits. The electric power supplied to the coupling eld is equal to the input power, minus the power lost in the stator and eld resistances, minus the power that supplies the stator leakage eld, minus the power that is dissipated or stored inside the rotor admittance block. Using the equivalent circuit of Fig. 2 and (27), (40)

Since the eld is assumed to be conservative, the integration may be performed over an arbitrary trajectory. Assume that the initial energy is , integrate the rst term from to , while the uxes are maintained at zerowhich forces to be zero as well. This transition does not change the eld energy. Then, consecutively integrate each ux from zero to an arbitrary nal value, while keeping and the other ux constant. Recall that the transformation to the rotor reference frame eliminates the dependence of the magnetizing inductances from the rotor position. The magnetizing currents are independent of as in (28) and

ALIPRANTIS et al.: SYNCHRONOUS MACHINE MODEL WITH SATURATION AND ARBITRARY ROTOR NETWORK REPRESENTATION

589

(29). Therefore, the nal value of eld energy is independent of the angular position , that is (41) This observation coupled with (37) yields the following wellknown expression for the EM torque: (42) For generator action (and ), the torque will be negative.

Next, the time derivatives of the stator ux linkages are calculated from (4) (49) (50) as functions of state variables and model inputs (the stator voltages). However, they are only evaluated as an intermediate calare not states. culation; they are not integrated since The differentiation with respect to time of (27) yields an alternate expression for the derivatives of the stator ux linkages [cf. (49) and (50)] (51) (52) The derivatives of the rotor and magnetizing currents may be evaluated using (10) and (11) and (31), so

G. Restrictions on the Inverse Magnetizing Inductances In (28) and (29), the inverse magnetizing inductances were dened as any arbitrary function of ux. However, to be consistent with the assumption of a lossless coupling eld, certain modeling restrictions must be imposed [41], [42]. Specically, the coupling elds energy expression, which in view of (41), has become

(43) must satisfy the requirements of a conservative eld. A necessary and sufcient condition for this is [43] (53) (44) Substitution of (28) and (29) into (44) yields (45) Canceling common terms and integrating both sides yields (46) This restriction has to be enforced during the magnetizing characteristics curve-tting procedure. Note that it also renders the incremental inverse inductance matrix symmetric. The method for determining the constants and is described in detail in [17]. H. Model Integration . The The state variables are selected as goal of the ensuing analysis is the formulation of equations for the time derivatives of the state variables. First, note that both the rotor and magnetizing currents have been previously expressed as functions of the states: the rotor currents are given in terms of the rotor admittance states from (6) and (9); the magnetizing currents depend on the magnetizing ux states, as seen from (28)(30). Hence, the stator currents (47) (48) are also functions of state variables. (55) (54) The following linear system of equations may therefore be formulated:

(56) and are obtained where the quantities from (49) and (50). The solution of (55) and (56) yields the time derivatives of two of the state variables (the magnetizing and ). Apart from being integrated themux linkages selves, these derivatives are also inserted in (7) and (8)as and for calculating the the magnetizing voltages and . derivatives of the remaining state variables

590

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 20, NO. 3, SEPTEMBER 2005

I. Model Summary The proposed model is of the voltage-in, current-out type; inputs are the armature and eld winding voltages, as well as the rotor angular position and speed, whereas outputs are the armature and eld winding currents, and the machines developed EM torque. The models computational structure may be summarized as follows. 1) Transform the stator voltages to the rotor reference frame using (1). 2) Compute the magnetizing currents, using (28) and (29). 3) Compute the rotor currents using (6) and (9). 4) Compute the -axis armature currents using (47) and (48). 5) Calculate the armature currents by applying the inverse of transformation (1). 6) Calculate the EM torque from (42). 7) Compute the derivatives of the stator ux linkages, using (27), (49), and (50). 8) Compute the incremental inverse inductance matrix (32). 9) Substitute known quantities into (55) and (56), and solve the 2 2 linear system for the derivatives of the magnetizing ux linkages. 10) Compute the derivatives of the rotor network states using (7) and (8). 11) Interface the synchronous machine model with the rest of the system that is simulated, numerically integrate the state equations, and repeat from step 1). The model is not computationally intensive, since it only involves simple numerical computations at each time step. Thus, it can be readily implemented in a dynamic simulation environment, such as Matlab/Simulink [20]. In order to initialize the model from a load-ow study where the machine is normally connected to a generator bus or a swing bus, the model states may be calculated by solving a set of nonlinear equations. Specically, it can be seen from (7), (8), and the proposed realization (75) that in the steady-state, all rotor network states must be equal , which is directly related to the to zero; the exception is . The eld winding voltage through mathematical expressions for the prespecied quantities ( , , or ) may be readily manipulated and written in terms of and . The nonlinear system may then be solved for the magnetizing ux states. J. External Impedance Incorporation The model is exible enough to allow the incorporation of an external resistance and inductance, connected in series with the eld winding. For example, such would be the case of a brushless excitation system, when the detailed model of the exciter-rectier system is replaced by a nonlinear average-value model. Essentially, this simplication produces a voltage behind reactance representation of the exciter. As was shown in [44], the effective voltage at the generator eld has the form (57) (62) which was obtained by a procedure analogous to the one described in the previous section. III. EXPERIMENTAL VALIDATION The experimental setup is depicted in Fig. 3 and contains a LeroySomer brushless synchronous generator, model LSA 432L7. This is a salient four-pole machine rated for 59 kW,

Fig. 3. Schematic of experimental setup; the brushless synchronous generator is feeding a nonlinear rectier load.

where is the voltage behind the effective resistance and the transient inductance ; the primes denote that all quantities have been referred to the stator. These voltage drops may be incorporated in the proposed model in a straightforward manner. In particular, the voltage at the eld side of the -axis admittance block may be written as (58) Using (5) and (6), this equation becomes

(59) and solving for yields

(60) . After the substitution of (60) where into (5), the following modied state equation is obtained:

(61) where denotes the identity matrix of dimension . The derivatives of the magnetizing ux linkages are given by the solution of the linear system of equations formed by (56) and

ALIPRANTIS et al.: SYNCHRONOUS MACHINE MODEL WITH SATURATION AND ARBITRARY ROTOR NETWORK REPRESENTATION

591

600 V, at 1800 r/min. The generators prime mover is a Dyne Systems 110-kW, 590 m, vector-controlled induction motorbased dynamometer that is programmed to maintain constant rated speed. The voltage regulator uses a proportional-integral control strategy to maintain the commanded voltage (560 V line-to-line, fundamental rms) at the generator terminals; the brushless exciters eld current is controlled with a hysteresis modulator. The rotors angular position and speed are measured with an optical position encoder that is tted on the machine shaft. The generator is loaded with an uncontrolled rectier that lter. feeds a resistive load through an The complete methodology for characterizing the main synchronous machine is described in detail in [17]. The obtained , parameters and functional forms are as follows: , ,

(63) (64) (65) (66) (67) (68) and , . Note that where is valid for the expression for (which is above the knee of the saturation curve). For greater values, the rational function diverges, and an appropriate continuation is dened so that the ux linkage increases linearly ). The load paramewith current (with the same slope as at mH, F, . With ters are this load, the machine operates at the region of the knee of the [17]). The brushless exmagnetizing curve ( citer is represented by an average-value model that incorporates magnetic hysteresis using Preisachs theory; this model is discussed in detail in [45] and a forthcoming publication [46]. The prime mover models output is the mechanical torque, which is inserted in the well-known equation of motion. The mechanical rotor speed may thus be computed by numerically inte. The electrical rotor pograting sition is then calculated by integrating the rotors speed (i.e., ). The prime movers model is documented in [45]. For the rst experiment, the switch is closed so that the . The generators voltage total resistive load is reference is modied according to the prole shown in Fig. 4. This series of commanded voltage steps creates an extended period of signicant disturbances, and tests the validity of the model for large-transients simulations. The actual voltage exhibits an overshoot, which is more pronounced for the faster slew rate steps. Moreover, due to the exciters magnetically hysteretic behavior, it does not fall to zero. The predictions

Fig. 4. Plots of the commanded and actual line-to-line voltage envelope, = which is computed from the synchronous reference frame voltages v [3(v + v )] . Each of the seven trapezoid shaped blocks is characterized by a different slope (the same for rise and fall) and peak voltage: (1) 20 000 V/s, 560 V; (2)(4) 2000 V/s, 560 V, 420 V, 280 V, respectively; (5)(7) 400 V/s, 560 V, 420 V, 280 V, respectively. (Note: the above voltage values correspond to rms quantities.)

Fig. 5. Shaft speed variation around the commanded value of 1800 r/min.

for mechanical speed (Fig. 5) are in excellent agreement with the experimental results. Finally, detailed voltage and current waveforms are depicted in Fig. 6. Since the proposed model is based on the -axes theory, higher harmonics attributable to the machines design are not represented. This reects on

592

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 20, NO. 3, SEPTEMBER 2005

IV. CONCLUSION This work presents an integrated perspective on synchronous machine modeling, using arbitrary transfer function representations that replace the rotors equivalent circuit structures. This approach offers several advantages, such as the direct incorporation of frequency response results into the modelwithout further consideration of equivalent circuit parameter identicationand accurate representation of magnetic saturation effects. The model retains the computational efciency of the -axes theoretical framework and is suitable for small- and large-signal time-domain simulations of power systems. APPENDIX REALIZATION ALGORITHM The following algorithm produces a realization with matrix diagonal [40]. It is valid only for the case where the roots of the minimal polynomial are distinct. It is assumed that the transfer function matrix elements are (19)(21). 1) Compute the roots of the minimal polynomial (69) 2) Expand into partial fractions (70) The 2 2 residue matrices may be computed by (71)
Fig. 7. Line-to-line voltage envelope. (Lowpass ltered;  = 2:65 ms.)

Fig. 6.

Steady-state voltage and current waveforms (560 V, line-to-line, rms).

and are of full rank; however (72) is a matrix of rank 1. 3) Write (73) for example, by computing the LU decomposition. The , are 2 2. may be factored as matrices

Fig. 8.

Shaft speed variation around the commanded value of 1800 r/min.

(74) the voltage waveforms of Fig. 4, wherein the experimental waveform contains more ripple than the simulated waveform. However, the harmonics that are caused by the nonlinearity of the load are predicted accurately. The second experiment involves sudden load changes. Iniis open; at , it is closed, and at tially, the switch , it is opened again. In Fig. 7, a lowpass-ltered version of the line-to-line voltage envelope is depicted. On average, the simulated and experimental waveforms are similar. As discussed above, the experimental voltage includes higher-order harmonics caused by slot effects. The mechanical speed waveforms are illustrated in Fig. 8. 4) The realization is given by

..

. . .

(75)

ALIPRANTIS et al.: SYNCHRONOUS MACHINE MODEL WITH SATURATION AND ARBITRARY ROTOR NETWORK REPRESENTATION

593

REFERENCES
[1] P. C. Krause, O. Wasynczuk, and S. D. Sudhoff, Analysis of Electric Machinery. New York: IEEE Press, 1995. [2] Guide for Synchronous Generator Modeling Practices in Stability Analyzes, IEEE Std. 1110, Nov. 1991. [3] C.-M. Ong, Dynamic Simulation of Electric Machinery Using Matlab/ Simulink. Englewood Cliffs, NJ: Prentice-Hall, 1998. [4] I. M. Canay, Causes of discrepancies on calculation of rotor quantities and exact equivalent diagrams of the synchronous machine, IEEE Trans. Power App. Syst., vol. PAS-88, no. 7, pp. 11141120, Jul. 1969. [5] R. P. Schulz, W. D. Jones, and D. N. Ewart, Dynamic models of turbine generators derived from solid rotor equivalent circuits, IEEE Trans. Power App. Syst., vol. PAS-92, no. 3, pp. 926933, May/Jun. 1973. [6] H. Bissig, K. Reichert, and T. S. Kulig, Modeling and identication of synchronous machines, a new approach with an extended frequency range, IEEE Trans. Energy Convers., vol. 8, no. 2, pp. 263271, Jun. 1993. [7] I. M. Canay, Determination of the model parameters of machines from the reactance operators x (p); x (p) (evaluation of standstill frequency response test), IEEE Trans. Energy Convers., vol. 8, no. 2, pp. 272279, Jun. 1993. [8] A. Keyhani and H. Tsai, Identication of high-order synchronous generator models from SSFR test data, IEEE Trans. Energy Convers., vol. 9, no. 3, pp. 593603, Sep. 1994. [9] J. L. Kirtley, On turbine-generator rotor equivalent circuits, IEEE Trans. Power Syst., vol. 9, no. 1, pp. 262271, Feb. 1994. [10] M. R. Harris, P. J. Lawrenson, and J. M. Stephenson, Per-Unit Systems With Special Reference to Electrical Machines. Cambridge, U.K.: Cambridge Univ. Press, 1970. [11] J. Verbeeck, R. Pintelon, and P. Guillaume, Determination of synchronous machine parameters using network synthesis techniques, IEEE Trans. Energy Convers., vol. 14, no. 3, pp. 310314, Sep. 1999. [12] J. Verbeeck, R. Pintelon, and P. Lataire, Relationships between parameter sets of equivalent synchronous machine models, IEEE Trans. Energy Convers., vol. 14, no. 4, pp. 10751080, Dec. 1999. [13] I. Kamwa and P. Viarouge, On equivalent circuit structures for empirical modeling of turbine-generators, IEEE Trans. Energy Convers., vol. 9, no. 3, pp. 579592, Sep. 1994. [14] I. M. Canay, Physical signicance of sub-subtransient quantities in dynamic behavior of synchronous machines, Proc. Inst. Elect. Eng. B, vol. 135, no. 6, pp. 334340, Nov. 1988. [15] P. L. Dandeno and M. R. Iravani, Third order turboalternator electrical stability models with applications to subsynchronous resonance studies, IEEE Trans. Energy Convers., vol. 10, no. 1, pp. 7886, Mar. 1995. [16] S. D. Sudhoff, D. C. Aliprantis, B. T. Kuhn, and P. L. Chapman, An induction machine model for predicting inverter-machine interaction, IEEE Trans. Energy Convers., vol. 17, no. 2, pp. 203210, Jun. 2002. [17] D. C. Aliprantis, S. D. Sudhoff, and B. T. Kuhn, Experimental characterization procedure for a synchronous machine model with saturation and arbitrary rotor network representation, IEEE Trans. Energy Convers., to be published. [18] L. Salvatore and M. Savino, Experimental determination of synchronous machine parameters, Proc. Inst. Elect. Eng., B, vol. 128, no. 4, pp. 212218, Jul. 1981. [19] I. M. Canay, Modeling of alternating-current machines having multiple rotor circuits, IEEE Trans. Energy Convers., vol. 8, no. 2, pp. 280296, Jun. 1993. [20] The MathWorks, Inc., Simulink Reference, Natick, MA, 2002. [21] AEgis Technologies Group, Inc., Advanced Continuous Simulation Language (ACSL) Reference Manual, Huntsville, AL, 1999. [22] R. G. Harley, D. J. N. Limebeer, and E. Chirricozzi, Comparative study of saturation methods in synchronous machine models, Proc. Inst. Elect. Eng. B, vol. 127, no. 1, pp. 17, Jan. 1980. [23] G. R. Slemon, Analytical models for saturated synchronous machines, IEEE Trans. Power App. Syst., vol. PAS-90, no. 2, pp. 409417, Mar./Apr. 1971. [24] J. E. Brown, K. P. Kovcs, and P. Vas, A method of including the effects of main ux path saturation in the generalized equations of A.C. machines, IEEE Trans. Power App. Syst., vol. PAS-102, no. 1, pp. 96103, Jan. 1983. [25] R. S. Ramshaw and G. Xie, Nonlinear model of nonsalient synchronous machines, IEEE Trans. Power App. Syst., vol. PAS-103, no. 7, pp. 18091815, Jul. 1984. [26] G. Xie and R. S. Ramshaw, Nonlinear model of synchronous machines with saliency, IEEE Trans. Energy Convers., vol. 1, no. 3, pp. 198204, Sep. 1986.

[27] F. P. de Mello and L. N. Hannett, Representation of saturation in synchronous machines, IEEE Trans. Power Syst., vol. 1, no. 4, pp. 818, Nov. 1986. [28] J. O. Ojo and T. A. Lipo, An improved model for saturated salient pole synchronous motors, IEEE Trans. Energy Convers., vol. 4, no. 1, pp. 135142, Mar. 1989. [29] J. Tamura and I. Takeda, A new model of saturated synchronous machines for power system transient stability simulations, IEEE Trans. Energy Convers., vol. 10, no. 2, pp. 218224, Jun. 1995. [30] S.-A. Tahan and I. Kamwa, A two-factor saturation model for synchronous machines with multiple rotor circuits, IEEE Trans. Energy Convers., vol. 10, no. 4, pp. 609616, Dec. 1995. [31] K. A. Corzine, B. T. Kuhn, S. D. Sudhoff, and H. J. Hegner, An improved method for incorporating magnetic saturation in the q-d synchronous machine model, IEEE Trans. Energy Convers., vol. 13, no. 3, pp. 270275, Sep. 1998. [32] H. Tsai, A. Keyhani, J. A. Demcko, and D. A. Selin, Development of a neural network based saturation model for synchronous generator analysis, IEEE Trans. Energy Convers., vol. 10, no. 4, pp. 617624, Dec. 1995. [33] S. Pillutla and A. Keyhani, Neural network based saturation model for round rotor synchronous generator, IEEE Trans. Energy Convers., vol. 14, no. 4, pp. 10191025, Dec. 1999. [34] H. B. Karayaka, A. Keyhani, B. L. Agrawal, D. A. Selin, and G. T. Heydt, Identication of armature, eld, and saturated parameters of a large steam turbine-generator from operating data, IEEE Trans. Energy Convers., vol. 15, no. 2, pp. 181187, Jun. 2000. [35] S. Pillutla and A. Keyhani, Neural network based modeling of round rotor synchronous generator rotor body parameters from operating data, IEEE Trans. Energy Convers., vol. 14, no. 3, pp. 321327, Sep. 1999. [36] H. B. Karayaka, A. Keyhani, G. T. Heydt, B. L. Agrawal, and D. A. Selin, Neural network based modeling of a large steam turbine-generator rotor body parameters from on-line disturbance data, IEEE Trans. Energy Convers., vol. 16, no. 4, pp. 305311, Dec. 2001. [37] J. Verbeeck, R. Pintelon, and P. Lataire, Inuence of saturation on estimated synchronous machine parameters in standstill frequency response tests, IEEE Trans. Energy Convers., vol. 15, no. 3, pp. 277283, Sep. 2000. [38] N. Dedene, R. Pintelon, and P. Lataire, Estimation of a global synchronous machine model using a multiple-input multiple-output estimator, IEEE Trans. Energy Convers., vol. 18, no. 1, pp. 1116, Mar. 2003. [39] Test Procedures for Synchronous Machines, IEEE Std. 115, Dec. 1995. [40] P. J. Antsaklis and A. N. Michel, Linear Systems. New York: McGrawHill, 1997. [41] J. A. Melkebeek and J. L. Willems, Reciprocity relations for the mutual inductances between orthogonal axis windings in saturated salient-pole machines, IEEE Trans. Ind. Appl., vol. 26, no. 1, pp. 107114, Jan./Feb. 1990. [42] P. W. Sauer, Constraints on saturation modeling in AC machines, IEEE Trans. Energy Convers., vol. 7, no. 1, pp. 161167, Mar. 1992. [43] G. B. Thomas, Calculus and Analytic Geometry, 4th ed. Reading, MA: Addison-Wesley, 1968. [44] S. D. Sudhoff, K. A. Corzine, H. J. Hegner, and D. E. Delisle, Transient and dynamic average-value modeling of synchronous machine fed loadcommutated converters, IEEE Trans. Energy Convers., vol. 11, no. 3, pp. 508514, Sep. 1996. [45] D. C. Aliprantis, Advances in electric machine modeling and evolutionary parameter identication, Ph.D. dissertation, Purdue University, West Lafayette, IN, Dec. 2003. [46] D. C. Aliprantis, S. D. Sudhoff, and B. T. Kuhn, A brushless exciter model incorporating multiple rectier modes and Preisachs hysteresis theory, IEEE Trans. Energy Convers., to be published.

Dionysios C. Aliprantis (M04) received the electrical and computer engineering diploma from the National Technical University of Athens, Athens, Greece, in 1999. He received the Ph.D. degree in electrical and computer engineering from Purdue University, West Lafayette, IN, in 2003. Currently, he is serving in the armed forces of Greece. His interests include the modeling and simulation of electric machines and power systems, and evolutionary optimization methods.

594

IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 20, NO. 3, SEPTEMBER 2005

Scott D. Sudhoff (SM01) received the B.S. (Hons.), M.S., and Ph.D. degrees in electrical engineering from Purdue University, West Lafayette, IN, in 1988, 1989, and 1991, respectively. Currently, he is a Full Professor at Purdue University. From 1991 to 1993, he was Part-Time Visiting Faculty with Purdue University and as a Part-Time Consultant with P. C. Krause and Associates, West Lafayette, IN. From 1993 to 1997, he was a Faculty Member at the University of Missouri-Rolla. He has authored many papers. His interests include electric machines, power electronics, and nite-inertia power systems.

Brian T. Kuhn (M93) received the B.S. and M.S. degrees in electrical engineering from the University of Missouri-Rolla in 1996 and 1997, respectively. He was a Research Engineer at Purdue University, West Lafayette, IN, from 1998 to 2003. Currently, he is a Senior Engineer with SmartSpark Energy Systems, Inc., Champaign, IL. His research interests include power electronics and electrical machinery.