Harmonic modelling of synchronous machines P.M. Hart, M.EngSc W.J. Bonwick, PhD, FlEE Indexing terms: Power systems and plant, Synchronous motors Abstract: The connection of power converter and other nonlinear loads to the electrical power system results in other system components being subject to voltage waveform distortion. In order to predict the resultant heating and harmonic current flows, it is necessary to have fairly accurate models for components such as synchronous machinery. The paper presents a theoretical and experimental study into the electrical behaviour of synchronous machines when subjected to harmonic voltage disturbance at the terminals. It is shown that the time-varying nature of the machine inductances results in current flow at both the applied and at an associated harmonic frequency. Consequently, it is inadequate to model the machinery by a pure reactance at each harmonic frequency. List of symbols phase indices index = harmonic number (always positive) = current and voltage at harmonic frequency, no = current and voltage at harmonic frequency, (n T 2)o = angle of field winding at t = 0 (see eqn. 3) a, b, c = f = field n ' n y Vn - . - 1n 9 u n a i La, L f f L&, Lb L&',Lb' zn z Zn TI *XY *x N 9: 0 PPS nPs 1 = machine = appaient inductances impedance at frequency, n o = apparent impedance at frequency, (n T 2)w = supply impedance at frequency, n o = supply impedance at frequency, ( n f 2 ) o = flux linking winding x due to current in winding, y = total flux linking winding, x = denotes phasor = denotes phase angle = positive phase sequence = negative phase sequence small to medium sized distorting loads if they are within preset size limits, for large loads a detailed harmonic penetration study is necessary [l]. The study will usually extend back to the zone substation for strong systems, but may need to extend to generators for small systems. Power converter loads act as harmonic sources of current which propagate through the distribution system. Shunting loads within the distribution system provide harmonic current paths, and thus affect the overall current flow in the system. At supply frequency it is usual for the impedance of shunting loads to greatly exceed that of the system impedance as seen at the load. This is not necessarily true at harmonic frequencies because resonances in the power system may occur, expecially if capacitors are connected nearby. If an accurate harmonic penetration study is to be achieved it will be necessary that system components including shunting loads are accurately modelled. The study is made more difficult because of changing loads and changing circuit arrangements. Consequently, the study should include likely extremes of network loading variation and anticipated network changes. The paper considers the harmonic modelling of the synchronous machine, and presents an analysis of the harmonic currents and voltages that result at the machine terminals when the machine is subjected to harmonic disturbance of the supply. The effects on shaft torques are not considered here, but they have been treated by Williamson 161. Basic plant harmonic item models are given in References 2, 3,4 and 5. A second major concern is whether a machine will overheat when subjected to disturbance. This can be predicted if the machine impedance at harmonic frequencies is known. The losses in machines drawing nonsinusoidal current wneforms have been studied by Williamson et al. [6, 71. It is usual to model synchronous machinery by a reactance which is based on either the negative sequence or the transient reactances [2, 31, and induction machines by an inductance value based on the direct on-line current. Losses may be included by adding resistances [5] as shown in Fig. 1. In the paper, the authors report Introduction The increased use of power converters in industry raises concern about harmonic disturbance levels in the power system. Whilst the supply authority generally accepts b Paper 5797B (Pl), first received 5th January and in revised form 26th June 1987 The authors are with Monash University, Department of Electrical Engineering, Clayton, Victoria 3168, Australia 52 Fig. 1 Harmonic equivalent circuits for synchronous and induction machines a Circuit based on negative sequence impedance b Formal equivalent circuits IEE PROCEEDINGS, Vol. 135, P I . B, N o . 2, M A R C H 1988 on their own theoretical work and discuss supporting evidence. It is shown that the model of Fig. l a is generally inappropriate for synchronous machine modelling. 2 Synchronous machine and 2.1 Application of a single current harmonic The stator voltage components which result when balanced harmonic currents of harmonic number n are applied to a 3-phase synchronous machine are determined. The synchronous machine has a single field winding on a round rotor and no subtransient properties. Subtransient affects are considered in Section 2.3. In complex exponential notation the applied currents are + 3: in] i, = in exp j [ n o t + 3: in T 27c/3] i, = in exp j [ n o t + 3: in f 2n/3] i, = in expj[nwt (1) The upper signs in eqn. 1 apply for a positive-phase sequence applied-voltage set, and the lower for negative phase sequence. Expressions for voltage, current and flux may be obtained by taking the real part of each relevant complex equation. The flux linking the rotor field is $y = in L,,(expj[nwt where the inductances L & - Lh - (7) 4 Lrr This equation is valid for a round rotor machine only as saliency effects on the inter-phase inductances Lab and La, are ignored. Because there is a single field winding, the rotor flux pulsates along the field axis. Such a flux can be represented as the resultant of two counter rotating waves (relative to the field winding) of constant amplitude, one at rotational frequency n o and the other at (n f 2)w with respect to the stator. This is the interpretation of eqn. 6. The a-phase stator voltage is obtained by differentiating eqn. 6 and is 2 v, = injno( L d \ + 3: in] COS (wt + a) + exp j[nwt + 3: in T 2n/3] x COS (wt + 2~/3) + exp j[nwt + 3: in & 2n/3] x cos ( o f + a + 2x/3)} = $ L a f i nexp j [ ( n 1)wt T u + 3: in] ub = injnw( La, COS (ut + a). (3) Inductance symbols are defined in Reference 8. The field winding is assumed to be rotating at synchronous speed. Thus the field current induced is at frequency ( n T l ) w and is + x exp j [ ( n f 1)wt T a 3: in + A] (4) where i.= arctan (rr/[n & l ) o L r r ] ) . This equation is not valid for the fundamental ( n = 1 ) pps supply as there is no relative motion of field and flux wave under this condition, and hence no induction. Considerable simplification results if (n T l)wLj, % r, (5) This will be assumed to apply, and is well founded provided losses do not rise too rapidly with frequency. For laminated rotor machines, eddy current losses will be small, whereas in solid pole machines, eddy currents may dramatically increase losses, and decrease the effective inductance L,, and La,. The flux linking the stator a-phase has contributions from the field and the b and c phase windings and is IEE PROCEEDINGS, Vol, 135. Pt. B, No. 2, M A R C H 1988 ' exp j[nwt + 3: in] / (6) .) L&+ L' + inj ( n T 2)w (L' (2) where the mutual inductance between the field winding and the a-phase winding is + 3: in] lL') Using a similar analysis the b-phase voltage is - x exp j [ ( n T 2)wt T 2u 3 L:, exp jCnwt T 2x13 2"9 + 3: in] + x exp j [ ( n f 2)ot & 2a T 2x13 3: in] (9) Comparing Eqns. 8 and 9, it is seen that the additional harmonic voltage term has the opposite phase sequence to that which is applied. For example, if n = 7 with positive-phase sequence, the additional harmonic component has harmonic number 5 with negative-phase sequence. This is the usual situation for a balanced system. If, however, n = 7 with negative-phase sequence, the additional component has harmonic number 9 and is positive-phase sequence : not zero-phase sequence as is usually expected. Consequently, the phase voltage set will not be balanced. The analysis is also valid for the application of negative-phase sequence current at fundamental frequency in which case the additional component has harmonic number 3, and is positive phase sequence [ l o ] . 2.2 Application of a single voltage harmonic If a balanced set of harmonic voltages is applied to a synchronous machine, an additional harmonic current with harmonic number n f 2 is anticipated, based on the previous results. The proposed solution is i, = in exp j[nwt + 3: in] + i, exp j [ ( n T 2)wt + 3: i,] + 3: in T 2x131 + exp j [ ( n f 2)wt + 3: i, i 2+] i, = in exp j[nwt + 3: in k 271131 + < exp j [ ( n T 2)wt + 3: k 2n/3] i, = in exp j[nwt (10) The barred quantities are the current components at frequency ( n f 2)w. The phase sequence of these com53 ponents is the reverse of that at the applied frequency, nw. Using the results of the previous section, the a-phase voltage is u, = in jnw( + L& L' q) exp jrnwt + 9: i n ] + in j ( n T 2)w( L b 2 " i ) exp j [ ( n T 2) x wt T 2a + 9: in] Voltage equations for the general case are obtained by applying a voltage mesh equation at each frequency nw and ( n f 2 ) o . The voltage components applying across the machine are those given by eqn. 11 in response to the currents given by eqn. 10. Hence at frequency n o u, exp j 9: u, = - i n z , exp j 9: z , in jnm(L' ') exp j 9: in and at frequency (n T 2 ) o u, exp j 9: ir;; - i n z , exp j 9: U;; - For an applied voltage v, = v, = i,,j(n T 2)-( L:, + L' q) expjxi, exp jnwt (12) comparing eqns. 11 and 12 gives the following equations u, = i,jnw(' 2 ' i ) + i,jnw(L:, 'i) + i , , j ( r 1 k 2 ) o ( ~ ; ';)expj[T2a+ <in] (16) These equations may be manipulated to give equations for inand inas follows in exp j 3: in = u , c exp j 9: u, - F, exp 9: i j ~ j n (L' w 2' i ) exp +2aj Eqns. 13a and b may be rearranged to give u, = j n w in 2L&L; L& L; + Comparing eqns. 9 and 14, it is seen that the effective inductance at the applied frequency is different for the voltage and current sourced cases. If harmonic voltages are applied at both frequencies n o and (n f 2 ) o , then the eqn. 13 will be modified by the inclusion of and 3: G. The currents at both frequencies will consequently be affected. and - - in exp j 9: in = 2.3 The general case In the general case, a machine will be subjected to applied voltages v, and y;; via system impedances z, and z;; as shown in Fig. 2. The system model may be viewed In these equations 5, = z , exp j 9: z , +j n w (Ld and - 5, Fig. 2 System in general case; balanced representation as a Thevenin model and consequently linearity is being assumed. It will be assumed at this stage that the applied harmonic voltages have opposite phase sequences, one positive and the other negative. This is the condition that would apply if the distorting load responsible for the harmonic disturbance draws balanced currents. Consequently, a single phase model may be used. 54 = exp j < + j ( n f 2)w Eqns. 17 and 18 show that the effective voltages at each frequency involve the voltage sources at both frequencies, there being a phase term which depends on angle a. The resonance condition for the equation is under which condition currents at both frequencies become large. Eqns. 15 and 16 suggest the equivalent circuit shown in Fig. 3. The interaction between the two sides of the circuit is shown as a mutual inductance, but this is an over simplification. The mutual inductance is nonIEE PROCEEDINGS, Vol. 135, P t . B, N o . 2, M A R C H 1988 reciprocal, and the frequency difference between sides has been ignored. The model may be used in schemes to calculate harmonic current flow. The results presented here Fig. 3 Equivalent circuit,forgeneral case show that the machine cannot be modelled by an impedance (Fig. l(a)) if significant voltage sources exist at each frequency. In the case where one harmonic voltage source is large and the other small so that q N 0, the apparent machine impedance to current flow is f The eqns. 17 and 18 are not valid for the application of positive phase sequence voltage at fundamental frequency ( n = 1) as there is no induced field current, and V, and are zero. For negative sequence voltages with n = 1, the equations are valid and the additional frequency harmonic number is 3. If there is no background voltage at this frequency then the machine impedance offered to negative sequence fundamental current is given by eqn. 21 and contains terms dependent on the system impedance at the third harmonic. The machine model leading to eqns. 17 and 18 is linear and time varying. If several harmonic frequency sources exist, the total current flow will be the superposition of those given by the theory, when applied to each harmonic n and its associated harmonic ( n T 2). If the rotor carries two windings, one on the d-axis and one of the q-axis, then the analysis is altered. If the two windings are electrically identical, the net gap flux wave resulting from induced current in both windings will rotate at frequency no. The previously discussed wave at frequency ( n T 2)w will be suppressed, and consequently the machine impedance at frequency nw is Z, =jwn L&+ L; ~ f However, if the windings are not electrically identical, a counter rotating flux wave will occur with rotational frequency ( n T 2)w and a magnitude dependent on the extent of the electrical difference between windings. In the case of two rotor windings the substitution is invalid. However, the equations presented previously which are expressed in terms of L&and Lb are valid. I E E PROCEEDINGS, Vol. 135, Pt. E , No. 2, M A R C H 1988 The results of this section are clearly associated with the transient response of the machine. If the machine has subtransient properties, application of harmonic disturbances will induce currents in the damper windings or pole face which will partly shield the field windings from the applied flux wave. The machine response will be given by eqns. 17 and 18 with the following replacements L:,+ L: and Lb+ Li (24) Because of the variation of flux penetration depth with frequency, these machine impedances will also be a function of frequency. Subtransient saliency will occur if the damper windings are incomplete (i.e. confined to the pole faces), if the dampers have a different construction between poles to that used on the pole axis [9] or if eddy currents in the pole faces of salient pole machines contribute to the subtransient response. Incomplete shielding of the field windings of a machine fitted with dampers subjected to harmonic disturbance has been demonstrated by Meyer et al. [9] for a 4.6 MW synchronous motor. For large to medium sized machines it is expected that the subtransient response will dominate. The additional frequency generated from within the machine, as demonstrated for example in eqn. 8, will then depend on the level of subtransient saliency. The results of Meyer indicate that the subtransient saliency can be significant. 2.4 Non -sinusoidal MMF effects Non-sinusoidal gap flux wave effects can be accounted for by representing the mutual inductances between the field and stator, and between the stator windings by a trigonometric series in the rotor angle 8. The resulting mathematics is extremely complex. It is found that the inclusion of non-sinusoidal inductance variation causes a spreading of the spectrum such that injection of a single harmonic current results in voltage arising at all odd harmonic numbers, including the fundamental. The extent to which the energy is spread across the spectrum depends upon the pole shape, the numbers of stator and rotor slots, and the damper design. The spectral spread is a second order effect compared to the generation of the associated voltage at frequency n f 2. 3 Measurements on synchronous machines To test the preceding theory a series of measurements was taken on two 5 kVA, 415 V, 4-pole synchronous machines (A and B). Full details of the machines used for measurement are given in Appendix 7. Machine A has an incomplete pole face damper winding built in and so exhibits subtransient saliency. Machine B exhibits transient saliency only and is known to have a voltage waveform with very little harmonic corruption. 3.1 Current spectrum for an applied 7th harmonic voltage, machine A A 350 Hz ( n = 7) 3-phase voltage was applied to machine A and the current spectrum was measured using a spectrum analyser applied across a resistive shunt. Measurements were taken for both negative and positive-phase sequence supply. Measured values are given in Table 1. These results are in accord with the theory presented in Section 2 where it was concluded that for n = 7 pps, a strong signal is expected at the 5th harmonic and for n = 7 nps a strong signal occurs at the 9th harmonic. The spectrum shows the spreading which is associated with non-sinusoidal inductance variations. The flux 55 Table 1 : Measured current spectrum f o r 350 Hz voltage aodication. Machine A. Positive-phase sequence Frequency, Hz Harmonic No. RMS current, mA ~~ 50 250 350 450 550 650 850 1150 1750 1850 1 5 7 (applied) 9 11 13 17 23 35 37 19.7 143.1 324.0 0.8 8.2 3.3 0.3 0.2 1.o 4.9 machine cored N I v Negative-ohase seauence 50 150 250 350 450 550 650 750 850 1650 1750 1 3 5 7 (applied) 9 11 13 15 17 33 35 26.3 6.6 0.6 329.2 141.4 0.1 3.3 1.o 0.3 0.3 2.3 waveforms arising from induced damper currents exhibit significant harmonic corruption and so the spreading of the spectrum is more extensive than would have been expected based on the non-sinusoidal inductance variation measureable under static conditions. It is seen in Table 1 that the ratio of currents i,,/< at frequencies no and (n f 2)o is close to equal for both positive and negative sequence cases. This is expected from eqn. 13 since - i,,/i, = L:, + L; L:, - L; ~ which is independent of frequency or phase sequence. 3.2 Measurement of apparent impedance, machine A The apparent impedance at harmonic frequencies of machine A was measured using the arrangement shown in Fig. 4. The bridge rectifier draws odd, non-triplen harmonic currents, some of which flow into the machine because the supply impedance has been artificially increased by adding X , . Using a special purpose, dual channel frequency unit, the voltage and current at each harmonic frequency can be displayed, and the apparent The measured apparent normalised impedance against frequency is given in Fig. 5. The values are seen to oscillate in a saw-tooth manner with frequency. The voltage component at each harmonic frequency has two components, one being the first term in eqn. 8 and the other associated with current flow at the associated frequency (i.e. second term in eqn. 8) which has phase angle dependence on a. These two components may either reinforce or cancel resulting in larger or smaller normalised impedances as observed. The results of Table 2 are generally in accord with theory, although some significant discrepancies d o occur. Factors which are not accounted for in the theory are machine losses (the variation with frequency which is not known), accurate modelling of the source machine, and incomplete maintenance of field flux linkages. These results do confirm that the apparent machine impedance at the applied harmonic frequency varies significantly when the system impedance is varied. Oscillographs of the current waveform drawn by the machine during test 1, table 2 are shown in Fig. 7a. The waveforms exhibit beating affects as expected from the two associated frequency terms. Machine B was also machine 60V 350Hz 0 +d Fig 6 Measured circuitfor machine B Quiet reactances are at 350 Hz mochiine 50HZ hormonic number, n 56 Fig. 5 Measured normalised appurrnt impedance against frequency, machine A 1 Noload 2 Reluctance m/c, field s.c.. 2.9 A. 193 Ci$ 3 Reluctance m/s, field o.c., 2.9 A, 193 V+ 4 Synchronous m/c. 6.0 A. 200 V+ IEE PROCEEDINGS, Vol. 135, Pt. B, N o . 2, M A R C H 1988 Table 2: Measured and theoretical apparent impedance Machine B. Circuit connection is shown in Figure 6. for Circuit description Sequence Measured, fl Theoretid, fl 1 n = 7 pps (nT2)=5 29 3 @ 78" 3 1 @82" 27 9 (a 90" 6 1 (g 90" n = 7 nps (n T 2) = 9 29 5 (a 80" 6 3 @ 77' 27 9 (a 90" 11 0 (a 90" n=7pps (nT2)=5 4 5 7 (a 79" 635@81" 465 558 2 -n=7nps 'L 3 4 5 3 (& 792" (nT2)=9 1059@846" n = 7 pps 69 7 6 72 5" (n T 2) = 5 203 1 @ 63' 4 6 5 @ 90" 1005@90" 60 5 (a 90" 296 0 @ 90" 21 7 (a 71 8' 1 3 2 ( a -785" 19 7 (a 90" 5 8 6 -90 n = 7 nps (nT2)=9 II 25pF 4 Conclusions v I v Fig. 7 Oscillographs Current (i) and applied voltage (V) waveforms. (i) and (ii) positive sequence applied harmonic. (iii) and (iv)negative sequence applied harmonic. b Waveforms with the quadrature field short-circuited, machine B. (V) armature voltage, (i) current, (i,) field current waveforms for machine B with 350 Hz voltage applied a fitted with a rotor winding on the quadrature axis which was left open during all previous tests. When this winding is short-circuited, the machine becomes zero transient salient with X &= X b . The current waveform is a pure 7th harmonic as shown in Fig. 7b. The measured machine impedance in this case is 13.3 R which is close to the calculated value of noL& (14.3 R). The measured results illustrate that a synchronous machine which exhibits transient saliency cannot be modelled by a simple equivalent circuit like those given in Fig. la. IEE PROCEEDINGS, Vol. 135, P t . B, N o . 2, MARCH I988 When a synchronous machine IS subjected to a harmonic voltage disturbance at frequency n o , harmonic current components are drawn at n o and at the associated frequency ( n & 2)o. The upper sign applies for positive and the lower sign for negative-phase sequence. Current flow at the associated frequency occurs because the machine is a time-varying electrical system. In a similar way, a synchronous machine fed by harmonic current at frequency n o develops a voltage across its terminal at both the applied and the associate frequency ( n f 2)w. In consequence of this behaviour, the machine cannot be modelled by an impedance applying at each harmonic frequency. The harmonic behaviour is governed by a set of coupled equations, eqns. 17 and 18, which involve machine and system parameters at both frequencies. Harmonic behaviour is also affected by non-sinusoidal inductance variation with rotor position and by the extent to which induced damper winding currents shield the field winding from the harmonic gap flux waves. 5 Acknowledgment The authors gratefully the of the Electrical Research Board of Australia. 6 References 1 'Disturbances in mains supply networks', AS227Y, Pts. 1 & 2 1979, Standard Association of Australia 2 PILEGGI, D.J., H A R K , H., CHANDRA, N.. and EMANUEL, A.E. : 'Prediction of harmonic voltages in distribution systems', Trans. PAS., 1981,100, pp. 1307-1315 3 C E D D E Y , D.K.: 'Computer calculation of harmonic voltage levels in high voltage transmission systems'. Seminar on miniqg and industrial loads, Sydney, August 1983 4 IEEE Working group on power systems harmonics: 'The effects of power system harmonics on power system equipment', IEEE PAS., 1985, 104, pp. 2555-2563 ~ 51 5 CIGRE Working group 36-05 : ‘Harmonics, characteristic parameters. methods of study, estimates of existing values in the network’, Electra, 1981, pp. 35-54 6 WILLIAMSON, A.C.: ‘The effects of system harmonics upon machines’, I J E E E , 1982,19, pp. 145-155 7 WILLIAMSON, A.C., and URQUHART, E.B.: ‘Analysis of the losses in a turbine-generator rotor caused by unbalanced loading’, IEE Proc. B, 1976, 123, pp. 1325-1332 8 FITZGERALD, A.E., and KINGSLEY, C.: ‘Electric Machinery’, Chapter 5 , McGraw-Hill, 1961 9 MEYER, A., and ROHRER, H.J.: ‘Calculations and comparative measurements for converter-fed synchronous motors’, Brown Boveri Review, 1985,72, pp. 71-77 10 KIMBARK, E.W., ‘Direct Current Transmission - Vol. l’, WileyInterscience, N.Y., 1971 7 Appendix (a) Machine A Salient pole synchronous machine 5 kVA, 4 15 V, 4 pole, 50 Hz, 1500 rev/min X, = 45.5 Q, R 1.83 R (at d.c.), X , = 15.0 0. ( b ) Machine B Round rotor synchronous machine, 5 kVA, 415 V, 4 pole, 50 Hz, 9 A, 1500 rev/min (c) Data for Machine B Single-field mode ZTS mode 5x = 54 mH 6.5 mH L,, = 4.9 H La, = 0.40 H Ld = 54 mH Lh = 6.5 mH L,, = 3.6 H La, = 0.34 H Ld = Li = I E E P R O C E E D I N G S , Vol. 135, P t . E , N o . 2, M A R C H 1988