IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO. 3, SEPTEMBER 2002 1 A New Synthetic Loading for Large Induction Machines With No Feedback Into the Power System Jafar Soltani, Member, IEEE, Barna Szabados, Senior Member, IEEE, and Gerry Hoolboom, Member, IEEE Index Terms—Measurement of losses, phase modulation, synthetic loading. I. INTRODUCTION H to the shaft. The two-frequency method [1] has major advantages over the other methods. It is applicable to both wound rotor and squirrel cage machines as opposed to Romeira’s method [2] which can be applied to wound rotor machines only. Furthermore, it does not require that all six leads of the rotor be brought out as in Fong’s method [3]. In recent work [7] our group has established that the two-frequency method did not lead to good results because of the very high fluctuations of stator voltage applied and, therefore, the degree of acceptance of the method is dubious. Moreover, a geared generator and a multiphase transformer is usually required both rated at the test machine rating, and the full rated power oscillates at a low frequency into the power system. The investigation of three synthetic loading methods at full-load temperature evaluation using calorimetric methods has shown the major drawbacks of the two-frequency method. Results have been compared with the conventional direct loading method [10]. Unfortunately, these results are limited to low capacity induction motors due to the power swing in the power system. In this paper, a new synthetic loading is proposed. The assumption made is that motor manufacturers prefer to build rotating machines used for the test rig rather than buying special purpose large transformers. The method is based on a bang–bang phase modulation technique to generate a variable frequency voltage generated by a synchronous generator. The use of two systems in parallel, one system with the test motor and one system with a “recovery” unit, limits the power swing with the power system to the total losses of the five machines used. The method has been proven first by simulation whereby the constraints of the method have been identified, and it has been experimentally verified on a test set up in the laboratory. IE E Pr E oo f Abstract—Full-load testing of large induction machines is constrained by the limitations in the power-supply and loading equipment of the manufacturer’s facilities, resulting in costly set up time. A new synthetic loading method is proposed based on a bang-bang phase control strategy. The rated power oscillation created is routed to an auxiliary system and the source hydro has to provide only the total losses of the system, without seeing the excessive power swings observed in other synthetic loading techniques. In this technique, only induction machines are used which would enable motor manufacturers to build the test rig in-house. The control stage is very simple to implement and requires only unregulated dc supplies for the excitation windings. The method is suitable for any induction machine and does not requires any set up time. It is possible to strictly maintain constant DEFINE RMS voltage and current at rated values for the duration of the heat runs. EAT RUN tests performed on electric machines are extremely important to both manufacturers and users. The test results verify the predictive performance calculation methods used, therefore lowering business risks during the tendering process. On the other hand, the test results demonstrate to the user that the contractually agreed upon performance has been met. Specifically, it can be verified that at full load the machine does not exceed the temperature insulation class limit. Established business patterns have demonstrated that the number of motors produced decreases as the machine power output increases, and that the provision of motor test plant follows the opposite trend [9]. Direct methods suffer from high costs due to the complex coupling and loading mechanisms, as well as the sophisticated sensor and measurement instrumentation required. Indirect methods suffer from large inaccuracies and mainly cause unacceptable power swings on the power system. There is obviously a need for alternate methods to produce full-load heat runs economically. A number of proposals have been devised [1]–[3], [5], [9] which attempt to produce accurate performance verification of induction machines without having to attach mechanical loads Manuscript received July 25, 2000; revised February 7, 2002. This work was supported in part by the U.S. Department of Commerce under Grant BS123. J. Soltani was with McMaster University, Hamilton, ON, Canada, L8S 4K1. He is now with Isfahan University of Technology, Isfahan, Iran (e-mail: j1234sm@cc.iut.ac.ir). B. Szabados and G. Hoolboom are with the Power Research Laboratory, McMaster University, Hamilton, ON, L8S 4K1 Canada (e-mail: szabados@mcmaster.ca). Publisher Item Identifier 10.1109/TEC.2002.801728. II. PRINCIPLE OF EQUIVALENT LOADING The internal air-gap voltage of an unloaded induction machine is very close in magnitude and phase angle to the apand the no load current is small plied armature voltage [Fig. 1(a)]. As the machine becomes loaded, the load angle increases and a larger armature current is produced [Fig. 1(b)]. The principle of equivalent loading is to increase the internal load angle without connecting any mechanical load onto the shaft. The two-frequency method [1], [2], [7] essentially uses one voltage at 60 Hz in series with another voltage at 50 Hz that is applied to the armature of the machine under test. This voltage will have a modulated phase angle, causing the rotor speed of the machine under test to try to follow. The inertia of the rotor acts as an energy storing device. 0885-8969/02$17.00 © 2002 IEEE 2 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO. 3, SEPTEMBER 2002 Fig. 2. Dual-field winding control. (a) Excitation windings. (b) Resultant flux control in quadrature. IE E Pr E oo f Fig. 1. Principle of load angle control. It has been shown [7] that the two frequency method sees variations of terminal voltage between 140% and 60% of rated value. This range drives the machine into saturation as well as undervoltage, introducing major inaccuracies in the resulting measurement. Furthermore, the full rated power swings at the modulation frequency between the power grid and the test motor. III. NEW PHASE MODULATION A. Principle The principle of the new phase modulation method is to generate a voltage using a three-phase synchronous machine. The synchronous machine needs to have a rotor with two windings with a spatial phase shift, as shown in Fig. 2(a). By appropriately exciting each winding, any phase angle within the range and shown in Fig. 2(b) can be achieved for the excitation field. The modulation in amplitude and phase can be achieved by appropriately modulating the excitation currents and of the field windings. A sinusoidal modulation would requires two precisely controlled dc sources, which is very difficult to implement especially since large inductive values are present in the field windings. B. Practical Implementation A synchronous generator using a wound rotor induction machine is created as shown in Fig. 3. Winding (A) of the rotor is excited by a controlled source DC1. Windings (B) and (C) are connected in series and excited with another fixed supply DC2. Fig. 4 shows that the resultant dc flux created by the rotor windings, represents the direct axis in the synchronous frame of the machine. By appropriately controlling the current in winding (A), one can modulate the phase w.r.t. the utility source phasors. Practically, the source DC1 is switched in a bang-bang mode, Fig. 3. Creating a controlled dc field in the rotor of an induction machine. Fig. 4. Creating a modulated synchronous field in the rotor of an induction machine. and due to the reactance of the field winding, the excitation current will vary exponentially. The vector diagram of Fig. 4 also illustrates the range of the phase shift when a bang–bang modulation is impressed on winding (A). The phase modulating synchronous generator implemented with this scheme is driven by a “driver” motor fed from the utility. The output of this synchronous generator feeds the test motor which, in this single rig implementation, would see the total power swing go through to the utility connection as described in a previous disclosure [11]. We propose here to use two systems similar to the one described in [11]. Both systems share the “driver” motor (D0) on the same shaft as shown in Fig. 5. Generator G1 of of the System(1) is rated at the maximum power rating test rig (highest motor rating to be tested), and feeds the motor M1 under test. The test motor could be either a wound rotor machine or a squirrel cage machine. Both the generator G2 and . the “recovery” machine M2 of System(2) are rated at If the field modulation of each generator is in opposition of phase, the power generated by each system is also in opposition of phase and, therefore, when one system absorbs power, the other generates it and vice-versa. By adjusting the magnitude of the excitation swing of generator G2, one can adjust the SOLTANI et al.: A NEW SYNTHETIC LOADING FOR LARGE INDUCTION MACHINES WITH NO FEEDBACK INTO THE POWER SYSTEM 3 power exchange with System(2) to exactly match the power swing of System(1). When this equilibrium is reached, the driver motor (D0) needs to provide only the losses in all five machines. Driver D0 is preferably a synchronous motor, but can be a dc motor or an induction motor with low slip. In the latter case, the motor would have to be slightly over rated in order to perform the test at close to synchronous speed. There is a set of performance curves describing the magnitude of the modulation as well as the depth of the modulation (frequency swing). It has been shown [11] that the optimum performance to reach full rated load is achieved with a 10-Hz modulation depth (55 to 65 Hz). C. System Simulation Fig. 5. Test set up. The system shown in Fig. 5 has been simulated with G1 and G2 being identical three phase, 2–kW wound rotor induction machines, and M1, M2 and D0 are 2–kW squirrel cage motors, with the corresponding parameters shown. kg-m IE E Pr E oo f Motors squirrel cage) Induction Generators poles KW Hz V Nm-sec/rad (windage iron) kg-m [with for resistance, for leakage reactance, for magnetizing reactance and for moment of inertia; subscripts and for stator and rotor]. The simulation uses the standard Park’s equations to convert three phase into the conventional DQ axis model [8]. The Appendix shows the relevant equivalent circuits used for each one of the interconnected systems (1) and (2) as well as the main equations used. The set of flux equations is solved using a static Runge–Kutta fourth-order method. Fig. 6 shows the results of the simulation, providing the modulated phase voltage applied to motor M1 and the corresponding phase current. One can note that although there is a definite phase modulation between current and voltage, the magnitude of the voltage is nearly invariant and at near rated value, while the magnitude of the current varies considerably with the modulating frequency. The instantaneous power is computed as the instantaneous products of current by voltage. The active power is computed as the average of the instantaneous power over one pseudo cycle of the 60–Hz-modulated signal. Finally, the total losses in the system are computed as the average of the active power over the modulation interval [11]. Fig. 7 shows the power flow at the terminal of machine M2 and illustrates the opposition of phase with M1 compared with Fig. 6. Fig. 8 clearly shows that the power flow through driver D0 is constant. Losses in M1 and M2 are found to be 0.22 PU each (base power 2 kW), while the total losses in the 5 machines show 1.2 PU, which is commensurate with the name-plate data provided by the manufacturer. Fig. 6. Power flow in machine M1 (test motor). D. Experimental Results Fig. 9 shows the experimental values of phase voltages and phase currents in machines M1 and M2. It clearly shows that both terminal voltages are modulated with the same frequency, although slightly out of phase are nearly constant in magnitude. However, the currents in the two machines are of different amplitudes, and also show definite phase differences at each instant in time, reflecting the power oscillation between the two machines. The computed instantaneous power and the computed average power or losses in the test machine M1 are shown in 4 Power flow in machine M2. a phase shift exists between the voltages V1 and V2 and the instantaneous power generated by one machine is not matched by an opposing absorbed power in the other machine. We have proven this in the simulation by including a phase difference between the direct axis on machines G1 and G2, and the result showed clearly that a power oscillation occurs and its magnitude depends upon the phase angle. In our experimental set up, the two machines were connected through a keyed mechanical coupling with a cogged rubber ring. The cogs allowed only an axis adjustment within steps of 40 . We verified that power oscillations vary with realigning the two shafts, but due to the physical key of the coupler constraint we could not arrive at better alignment than about 15 . In order to achieve perfect alignment, we would need a slanted keyed coupler which, of course, could easily be constructed on a final test rig. control Previous studies [7] advocate that a constant strategy should provide the best loss measurement together with a sinusoidal modulation. In this method, we do not use these constraints. The experimental results showed clearly that the use of the phase-modulation technique draws advantageously on the armature reaction of the generators which tend to maintain the voltage of the generator output as constant as possible. In fact, the phase modulation means that we actually modulate the angle between the internal voltage of the generator and the motor effectively, realizing a quasi current source. The results of the simulation show that there is no power swing between the utility supply and the test rig when System(2) is adjusted exactly to match the power swing of System(1). Although the experimental set up prevented us from creating this perfect match, the power oscillation was minimized, and certainly would be acceptable in a practical platform. We fully realize that this technique uses five machines. However, according to machine manufacturers we are working with, they much prefer constructing extra machines, rather than purchasing bulky transformers and gear boxes. Furthermore, one really needs only the driver D0 and the two generators G1 and G2 as fixed installation with carefully aligned shafts. Nothing prevents the use of two motors M1 and M2 to be tested at the time. There is always one extra motor in the plant to act as a recovery unit, and since there is no need for mechanical coupling, this is an easy installation. The extra two machines required (instead of a multiwinding transformer and a gear box) is considered a small investment versus achieving no power swings in the power system. IE E Pr E oo f Fig. 7. IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO. 3, SEPTEMBER 2002 Fig. 8. Power flow in driver D0 and total losses. Fig. 10. Finally , Fig. 11 shows the instantaneous power as well as the average active power flowing through driver D0, representing the total losses in the five machines. This is found to be slightly different than the simulated results. One has to note that the value of 1.5-PU power flow through the driver motor is misleading. This is true for this small system where the losses in each machine are approximately 20%. In industrial applications, this would be only around 5%. The speed of the test motor was measured with a tachogenerator. The signal obtained showed tooth ripple and brush commutation noise well in excess of the expected small variation of speed. IV. DISCUSSION Fig. 11 shows clearly a power oscillation through driver D0 while the average power still represents the total losses of the system. This oscillation has been traced to two main effects. The first cause comes from the phase modulation itself. The excitation currents of G1 and G2 have to be exactly in opposition of phase. While it is possible to apply phase opposition voltages to the windings, it is difficult to guarantee that both time constants are identical. Hence, the two excitation fluxes will not necessarily establish their final values symmetrically. We have tried to minimize the power ripple due to this cause by adjusting the relative excitation voltage on G2. The second main cause for the power ripple comes from the misalignment of the electrical axis between machines G1 and G2. If the two machines do not have their direct axis aligned, V. CONCLUSIONS We have demonstrated the viability of a new phase shifting method to generate a modulated three-phase feed to a test machine. This method can be easily implemented using a conventional three-phase wound rotor induction machine excited by two uncontrolled dc sources. Using a simple bang–bang control on one of the dc sources provides the modulation to the output voltage. Furthermore, if two similar systems run back to back, a rated power swing between the two systems can effectively be achieved. Only the total losses of the five machines have to be provided by the source. SOLTANI et al.: A NEW SYNTHETIC LOADING FOR LARGE INDUCTION MACHINES WITH NO FEEDBACK INTO THE POWER SYSTEM Fig. 11. 5 Instantaneous power in D0 and total losses. Fig. 9. IE E Pr E oo f when the test motor rating is much lower than the rating for the recovery system, since one has to match the instantaneous power in both machines. Even if the direct axis of both generators was perfectly aligned with the open-circuit voltage test, during loading, the internal angles will be different in the machines because of the different parameters of the attached motors. Therefore, some power oscillation cannot be prevented in practical applications without a sophisticated controller. However, as shown in our experimental verification, this power swing through driver D0 will be small compared to the rated power of the machines For practical implementations it would still be acceptable. Voltages and currents in machines M1 and M2. APPENDIX SYSTEM SIMULATION Fig. 10. Computed power and losses in test machines M1. Using simulation and experimental verification, we have constant requirement is not a necessity. It shown that the is possible to maintain constant RMS voltage and current in the test motor by adjusting the two field winding currents of the generators. The simulation and experimental verification proves that the main criterion of the test is to maintain the RMS quantities of current and voltage to the rated values computed over the modulation cycle. The main problem identified during the experimental part was the need for a closed-loop control to maintain the constant ratings on the test motor and the matching power exchanged with the recovery system. This problem becomes quite involved Fig. 12 shows the usual equivalent circuits with the conventional naming of parameters used for each one of the generators and motors in the interconnected systems (1) or (2). The left side of the equivalent circuit contains the lumped stationary compoand and the right side the rotor quantities of nents of referred to the stator. Indices and , respectively, refer to direct quadrature axis and zero sequence, and indices and refer to stator and rotor quantities, while refers refers specifically to the to the excitation field. The subscript excitation of the induction machine working as a synchronous in the parameters. generator, generally indicated with indice Furthermore, is the base angular frequency of the shaft while and are the rotor angular velocities of the driver motor , respectively. The voltages are calculated and motors from the equivalent circuit models of the stator of the generators. Fig. 13 shows the equivalent circuits for the rotors of the working as synchronous generators. machines The fluxes shown on the models are (A1) and (A2) The induction generator flux equations follow as 6 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO. 3, SEPTEMBER 2002 (A3) The field voltages are defined as (A4) (a) leading to (A5) (b) G M ). (a) The torque developed for each machine is IE E Pr E oo f Fig. 12. Equivalent circuit for stationary parts and rotor for ( ) and ( Direct axis. (b) Quadrature axis. (A6) REFERENCES (a) (b) [1] A. Meyer and H. W. Lorenzen, “Two-frequency heat run—A method of examination for three-phase induction motors,” IEEE Trans. Power App. Syst., vol. PAS-98, pp. 2338–2347, Nov./Dec. 1979. [2] M. P. Romeira, “The superimposed frequency for induction motors,” Proc. IEEE, vol. 36, pp. 952–953, 1948. [3] W. Fong, “New temperature test for polyphase induction motors by phantom loading,” Proc. IEEE, vol. 60, pp. 883–887, July 1972. [4] Std. Jpn. Electrotechn. Comm., JEC-37-1979, 1979. [5] H. R. Schwenk, “Equivalent loading of induction machines for temperature test,” IEEE Trans. Power App. Syst., vol. PAS-96, pp. 1126–1131, July/Aug. 1977. [6] C. Grantham, E. D. Spooner, and M. Sheng, “Synthetic loading of machines using power electronics,” Elect. Eng., vol. 67, no. 8, pp. 60–68, Aug. 1990. [7] A. Mihalcea, “Equivalent loading methods for determining total power loss in induction motors,” Master’s thesis, AU: At what University was this thesis written, CITY?, COUNTRY, Sept. 1999. [8] P. Krause, Analysis of Electric Machinery. New York: McGraw-Hill, 1986. [9] D. H. Plevin, C. N. Glew, and J. H. Dymond, “Equivallent load test for induction machines—The forward short circuit test,” IEEE Trans. Energy Conversion, vol. 14, pp. AU: Page numbers?–, Sept. 1999. [10] B. Szabados, J. Soltani, and G. Hoolboom, “A new synthetic loading of induction machines based on phase modulation,” in Proc. INDUSCON Conf., Porto-Alegre, Brazil, Nov. 2000, pp. 7–11. [11] A. Mihalcea, B. Szabados, and J. Hoolboom, “Determining total losses and temperature rise in induction motors using equivalent loading methods,” IEEE Trans. Energy Conversion, vol. 16, pp. 214–219, Sept. 2001. (c) Fig. 13. Equivalent circuit for the rotor of the generators. (a) Zero sequence. (b) Direct axis. (c) Quadrature axis. Jafar Soltani received the B.Sc. degree from the University of Tabriz, Tabriz, Iran and the Master’s and Ph.D. degrees from University of Manchester Institute of Technology (UMIST), Manchester, U.K. Currently, he is Associate Professor at Isfahan University of Technology, Teheran, Iran, and was on leave at McMaster University, Hamilton, ON, Canada. His main area of research is electrical machines and drives. SOLTANI et al.: A NEW SYNTHETIC LOADING FOR LARGE INDUCTION MACHINES WITH NO FEEDBACK INTO THE POWER SYSTEM Jerry Hoolboom (M’60) received the B.Sc. and M.Sc. from the University of Delft, Delft, The Netherlands. His expertise in electrical machine design was achieved while he was with McMaster University, Hamilton, ON, Canada, as Assistant Professor, and Westinghouse Canada, AU: Location of Westinghouse Canada?, where he held the position of Director of Technology. He is a registered power engineer in the province of Ontario. IE E Pr E oo f Barna Szabados (SM’87) received the B.Sc. degree from Grenoble University, Grenoble, France, the Master’s and Ph.D. degrees from McMaster University, Hamilton, ON, Canada. He is currently Professor of electrical and computer engineering at McMaster University and is the Director of the Power Research Laboratory. His main interests are power electronics and power apparatus in the field of control, measurement, and modeling of machines and transformers. Dr. Szabados is a registered P.Eng. in the Province of Ontario. 7