R. Bonert S. Rajakaruna Indexing terms: Induction generator, Stand-alone generation, Load governing Abstract: A capacitor-excited induction generator used with a hydraulic turbine in a stand-alone generating system can provide a high quality voltage and frequency control not matched by other small generating units. This is achieved without a turbine governor by using a controllable additional impedance on the load side. The control is achieved using a static converter. The analysis of the system and the design of the power and control system are presented. Measurements from an experimental unit are provided to verify the predicted performance. List of symbols List of symbols and their definition, units are metric using the SI unit syslem. Variables C capacitance i current Im imaginary part of .i complex component J inertia k , m constant L inductance R resistance t time Tsh, To torque; shaft, constant V voltage a control angle of bridge 6 pulse width (of chopper Y flux linkage w frequency in radians Q mechanical $,peed in radians Indices ct controller 0 IEE, 1998 IEE Proceedings online no. 19981680 Paper first received 10th October 1996 and in revised form 26th August 1997 R. Bonert is with the Department of Electrical Engineering, University of Toronto, Toronto, Ont., M5S 3G4, Canada S. Rajakaruna is with the Department of Electrical Engineering, Univer- sity of Moratuwa, Katubedda Moratuwa, Sri Lanka R rotor S stator a orthogonal component a P orthogonal component P 1 Introduction The proposed generator system consists of a capacitor- excited induction generator driven by an unregulated hydraulic turbine without a speed governor as shown in Fig. 1. The generator supplies an isolated electric system indicated as an inductive resistive load in Fig. 1. The voltage and frequency control of the system is achieved by an impedance controller at the terminals of the generator. This principle is sometimes called electric load governing. n gate open/close %' Fig. 1 1 p h a s e I controlled I bridge impedance controller I chopper I I _ _ - - - - - - Turbine generator system and electric circuit As can be seen in Fig. 1 the impedance controller consists of a phase controlled bridge and a chopper switch connected to a resistor. This controller can con- sume real and reactive power. The amount of power is controlled by control of the bridge and the pulse width control of the chopper. The principle of operation is that the impedance controller picks up the real power and reactive power not used by the load, so that the load or the impedance seen by the generator at ils ter- minals is always constant. The voltage and frequency at the terminals of the generator will then be constant as well. Any change in load is immediately compen- 33 IEE Proc.-Gener. Tvansm. Dixtrib., Vol. 145, No. I . Junuury I998 sated by the impedance controller thus holding the impedance and therefore voltage and frequency at the generator terminals constant. The advantage of this system is that an expensive and slow turbine governor is not needed and the impedance controller is a three-terminal device con- nected to the generator providing a high quality volt- age and frequency control. The disadvantage is the restricted use of the principle of this control for hydraulic generation at rivers or dams with sufficient water, where it does not matter whether water spills over the dam or some generated energy is dissipated in the resistor of the impedance controller. 2 system Dynamic model of t he turbine-generator Using the space phasor description and selecting a ref- erence system rotating synchronously with the stator voltage the induction machine can be described by the following differential equations [ 11. Using complex var- iables for voltage, current and flux linkage the stator and rotor equations and the mechanical equation can be written as: ( 3 ) with w the electrical frequency, C2 the mechanical speed, J the inertia, Tsh the torque and the indices R and S mark rotor and stator variables. For the symbols used see the list in the Appendix, Section 11. Using a flux linkage model with only one leakage inductance, the flux linkage equations can be written as: - - 9s =imL, (4) (5) 'ZR - Gs =ZRLR~ with Lm the magnetising inductance and L,, the leak- age inductance. A resistor representing the stator core losses may be added parallel to the magnetising inductance [2]. To model the induction machine for operation as a self- excited generator it is necessary to introduce saturation which can be assigned to the magnetising inductance in eqn. 4. The value of the magnetising inductance is modelled as depending on the magnitude of the stator flux linkage: Lm =f ( l Q ~ l ) (6) The load for the induction generator is modelled as a series to represent the true dynamic behaviour of such loads. The differential equations describing the load and the capacitor referred to the synchronous frame of the induction machine's stator voltage are: resistive inductive load with a resistor and inductor in ( 7) The complete equivalent circuit describing the dynamic 34 of the system including the capacitor for excitation of the induction generator and the load is shown in Fig. 2. n Fig. 2 Dynamic equivalent circuit of generator system The last component to be modelled is the turbine. It is assumed that the coupling to the generator is stiff; as a result the inertia of the turbine can be added to the inertia of the induction machine. The turbine is then modelled by a torque-speed characteristic. The torque- speed relation of water turbines operated with a con- stant flow and head can be expressed by the linear rela- tion [ 3] : where To and Qo is any point on the straight line torque-speed characteristic corresponding to the given input hydro power. The negative slope m varies between 0.6 and 1.1 depending on the type of turbine. The above equations describe the turbine generator system with a typical load and the excitation capacitor completely. Choosing to align the real axis of the refer- ence system with the stator voltage and splitting the complex differential equations into real differential equations the complete system of equations results in eight real differential equations written in state space form and ten algebraic equations. The equations are shown below. The algebraic equations result directly from the equivalent circuit developed above. The eighteen equa- tions contain eighteen unknown variables. Tsh =To - m(R - Ro) (9) differential equations algebraic equations IEE Proc.-Gener. Transm. Distrib., Vol. 145, No. I , January 1998 In the general case of variable speed the system of equations leads always to only one nonlinear transcen- dental equation for the frequency which has to be solved numerically. This holds true independently of the choice of the structure of the equation of the mag- netising characteristic. The decisive step to this general solution results from rewriting the steady state equa- tions such that the single phase equivalent circuit repre- senting the equations contains only elements parallel to the magnetising inductance, as shown in Figs. 4a and b and described in detail in [4]. c _ _ - - - - - - - - - - - - - - - 1 I I I 1 I I- - --- - LRS ' aca =--asa - a~~ a,ap =-%ssp - aLp The magnetising (characteristic from which L, is (26) (27) ( 28) (29) derived, is approximated by Qs =kl . i , for 0 5 i, 5 io kl b Qs =1%1 . io +- arctan(b(im - i o) ) for i o 5 i, The constants kl , 6, io are determined such that the derivative &ldi is cclntinuous at i , =io and the meas- ured curve is adequately approximated. This approxi- mation has the advantage that it matches most magnetising characteristics very accurately from the linear region well into the saturation region. Further- more the derivative falls monotonously and the inverse function is readily available. An example of the good approximation is shown in Fig. 3. 1.4 - 1.2 - 4- .- a, n & 1.0 - m Y ._ ; 0.8 - 8 0.6 - iii 3 c 4- 0.4 - -- 0 - . I 0 0.2 0.4 0.6 0.8 1 .O 1.2 magnetising current per unit Fig. 3 k, =2.87, io =0.11, b =3.167 ~ measured _ _ _ _ approximated Approximation of the magnetising characteristic 3 The 18 steady state equations describing the system behaviour are derived from eqns. 10-27 using eqns. 28 and 29. The resulting system of equations is nonlinear. To solve these equations a new approach was devel- oped which results In a very elegant way to find the steady state solution [4]. This approach includes earlier solutions [ S , 61 but is more general and the mathemati- cal formulae are much simpler. Steady state model and solution IEE ProcGener. Tuansm. Distrib., Vol. 145, No. I , January 1998 Fig.4 Steady state equivalent circuit a Full b Reduced Based on the steady state solution it is possible to calculate the operating conditions for any set of param- eters. On the other hand it is also possible to determine the required settings of the torque Tsh, the speed Q, the value of the excitation capacitor and the value of the load impedance to achieve a specific desired operating point for voltage, frequency and stator current. These calculations provide the base to determine the rating of the impedance controller. 4 Several solutions to build an impedance controller are possible. The controller shown in Fig. 1 was proposed in [7] and was chosen for this project. The controller offers a wide range of fast impedance control and pro- vides a proven reliable design. To keep the voltage har- monics injected by the impedance controller low, the chopper is switched synchronised to the pulses of the phase controlled bridge at five times the pulse fre- quency of the bridge. The remaining higher frequency harmonics are reduced by the capacitor on the AC- side, and as a result the remaining total harmonic dis- tortion in the voltage is always less than 5%. Modelling the selected controller is difficult, as differ- ent modes of continuous and discontinuous conduc- tions have to be analysed. The choice was made to describe the controller as a device providing real and imaginary current components at a given AC voltage as function of the control angle a of the bridge and the pulse width 6 of the chopper. Since the harmonics are limited, an approach which analyses only the fundamental components of the con- Model of the impedance controller 35 troller current is justified. But even this approach results in very complex formulae, too complex to yield a reasonable controller model, as reported in [8]. For- tunately closer examination of the complex equations and simulations suggests an approximate model describing the controller current as a space-phasor specified in magnitude and phase. It is then shown in [8], that the phase is primarily determined by the con- trol angle a of the phase controlled bridge, and for a given control angle of the bridge the amplitude of the current is proportional to the chopper pulse width 6. As a result the real and imaginary component of the controller current iCt, and iCtp can be expressed as a function of the bridge control angle a and the chopper pulse width 6 as: where imax(a) is the maximum possible current for a given delay angle a(6 =1). The constant numerical val- ues in the equations have been chosen to achieve a close match between the exact and the approximate model. The two equations eqns. 30 and 31 are used for the actual control. They model the steady state behaviour of the proposed impedance controller. The dynamic model for the impedance controller can be obtained by assigning a first-order delay to each of the converters. The delay for the bridge is modelled as usual with a time constant of half the pulsewidth at 60Hz which is 1.3ms. The delay for the chopper is modelled by half the pulsewidth which is 0.21 ms. Based on the above model the control range of the controller in terms of real and imaginary current com- ponents can be calculated and is shown in per unit as fbnction of the control angle a and the pulse width P W in Fig. 5. I I 0.8 3 C l c C -- 0.6 2 0.4 ? 8 2 0.2 B .- E - 0 0 0.4 0.8 1.2 1.6 2.0 real component, pu Fig.5 Control range of the controller The results of the steady state analysis provide the calculation of the required controller currents which together with the load impedance result to the desired operating point of frequency and voltage of the gener- ating system. 5 Concept of t he closed-loop frequency and voltage control To develop the controller structure, typical distur- bances have to be considered. The main disturbances are fast changes of the load. The load changes its real power as well as the power factor. Further, but small and slow-changing disturbances, can be expected from the water situation, e.g. small changes in the head, and from parameter changes due to temperature. As a result of these considerations the control effort should first be directed at compensating the large and fast load changes to maintain constant voltage and frequency. Since the major disturbance is the change in load impedance, the proposed impedance controller could be used to rapidly compensate any impedance change in the load. The difficulty in this approach is to deter- mine the actual load impedance and to derive then the required controller signals. . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fig. 6 _ _ _ - structure of plant . . . . . . . . . structure of feedforward control Control structure of the voltage andfiequency control A successful strategy for the controller system design was found by assuming that a perfect voltage and fre- quency control had been achieved, after which the task to control impedance becomes the task of controlling the current. Based on this strategy a principle diagram of the control structure of the system can be drawn in terms of current components as shown inside the dashed box in Fig. 6. The currents shown in Fig. 6 are all space-phasors. The dynamic of the blocks shown in Fig. 6 is described by the equations discussed earlier. Neglecting the slow turbine influence, Fig. 6 shows that for required constant voltage and frequency the generator current iG has to be constant. Since the capacitor current i c is constant the generator-capacitor current iGC has to be held constant to maintain voltage and frequency constant. As a result the control task is to compensate any change in load current iL immedi- ately by a controller current i,, such that the resulting current iGC is constant in magnitude and phase. The required controller is a fast feedforward control system shown in Fig. 6 inside the dotted box. The sys- tem measures the instant space phasor of the load cur- rent and compares this current with the desired generator-capacitor current i*Gc to provide the com- mand value of the controller current isct. The control algorithm uses the formulae eqns. 30 and 31 to deter- mine the control angle a and the pulse width 6 which fed to the controller provide the desired controller cur- rent to keep the current iGC constant and therefore keep the voltage and frequency constant. To complete the idea of just compensating currents, one has to keep in mind that the controller can only consume real power and that it can only provide induc- tive reactive power. The smallest possible controller current is zero. If the controller current is zero, the IEE Proc -Gener Transm Distrrb , Vol 145, No I January 1998 36 bad current is at its possible maximum current iLmax = i*GC which is equal i o the generator-capacitor current ice If the load current is smaller than the maximum, the controller currenl required to keep the current iGC and therefore the voll age and the frequency constant is Z C t =ZLmax - ZL Fig. 7 shows the space-phasor diagram of the currents to visualise the proposed feedforward control of com- pensating for the fast-changing load current. - - - (32) Fig. 7 Feedforward comptnsation of loud current - - As with all feedforward controls a precise knowledge of the system and its parameters is required to compen- sate completely for the disturbance. As parameters may change or are not known accurately and since the feed- forward algorithm as proposed is only approximate, there will always be a small deviation from the desired steady state value. To solve this problem a closed-loop feedback control for voltage and frequency is required. This control has to correct the command of the maxi- mum load current iLmax such that deviations in fre- quency and voltag? become zero. This required feedback control is shown in Fig. 6. To determine the parameters of the feedback regula- tor a small-signal analysis was carried out. The result showed a dominant pair of poles. Based on this knowl- edge and keeping in mind that the feedback loop has to handle only slow disturbances, it was decided to use only two simple and slow PI-type regulators and pro- vide a static decoupler only. The static decoupler was based on the gains of the system and the PI regulator time constants were selected slower than the time con- stant of the dominant pair of poles. The gain of the regulators was selected to be considerably smaller than one. The choice of slow regulators separates the fast transients of the feedforward control very clearly from the feedback control and avoids that these two systems are fighting each other. 6 Simulation of the power system and the proposed controls The purpose of the simulation was to verify the con- cept of the fast feedforward control and the simplified and slow corrective control as proposed. The impact of the different delays of the impedance controller and the approximation of the impedance controller had to be studied. The complele system consists of 12 nonlinear differential equations resulting from the proposed model, the delays used in modelling the behaviour of the proposed impedance controller and the two PI-reg- IEE Proc -Gener Transm Du ri b, Vol 145, No 1, January 1998 ulators. Included in the simulation are the algebraic equations as defined by the system model, by the approximation of the impedance controller and by the static decoupler. The simulation was carried out using the interactive simulation system ISIPC at the Univer- sity of Toronto [9]. The simulation confirmed the proposed concept and predicted excellent results for the frequency and voltage control. The feedforward current compensation elimi- nates all fast transient disturbances resulting from the change in load current. Using only the feedforward strategy to compensate the change in load current small steady state errors occur as expected. These errors can mainly be traced to the approximations used to describe the impedance controller. Fig. 8 3 kVA experimental generator system I , chopper pulse \ Fig.9 Signal electronic and instrumentation for the experimental gener- ator system 7 verification by measurements To test the controller a power system with a 3kVA induction machine consisting of the required compo- nents was set up as shown in Fig. 8. The hydraulic tur- bine was implemented using a DC machine with adjustable armature voltage, adjustable field and a larger adjustable armature circuit resistor. This set-up permits adjustment of the DC machine characteristic to any desired slope, torque To and speed Qo to resemble the desired turbine characteristic. The circuits for the proposed controllers and the elec- tronic circuits to control the bridge and the chopper were built using the microcontroller MC68332 as shown in Fig. 9. The gating pulses for the chopper and the variable frequency phase controlled bridge are gen- Implementation of the proposed controls and 37 erated in the microcontroller. The complete software for the gating control, the control and regulator algo- rithms, the analog I/O and the keyboard I/O is written in the C-language. A 16 bit fixed point number system is used for the arithmetic. The control is implemented as a dual rate sampled data system, with a sample rate of 3kHz in general and a slow sample rate of 60Hz for the frequency measurement and the slow closed loop frequency regulator and voltage regulator. To demonstrate the system performance several tests with different load changes have been carried out; three of them will be discussed. +30y0 r------- - _ _ _ 1.2 1.06 I steadv - I 2 1.04 I 0.6 : 4 1.06 1.04 ; time of change 0.4 , , /, I 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 - I - I I 1.04 : 0.96 0.94 0.98 1 If m I - I c j,J I I I I ' time of change 0.96 0.94 In the first test a load operating at 90% of rated cur- rent at power factor 0.8 is changed by sudden switching to a load at 45% of rated current at about the same power factor of 0.76. This is a very large change in load and rather untypical for such a system, but smaller changes are so well controlled, that not much can be observed. Figs. 10a and b show the change in voltage and frequency in response to this large change in load. Voltage and frequency are shown in per unit (pu) of the rated values. The values shown are the mag- nitude and frequency of the space-phasor of the stator voltage as measured and calculated by the microcon- troller.The uncontrolled transient is indicated in these graphs with a dashed line. The results are very good and confirm the predictions of the simulation. The second test shows a resistive load running at about 80% of rated current being switched to about 8% of rated current. This is one of the harshest conditions, as pure resistive loads cause the fastest load current changes possible. This kind of test indicates the behav- iour of the system at transients such as load shedding or reclosure after the trip of a switch. Figs. l l a and b show the change of voltage and frequency in response to the second test transient. The dashed line indicates I - I c j,J I I I I ' time of change the disturbance for the unregulated generator. The offset which can be seen is due to the fact that the displayed traces are the response with only the feedfor- ward control active. 0.96 0.94 2 +23% . I : time of change . , :J, ' time of change 0.94 :f I I I I I I 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 time,s ,+84% a 3 1.02 1.00 0.98 Q - b time$ b Fig .I 1 Measurements test 2 Variation of a generator voltage and b frequency 70% load change, feedforward control only __ controlled response _ _ _ _ uncontrolled response "bc 100 V 0 -1 00 Fi y. 12 50A load change Measurements: variation of the generator line-to-line voltage at Both examples demonstrate the superior performance of the proposed control. Looking at the actual line-to- line voltage of the stator the small change in space- phasor amplitude and frequency can hardly be seen. Fig. 12 shows the stator line-to-line voltage Vb, and a load transient of 50% from full load to half load at about constant power factor. 8 Nonsymmetrical loads and automatic start up Using the experimental set-up and the established sim- ulation further studies were carried out [IO]. The first study concentrated on the feasibility of handling non- symmetrical loads and compensating by the impedance controller. The result of this study is that due to the fast switching chopper, it is possible to reduce the impact of nonsymmetrical loads on the symmetry of IEE Proc -Gener Transm Distrib , Vol 145, No I , January 1998 38 the stator voltages considerably. The control strategy developed holds the voltage and frequency variations of the stator voltage within 2% at 25% nonsymmetry. The nonsymmetry is defined as the amplitude of the negative sequence current to the positive sequence cur- rent. This is a very good result. A second study developed a control strategy for automatic start up of the generator turbine set assum- ing no external pow1:r is available and the system is started by opening the gate of the penstock which causes the turbine to start up. The automatic system senses the sudden increase in stator voltage of the gen- erator as soon as the self-excitation process takes place and stabilises the voltage and frequency to the desired values upon which the load can be connected by the main switch to the generator system. 9 Conclusions The self-excited induction generator in a stand-alone generating system WI th a hydraulic turbine using the principle of electric load governing by an impedance controller can provide excellent voltage and frequency control. With state of the art electronics and the appro- priate control concept, it is possible to achieve a volt- age and frequency regulation not offered by other small generating systems. The principle of the proposed impedance controller is general and can be used for other impedance control tasks. The proposed system is also applicable to syn- chronous generators including generators with perma- nent magnet excitation. 10 References SLEMON, G.R.: Electric machines and drives (Addison-Wes- ley, 1992) MALIK, N.H., and HAQUE, S.E.: Steady state analysis and performance of an isolated self-excited induction generator, ZEEE Trans., 1986, EC-I, ( 3) , pp. 134139 GULLIVER, J.S., and ARNDT, R.E.A.: Hydropower engineer- ing handbook (McGraw-Hill, 1991), pp 4.404.43 RAJ AKARUNA, S., and BONERT, R.: A technique for the steady state analysis of a self-excited induction generator with variable speed, ZEEE Trans. Energy Convers., 1993, 8, (4), pp. ARILLAGA, J ., and WATSON, D.B.: Static power conversion from self-excited induction generators, Proc. IEE, 1978, 125, (8), MURTHY, S.S., MALIK, O.P., and TANDON, A.K.: Analysis of self-excited induction generators, ZEE Proc. C, 1982, 129, (6), BONERT, R., and HOOPS, G.: Stand alone induction generator with terminal impedance controller and no turbine controls, IEEE Trans. Energy Convers., 1990, 5, (I), pp. 28-31 RAJ AKARUNA, S.: Control of a stand-alone self-excited induc- tion generator driven by an unregulated turbine. PhD thesis, Department of Electrical Engineering, University of Toronto, 1993 757-161 pp~ 743-746 pp. 260-265 BONERT, R.: Interactive simulation of dynamic systems on a personal computer to support teaching, ZEEE Trans. Power 10 RUYTER, E.: Automatic start-up and unbalanced load behav- iour of an electronically controlled induction generator system. Diplomarbeit, University of Toronto and University of Karl- sruhe, Germany, 1995 Sy~t . , 1989, 4, (l), pp. 380-383 IEE ProcGener. Transm. Distrib., Vol. 145, No. 1. January 1998 39