In the following "synchronous machine" shall be taken to mean synchronous generator. Synchronous generators are used in electric power networks in the first place to supply active and reactive power in the "hour scale". Active power can also be controlled in the "second-minute scale" (frequency control), as well as reactive power (voltage control). Synchronous machines also provide suitable contributions in the "millisecond scale" to the fault currents, so that error states in the network can be quickly resolved in selective man- ner.

Synchronous machines are important production sources of reactive power in power systems. When the reactive power requirement increases in the system, this tends to lower the terminal voltage on the synchronous ma- chine. To keep the voltage constant, the field current is normally increased by means of the voltage regulator of the synchronous machine. The synchronous machine will thus produce the reactive power required to achieve reactive power balance at the desired terminal voltage.

The above-mentioned process applies as long as the power production corresponds to one point in the per- missible area in the capability graph of the synchro- nous machine, i.e. the graph of limits as regards reac- tive and active power, see Figure 1 showing the rela- tionship at overexcited operation. At overexcited op-

eration, i.e. when the synchronous machine is producing reactive power, the permissible operating area is lim- ited by thermally based rotor and stator current lim- its. The synchronous machines of today are normally dimensioned so that rotor and stator current limits in- tersect each other at a point in the capability graph corresponding to rated power at rated power factor, see Figure 1. The rated power factor for synchronous gen- erators is typically 0.8 - 0.95. At overexcited opera- tion, where the power factor is greater than the rated power factor, the limit for the capability graph of the synchronous machine consists of the stator current limit and, at overexcited operation, where the power factor is less than the rated power factor, the limit consists of the rotor current limit.

In conventional technology, if the stator or rotor cur- rent limits are exceeded current limiters, if such are installed and used, come into operation. These limiters reduce the currents by lowering the excitation. Since it takes a certain time before damaging temperatures are obtained, intervention of the current limiters of the stator or rotor is delayed several seconds before the current is lowered. The delay typically depends on the size of the current but it is usually less than one minute, see e.g. VERIFICATION OF LIMITER PERFORMANCE IN MODERN EXCITATION CONTROL SYSTEMS in IEEE Transaction on Energy Conversion, Vol. 10, No. 3, September 1995.

The current reduction is achieved by a decrease in the field current which results in a decrease in the termi- nal voltage and reactive power production of the gen- erator. The consequences for the part of the system in the vicinity of the machine are that the local reactive power production decreases and that it is more diffi- cult to import power from adjacent parts of the system, when the voltage drops.

If the transmission network is unable to transmit the power required at prevailing voltages there is a risk of the power system being subjected to voltage col- lapse. To avoid this it is advantageous for the power to be produced locally, close to the load. If this is not possible, and the power must be transmitted from other parts of the system, it is, as known, advanta- geous if this can be done at as high a voltage level as possible. When the voltage drops, the reactive power production (shunt capacitances) of the transmission lines decrease. Transformer tap-changers act in order to keep the voltages to the loads constant, and thus the power of the loads constant. If the power consump- tion of the loads is constant and the transmission voltage is lower than normally, the currents in the transmission lines will be higher and the reactive power consumption of the transmission lines will be greater (series inductances), see Cigré brochure 101, October 1995.

In many power systems, if current limiters come into operation for certain synchronous machines as described above, the reactive power production is limited and this may lead to a voltage collapse of the system.

In normal operation of the power system, with an essen- tially intact network, these situations are normally avoided by the installation of additional reactive power production resources, e.g. mechanically switched shunt capacitors and/or thyristor controlled static var compenstors (SVC), if necessary. However, as a wide- spread voltage collapse usually has severe consequences for the society, also abnormal operating conditions needs to be considered. If the network is weakened, due to e.g. faults or maintenance on important elements of

the network, the installed reactive power producing re- sources may no longer be sufficient, resulting in the above described situation which may lead to voltage collapse. The cost of installing additional controlla- ble reactive power producing resources, e.g. SVC de- vices, such that also these abnormal operating condi- tions can be handled is considerable. There is conse- quently a need for inexpensive controllable reactive power production reserves. These reserve resources should be capable of delivering reactive power such that voltage can be maintained at prescribed levels for at least 10 to 20 minutes giving the system operators a chance to take preventive actions, such as e.g. start- ing gas turbines or shedding load.

In power systems known today, or in power plants, the energy conversion usually occurs in two stages, using a step-up transformer. The rotating synchronous machine and the transformer, each have a magnetic circuit. It is known that manufacturers of such equipment are cau- tious and conservative in their recommendations for the set values in the limit devices, see Cigre brochure 101, October 1995, section 4.5.4., page 60. Coordina- tion is required and a certain risk of conflict thus exists in dimensioning and protecting generators and transformers. The step-up transformer has no air gap and is therefore sensitive to saturation as a result of high voltage or geomagnetic currents. The transformer also consumes part of the reactive power of the genera- tor, both at normal and abnormal operation. The major- ity of the active losses appear in the conductors of the armature circuit and the step-up transformer, while the core losses are relatively small in both devices.

One complication here is that the losses are normally developed at medium and high voltage and are therefore

more difficult to cool away than if they had been de- veloped at earth potential.

The object of the present invention is to achieve a synchronous machine for power and/or voltage control and a method for power and/or voltage control in order to avoid voltage collapse in power systems.

This object is achieved by a method and a synchronous machine of the type described in the introduction, with the features defined in claims 1, 26 and claim 30, re- spectively.

According to the invention, thus, the synchronous ma- chine is designed so that the thermally based rotor current limit is raised with respect to the thermally based stator current limit such that either the inter- section with the thermally based stator current limit in the capability graph is at a power factor value con- siderably below the rated power factor value, or the rotor current limit is raised above the stator current limit such that the two limits do not intersect. If the rotor and stator current limits intersect at the power factor zero in the capability graph as shown in Fig- ure 2, or if the rotor current limit is raised above the stator current limit, the stator current limit will be limiting for all overexcited operation.

In the following "cable" shall refer to high-voltage, insulated electric conductors comprising a core having a number of strand parts of conducting material such as copper, for instance, an inner semiconducting layer surrounding the core, a high-voltage insulating layer surrounding the inner semiconducting layer, and an outer semi-conducting layer surrounding the insulating layer. A synchronous machine with a stator winding

which comprises this type of cable can be designed for direct connection to the power network at higher volt- ages than with conventional machines, thus eliminating the need for a step-up transformer. In the case of re- active power production it is advantageous to use a ma- chine designed for direct connection to transmission level, since the reactive power consumed in the step-up transformer in the conventional plant instead can be delivered to the power network with a machine according to the invention.

The advantages of the invention are particularly no- ticeable in a machine wound with a cable of the type described above, particularly a cable having a diameter within the interval 20-200 mm and a conducting area within the interval 80-3000 mm2. Such applications of the invention thus constitute preferred embodiments thereof.

Raising the rotor current limit has a number of advan- tages for a synchronous machine. It enables direct measurement of limiting stator temperatures, for in- stance. This is considerably more difficult if the limiting temperatures are located in the rotor since it is difficult to measure, or in any other way communi- cate with a rotating object. Furthermore, reducing ac- tive power enables more reactive power to be produced.

This is also possible with conventional rotor dimen- sioning but more MVAr per reduced MW results in this case, as can be seen in the curves in Figures 1 and 2.

A number of other advantages are also gained by raising the rotor current limit, specific to this type of ma- chine. The time constants for heating (and cooling) the stator are large in comparison with a conventional machine. This means that the machine, with conven-

tional stator current limiters, can be run overloaded for longer than a conventional machine without damaging temperatures being reached. Simulations indicate that the stator safely can be overloaded 80 % for 15 minutes in some cases. This extended time period can be util- ized to take action either to reduce the system's need for reactive power, or to increase the production of reactive power. It is also easier to implement forced cooling of the stator of the machine. A machine of this type has a degree of efficiency comparable with that of a conventional machine, i.e. the stator losses are approximately equivalent. While a conventional ma- chine has primarily conductor losses, this type of ma- chine has less conductor losses and more core losses.

Since the core losses are developed at earth potential they are easier to cool away. A cooling machine can be used, for instance, for forced cooling in situations with high core temperatures.

With conventional current limiters the time period con- tributed by the time constant for heating, can be util- ized to reduce the active power and thus enable in- creased and/or prolonged production of reactive power.

The need for reducing the field is thus less and, in the best case, is eliminated.

With direct temperature measurement or temperature es- timation (or a combination thereof) we can pass from using the term "stator current limit" to talking about stator temperature limit(s) . Since it is the stator temperature (in critical points), and not the stator current, that is limiting, this offers a number of ad- vantages. The general tendency to set the limiter con- servatively can thus be lessened since it is the pri- mary quantity that is known and not a derivative. With a conventional current limiter no consideration can be

taken to the temperature of the machine when the cur- rent limit is exceeded, i.e. no consideration can be taken to the fact, for instance, that the machine was started shortly before the current limit was exceeded, or that the load was low shortly before. This can be avoided by using stator temperature limit(s) instead.

Cooling of the machine is dimensioned so that the sta- tor in continuous rated operation does not exceed a certain temperature - let us call this the rated tem- perature. This temperature is consciously set conser- vatively, i.e. the stator (insulation) can withstand higher temperatures for long periods of time. If the temperature in the critical points is known the machine can be run above rated operation for relatively long periods.

Dimensioning the rotor with salient poles (hydro- electric generators) in synchronous machines according to the invention is facilitated by the fact that the inner diameter of the stator can be made larger than in conventional machines since the stator winding is com- posed of cable in which the insulation takes up more space. It is thus possible to design the stator for this type of synchronous machine in accordance with conventional dimensioning procedures and change only the design of the rotor so that the rotor current limit is raised in the desired manner.

For a synchronous machine incorporating an air-cooled rotor with salient poles, this can be done, for in- stance, by utilizing the extra space to wind extra turns of the field winding in order to increase the magnetic pole voltage. A certain number of turns in the field winding then consist of cooling turns, thus increasing the cooled surface of the field winding. If the extra turns are provided with the same proportion

of cooling turns, as the other turns the temperature increase in the field winding can be kept at the same level as in a conventional dimensioning procedure, de- spite the magnetic pole voltage being raised.

For a synchronous machine with cylindrical rotor (turbo-rotor) the rotor current limit can be increased by making the machine longer, for instance.

The invention will now be explained in more detail in the following with reference to the accompanying draw- ings in which Figures 1 and 2 show capability graphs for overexcited synchronous machines with conventional dimen- sioning and in accordance with the invention, respectively, Figure 3 shows a cross section through the cable used for the stator winding in the synchronous ma- chine according to the invention, Figures 4 and 5 show two embodiments of a temperature estimator in the synchronous machine according to the invention, Figure 6 shows an example of a temperature-monitoring circuit that emits an output signal for further control, and Figures 7 - 9 show various circuits for control of the synchronous machine according to the invention.

Figure 3 shows a cross section through a cable used in the present invention. The cable is composed of a con- ductor consisting of a number of strand parts 2 made of copper, for instance, and having circular cross sec- tion. This conductor is arranged in the middle of the cable 1 and around the conductor is a first semicon- ducting layer 3. Around the first semiconducting layer

3 is an insulating layer, e.g. XLPE-insulation, and around the insulating layer is a second semiconducting layer that is normally earthed.

In the machine according to the invention the windings are thus preferably cables of a type having solid, ex- truded insulation, such as those used nowadays for power distribution, e.g. XLPE-cables or cables with EPR-insulation. Such cables are flexible, which is an important property in this context since the technology for the device according to the invention is based pri- marily on winding systems in which the winding is formed from cable which is bent during assembly. The flexibility of a XLPE-cable normally corresponds to a radius of curvature of approximately 20 cm for a cable 30 mm in diameter, and a radius of curvature of ap- proximately 65 cm for a cable 80 mm in diameter. In the present application the term "flexible" is used to indicate that the winding is flexible down to a radius of curvature in the order of four times the cable di- ameter, preferably eight to twelve times the cable di- ameter.

Windings in the present invention are constructed to retain their properties even when they are bent and when they are subjected to thermal stress during opera- tion. It is vital that the layers retain their adhe- sion to each other in this context. The material prop- erties of the layers are decisive here, particularly their elasticity and relative coefficients of thermal expansion. In a XLPE-cable, for instance, the insulat- ing layer consists of cross-linked, low-density poly- ethylene, and the semiconducting layers consist of polyethylene with soot and metal particles mixed in.

Changes in volume as a result of temperature fluctua- tions are completely absorbed as changes in radius in

the cable and, thanks to the comparatively slight dif- ference between the coefficients of thermal expansion in the layers in relation to the elasticity of these materials, radial expansion can take place without the adhesion between the layers being lost.

The material combinations stated above should be con- sidered only as examples. Other combinations fulfill- ing the conditions specified and also the condition of being semiconducting, i.e. having resistivity within the range of 10-1-106 ohm-cm, e.g. 1-500 ohm-cm, or 10-200 ohm-cm, naturally also fall within the scope of the invention.

The insulating layer may consist, for example, of a solid thermoplastic material such as low-density poly- ethylene (LDPE) , high-density polyethylene (HDPE) polypropylene (PP), polybutylene (PB), polymethyl pen- tene (PMP), cross-linked materials such as cross-linked polyethylene (XLPE), or rubber such as ethylene propyl- ene rubber (EPR) or silicon rubber.

The inner and outer semiconducting layers may be of the same basic material but with particles of conducting material such as soot or metal powder mixed in.

The mechanical properties of these materials, particu- larly their coefficients of thermal expansion, are af- fected relatively little by whether soot or metal pow- der is mixed in or not - at least in the proportions required to achieve the conductivity necessary accord- ing to the invention. The insulating layer and the semiconducting layers thus have substantially the same coefficients of thermal expansion.

Ethylene-vinyl-acetate copolymers/nitrile rubber, butyl graft polyethylene, ethylene-butyl-acrylate-copolymers and ethylene-ethyl-acrylate copolymers may also consti- tute suitable polymers for the semiconducting layers.

Even when different types of material are used as base in the various layers, it is desirable for their coef- ficients of thermal expansion to be substantially the same. This is the case with combination of the materi- als listed above.

The materials listed above have relatively good elas- ticity, with an E-modulus of E<500 MPa, preferably <200 MPa. The elasticity is sufficient for any minor differences between the coefficients of thermal expan- sion for the materials in the layers to be absorbed in the radial direction of the elasticity so that no cracks appear, or any other damage, and so that the layers are not released from each other. The material in the layers is elastic, and the adhesion between the layers is at least of the same magnitude as the weakest of the materials.

The conductivity of the two semiconducting layers is sufficient to substantially equalize the potential along each layer. The conductivity of the outer semi- conducting layer is sufficiently great to enclose the electrical field in the cable, but sufficiently small not to give rise to significant losses due to currents induced in the longitudinal direction of the layer.

Thus, each of the two semiconducting layers essentially constitutes one equipotential surface and the winding, with these layers, will substantially enclose the elec- trical field within it.

There is, of course, nothing to prevent one or more ad- ditional semiconducting layers being arranged in the insulating layer.

As mentioned above, the stator current limit is ther- mally restricted in the present invention. It is the insulation 4 that sets the limit in the first place.

If a cable with XLPE-insulation is used, the tempera- ture of the layer between the conductor and the insula- tion should not exceed 900C, which is the maximum tem- perature at rated operation and normal location in earth, for instance, i.e. the insulation can withstand this temperature for several hours and it may be briefly somewhat exceeded. The temperature of the sur- face layer between the insulation and the iron in the stator should not exceed a temperature limit of typi- cally 550C, i.e. the temperature difference over the insulation will be at least 350C.

A synchronous machine according to the invention may be dimensioned for a temperature of 70-800C in the conduc- tor and a core temperature of 40-500C at rated opera- tion. These temperatures are extremely dependent on the temperature of the coolant. A cooling machine may be used to lower this temperature although in normal operation this has a negative effect on the degree of efficiency. On the other hand, connection of such a machine may be justified in an emergency situation, al- though it must be taken into consideration that it may take a relatively long time to start up.

In order to make maximum use of the thermal inertia in the stator in a synchronous machine according to the invention it is desirable for the surrounding conductor and iron temperatures to be determined in the part of the insulation most critical from the heating aspect.

This can be achieved by direct measurement using meas- uring devices, or with a temperature estimator of the type shown in Figure 4. It is also possible to combine temperature measurement and temperature estimation ac- cording to Figure 5.

In Figure 4 losses in conductors caused by the stator current, and thus dependent on the machine's loading, are represented by a current source PLE, and the core losses caused by the flux (voltage), which are more or less constant irrespective of the load, by a current source PFE. The temperature of the coolant is repre- sented by the voltage source TKy. RR+S represents thermal resistance for cooling tubes and silicon fill- ing, RISO thermal resistance for the insulation and CLE, CISO and CFE, the thermal capacitance for conduc- tor, insulation and core. TLE in point 54 represents the temperature in the conductor and TISO in point 52 the mean temperature of the insulation. The model shown in Figure 4 can be calibrated by comparison of TFE with directly measured iron temperature. The tem- perature TLE is relatively difficult and expensive to measure directly since the conductor is normally at high potential.

The model shown in Figure 4 can also be refined by di- viding the thermal resistance between conductor and iron into several resistances connected in series, which would correspond to different radii of the insu- lation. By placing a capacitance from a point between each consecutive resistance and a reference tempera- ture, OOC, any temperature dependence of the thermal capacitance of the insulation can be modelled more pre- cisely. Since a temperature gradient exists in the in- sulation, such a division would result in a somewhat improved result.

In Figure 4 the temperatures TLE, TISo and TFE are considered as states whereas TKY, PLE and PFE are considered as input signals. The initial state values are needed to start the temperature estimator and the estimator is normally started simultaneously with the machine, i.e. from cold machine.

The number of nodes can of course be increased, but the embodiments described in connection with Figure 4 and below in connection with Figure 5 are to be preferred.

Figure 5 shows a modification of the temperature esti- mator in Figure 4, in which the iron temperature TFE is measured directly. The iron temperature will then be represented by a voltage source TFE in the thus simpli- fied diagram, and serves as input signal, together with PLE . The temperatures TISO and TLE constitute states and are obtained in the points 52 and 54 in the same way as in Figure 4.

The copper losses are dependent on the stator current and thus on how heavily loaded the machine is. The iron losses are dependent on the flux, which is more or less constant at terminal voltage, depending on the load. The time constant for the temperature increase and cooling of the core circuit is, on the other hand, extremely large in this type of machine and the machine therefore has greater overload capacity if it has just been started.

Both the iron losses and the copper losses will de- crease if the field is reduced.

An advantage of the synchronous machine according to the invention in comparison with a conventional machine

is that the electric losses are more associated with the flux in the core than with currents in the conduc- tors in the armature circuit. The core losses are de- veloped at earth potential, which facilitates normal cooling and even forced cooling with cooling machines.

The conductors of the armature circuit have relatively low current density and the losses on the high-voltage potential are relatively small.

The time constant for heating - and thus cooling - the core circuit is extremely large. Calculations show that the adiabatic temperature increase occurs in the order of hundredths of OK/s. The temperature increase in the armature circuit is also somewhat elevated as a result of the great thermal resistance in the solid in- sulation of the winding cable. At the current densi- ties in question the adiabatic temperature increases by 1/30 to 1/100 OK/s, while conventional machines have an adiabatic temperature increase in the order of 1/10 OK/s.

Both the temperature in the conductor TLE, and in the core TFE must be monitored and Figure 6 illustrates an example of a monitoring circuit that emits an output signal for further control. This circuit thus com- prises a temperature estimator 2 according to Figure 4, to which the input magnitudes I (stator current) , U (terminal voltage) and TKy are supplied. The output signals TLE, and TFE are obtained from the estimator 2, these being compared at 4 and 6, respectively, with pre-set limit values TL,LE, and TL,FE , as mentioned above, and the result of the comparison is supplied to a gate 8 (Lowest Value Gate). This gate emits a con- trol signal at its output constituting the temperature difference between temperature and temperature limit which is greatest in absolute terms.

If TFE is measured directly, only TLE, need be deter- mined from I and TFE with the aid of the temperature estimator. If both TFE and TLE, are measured directly, no temperature estimator is required and the measured temperatures are instead compared directly with the limit values.

Figure 7 shows in block diagram form an example of a control circuit for reducing the active power if the stator current exceeds a maximum permissible limit value.

A synchronous generator G is connected to a power net- work via a breaker 10. The generator G is excited via a thyristor-rectifier 12. The voltage U is supplied via a voltage transformer PTS to a measured value con- verter 14, a unit IL"Prod" for determining of the ac- tual stator current limit IL, and to a unit AP"Prod" for generating a signal "AP order" for reducing the ac- tive power if the stator current exceeds the stator current limit. In the same way, the current I is supplied via a current transformer CT to the units IL "Prod" and AP"Prod". In the unit IL "Prod" the direc- tion of the reactive power, voltage drop and initial time delay allowed for reducing the field are taken into consideration when determining the stator current limit. The stator current limit is based on the stator temperature at rated operation (TLE, = 70-800C and TFE = 40-500C with XLPE-insulation) . The rate of reduction and maximum range for the reduction of the active power is also determined in the unit AP"Prod", as well as a function, if any, for returning to the active power production the synchronous machine had before the sta- tor current limit was exceeded, if the reactive power requirement of the system again decreases.

The maximum reactive power the synchronous machine in the embodiment described can produce in steady state operation is equivalent to 100% of rated power and is obtained when the active power has been reduced to zero. However, there is cause to introduce a lower limit greater than zero for reducing of active power, since further reduction of active power gives little in return of increased ability to produce reactive power, see Figure 2. If more reactive power is required in steady state operation, this must be meet by a reduc- tion of the field after an appropriate time delay.

The output signal U from the network converter 14 is compared at 16 with a predetermined reference value UREF and the result of the comparison is supplied to an amplifier and signal-processing unit 18 before being supplied to a gate 20.

At 22 the stator current I is compared with the stator current limit IL generated in the unit IL"Prod", and the result of the comparison is supplied to an ampli- fier and signal-processing unit 24 and a subsequent block 26 with non-linear characteristic. The non-linear characteristic is such that a large output signal is obtained for positive input signals and an output sig- nal proportional to the input signal for negative input signals. The output signal from the block 26 is also supplied to the gate 20 which is a Lowest Value Gate, i.e. the signal that is lowest is obtained as output signal.

The output signal from the gate 20 is supplied to a signal-processing unit 28 with integrating action which is in turn connected to a trigger circuit 30 for the rectifier 12 of the excitation machine.

The control circuit in Figure 7 comprises essentially three main parts: an automatic voltage regulator, a stator current limiter and a system for reducing the active effect in order to increase the ability of the synchronous machine to meet the system's demand for re- active effect at the desired voltage level.

Reduction of the field current can be achieved in sev- eral ways according to the invention. A traditional limiter may thus be used that operates on the principle that if the stator current exceeds the stator current limit during a maximum permissible period, the field current is lowered until the stator current becomes equal to the stator current limit.

The actual control may be effected in various ways. In this case the initial time delay must be at least long enough to ensure that brief large currents arising out of error conditions in the system do not cause reduc- tion of the field because the current limit has been exceeded. Various methods of time delay are possible, e.g. a constant delay time irrespective of by how much the current exceeds the limit, or inverse time charac- teristic, i.e. the more the current exceeds the limit, the shorter the time delay. If the stator current limit has been exceeded, a period of time must be al- lowed for cooling. The type of synchronous machine un- der consideration has large time constants with regard to heating and cooling of the stator and the time delay can therefore be large in comparison with in the case of a conventional machine. This is because time is al- lowed either to reduce the system's demand for reactive power or increase the machine's ability to produce re- active power.

The dimensioning of the machine, together with reduc- tion of active power increases the machine's ability to produce reactive power.

According to the invention reduction of the field cur- rent is also possible starting from the temperature at the most critical points. The temperature of the con- ductors in the stator and the core temperature in the stator at the most critical points can be determined either through direct measurement, which may be diffi- cult in the case of conductor temperature, or with the aid of a temperature estimator with copper losses (stator current), iron losses (voltage) and coolant temperature as input signals, as discussed above. Two modes are thus possible for control, namely: 1) if the temperatures are below their maximum permissible temperature limits the field current is controlled so that the terminal voltage becomes equal to the desired operating voltage, and 2) if the terminal voltage is less than the desired operating voltage, the field current is con- trolled so that the conductor temperature or core tem- perature becomes equal to the maximum permissible tem- perature limit and the other temperature is below its limit.

The transition point where the stator temperature is equal to the maximum permissible stator temperature and the terminal voltage is equal to the desired operation voltage can be realized with a Lowest Value Gate, as described in connection with the figure.

Mode 1 above corresponds to normal voltage control, whereas mode 2 protects the machine against high tem- peratures since terminal voltage and stator temperature decrease when the field current decreases.

Figure 8 shows a control circuit for achieving control of the above-mentioned type.

Besides the current I and the voltage U , the unit AT"Prod" is also supplied with the temperature TKy of the coolant. The output signal from the unit AT"Prod" is supplied to an amplifier and signal-processing unit 40 and the block 26 with non-linear characteristic, as described earlier, for supply to the gate 20 together with the processed and amplified output signal from comparison of the voltage U with desired operation voltage Uref. Depending on the output signal from the gate 20, control of the machine is then carried out in a manner corresponding to that described in the embodi- ment according to Figure 7.

If the limiting temperature (TLE or TFE) approaches its maximum temperature limit (e.g. TL,LE, = 900C and TL,FE = 550C with XLPE-insulation) with a time derivative greater than zero, the above control may result in an "over-swing" in the temperature. If this over- temperature is brief, and providing it is moderate, it does not constitute a serious risk to the insulation.

However, it may result in a temporary voltage drop that may upset the stability of the powder system, as a re- sult of the control circuit attempting to counteract the over-temperature by reducing the field.

To avoid this, the control circuit may be supplemented with a temperature predicting circuit, e.g. based on the time derivative of the temperature, so that even before maximum temperature is reached, the voltage is permitted to gently start falling. The "over-swing" in temperature will then be slight, or altogether elimi- nated.

The voltage will thus commence falling earlier, but not so quickly.

A comparison between a traditional current limiter ac- cording to Figure 7 and a stator temperature limiter according to Figure 8 shows the latter to have the ad- vantage of allowing overload over a long period of time, in the order of hours, whereas the traditional current limiter only permits overload for a short pe- riod of time, in the order of seconds - minutes.

If the machine is equipped with stator temperature lim- iters, however, a warning signal should be sent to the operating centre as soon as the temperature for rated operation is exceeded, since this indicates that an overload situation exists and should be remedied.

Figure 9 shows a further development of the control circuit in Figure 7. Here a restricted control based on the temperature, aimed at maintaining the terminal voltage at as acceptable a level as possible for as long as possible by utilizing the thermal capacity of the stator to the maximum, is combined with a control of active and reactive power.

An output signal is thus generated in the unit AT"Prod" in the same way as in the circuit according to Figure 8. This signal is supplied to the amplifier and sig- nal-processing unit 40, block 26 and gate 20 to achieve the same limiting control as in Figure 8. The output signal from the unit AT"Prod" is also supplied to the unit AP"Prod", together with the voltage U- , whereupon a control signal AP order is obtained as output signal from the unit AP"Prod" in order to reduce the active power to U=ref, i.e. the terminal voltage equal to de-

sired operating voltage or until the active power reaches a predetermined minimum power limit, as men- tioned earlier. The reduction of active power is pref- erably commenced when either the core or the conductor temperature exceeds the temperatures the machine is di- mensioned for.

Yet another control possibility is based on starting a cooling machine to lower the iron and copper tempera- tures when either a current or temperature limit is reached. This enables the machine to be loaded fur- ther.