FI116430B - Synchronous machine procedure - Google Patents

Description

1 U:.; Hj

Method in conjunction with a synchronous machine

BACKGROUND OF THE INVENTION The present invention relates to a method according to the preamble of claim 1 in connection with a synchronous machine.

5 As with other rotating electrical machines, synchronous machines are controlled as accurately and reliably as possible depending on the application. In the past, it has been known to implement control of a synchronous machine based directly on a feedback field from the rotor axis of the synchronous machine. However, the implementation of the feedback requires a mechanical sensor with wiring. A mechanical sensor is a critical part of such 10 motor drives, where sensor failure necessarily means stopping operation and installing a new sensor. In terms of cost, the best solution for electric motor use is a sensorless system.

Typical modern engine control methods are based on mathematical modeling of the machine. Based on such a model and the resulting mouths, it is also possible to determine the rotor position angle and angular velocity, whereby the control of the machine can be implemented without direct feedback. One problem in such sensorless systems is the accurate estimation of the stator winding flux. The stator winding flux is an internal quantity of an electrical machine, the value of which is of particular importance in methods based on vector control or direct torque control.

·; · /; Estimating the direction and magnitude of the stator winding flux is problematic, especially at low speeds due to possibly incorrect engine model parameters. The stator winding flux is typically estimated by the stator • ·; t as a time integral of voltage and resistive voltage loss • · 25 ^ s = iUs-4 £ s) d '+ ^ s0- (1) • f: where us is the stator voltage, Rs is the stator resistance estimate, is is the stator current and ψ is the initial value of the stator winding flux. At low speed, the stator voltage and the resistive loss R ss are of the same magnitude, and the error in the resistance estimate Rs results in an incorrect stator winding estimate. This can lead to stability problems, or to a large error in the torque estimate calculated from the measured current and the stator winding flux.

1 1 i: Ο ϋ 2

WO 9949564 discloses a method for correcting the center of a stator flux estimate. The method of the publication is based on the stator flux and the stator current dot product, and the disadvantage of this method can be considered to be the slowness caused by the low pass filtering of the method, especially in the case of torque steps. In addition, the method of publication is not suitable for use at zero speed or low speed, since the flow and currents are then nearly equilibrium, so that the current and stator flow vectors do not form a circle whose center can be corrected according to the method.

One way to determine the rotor position angle is to inject a signal 10, which means adding a separate signal from one of the original signals fed to the machine. The signal can be injected, for example, into phase voltages or phase currents supplied to the machine. Signal injection aims to improve positionless sensor control by influencing motor behavior to facilitate rotor angle estimation.

15 However, the machine control method has its limitations for signal injection. When vector control is used, signal injection is generally accomplished by injecting the signal into either phase currents or voltages of the machine. However, this solution is not applicable in the case of methods using direct torque control.

BRIEF DESCRIPTION OF THE INVENTION It is an object of the present invention to provide a process which

• «· I

: w avoids the above disadvantages, and allows you to determine the rotor angle and * * * t: angular velocity of a synchronous machine in a simple and reliable manner. This * · ... the object is achieved by a method according to the invention which is characterized in what is stated in the characterizing part of the independent claim 1.

In direct torque control, the inverter output voltages are selected in a known manner using a switching table. The resultant of the connection table is the flow and torque instructions. In the method of the invention, injection instructions are added to these instructions, the behavior of which in an open pole machine is determined by stator current and stator flux. static

1 I

'! Further, the difference between the actual torque current and the stator flux injection frequency response can be determined by:: providing a difference between the actual rotor position of the machine and the estimated position angle, and implementing a control circuit that corrects this difference at all angular velocities.

116430 3

The method of the invention reliably and simply corrects the estimate of the rotor position angle, whereby the adjustment of the synchronous machine itself can be achieved using a method based directly on torque control without the problems of low speeds.

5 Brief Description of the Drawings

The invention will now be further described in connection with preferred embodiments, with reference to the accompanying drawings, in which:

Figure 1 shows triangular injection signals;

Figure 2 shows the measurement times during injection over one period; Figure 3 shows the formation of the injection angle;

Figure 4 shows the change in stator current and flux caused by injection signals;

Figure 5 shows the injection angle used in the inductance model;

Figure 6 shows an inductance model; Figure 7 shows the fit of the measured inductances to an inductance model;

Figure 8 shows amplitude modulated injection signals;

Fig. 9 is a block diagram of a method employing signal injection; and: * * ': FIG. 10 shows a signal injection flow diagram.

Detailed Description of the Invention

| I

* *. The method according to the invention is based directly on the torque control, in which the control of the electrical machine is carried out in a manner known per se: · 'utilizing the so-called optimum switching table. This basic II · 25 table selects the switching vector S * to be used at any given time, that is, the inverter output switch combination. Figure 9 illustrates a process according to the invention in combination with the entire control system. In this system, the coupling vector S * is generated using the stator current vector iss, the stator voltage vector uss, the angular velocity estimate <vr, and the torque and roll - *. ' as described below. With these output data, in systems based directly on torque control, the logic circuit 91 thus determines the estimator of the coupling vector S * and the stator flux vector. The switching vector S * is further implemented by inverter 92 which controls the motor 93. connected to the system. In the embodiment of FIG. 9, the torque reference Tref is obtained from the speed controller 911, which inputs the difference

5 When direct torque control is used as a machine control method, the signal injection may be performed in a manner different from the vector control method. According to the method of the invention, a torque injection signal 7inj is generated and this injection signal is added to the torque reference Tref to provide a torque control value Tcnt. Further, in accordance with the invention, a winding injection signal ^ inj is generated and this signal is added to the winding flow reference value ref ref to obtain a winding flow control value ψοη [.

In direct torque control, the control of the machine is accomplished by the flow and torque instructions ^ ref, Tref, which the control system seeks to implement by selecting appropriate voltage vectors. Because the current and voltage drawn by the machine vary according to the setpoint values, it is possible to add the injection signals directly to the setpoint values, so that the machine control system will try to follow the injection instructions added to the instructions. In this case, the current drawn by the machine and the voltage change with the set values added by the injection signals, i.e. the control values ^ cnt,

Tcnt, following.

When using the signal injection method, the shape of the injected signal: t:: is important. The signal shall be such that the control system is capable of monitoring changes in the signal caused by the signal. In this case, the frequency of the signal becomes an important factor. If the frequency of the signal is too high, the control system will not be able to adjust the machine with sufficient accuracy, thus losing the benefit of 25 injections. Therefore, changes in the signal must occur at a frequency such that the control system is capable of controlling the machine. Thus, in accordance with the invention, the preferred alternative to the injectable signal is a triangular wave signal having a frequency greater than the nominal frequency of the machine but still small enough to successfully control the machine: ··· * 30. Figure 1 shows the triangular wave injection signals Αψτβί, ATref to be added to the machine flow and torque instructions. The production of these signals is shown in Figure 9 in simplified form in the form of signal generators 94, 95.

In accordance with a preferred embodiment of the invention, the injection signals 3535 are triangular waves because the implementation of the triangular wave is simple and the control system is easy to follow ramp-like changes in the signal. Since the 116430 5 signal injection is implemented as part of the control system, attention must be paid to the amount of computation required, which makes the triangular wave signal a good choice.

Throughout this specification, reference is made to Figure 5 of the flowchart illustrating the method of the invention and to the numbering thereof, with the numbers placed in parentheses. Application of the method of the invention is initiated by initiating injection (101). This can be done at the same time as the synchronous machine control is started. Following the initiation of the injection, we proceed to the step of continuously monitoring the stator current and stator flux of the machine 10 (102).

According to an embodiment of the invention, currents (103) are used to measure the rotor angle estimate at the peak of the annual signal from the machine during injection. The peak value of the annual signal is simply determined by comparing two consecutive measured values, and from these values, the peak value of the annual signal is deduced.

Selecting a triangular wave as the injection signal can still influence the frequency and the peak value of the signal. In the flow and torque instructions, a separate signal is added to each, whereby the frequency and peak value of both signals are selectable. The frequency-20 index of the signal to be added to the torque is selected at half the frequency of the signal to be added to the flux.

: When performing a signal injection via direct torque control: the injection signals added to the reference values cause a change in the phase currents taken by the machine. Because the machine is modeled using space vectors, the change in current drawn by the machine in the composite form from the stator current is measured at the peak values of the injection signals. According to the invention, the measured current components are converted from the stator coordinate system to the rotor coordinate system, since the current and flux vectors represented in the static coordinate system change sinusoidally as a function of time. Thus, in the stator coordinate system, the current vector measured at time t is • at a different angle to the α-axis of the coordinate system than the current vector measured at time * 2 mi-30. When converting a current vector to a rotor coordinate system, the original sinusoidal current vector can be represented as a constant. In addition, the current converted to the rotor coordinate system shows the changes caused by the injection signal, added to the constant value in the form of an injection signal.

Another variable needed to calculate the rotor angle estimate is: 35 stator winding flux estimates u / s. The estimate of the stator winding current can be calculated from the currents and voltages measured by equation (1) in I to s.

6

1164iC

Since the current is measured according to the invention during injection, the changes caused by the injection signal are also visible in the winding flux. The winding flux is first calculated in the stator coordinate system, after which it is converted to the rotor coordinate system. In this case, the coil flow becomes a standard quantity if changes in injection are observed.

Figure 9 shows the determination and recording of peak values. The stator current of the synchronous machine 93 is continuously measured, and this is converted by the transform member 99 into the rotor coordinate system to provide a stator current vector in the rotor coordinate system irs. Likewise, the estimates of the stator winding flux) and if / ss are continuously determined, and this quantity is also transferred to the rotor coordinate system using the transform element 901 to provide an estimate of the stator winding flux in the rotor coordinate system ψτ. The recording element 902 determines the instant of the peak flux and stores the flux and current values at that instant (104).

According to the invention, it is possible to calculate the stator inductance, or transient inductance, of the change state in different directions with respect to the rotor angle, by means of the quantities converted into the rotor coordinate system. In this case, calculate the difference between the two measured points, ie the change in current and flux from the reference to the peak value of the injection signal, both positive (when increasing the reference) and. ; : 20 in the negative (set point reduction) direction. Figure 2 shows the measurement: ·, points, and changes in the signals between the measurement points. In Figure 2, the injection signal is suitably plotted to improve the perception of the measuring points. According to the invention, the recording element 902 .., stores at least two values of both the change in flow and the current, thus changing * * ·; 25 tos can be calculated.

'* ··' When calculating the change in positive direction, the change in both flux and current between points 1 and 2 is calculated. Similarly in the negative direction

t I

,; ; the changes in current and year during the change in the current are calculated between points 3, 4 of FIG. 2.

, v. 30 The change current and the change flow can be used to calculate the value of the inductance in the component form in the direction of the change. There may be a phase difference between phase currents and voltages. In this case, the flow may change in a different direction from the current, t · ': so the year must be projected in the direction of the current change. In this case, it is possible to calculate the induction in that direction. The mathematical definition of projection is 7 116430 (2) “^" | * | "/ where 47 and b are vectors and γοη is the angle between them, a is the projection of vector 5 into a vector b.

By calculating the change inductance (105) in the direction of current change, the projection gives: Αψ Δψ Δ / Jcosr - · (3) ^ | A £ S | 2 10 This calculation of the change inductance is shown in FIG. 9 at block 903. Similarly, the change inductance can in the direction of change.

Since, according to the embodiment, the calculation of the inductance is performed in the direction of the current change, it is necessary to control the direction of the current change. The direction of the current change can be influenced by injecting signals of different magnitude into the flux and torque ferrule. The signal ratio can be used to control the current change in the desired direction. If the signal injection is only applied to the flow instruction, the result is a change in current and flux along the stator vector v20. Correspondingly, a mere injection into the torque reference causes 90 ° • · '*. change relative to the stator vector. However, the injection of the torque signal does not go directly to the flux, but when calculating the change in the year, it should be noted that the proportion of the flux q component is calculated from the torque injection. This proportion can be calculated by solving the known formula of torque 25: ···: Te = ^ p ('l / sdis <r ¥ sllisd) 3 (4),', - 3Γ pi} PPM hq ~ i ^ sq ^ sd ^ sd hq ) i · 2.

Λ the proportion of the q-direction component: • · 30 116430 8 Ψη = τ-7 --- V (5) όρ Vsd Ψ * Α ~ Ψ? Μ.

^ Lq Ld, where p is the number of poles of the machine, Te is the electric torque, Lq is the specified inductance in the q direction of the machine, Ld is the inductance in the d-5 direction of the machine and y / PU is the winding current produced by the permanent magnets. Placing the torque signal change A7jnj in the formula (5) and the annual signal change in the d-direction component of the flux can solve the torque signal change in the stator flux.

While the frequency of the signal injected into the torque reference is, according to a preferred embodiment of the invention, half the frequency of the signal injected into the flux reference, the positive peak of the flux reference is positive every other time and negative every other time. Thus, the change in current is positive every other time for one signal to be added to the torque and negative every other 15 times. In this case, the change in the inductance takes place alternately in the positive and negative directions. In the method according to the invention, the frequency of the injection signal may be, for example, 200 Hz for flux injection and 100 Hz for torque injection.

Figure 3 illustrates the formation of the injection angle. Figure 3 illustrates a situation in which the position of the stator flux vector in the complex plane is in the direction of the α-axis. In this case, the amplitude of the signal to be added to the torque reference is Atreu, which causes a change in the flux vector calculated by equation (5). The amplitude of the signal to be added to the flow instruction is Δψχ &, which is transferred directly to the d-direction: component of the flow. The injection angle of the injection signals can be expressed by trigonometry as follows: * · <9in; = arctan -— (6)

O \ ^ sdJ

This calculation of the injection angle is shown in block 904 of the block diagram of FIG. 9.

Thus, during a single period of the signal injected into the torque reference, a total of two measurements of inductance change are obtained. The number could easily be increased to four if the direction of injection were also lowered as the signal changed from peak to zero. Since there is no significant change in signal behavior in this situation, the amount of computation required can be reduced by computing only two injection data per cycle.

5 When considering the effect of the injection signals on the normal current and flow vectors rotating in the stator coordinate system, this can be clarified with reference to Figure 4. In Fig. 4, the stator vector is rotated by an angle θτ on the α-axis of the stator coordinate system, i.e. the estimated rotor angle is θτ. The current injection of Figure 4 would cause the changes in the flow and flow vectors shown in the figure. When considering the situation, the rotational motion of the stator current and flux in the coordinate system must be taken into account as a function of time, so that the representation and calculation of the situation in the rotor coordinate system is justified.

The values of the change inductances calculated by injection can, according to the invention, correct any error in the angular estimate of the rotor. According to the method of the invention, using an inductance model, calculating the inductance value in the direction of the stator current or stator flux in the direction of the change vector and comparing the change inductance with the value calculated from the inductance model. Further according to the invention, the rotor angle estimate of the synchronous machine is corrected based on a comparison of said inductances.

Inductance * calculated according to a preferred embodiment of the invention. it is fitted (109) to a so-called inductance model prepared in advance from the machine. The inductance model can be thought of as the variation of the ratios of the inductances in the d and q directions as a function of rotor angle. When sensed, injection direction • * * '. and the inductance calculated in this direction, the value of the inductance can be compared with the inductance value calculated with 25 assumed angular data. As the values differ, the angle estimate is corrected so that the angle estimate approaches the correct value.

The inductance model can be represented in a known manner by the formula I »f t; $ 30 = Ls0 + Ls2cos20 [n ^, (7)!.! where /..,3'·'- (8) 35 Ls2 = ^ i (9) 116430 10 and flnj is the angle between the d-axis of the rotor coordinate system and the direction of injection, as shown in Figure 5.

In the inductance model, the d-axis is fixed to the lowest value of the model, 5 which is the value of the longitudinal inductance. The maximum point is at a transverse inductance value 90 ° from the longitudinal inductance. From the formula (7), the inductance model can be plotted as a function of angle according to Figure 6.

If necessary, the calculation of the inductance model can be simplified or refined. Simplification is achieved by considering the parametric parameters Lso and Ls2 in the model as follows: Lso is the average of one cycle and Ls2 is half the difference between the minimum and maximum values. The model can be refined by taking into account the noise in the measurements and calculating the angular error using, for example, the non-linear square sum method. The model formed by the square sum method can be further optimized by various optimization means.

The rotor angle estimate is corrected in accordance with the embodiment by fitting the calculated inductance values in the desired direction to the values of the model calculated in the same direction. Since the rotor angle is initially assumed to be correct, ie the rotor angle has been reliably estimated by some method, the so-called "angle" can be calculated from this angle information. values of the correct 20 inductance models in the injection directions. When calculating the values of the -.... inductance model, the true direction of the injection, ie the values of the model, must be taken into account. in the same direction as the actual injection direction. If the actual V injection angle changes from the assumed injection angle, the calculated inductance: values must be updated during the run. This is done by calculating the realized injection angle from the measured current and flux values and using this angle to calculate the model values.

Preferably, low pass filtering is used in the calculation of the inductances, i.e. the new calculated value is not used directly, but the new value is calculated by the previous filtered value. The value of the filtered change inductance can be calculated (106) from the unfiltered value using a low-pass filtering method known per se: ^ <9, fi (^) = ^ <9, filtr - 1) + (^ <9 (^) —Filt ~ O) > 00) ,, ..: τ 35 where k denotes the moment of measurement and k-1 the last moment of measurement. The low pass filtering is shown in the block diagram of FIG. 9 at block 905.

11643C

11

It is possible to improve the low pass filter used in the inductance model by methods known per se. Applying the known phase-loss filter equation to the inductance equation (10) gives: 4?, Filtr W = A?, Fm (^ ~ l) + (k) ~ -Vfiit '0) 5 Γι (11) + ^ (Le (k) - Le (k \)).

Γι

When fitting the measured inductance values to the model, the periodicity of the inductance model must be taken into account. The periodicity can be expressed as 10 - - ± ηπ <θ; η; <- ± ηπ, (12) 2 J 2 where n is a positive integer. Hereby, all injection angles can be adjusted to a given range such that the injection angle values range from -π / 2 to π / 2 according to formula (12). Now, by influencing the injection angle, any value from the curve implementing the inductance model can be calculated and fitted to the model.

Fig. 7 shows the fitting of two measured values to the inductance ': model. The injection is implemented in Fig. 7 at angles <9pos and 0neg. Measurement values: error relative to original curve are represented by ALp0S and AZ, neg and j ': 20 rotor angle error by 0err1 and <9err2.

When fitting the calculated inductance values to the inductance curve, pay attention to the required accuracy. During one cycle, two or even four measurement data are obtained from the currents used to calculate the inductances.

However, in these situations, the injection angle is fixed at a maximum of: 25 for two different values, if the injection in the positive direction is different, ·. in a direction rather than in a negative direction. In this case, the amplitude of the positive • · half-period of the triangle wave is different from that of the negative half-wave. For reliable and accurate calculation of the root torque angle information, inductance values are preferably calculated at various injection angles during signal injection and fitted to the original model.

.,: Injection signals to be added to the reference values can be simply amplitude modulated so that the angle of the signal injection can be changed during injection. By changing the amplitude of the signals at the end of each cycle, the angle of injection during the next period is changed. By increasing the amplitude of the signal injected into the flow guide in a different direction from that of the signal injected into the torque reference, a desired series of injection angles can be formed between cycles. This is accomplished by changing the amplitudes of the 5 (108) signals in predefined cycles.

The amplitude modulated injection signals are shown in Fig. 8. The advantage of the amplitude modulated signal injection according to the embodiment is the increasing accuracy, since more measurement values are obtained which are fitted to the model. The downside to the method is the time it takes, as calculating induk-10 dances at different injection angles will require several cycles before any correction can be made. However, the waiting time can advantageously be reduced by changing the method so that the number of measurements to be used is determined and the oldest measurement value is discarded each time a new measurement value comes. In this situation, it is easy to select the required amount of measurement data to be used to correct the rotor angle estimate. When using the method, the longest waiting time is at the start of the injection, which means waiting for the required amount of measurement data to be collected.

When the rotor angle estimate becomes erroneous, it is possible to perform an error correction such that the angular error values obtained from the inductance model are simply summed and the average error is corrected to obtain an angle correction. In this case, there may be a situation where the angular estimate reaches almost the correct rotor angle value ;; V as in the original model, but there is a phase shift between the corrected inductance model and the original: · 'model. This phase shift is due to the fact that it is not possible to compare the possible phase difference between the corrected angular ground and the inductance model by calculating the average of the * 25 cells only. A possible phase difference is due to the drift of the mean to zero, but the individual terms in the mean are not zero. In such a case, the correction method in question is not valid and, for example, a correlation correction must be used, which can also correct any phase difference error.

Correlation can be used to measure delays and phase shifts in electric motor drives and to identify operating parameters. The mathematical definition of the correlation mate- '· · ·' is 116430 13 Σ (χi-χhi-y) r = ~ Tt1 = 1 - Ί k ----- '(13) »1 = 1 V / = i where x is is the average of k samples of the x signal and y is the average of k samples of the y signal.

5 The value of the correlation r lies in the range When the correlation is one, the signals x and y are exactly the same for their exchange components. If the correlation is minus one, there is a 180 ° phase difference between the signals, but otherwise the signal exchange components are the same. With a correlation of zero, the signal exchange components have nothing to do with each other. For sine-10 shaped signals, this means that the signals have a phase difference of ± 90 °. The correlation is calculated (110) according to an embodiment of the invention in the block diagram of FIG. 9 by block 907.

The correlation formula does not always need to be used in full, but the application of the formula depends on the situation. In all situations, for example, the scaling term 15 as a divisor is not necessary, since the value of the correlation indicator may be known to be well within the appropriate range without even scaling. This is also the case with the method of the invention, since the v: value of the correlation is driven to zero, whereby the lack of scaling causes only a gain of jj in the phase angle correction.

20 Applying the correlation to the rotor angle estimation, the signals x and y can be defined to have the following form: x = a \ + b \ cos (20jnj) (14) • ·; 25 y = a2 + 62 sin (26 »inj) (15) r t i

The signal x corresponds to the filtered inductance value Le, fit according to formula (10): and the signal y is thought to be 90 ° in phase shift with the signal x. Since signal x is thought to implement an inductance model, signal y 30 must be a sinusoidal curve calculated at the same injection angle generated in block diagram 906 in FIG. 9. In this case,

11643C

14 if the rotor angle estimate is incorrect, the correlation value is non-zero. In this case, the correlation indicates the direction and magnitude of the angular error.

Based on the measurement data, the difference of the inductances from the original inductance curve is calculated and a correlation-5 th r is calculated using a sinusoidal signal. As the correlation r deviates from zero, the difference towards zero is corrected by the P1 controller 97 (111), whereby the error in the rotor angle can also be corrected. In other words, if the correlation is zero (112) the angle need not be corrected as it is correct. If the correlation deviates from zero, the angle is corrected (113) by the P1 control. The output of the P1 controller is the error θβπ in the estimated rotor angle 10. Then the rotor angle can be expressed as: θτ = θτ + θ, η. (16)

The whole of the signal injection method described above can best be understood by looking at Figure 9, which shows the control system, motor and inverter used in their entirety. The figure shows the dotted line 96 of the signal injection.

In the embodiment shown in Figure 9, in the control system, the rotor angle? R is estimated as the difference between the estimated stator flux angle Gs and the estimated 20 pole angle Ss in block 98. The pole angle Ss is obtained by the known: *: '; estimated in block 908 by the equation • »t I · • I · ί V tmSs = -, (17) ·: ··: \ is \ ~ Lsqts'1-s« · «• · # • ·» t 25 where L is the relative value of the torque. The torque itself is estimated in block 910 in a known manner by the cross product of the flow and current vector. : Estimation of the stator flux angle needed to calculate the rotor angle estimate (Il * country §s is calculated in block 912 from the stator flux estimate ψ *.

The rotor angle is derivatized by the derivative 913 to obtain an estimate of the electric rotor angular velocity ώχ. The electric angular velocity is converted to a mechanical angular velocity by dividing it by: * p 909.

': In the described embodiment of the invention, the frequency of the flux injection signal has been higher than that of the torque injection signal. However, it is clear that

11643C

The method of the invention may also be implemented by selecting a frequency higher than the flux injection signal as the frequency of the torque injection signal.

It will be obvious to a person skilled in the art that as technology advances, the basic idea of the invention can be implemented in many different ways. The invention and its embodiments are thus not limited to the examples described above, but may vary within the scope of the claims.

<I t * t · i r i ti »t ♦ l • ·» ·

Iti »· ♦» · il »• · f · *» · I * * »»

It * I 4 · · lit t · 1 · 1

Claims (12)

A method in connection with a synchronous machine, which synchronization machine is controlled by a method based on direct control of a torque, whereby a guide value (Tref) for the torque and a guide value (ψΓβ1) for interconnected flow are generated. , a stator current vector (iss) and a stator voltage vector (uss) for the control machine are determined, an angular velocity estimate (ώΓ) for the rotor is established, an inductance model for the rotor of the control machine is formed, and a torque injection signal (Tinj) is formed, characterized by where the torque injection signal (7ini) is added to the torque (7ref) of the torque to obtain a control value (7cnt) of the torque, an injection signal (inj inj) for interconnected flow is created ref) for interconnected flow to obtain a control value (ψ ^) for interconnected flow, the synchronous machine is controlled by the method based on direct control of the torque by using as instator the stator current vector: ': (Q and stator voltage vector (m *), the angular velocity estimate (ώ,.) of the rotor,: the control value (Tcnt) of the torque and (in // cnt) for interconnected flow:: de,: 1,: an estimate (ws) for the vector of the stator flow is determined,. ». * ··· 'The estimate (ψ *) of the stator flow vector is converted into the rotor, "coordinate system {ψ [)," ···, "the stator current vector (iss) is converted to the rotor coordinate system (irs), a change vector (Δ ^) for one of the injection signals generated t> ·; stator flow is determined, ': * * a change vector (AfJ of one of the injection signals generated': stator current is determined, ·: ·: the direction (Qinj) of the stator current or stator flow is calculated, the stator current of the change inductor (Le) is calculated, the direction of the change vector (0inj) on the basis of the change vector (Δ ^ ') of the stator flow and the change vector (Δ /') of the stator current, by means of an inductance model, an inductance value (L) is calculated in the direction of the stator current or the stator flow (end inu) of the stator current (0inj) is compared with the inductance value (L) calculated from the inductance model and an estimate (0) for the rotor angle of the synchronous machine is corrected on the basis of the comparison of the inductances (Le, L). Method according to claim 1, characterized by the reduced torque injection signal (7] nj) is a soft-tooth signal. 3. A method according to claim 2, characterized in that the interconnected flow injection signal (jnj) is a soft-tooth signal. Method according to any one of claims 1 to 3, characterized in that the frequency of the interconnected flow injection signal is a multiplicity of the torque injection signal frequency. Method according to any one of claims 1-3, characterized in that the frequency of the torque injection signal is an amount of the frequency of the interconnected flow injection signal. . ; , 20 Method according to claim 4 or 5, characterized in that; , ·, The relationship between the frequencies of the injection signals is two. Method according to any one of claims 1-6, characterized in that the change vector (Δij / rs) for that of the injection signals generated; v. stator flow is determined by calculating the value of stator flow over two time points; 25 and by removing the first value from the latter value. Method according to any of claims 1 to 6, characterized in that the change vector (Δ / ') for the injection signal generated by: the stator current is determined by calculating the value of the stator current under two * ·; · * Times and by removing the first value from the later value. : V: 30 Method according to claims 7-8, characterized in that the change vector for the stator flow generated by the injection signals and '. the stator current change vector is calculated over the same time period. Method according to any of claims 1-9, characterized in that the comparison of the change inductance (Le) with the inductance value (L) calculated from the inductance model comprises a step where the correlation between said inductances is determined. Method according to any of claims 1 to 10, characterized in that the correction of the estimate (Θ) of the rotor angle of the synchronous machine on the basis of the comparison of the inductances (Lg, L) comprises a step in which the error of the rotor angle (0r) is summed. rotor angle estimate (0) to correct the estimate. Method according to any one of claims 1 to 11, characterized in that the error of the rotor angle (0rr) is formed from the output of a PI controller (97), in whose input the result (s) of the correlation calculation of the change inductance (Lg) and the calculated inductance (L) is input. • »·» · * · 1 * · · »· I I» t> * »>»