Method of estimating magnetic pole position in synchronous motor

This application is based upon and claims the benefit of priority of the prior Japanese Patent Application 2005-280410 filed on Sep. 27, 2005 so that the contents of which are incorporated herein by reference.

1. Field of the Invention

The present invention relates to a method of estimating a position (e.g., magnetic pole position) of a rotor in a synchronous motor wherein permanent magnets are used for the rotor.

2. Description of Related Art

In a synchronous motor such as a surface permanent magnetic synchronous motor (SPMSM) or an internal permanent magnetic synchronous motor (IPMSM), permanent magnets are used for a rotor, and S and N magnetic poles formed by the permanent magnets are alternately disposed around a shaft of the rotor. During the operation of this motor, an alternating current voltage is applied to windings of a stator, and an alternating current generated from the applied voltage flows through the windings. A phase of the voltage is controlled in synchronization with a position (e.g., magnetic pole position) of a rotor. Therefore, a position detector for detecting the magnetic pole position of the rotor is required. As this detector, a Hall element, an encoder or a resolver is used. However, it is difficult to downsize a motor with a position detector, and it is required to connect the detector with a controller for controlling the motor through connection lines.

Therefore, to control a synchronous motor without using any position detector, a magnetic pole position of a rotor in a motor has recently been estimated by using an induced voltage of the motor. During the operation of the motor, the motor itself induces a voltage changed with the magnetic pole position. Therefore, the magnetic pole position can be estimated from the induced voltage. For example, Published Japanese Patent First Publication No. 2001-251889 discloses a motor model approximated on a dq rotational coordinates system having a d-axis and a q-axis to estimate a magnetic pole position based on an induced voltage. The d-axis is set so as to be directed from an S magnetic pole to an N magnetic pole in a rotor. The q-axis is set to be orthogonal to the d-axis on a plane perpendicular to a rotation axis of the rotor. An original point of the coordinates system is placed on the rotation axis.

In the estimation based on this motor model, a shaft shift Δθ of an estimated pole position from a true pole position (i.e., d-axis) in a synchronous motor is calculated from a component Idc (hereinafter, called d-axis component) of a detected direct current along the d-axis, a component Iqc (hereinafter, called q-axis component) of the detected direct current along the q-axis, a d-axis component V*d of a voltage applied to the motor, a q-axis component V*q of the applied voltage and an instructed rotational speed ω*r of the rotor. An estimated magnetic pole position of the rotor is calculated from the shift Δθ.

However, in an actual motor, an alternating current voltage of three phases is applied to windings of a stator of a motor, and a three-phase alternating current flowing through the windings of the stator is detected. To obtain the components Idc and Iqc of the direct current expressed on the d-q rotational coordinates system, it is required to transform three-dimensional coordinates system for the actually detected alternating current into two-dimensional rotational coordinates system, based on an estimated magnetic pole position. In this case, the components Idc and Iqc include error derived from position error in the estimation of the position. Because a next magnetic pole position is estimated based on the components Idc and Iqc including error, precision of the estimation is inevitably lowered. As a result, when the motor is controlled according to the estimated position, the motor cannot stably be operated.

To solve this problem, direct estimation of a magnetic pole position based on a motor model using an extended induced voltage has been proposed in a paper of “An Extended E.m.f Observer for Salient-Pole Brushless DC Motor's Sensorless Control” in the national meeting in 1999, the Institute of Electrical Engineer of Japan (IEEJ), No. 1026.

In this paper, a magnetic field observer is constituted based on the model defined on an αβ stationary coordinates system representing a two-phase alternating current coordinates system. Each of an alternating current voltage applied to a synchronous motor and an alternating current detected from the motor is expressed by a vector having two components along α- and β-axes of the αβ coordinates system. When giving, to the observer, α- and β-axes components of the applied alternating current voltage and α- and β-axes components of the detected alternating current, a magnetic pole position can be directly estimated. This position is expressed as an angle between the α-axis and a line connecting an original point and a magnetic pole of a rotor in the αβ coordinates system.

In this paper, precision of the estimation of a magnetic pole position can be improved by the motor model using the alternating current voltage and the alternating current defined on the two-phase alternating current coordinates system. However, because the model includes a differential term to calculate an induced voltage in detail, an amount of calculation required for the estimation is enormously increased.

An object of the present invention is to provide, with due consideration to the drawbacks of the conventional estimation of a magnetic pole position, a method wherein a magnetic pole position of a rotor in a synchronous motor is correctly estimated while reducing an amount of calculation required for the estimation without substantially lowering precision in the estimation.

According to an aspect of this invention, the object is achieved by the provision of a method of estimating a magnetic pole position in a synchronous motor, comprises the steps of (1) calculating an induced voltage of the motor, to which an alternating current voltage is applied and from which an alternating current is detected, from components of the applied alternating current voltage in a two-phase alternating current coordinates system and components of the detected alternating current in the two-phase alternating current coordinates system, and (2) estimating a magnetic pole position of the rotor from the induced voltage. A differentiated value of a wave height of the alternating current with respect to time is substantially set at zero.

Although a wave height of the alternating current is slightly changed with time, this wave height is approximated to or replaced with a wave height substantially not depending on time. In this case, differential terms in the calculation of the induced voltage from the applied alternating current voltage and the detected alternating current can be removed. Therefore, an amount of calculation required for the estimation can be considerably reduced, and precision in the estimation is not substantially lowered.

As compared with an induced voltage model constituted on the dq rotational coordinates system, precision in the estimation can be heightened.

FIG. 1 is a block diagram of a control unit for a synchronous motor wherein a method according to a first embodiment of the present invention is performed;

FIG. 2 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to the first embodiment;

FIG. 3 shows wave shapes of detected alternating currents having a constant wave height and a wave shape of a magnetic pole position;

FIG. 4 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to a second embodiment of the present invention;

FIG. 5 is a block diagram of a position and speed estimating unit of the control unit wherein the method shown in FIG. 4 is performed;

FIG. 6 is a block diagram of a position and speed estimating unit of the control unit according to a modification of the second embodiment;

FIG. 7A shows a wave shape of an extended induced voltage with noises in an αβ coordinates system according to a third embodiment of the present invention;

FIG. 7B shows a wave shape of an extended induced voltage with noises in a dq coordinates system according to the third embodiment;

FIG. 7C shows a wave shape of a filtered extended induced voltage in the dq coordinates system;

FIG. 7D shows a wave shape of a filtered extended induced voltage in the αβ coordinates system;

FIG. 8 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to the third embodiment;

FIG. 9 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to a fourth embodiment;

FIG. 10 is a block diagram of a position and speed estimating unit wherein the method shown in FIG. 9 is performed;

FIG. 11 shows a positional relation between a voltage vector and a position vector of an estimated position;

FIG. 12 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to a fifth embodiment;

FIG. 13 is a block diagram of a position and speed estimating unit wherein the method shown in FIG. 12 is performed;

FIG. 14 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to a first modification of the fifth embodiment:

FIG. 15 is a block diagram of a position and speed estimating unit wherein the method shown in FIG. 14 is performed; and

FIG. 16 is a block diagram of a position and speed estimating unit wherein a second modification of the fifth embodiment is performed.

Embodiments of the present invention will now be described with reference to the accompanying drawings.

FIG. 1 is a block diagram of a control unit for a synchronous motor wherein a method according to a first embodiment of the present invention is performed. FIG. 2 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to the first embodiment.

A synchronous motor 10 shown in FIG. 1 has windings of a stator and a rotor (not shown) made of permanent magnets.

When a three-phase alternating current voltage is applied to the windings, a three-phase alternating current composed of phase currents of U, V and W phases flows through the windings so as to generate a changeable magnetic field due to electromagnetic induction, and the rotor having the permanent magnets is rotated on its rotation axis in response to the magnetic field. The rotational force is outputted through a transmitter such as a pulley and a belt.

A phase of the current is controlled in a control unit shown in FIG. 1 to stably rotate the rotor at a desired rotational speed.

As shown in FIG. 1 , data of an instructed (or desired) rotational speed ω* of the rotor (i.e., speed of a magnetic pole) is repeatedly given to the control unit. Each time data of the speed ω* is given to the control unit, a first difference calculator 1 calculates a difference between the speed ω* and an estimated rotational speed ω es of the rotor. A speed controller 2 calculates values of both a d-axis component i d * and a q-axis component i q * of an instructed alternating current i ac * defined in a dq rotational coordinates system from the difference ω*−ω es such that the estimated rotational speed ω es approaches the instructed rotational speed ω*. As described later, the speed ω es is derived from the components i d * and i q *.

In the dq rotational coordinates system, the d-axis is directed from an observed S magnetic pole to a corresponding N magnetic pole of the rotor, the q-axis is orthogonal to the d-axis on a plane perpendicular to a rotation axis of the rotor, and an original point is set at a position on the rotation axis. Therefore, a true magnetic pole position is placed on the d-axis.

A second difference calculator 3 calculates a d-axis difference Δi d between the d-axis component i* d and a d-axis component i d of a detected current and calculates a q-axis difference Δi q between the q-axis component i* q and a q-axis component i q of the detected current. The current is detected from stator windings of a motor 10 . A current controller 4 calculates values of both a d-axis component V* d of an instructed voltage and a q-axis component V* q of the instructed voltage from the differences Δi d and Δi q such that each of the differences Δi d and Δi q approaches zero.

A dq/αβ coordinate transformer 5 transforms the dq rotational coordinates system into an αβ coordinates system, based on a final estimated magnetic pole position θ es of the rotor, to convert the components V* d and V* q of the instructed voltage into an α-axis component V* α and a β-axis component V* β of the instructed voltage defined in the αβ coordinates system. The position θ es is expressed by a rotational angle of the rotor.

As is well-known, the αβ coordinates system is a stationary system and is defined by an α-axis and a β-axis orthogonal to each other on a plane perpendicular to the rotation axis of the rotor, and its original point is set on the rotation axis of the rotor to be the same as that of the dq coordinates system. Because the αβ coordinates system is a stationary system, voltage and current are, respectively, expressed by an alternating current signal and an alternating current voltage signal on the αβ coordinates system. Therefore, the αβ coordinates system represents a two-phase alternating current coordinates system.

When a magnetic pole position of the rotor is correctly estimated so as to coincide with the true position, a relative relation in phase (i.e., an angle between the α-axis and the d-axis) between the αβ stationary coordinates system and the dq rotational coordinates system can be specified based on the correctly estimated position.

Therefore, the coordinate transformation can be correctly performed based on the estimated position θ es .

An αβ/uvw coordinate transformer 6 transforms the αβ coordinates system into a UVW coordinates system defined by a U-phase axis, a V-phase axis and a W-phase axis to convert the components V* α and V* β of the instructed voltage into a U-phase component V*u, a V-phase component V*v and a W-phase component V*w. The three axes of the UVW coordinates system are spaced at equal intervals of 120 degrees in electrical angle and correspond to U, V and W phases, and an original point of the UVW coordinates system is set at the same position of that of the αβ coordinates system.

A pulse width modulation (PWM) signal generator 7 generates a PWM signal from each of the components V*u, V*v and V*w. The PWM signals are, respectively, transmitted to an inverter 8 to operate three switching elements corresponding to the stator windings. Therefore, an alternating current voltage having the components V*u, V*v and V*w is applied to the windings. In response to the applied voltage, a three-phase alternating current having a U-phase component i U , a V-phase component iv and a W-phase component i W flows through the windings. Further, the motor 10 itself generates an induced voltage depending on a magnetic pole position. Therefore, the alternating current depends on the applied voltage and the induced voltage.

A current sensor 9 detects values of the components i U , i V and i W of the alternating current. The sensor 9 may detect only two of the three component values to calculate the other component value from the detected values. A uvw/αβ coordinate transformer 11 transforms the UVW coordinates system into the αβ coordinates system and converts the components i U , i V and i W of the detected alternating current to an α-axis component i α and a i-axis component i β of the detected alternating current.

The components i α and i β are received in an αβ/dq coordinate transformer 12 and a position and speed estimating unit 13 . The transformer 12 transforms the αβ coordinates system into the dq coordinates system, based on a final estimated magnetic pole position θ es outputted from the estimating unit 13 , to convert the component i α and i β of the detected alternating current to the d-axis component i d and the q-axis component i q of the detected current. The components i d and i q of the detected current are transmitted to the calculator 3 .

Estimation of position and speed of a magnetic pole (or a rotor) performed in the unit 13 is described with reference to FIG. 2 .

An extended induced voltage calculator 14 of the unit 13 calculates values of an α-axis component e α and a β-axis component e β of the induced voltage from values of the components V* α and V* β of the instructed voltage and values of the components i α and i β of the detected current (step S 11 ). This calculation is performed based on an approximate induced voltage model which is obtained as approximation of an extended induced voltage model. The extended induced voltage model according to the prior art is expressed in the αβ coordinates system according to a first formula:

[ V α V β ] = [ R + pL d ω re ⁡ ( L d - L q ) - ω re ⁡ ( L d - L q ) R + pL d ] ⁡ [ i α i β ] + { ( L d - L q ) ⁢ ( ω re ⁢ i d - i q ) + ω re ⁢ ⁢ K E } ⁡ [ - sin ⁢ ⁢ θ re cos ⁢ ⁢ θ re ]

by using both an a -axis component V α and a β-axis component V β of an instructed voltage applied to a motor, a resistance R of a stator, a rotational speed ω re of a rotor in electrical angle obtained by differentiating a position of the rotor estimated in a previous estimation cycle with respect to time, a d-axis inductance L d in the motor, a q-axis inductance L q in the motor, a differential operator p, both an α-axis component i α and a β-axis component i β of a current detected from the motor, and a constant K E of an induced voltage.

In this formula, the second term of the right side is defined as an extended induced voltage. In this model, in addition to an induced voltage based on permanent magnets of the rotor, an inductance difference L d −L q generating a reluctance torque is considered as a component of an additional magnetic flux. When a magnetic field observer is constituted based on this model, both an α-axis component e α and β-axis component e β of an extended induced voltage can be directly calculated according to a second formula:

[ e α e β ] = s s + α ⁢ { [ V α V β ] - [ R + pL d ω re ⁡ ( L d - L q ) - ω re ⁡ ( L d - L q ) R + pL d ] ⁡ [ i α i β ] }

by additionally using a differential operator s and an observer pole α.

Because this model includes a differential term s/(s+α), an amount of calculation required for the estimation is enormously increased. To reduce the amount of calculation in this embodiment, an approximate induced voltage model not including any differential term is used for estimation of a magnetic pole position, in place of the extended induced voltage model.

More specifically, a wave height Ia(t) of both the α-axis component i α (i α =−Ia (t)·sin θ, θ=ω re t+c, c is a constant) and the β-axis component i β (i β =Ia(t)·cos θ) in the detected current is changed with time. This time depending wave height Ia(t) is approximated to or replaced with a wave height Ia substantially not depending on time t (d/dt·Ia≈0). That is, a differentiated value of a wave height of the alternating current with respect to time is substantially set at zero (step S 12 ). In this case, a differential term for the α-axis component i α in the extended induced voltage model can be rewritten according to a third formula.

In the same manner, a differential term for the β-axis component i β in the extended induced voltage model can be rewritten according to a fourth formula.

pL d ·i β =L d ·ω re ·i α

Therefore, the second formula can be simplified and rewritten, and a fifth formula expressing the approximate induced voltage model according to embodiments of the present invention is obtained by inserting the third and fourth formulas to the second formula.

=

In this fifth formula, the approximate induced voltage model for calculating the components e α and e β is expressed by the stator resistance R, the rotational speed ω re and the q-axis inductance Lq. Therefore, when it is regarded that the differentiated value of the wave height Ia with respect to time is substantially equal to zero, that is, the wave height is substantially constant, differential terms included in the extended induced voltage model can be removed in the approximate induced voltage model. When the components e α and e β are calculated based on the approximate induced voltage model not including differential terms, an amount of calculation can be considerably reduced.

An estimated position calculator 15 calculates a first estimated magnetic pole position θ es1 from the components e α and e β calculated in the calculator 14 (step S 13 ).

Because a voltage vector determined by differentiating a magnetic flux (i.e., magnetic pole position) of a rotor with respect to time is advanced by 90 degrees from the magnetic flux, the relation of tan θ es1 =−e α /e β is satisfied.

Therefore, the position θ es1 is obtained by an arctangent calculation according to a sixth formula.

=

When no noises are superimposed on the induced voltage, the position θ es1 is outputted to the transformers 5 and 12 as a final estimated magnetic pole position θ es . A differential processing unit 16 differentiates the position θ es1 with respect to time according to a seventh formula to obtain the estimated rotational speed ω es of the rotor (step S 14 ).

ω es = s 1 + τ · s · θ es ⁢ ⁢ 1

Here, a differential operator s and a constant τ are used in this formula.

The position θ es1 denotes an alternating current component corresponding to the induced voltage, and this component periodically changes with time in response to the components e α and e β . When the position θ es1 is differentiated, the position θ es1 denoting the alternating current component can be converted into the speed ω es denoting a direct current component which is constant with respect to time.

The inductances Ld and Lq are changed with time due to dead time generated by the actuation of the motor 10 and/or salience of magnetic poles. Therefore, when the induced voltage components e α and e β are calculated, high frequency components are superimposed on the induced voltage as noises. When the position θ es1 is derived from the induced voltage, the high frequency components inversely influence on the position θ es1 . Therefore, precision of the estimation of a magnetic pole is degraded by the high frequency components.

A low pass filter 17 performs a low pass filtering for the speed ω es to remove the high frequency components from the speed ω es (step S 15 ). Because the speed ω es is expressed by a direct current component, the high frequency components can be reliably removed from the speed ω es .

FIG. 3 shows a wave shape of each of detected alternating currents i α (i α =−Ia·sin θ) and (i β =Ia·cos θ) and a wave shape of a magnetic pole position θ indicated by a saw tooth wave. As shown in FIG. 3 , high frequency components can be removed from an estimated position θ without loosing information of start and end positions (θ=0, 2π) of a magnetic pole. The start and end positions are placed on the α-axis. Therefore, precision of the estimation of a magnetic pole position can be improved when the coordinates transformation is performed in the transformer 5 based on a final estimated pole position not including high frequency components.

The filter 17 outputs the speed ω es to the calculator 1 and an integrator 18 . The integrator 18 integrates the speed ω es with respect to time to obtain a second estimated magnetic pole position θ es2 (step S 16 ). The position θ es2 denotes a position from which the high frequency components included in the position θ es1 are removed.

When the position θ es2 is obtained by filtering the position θ es1 in the filter 17 , a phase delay occurs in the position θ es2 . To compensate this phase delay, a phase compensating adder 19 adds a first phase compensating value outputted from a phase compensator 20 to the position θ es2 to obtain a final estimated magnetic pole position θ es (step S 17 ).

To calculate the phase compensating value in the compensator 20 , a third difference calculator 21 calculates a positional difference Δθ es (=θ es1 −θ es ) between the estimated positions θ es1 and θ es (step S 18 ), and the compensator 20 multiplies the difference Δθ es by a first proportion and integration (PI) gain set at a value of Kp 1 +Ki 1 /s (1/s denotes an integration with respect to time) to obtain the first phase compensating value (step S 19 ).

Because a phase lock loop (PLL) is constituted by the adder 19 , the compensator 20 and the calculator 21 , a phase of the position θ es can approach a phase of the position θ es1 . That is, the position θ es synchronized in phase with the position θ es1 can be obtained (step S 20 ). Therefore, although the phase delay is caused to remove the high frequency components in the filtering, a phase delay in the position θ es can be prevented by adding the first phase compensating value to the position θ es2 .

The final estimated position θ es is outputted to the transformers 5 and 12 (step S 21 ). Therefore, the estimated position θ es approaches a true magnetic pole position (i.e., d-axis) of the rotor, and the transformation between the dq coordinates system and the αβ coordinates system can be correctly performed in the transformers 5 and 12 . That is, the motor can stably be operated.

Accordingly, because the time depending wave height of the detected current components i α and i β is replaced with the wave height Ia substantially not depending on time t, a differential term for the components can be removed from the extended induced voltage model. Therefore, an amount of calculation required for the estimation can be reduced, and precision of the estimated position θ es can be improved, as compared with a case where estimation is performed based on an induced voltage model constituted on the dq rotational coordinates system.

When the first PI gain is excessively heightened, the high frequency components are undesirably included in the position θ es . In this case, because precision of the estimation of a magnetic pole position is lowered, it is required not to excessively heighten the first PI gain.

In this embodiment, high frequency noises are removed in the filter 17 . However, when adverse influence of the noises on the first estimated position θ es1 is low, the position θ es1 may be outputted to the transformers 5 and 12 without performing the low pass filtering for the position θ es1 , and the speed ω es obtained in the unit 16 may be outputted to the calculator 1 .

FIG. 4 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to a second embodiment of the present invention. FIG. 5 is a block diagram of another estimating unit 24 wherein the method shown in FIG. 4 is performed.

As shown in FIG. 4 and FIG. 5 , an estimating unit 24 has a phase compensating adder 22 arranged between the filter 17 and integrator 18 , in place of the adder 19 . The adder 22 adds a second phase compensating value outputted from the compensator 20 to the speed ω es outputted from the filter 17 to obtain a phase compensated speed ω es (step S 41 ). The integrator 18 integrates this speed ω es with respect to time to obtain a final estimated magnetic pole position θ es (step S 42 ). The calculator 21 calculates a positional difference Δθ es (=θ es1 −θ es ) between the estimated positions θ es1 and θ es (step S 43 ). The compensator 20 multiplies the difference Δθ es by a second PI gain set at a value of Kp 2 +Ki 2 /s to obtain the second phase compensating value (step S 44 ).

A phase lock loop (PLL) is constituted by the calculator 21 , the compensator 20 and the adder 22 . Therefore, a phase synchronization of the final estimated position θ es with the first estimated position θ es1 is performed in the PLL, and the position θ es synchronized in phase with the position θ es1 can be obtained (step S 45 ). The position θ es synchronized with the position θ es1 is outputted to the transformers 5 and 12 (step S 46 ). Accordingly, in the same manner as in the first embodiment, a phase delay in the position θ es can be prevented, and the same effects as in the first embodiment can be obtained.

FIG. 6 is a block diagram of an estimating unit 25 according to a modification of the second embodiment.

As shown in FIG. 6 , an estimating unit 25 may have both the adders 19 and 22 . The compensator 20 outputs the first and second phase compensating values to the adders 19 and 22 , respectively. The adder 22 adds the second phase compensating value to the speed ω es outputted from the filter 17 to obtain a phase compensated speed ω es . The integrator 18 integrates this speed ω es to obtain a second estimated magnetic pole position θ es2 . The adder 19 adds the first phase compensating value to the position θ es2 to obtain a final estimated magnetic pole position θ es .

A phase lock loop (PLL) is constituted by the calculator 21 , the compensator 20 and the adders 19 and 22 , so that the position θ es synchronized with the position θ es1 can be outputted to the transformers 5 and 12 . Accordingly, a phase delay in the position θ es can be further prevented.

In the first embodiment, to remove high frequency components (i.e., noises) superimposed on extended induced voltage components e α and e β , the position θ es1 expressed by an alternating current component is converted into the speed ω es expressed by a direct current component. In contrast, in the third embodiment, the voltage components e α and e β expressed by alternating current components are directly converted into extended induced voltage components e d and e q expressed by direct current components to remove noises from the voltage components e d and e q .

FIG. 7A shows a wave shape of an extended induced voltage in the αβ coordinates system, FIG. 7B shows a wave shape of an extended induced voltage in the dq coordinates system, FIG. 7C shows a wave shape of a filtered extended induced voltage in the dq coordinates system, and FIG. 7D shows a wave shape of a filtered extended induced voltage in the αβ coordinates system. FIG. 8 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to the third embodiment.

Because the voltage components e α and e β are defined in the two-phase alternating current coordinates system (e.g., αβ coordinates system), the voltage components e α and e β change with time. As shown in FIG. 7A , when high frequency components denoting noises are superimposed on the extended induced voltage, it is difficult to remove the noises from the components e α and e β .

Therefore, as shown in FIG. 7B , the components e α and e β of the calculator 14 are converted into components e d and e q of the extended induced voltage by transforming the αβ stationary coordinates system into the dq rotational coordinates system (step S 81 in FIG. 8 ). The extended induced voltage is expressed by direct current components e d and e q , so that high frequency noises can be easily removed from components e d and e q . Then, a low pass filtering is performed for the components e d and e q to remove high frequency components from the extended induced voltage (step S 82 in FIG. 8 ). Therefore, as shown in FIG. 7C , the components e d and e q not including high frequency components are obtained. Then, the components e d and e q are converted into components e α and e β by transforming the dq rotational coordinates system into the αβ stationary coordinates system (step S 83 in FIG. 8 ). Therefore, as shown in FIG. 7D , the components e α and e β not including noises are obtained.

Thereafter, a final estimated magnetic pole position θ es not including noises is calculated from the components e α and e β according to an eighth formula (step S 84 in FIG. 8 ).

θ es = tan - 1 ⁢ - e α e β

Then, the position θ es is differentiated with respect to time according to a ninth formula to obtain an estimated magnetic pole speed ω es (i.e., rotational speed of rotor) (step S 85 in FIG. 8 ).

ω

es

=

The position θ es is outputted to the transformers 5 and 12 shown in FIG. 2 , and the speed ω es is outputted to the calculator 1 shown in FIG. 2 (step S 86 in FIG. 8 ).

Accordingly, because high frequency noises are removed from the direct current components of the extended induced voltage, the same effects as in the first embodiment can be obtained.

In this embodiment, phase compensation for phase delay caused in step S 82 may be performed for the estimated magnetic pole position in the same manner as in the first embodiment.

FIG. 9 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to the fourth embodiment, and FIG. 10 is a block diagram of a position and speed estimating unit wherein the method shown in FIG. 9 is performed.

As shown in FIGS. 9 and 10 , a position and speed estimating unit 30 calculates an estimated magnetic pole position from the extended induced voltage. More specifically, an inner product calculator 31 of the unit 30 calculates an inner product of an extended induced voltage vector e (e α ,e β ) and an estimated position vector θ es of an estimated magnetic pole position according to a tenth formula (step S 91 ).

e · θ es =  e  ⁢  θ es  ⁢ cos ⁡ ( π 2 + Δ ⁢ ⁢ θ ) = -  e  ⁢  θ es  ⁢ sin ⁡ ( Δ ⁢ ⁢ θ )

Δθ denotes a difference in angle between vectors θ es and θ c , and θ c denotes a true magnetic pole position placed on the d-axis.

FIG. 11 shows a positional relation between the vectors e and θ es . The voltage vector e is obtained by differentiating a magnetic flux Φ of a rotor 40 of the motor 10 with respect to time. The flux Φ is directed from an S magnetic pole to an N magnetic pole. Therefore, as shown in FIG. 11 , the voltage vector e is advanced by 90 degrees from the direction of the flux Φ. Further, a position (or angle) indicated by the flux is equivalent to a true magnetic pole position of the rotor. Therefore, when the vector θ es correctly indicates the true magnetic pole position of the rotor, an inner product of the vector e and the vector θ es becomes zero. In other words, when the vector θ es is corrected such that the inner product substantially becomes zero, the vector θ es correctly indicates the true magnetic pole position θ c of the rotor.

A speed converter 32 multiplies the inner product by a third PI gain set at a value of Kp 3 +Ki 3 /s to obtain an estimated speed ω es (step S 92 ). That is, the converter 32 does not output a speed calculated according to the seventh formula but outputs a speed depending on a positional difference Δθ between the vectors θ es and θ c . This means that the converter 32 can calculates the speed ω es such that a corrected position derived from the speed ω es approach the true position θ c and such that an inner product calculated from the corrected position substantially becomes zero. The speed ω es is outputted to the calculator 1 shown in FIG. 1 and an integrator 34 . The integrator 34 integrates the speed ω es with respect to time to obtain an intermediate estimated magnetic pole position θ esi (step S 93 ).

Further, a phase delay occurs in the position θ esi due to the integration. To compensate this phase delay, a phase compensator 33 multiplies the inner product by a fourth PI gain set at a value of Kp 4 +Ki 4 /s to obtain a compensating value (step S 94 ). An adder 35 adds this compensating value to the position θ esi to obtain the position vector θ es (step S 95 ). The vector θ es is outputted as a final estimated magnetic pole position to the transformer 5 shown in FIG.

1 and the calculator 31 (step S 96 ).

When a position (or angle) indicated by the vector θ es is shifted from a true magnetic pole position θ c by a position error Δθ, it is required that the converter 32 calculates a speed ω es , which is heightened or lowered based on the position error Δθ, such that an estimated position calculated from the speed ω es approaches the position θ c . Because a value of the inner product corresponds to the error Δθ, the converter 32 can calculates a speed ω es depending on the error Δθ such that an estimated position θ es obtained from the speed ∫ es substantially reaches the true position θ c . Therefore, the pole position θ es estimated in the unit 30 can substantially reach the true position θ c .

As shown in FIG. 11 , for example, when an estimated position vector θ es is delayed from the true position θ c (i.e., Δθ>0), the converter 32 calculates a speed ω es which is higher than a speed obtained by differentiating the position vector θ es with respect to time according to the seventh formula. Therefore, the heightened speed ω es is calculated based on the error Δθ, and the position vector θ es can approach the true position θ c .

Although high frequency components superimposed on the voltage vector e is transferred to the speed ω es , the components are smoothed in the integration of the speed ω es . Therefore, the integrator 34 substantially performs the filtering processing in addition to the integration.

As described above, because an estimated speed ω es obtained by multiplying the inner product by the third PI gain depends on the position error Δθ, the speed ω es can be heightened or lowered based on the position error Δθ. Accordingly, when an estimated position vector θ es is obtained from the speed ω es so as to substantially reduce the inner product to zero, the position vector θ es can be appropriately corrected and approach the true position θ c .

FIG. 12 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to the fifth embodiment, and FIG. 13 is a block diagram of a position and speed estimating unit wherein the method shown in FIG. 12 is performed.

As shown in FIGS. 12 and 13 , in a position and speed estimating unit 50 , a speed estimating unit 51 successively receives extended induced voltage vectors e(e α ,e β ) from the calculator 14 and calculates a first estimated speed ω es1 from the vectors e t1 and e t2 calculated at different times t 1 and t 2 in the calculator 14 (step S 121 ). That is, the speed ω es1 is calculated from a time interval between the times t 1 and t 2 and a positional difference between the vectors e t1 and e t2 .

High frequency components denoting noises are superimposed on each of the vectors e t1 and e t2 , so that the components exist in the speed ω es1 . A low pass filter 52 removes the components from the speed ω es1 (step S 122 ), so that noises existing in the speed ω es1 can be considerably reduced. In this low pass filtering, a phase delay occurs in the speed ω es1 .

Further, in the same manner as the calculator 31 and converter 32 in the fourth embodiment, an inner product calculator 54 calculates an inner product of the voltage vector e(e α ,e β ) and an estimated position vector θ es indicating an estimated magnetic pole position of the rotor 40 (step S 123 ), and a speed converter 55 multiplies the calculated inner product by the third PI gain to obtain a speed correction ω es2 (step S 124 ). The speed correction depends on a position error Δθ between an estimated position of the vector θ es and a true position θ c of the magnetic pole. An adder 53 adds the speed correction ω es2 to the estimated speed ω es1 to correct the speed ω es1 and to compensate the phase delay of the speed ω es1 (step S 125 ). Therefore, a final estimated speed ω es (=ω es1 +ω es2 ) having no phase delay is obtained. The speed ω es is outputted to the calculator 1 shown in FIG. 1 and an integrator 56 . The integrator 56 integrates the speed ∫ es with respect to time to obtain the vector θ es (step S 126 ). The position vector θ es is outputted as a final estimated magnetic pole position to the transformer 5 shown in FIG. 1 and the calculator 54 (step S 127 ).

The speed correction ω es2 is determined based on a position error between an estimated position and a true position θ c of the magnetic pole. Therefore, the speed correction ω es2 can be determined such that an inner product derived from the speed ω es is substantially set to zero, and the position vector θ es corrected by the speed correction can approach the true position θ c , in the same manner as in the fourth embodiment. Accordingly, in the same manner as in the first embodiment, a phase delay in the position θ es can be prevented, and the same effects as those in the first embodiment can be obtained.

FIG. 14 is a flow chart showing the procedure of a method of estimating a magnetic pole position according to a first modification of the fifth embodiment, and FIG. 15 is a block diagram of a position and speed estimating unit wherein the method shown in FIG. 14 is performed.

As shown in FIG. 14 and FIG. 15 , a phase compensator 57 may be used in place of the converter 55 . More specifically, the integrator 56 integrates the speed ω es1 outputted from the filter 52 with respect to time to obtain a first estimated position vector θ es1 (step S 141 ). The compensator 57 multiplies the inner product outputted from the calculator 54 by the fourth PI gain to obtain a phase correction (or position correction) (step S 142 ). To correct the position vector θ es1 and to compensate the phase delay caused in the filter 52 , an adder 58 adds the phase correction to the position vector θ es1 to obtain an estimated position vector θ es denoting a final estimated magnetic pole position (step S 143 ). The vector θ es is outputted to the transformers 5 and 12 (step S 144 ).

Because the phase correction is increased or decreased based on the inner product, the vector θ es can approach a true position θ c of the magnetic pole, in the same manner as in the fourth embodiment. Accordingly, a phase delay in the position θ es can be prevented, and the same effects as those in the first embodiment can be obtained.

FIG. 16 is a block diagram of a position and speed estimating unit wherein a second modification of the fifth embodiment is performed.

As shown in FIG. 16 , both the converter 55 and the compensator 57 may be used. More specifically, the adder 53 adds the speed correction ω es2 to the speed ω es1 to obtain a corrected speed ω es , and the integrator 56 obtains an estimated position θ es1 from the corrected speed ω es . The adder 58 adds the phase correction to the position θ es1 to obtain an estimated position vector θ es denoting a final estimated magnetic pole position and outputs the vector θ es to the transformers 5 and 12 .

These embodiments should not be construed as limiting the present invention to the methods of those embodiments, and the method according to this invention may be combined with that based on the prior art.

For example, a magnetic pole position is estimated in the αβ coordinates system representing the two-phase alternating current coordinates system. However, in placed of the αβ coordinates system, any two-dimensional coordinates system representing the two-phase alternating current coordinates system may be used. This two-dimensional coordinates system may be defined by two axes not orthogonal to each other and an original point placed on the rotation axis at a position different from that in the dq coordinates system.

Further, in the fourth and fifth embodiments, an inner product may be calculated from normalized vectors. More specifically, the position vector θ es is divided by an absolute value of the vector θ es , the voltage vector e is divided by an absolute value of the vector e, and an inner product of the divided vectors is calculated. Therefore, because the inner product depends on only an angular difference (or position error) between the vectors θ es and e, the estimated speed ω es can precisely be calculated.

Moreover, the fourth and fifth embodiments are not limited to the inner product, and a magnetic pole position may be estimated from a second position vector θ ±90es which is advanced or delayed from the position vector θ es by 90 degrees. More specifically, a second position vector θ ±90es different from the position vector θ es by an angle of +90 degrees or −90 degrees around the original point is calculated, and a length of an outer product of the vector θ ±90es and the voltage vector e is calculated according to an eleventh formula.

When an angle (or phase difference) between the vectors e and θ es substantially equals to 90 degrees, an angle between the vectors e and θ ±90es substantially equals to 0 or 180 degrees. Therefore, when the position vector θ es is corrected such that a value of the outer product approaches zero, the position vector θ es can indicate a true magnetic pole position.

Furthermore, in the embodiments, a position of a magnetic pole of a rotor in a synchronous motor is estimated.

However, a position (i.e., rotational angle) of an arbitrary portion of the rotor may be estimated.