DESCRIPTION CONTROL DEVICE FOR ALTERNATING CURRENT ROTARY MACHINE
TECHNICAL FIELD
[0001] The present invention relates to a control device
for AC rotary machine, capable of controlling an AC rotary
machine such as an induction machine or a synchronous machine
without using a position sensor for obtaining rotor position
information.
BACKGROUND ART
[0002] A sensorless control method performed based on induced voltage with use of an adaptive observer or the like is known as a method for controlling an AC rotary machine such as a synchronous machine or an induction machine. The sensorless control method based on induced voltage has a feature that a position sensor or a velocity sensor can be omitted. However, in a low-speed rotation region, induced voltage is small and therefore it is difficult to detect or estimate induced voltage. Thus, there is a problem that the drive characteristics deteriorate in a low-speed rotation region.
On the other hand, in a low-speed region, by performing control based on a result of position detection using saliency of inductance while superimposing voltage or

current having a frequency different from a fundamental frequency of an AC rotary machine, it becomes possible to perform driving in a low-speed region even by a position sensorless control. However, in the case of performing driving in a high rotation region by using saliency of inductance, there is disadvantage in operation efficiency, voltage usage rate, and maximum current because of occurrence of voltage or current other than a fundamental wave.
Considering the above, if an AC rotary machine is driven based on a result of position detection using saliency of inductance and the AC rotary machine is driven by sensorless control using induced voltage in a high rotation region, it becomes possible to provide a device that can be driven in the entire speed region with low cost. In this case, it is important to perform smooth driving over a range from a low-speed region to a high-speed region. [0003] For making it possible to smoothly drive an AC rotary machine over a range from a low-speed region to a high-speed region, the following invention is disclosed, for example.
In order to appropriately generate a phase in a rotational dq coordinate system required in a vector controller basic section without using a position sensor, a low-frequency region phase generator for generating a phase for low-frequency region and a high-frequency region phase

generator for generating a phase for high-frequency region are provided, and further, a phase synthesizer for synthesizing these two kinds of phases by averaging them in a weighted manner in terms of frequency is provided, whereby the synthesized final phase is used as a phase in a rotational dq coordinate system (for example, see Patent Document 1) .
[0004] In addition, the following invention is disclosed. In a sensorless control method for performing continuous control over a range from zero speed to a high-speed region of an AC electric motor, control is performed by using a velocity estimated value coest and a position estimated value 8 0 estimated by a mechanical mathematical expression model by a position and velocity estimation device, so that position error A 6 obtained by calculation of a cross product of a first magnetic flux vector calculated by using a rotor angle and a second magnetic flux vector calculated without using a rotor angle becomes zero (for example, see Patent Document 2) .
CITATION LIST
PATENT DOCUMENT
[0005] Patent Document 1: Japanese Laid-Open Patent
Publication No. 10-94298 {paragraph [0032], FIG. 1)
Patent Document 2: Japanese Laid-Open Patent

Publication No. 2006-158046 (paragraphs [0012] and [0013], FIGS. 2 and 3)
SUMMARY OF THE INVENTION PROBLEMS TO BE SOLVED BY THE INVENTION
[0006] In the invention disclosed in Patent Document 1, since two kinds of phases are synthesized by being averaged in a weighted manner in terms of frequency, it is necessary to obtain a phase for low-speed region and a phase for high¬speed region simultaneously, and therefore computing processing is performed in both the low-frequency region phase generator and the high-frequency region phase generator simultaneously, resulting in a problem of greatly increasing the amount of computing processing. In addition, in a low-speed region in which a frequency is low, it is necessary to perform driving based on, of the two kinds of phases, the phase of the low-frequency region phase generator, and therefore there is also a problem that, if estimation response of the low-frequency region phase generator is delayed, response performance such as velocity control response or torque control response cannot be maintained at a high level in a low-speed region.
[0007] In addition, in the invention disclosed in Patent Document 2, a rotor angle is needed for obtaining the first magnetic flux vector, and it is necessary to obtain the rotor

angle in advance by detection or computing processing, resulting in a problem of increasing the computing amount. In addition, if response of detection or estimation of a rotor angle needed for obtaining the first magnetic flux vector is delayed, response of position error A 8 obtained by calculation of cross product of the first magnetic flux vector and the second magnetic flux vector is also delayed, and response of the velocity estimated value west and the position estimated value 0 est calculated by the position and velocity estimation device is also delayed. As a result, a problem arises that response performance such as velocity control response or torque control response cannot be maintained at a high level.
[0008] The present invention has been made to solve the above problems, and an object of the present invention is to provide a control device for AC rotary machine that can smoothly drive an AC rotary machine without using a position sensor over the entire speed region and can realize reduction in the computing amount and improvement in response performance.
SOLUTION TO THE PROBLEMS
[0009] A control device for AC rotary machine according to the present invention includes: a current vector detection section for detecting a detected current vector from current

of an AC rotary machine; a magnetic flux vector detection section for detecting a rotor magnetic flux vector from the detected current vector of the AC rotary machine and outputting the rotor magnetic flux vector as a detected magnetic flux vector; an adaptive observation section for outputting an estimated current vector, an estimated magnetic flux vector, and an estimated position of the AC rotary machine; a control section for outputting a voltage instruction vector so that the detected current vector coincides with a current instruction vector; a voltage application section for applying voltage to the AC rotary machine, based on the voltage instruction vector; a deviation vector calculation section for outputting a current deviation vector which is a deviation between the estimated current vector and the detected current vector, and a magnetic flux deviation vector which is a deviation between the estimated magnetic flux vector and the detected magnetic flux vector; and a deviation amplification section for amplifying the current deviation vector and the magnetic flux deviation vector and outputting the amplified vectors as an amplified deviation vector. The adaptive observation section calculates and outputs the estimated current vector, the estimated magnetic flux vector, and the estimated position, based on the voltage instruction vector and the amplified deviation vector. Further, the control section outputs a

voltage instruction vector on which a high-frequency voltage vector having a frequency component different from a frequency for driving the AC rotary machine is superimposed, and the magnetic flux vector detection section calculates and outputs the detected magnetic flux vector, based on a magnitude of a high-frequency current vector having the same frequency component as the high-frequency voltage vector, included in the detected current vector, and on a magnitude of a rotor magnetic flux.
EFFECT OF THE INVENTION
[0010] Owing to the above-described configuration, the control device for AC rotary machine according to the present invention makes it possible to provide a control device for AC rotary machine that can smoothly drive an AC rotary machine without using a position sensor over the entire speed region and can realize reduction in the computing amount and improvement in response performance.
BRIEF DESCRIPTION OF THE DRAWINGS
[0011] [FIG. 1] FIG. 1 is a system configuration diagram according to a control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 2] FIG. 2 is a configuration diagram of a control section according to the control device for AC rotary

machine of embodiment 1 of the present invention.
[FIG. 3] FIG. 3 is a configuration diagram of a deviation amplification section according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 4] FIG. 4 is a configuration diagram of an adaptive observation section according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 5] FIG. 5 is an internal configuration diagram of the adaptive observation section according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 6] FIG, 6 is a configuration diagram of a magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 7] FIG. 7 is an internal configuration diagram of the magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 8] FIG. 8 is an internal configuration diagram of the magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 1 of the present invention.

[FIG. 9] FIG. 9 is an internal configuration diagram of the magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 10] FIG. ID is an explanation diagram of a rotor magnetic flux vector according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 11] FIG. 11 is a system configuration diagram of a modification according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 12] FIG. 12 is a configuration diagram of a deviation amplification section of the modification according to the control device for AC rotary machine of embodiment 1 of the present invention.
[FIG. 13] FIG. 13 is a configuration diagram of a magnetic flux vector detection section according to a control device for AC rotary machine of embodiment 2 of the present invention.
[FIG. 14] FIG. 14 is an internal configuration diagram of the magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 2 of the present invention.
[FIG. 15] FIG. 15 is an internal configuration diagram of the magnetic flux vector detection section

according to the control device for AC rotary machine of embodiment 2 of the present invention.
[FIG. 16] FIG, 16 is a system configuration diagram according to a control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 17] FIG. 17 is a configuration diagram of a control section according to the control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 18] FIG. 18 is an internal configuration diagram of the control section according to the control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 19] FIG. 19 is a configuration diagram of an adaptive observation section according to the control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 20] FIG. 20 is an internal configuration diagram of the adaptive observation section according to the control device for AC rotary machine of embodiment 3 of the present invention.
;nt 3 of the
iguration
[FIG. 21] FIG. 21 is a configuration diagram of a magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 22] FIG. 22 is an internal configuration

diagram of the magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 23] FIG. 23 is. an internal configuration diagram of the magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 24] FIG. 24 is an internal configuration diagram of the magnetic flux vector detection section according to the control device for AC rotary machine of embodiment 3 of the present invention.
[FIG. 25] FIG. 25 is an explanation diagram of a rotor magnetic flux vector according to the control device for AC rotary machine of embodiment 3 of the present invention.
DESCRIPTION OF EMBODIMENTS [0012] Embodiment 1
Embodiment 1 relates to a control device for AC rotary machine that calculates a magnetic flux deviation vector which is a deviation between a detected magnetic flux vector and an estimated magnetic flux vector, and a current deviation vector which is a deviation between a detected current vector and an estimated current vector, and drives an AC rotary machine by using an estimated velocity and an

estimated position outputted from an adaptive observation section based on an amplified deviation vector obtained by amplification of the calculated deviation vectors. Here, in calculation of a detected magnetic flux vector outputted from a magnetic flux vector detection section, a high-frequency current vector having the same frequency component as a high-frequency voltage vector is extracted from a detected current vector, and then, by using the magnitude of a component orthogonal to the high-frequency voltage vector, a difference A 8 between d axis on which the high-frequency voltage vector is applied and dm axis which is parallel to a rotor magnetic flux vector is calculated, and only A 6 and a magnitude 0 f of the rotor magnetic flux vector are used. [0013] Hereinafter, embodiment 1 of the present invention will be described based on FIG. 1 showing a system configuration diagram according to a control device 1 for AC rotary machine, FIG. 2 showing a configuration diagram of a control section, FIG. 3 showing a configuration diagram of a deviation amplification section, FIG. 4 showing a configuration diagram of an adaptive observation section, FIG. 5 showing an internal configuration diagram of the adaptive observation section, FIG. 6 showing a configuration diagram of a magnetic flux vector detection section, FIGS. 7 to 9 showing internal configuration diagrams of the magnetic flux vector detection section, FIG. 10 showing an explanation

diagram of a rotor magnetic flux vector, FIG. 11 showing a system configuration diagram of a modification, and FIG. 12 showing a configuration diagram of a deviation amplification section of the modification.
[0014] First, the entire system configuration including the control device 1 for AC rotary machine according to embodiment 1 of the present invention will be described.
In FIG. 1, the control device 1 for AC rotary machine is composed of a current vector detection section 3, a control section 4, a voltage application section 5, a deviation vector calculation section 6, a deviation amplification section 7, an adaptive observation section 8, and a magnetic flux vector detection section 9, in order to control an AC rotary machine 2.
It is noted that in FIG. 1, two diagonal lines (//), three diagonal lines (///), and four diagonal lines (////) indicate 2-dimensional, 3-dimensional, and 4-dimensional vectors, respectively.
[0015] The details of the configuration, function, and operation of the control device 1 for AC rotary machine will be described later. First, the summary of the function of each constituent part of the control device 1 for AC rotary machine will be described.
The current vector detection section 3 detects a detected current vector from current of the AC rotary machine

2. The magnetic flux vector detection section 9 detects a rotor magnetic flux vector from the detected current vector of the AC rotary machine 2 outputted from the current vector detection section 3, and outputs the rotor magnetic flux vector as a detected magnetic flux vector. The adaptive observation section 8 outputs an estimated current vector, an estimated magnetic flux vector, and an estimated position of the AC rotary machine 2 from the detected magnetic flux vector. The control section 4 outputs a voltage instruction vector so that the detected current vector coincides with a current instruction vector, and the voltage application section 5 applies voltage based on the voltage instruction vector, to the AC rotary machine. The deviation vector calculation section 6 outputs a current deviation vector which is a deviation between the estimated current vector and the detected current vector, and a magnetic flux deviation vector which is a deviation between the estimated magnetic flux vector and the detected magnetic flux vector. The deviation amplification section 7 amplifies the current deviation vector and the magnetic flux deviation vector and outputs them as an amplified deviation vector. [0016] In FIG. 1, in the present embodiment 1, the case where the AC rotary machine 2 is a permanent magnet synchronous machine of saliency type having three-phase windings will be described as an example. However, even in

the case of using a rotary machine of another type, the control device for AC rotary machine can be configured on the same principle.
[0017] Hereinafter, the configuration, function, and operation of each constituent part of the control device 1 for AC rotary machine will be described in order.
The current vector detection section 3 detects three-phase current flowing in the AC rotary machine 2, and outputs, as the detected current vector, d-axis current id and q-axis current iq which are respectively obtained by a coordinate converter 31 performing, by using an estimated position $ 0 described later, coordinate conversion of the three-phase current into a d-axis direction and a q-axis direction of d-q axes which are known as an orthogonal coordinate system rotating in synchronization with a rotor of the AC rotary machine 2.
[0018] Next, based on FIG. 2, the configuration, function, and operation of the control section 4 will be described.
The control section 4 is composed of an adder/subtractor 41, a current controller 42, a high-frequency voltage vector generator 43, an adder/subtractor 44, and a coordinate converter 45.
In the control section 4, the adder/subtractor 41 subtracts a detected current vector (ids, iqs) from a current instruction vector (id_ref, iq_ref) given from the outside.

The current controller 42 performs proportional integral control so that the deviation between the current instruction vector and the detected current vector, outputted from the adder/subtractor 41, becomes zero, and outputs a fundamental wave voltage vector (vdf, vqf). The high-frequency voltage vector generator 43 outputs a high-frequency voltage vector (vdh, vqh) on d-q axes.
It is noted that in the present embodiment 1, the high-frequency voltage vector is set as vqh = 0, and thus is alternating voltage to be applied only in the d-axis direction.
The adder/subtractor 44 acids the fundamental wave voltage vector (vdf, vqf) and the high-frequency voltage vector (vdh, vqh), and thereby outputs a voltage instruction vector (vd, vq) on d-q axes. Then, by using an estimated position & 0 described later, the coordinate converter 45 converts the voltage instruction vector (vd, vq) on d-q axes outputted from the adder/subtractor 44 into a voltage instruction vector (vu, vv, vw) in a coordinate system at rest from d-q axes, and outputs the voltage instruction vector (vu, w, vw) .
[0019] The voltage application section 5 applies three-phase voltage based on the voltage instruction vector (vu, vv, vw) outputted from the control section 4, to the AC rotary machine 2.

[0020] Next, the configuration, function, and operation of the deviation vector calculation section 6 will be described.
The deviation vector calculation section 6 is composed of an adder/subtractor 61 and an adder/subtractor 62.
In the deviation vector calculation section 6, the adder/subtractor 61 subtracts a detected current vector {ids, iqs) outputted from the current vector detection section 3, from an estimated current vector (idsO, iqsO) outputted from the adaptive observation section 8 described later, and thereby outputs a current deviation vector (eids, eiqs). In addition, the adder/subtractor 62 subtracts a detected
magnetic flux vector (<£drD, (/> qrD) outputted from the magnetic flux vector detection section described later, from
an estimated magnetic flux vector (<£drO, 0qrO) outputted
from the adaptive observation section 8 described later, and
thereby outputs a magnetic flux deviation vector (e^dr, etf>
qr).
[0021] Next, based on FIG. 3, the configuration, function,
and operation of the deviation amplification section 7 will
be described.
The deviation amplification section 7 is composed of gain matrix calculators 71 and 72 and an adder/subtractor 73.
The gain matrix calculator 71 multiplies (eids, eiqs)T (T means a transposed matrix) which is a transposed

matrix of the current deviation vector (eids, eiqs), by a matrix He, and outputs the result. The gain matrix calculator 72 multiplies the magnetic flux deviation vector (e^dr, e0qr)T by a matrix Hf and outputs the result.
Here, the matrixes He and Hf are gain matrixes defined-by expression (1). In expression (1), hll to h44 are amplification gains, and the values of hll to h44 can be freely set. [0022]
[0023] In FIG. 3, the adder/subtractor 73 adds the vector outputted from the gain matrix calculator 71 and the vector outputted from the gain matrix calculator 72, and thereby outputs an amplified deviation vector (el, e2, e3, e4)T.
It is noted that when the rotation speed of the AC rotary machine 2 is high, the adaptive observation section 8 described later can satisfactorily estimate the estimated velocity and the estimated position to be outputted, without using the magnetic flux deviation vector which is a deviation between the detected magnetic flux vector and the estimated magnetic flux vector. Therefore, in the case where the

absolute value of an estimated velocity is great, the values of hl3 to h44 are set at zero in the gain matrix calculator 72 so that output of the gain matrix calculator 72 in a high rotation region becomes zero. As a result, by stopping computing in the magnetic flux vector detection section 9, the computing amount can be reduced. In addition, by also setting at zero the high-frequency voltage vector (vdh, vqh) outputted from the high-frequency voltage vector generator 43 in the control section 4, occurrence of high-frequency current which would be caused by the high-frequency voltage vector (vdh, vqh) is prevented, whereby loss due to high-frequency current can be eliminated.
[0024] Next, the configuration, function, and operation of the adaptive observation section 8 will be described. FIG. 4 is a configuration diagram of the adaptive observation section 8, and FIG. 5 is a configuration diagram of a status observer 82 which is a main constituent element of the adaptive observation section 8.
In FIG. 4, the adaptive observation section 8 is composed of a coordinate converter 81, the status observer 82, and an integrator 83.
In FIG. 5, the status observer 82 is composed of gain matrix calculators 821, 823, 825, and 826, an adder/subtractor 822, an integrator 824, and a velocity estimator 827.

[0025] Next, the entire operation of the adaptive observation section 8 will be described.
With regard to the AC rotary machine 2, armature resistance is set at R, armature inductance in the d-axis direction is set at Ld, armature inductance in the q-axis direction is set at Lq, estimated velocity is set at corO, power-supply angular frequency is set at ds0 and #qs0, and a d-axis component and a q-axis component of a voltage instruction vector on d-q axes are respectively defined as vds and vqs. In this case, the estimated armature reaction vector {$ds0, tf>qs0) and the estimated magnetic flux vector ($drO, <£qr0) can be obtained as shown by expression (3). [0028]

[0029] In addition, where Laplace operator (differential operator) is defined as s, kp is defined as proportional gain,
and ki is defined as integral gain, the estimated velocity to rO which is an internal parameter of the matrix A in expression (2) is given by expression (4), using the current deviation vector (eids, eiqs) and the estimated magnetic flux vector ( $ drO, 4> qrO) . [0030]
[0031] The estimated position 0 0 can be obtained by integrating the estimated velocity o>r0 as shown by expression (5}. [0032]
[0033] In addition, the estimated current vector (idsO,

iqsO) can be calculated by expression (6).
[0034]
[0035] Similarly, the estimated magnetic flux vector (<£
drO, tf>qr0) can be calculated by expression (7).
[0036]
[0037] As described above, by using expressions (2) to (7), the estimated position 0 0, the estimated current vector
(idsO, iqsO), and the estimated magnetic flux vector {tf>dr0,
dsO, <£qsO, ^drO, $qrO)T. The above corresponds to the right-hand side of expression (3) . It is noted that the left-hand side of expression (3) corresponds to input to the integrator 824.
The gain matrix calculator 825 multiplies the vector (0dsO, <£qsO, <£drO, $qrO)T by the matrix Cl and thereby outputs the estimated current vector ($dsO, $qsO)T. This corresponds to expression (6).
Here, the vector (4>ds0, <£qsO, tfidrO, 0qrO)T is a stator/rotor estimated magnetic flux vector.
The gain matrix calculator 826 multiplies the vector (#dsO, 0qsO, <£drO, 0qrO)T by the matrix C2 and thereby outputs the estimated magnetic flux vector (<£>drO, $ qrO)T. This corresponds to expression (7).
The velocity estimator 827 calculates the estimated velocity corO by expression (4), using the current deviation vector (eids, eiqs) and the estimated magnetic flux vector
(0drO, 0qrO) .
The gain matrix calculator 823 receives the
estimated velocity co rO outputted from the velocity estimator 827, applies the matrix A to the vector {0dsO, 4>qs0, <£drO, $qrO)T outputted from the integrator 824, and outputs the result to the adder/subtractor 822.
[0040] The above is the description of the functions and operations of the entirety and each main constituent element

of the adaptive observation section 8. The feature of this adaptive observation section 8 is that the amplified deviation vector (el, e2, e3, e4)T includes e3 and e4 obtained by amplifying the magnetic flux deviation vector {e #dr, e$qr), whereby the estimated velocity corO and the estimated position 0 0 can be satisfactorily calculated in a low-speed region including zero speed.
[0042] Since the detected magnetic flux vector outputted from the magnetic flux vector detection section 9 described later is not influenced by constant error or voltage error in a region of zero speed to low speed, the adaptive observation section 8 of the present embodiment can satisfactorily estimate the velocity and the position even in a region of zero speed to low speed by calculating the magnetic flux deviation vector which is a deviation between the detected magnetic flux vector and the estimated magnetic flux vector and using e3 and e4 obtained by amplifying the magnetic flux deviation vector.
[0042] Next, the configuration, function, and operation of the magnetic flux vector detection section 9 will be described based on FIGS. 6 to 10.
First, the configuration of the magnetic flux vector detection section 9 will be described.
In FIG. 6 showing a configuration diagram of the magnetic flux vector detection section 9, the magnetic flux

vector detection section 9 is composed of a filter 91, an orthogonal component extractor 92, and a detected magnetic flux vector calculator 93.
As shown in FIG. 7, the filter 91 is composed of a notch filter 911 and an adder/subtractor 912. As shown in FIG. 6, the orthogonal component extractor 92 is composed of an orthogonal component selector 921 and an amplitude calculator 922. As shown in FIG. 9, the orthogonal, component extractor 92 is composed of a difference calculator 931, a cosine calculator 932, a sine calculator 933, and multipliers 934 and 935.
[0043] Next, the function and operation of the magnetic
flux vector detection section 9 will be described. First,
the entire operation of the magnetic flux vector detection
section 9 will be described, and then the function and
operation of each constituent element will be described.
[0044] A mathematical expression of a high-frequency
current vector flowing in the AC rotary machine 2 when the high-frequency voltage vector generator 43 in FIG. 2 outputs a high-frequency voltage vector (vdh, vqh) will be described.
As shown in FIG. 10, a coordinate system rotating in synchronization with the rotor of the AC rotary machine 2, formed by the control section 4, is set as orthogonal coordinate d-q axes. The same direction as a rotor magnetic flux vector Or of the AC rotary machine 2 is set as dm axis,

and a direction orthogonal thereto is set as qm axis. The difference between d axis of the orthogonal coordinate system and dm axis of the rotor magnetic flux vector is set as A 0 . It is noted that d axis is the direction of the estimated
position 6 0 outputted from the adaptive observation section 8. In stationary state, operation is performed such that d axis coincides with dm axis. FIG. 10 shows the case where a deviation of A 6 occurs momentarily.
In this case, a mathematical expression of the AC rotary machine 2 when the high-frequency voltage vector (vdh, vgh) is applied on d axis and q axis, respectively, can be represented as expression (8). Here, p in the expression is a differential operator. [0045]

[0046] As described above, the high-frequency voltage
vector is applied only in a region of zero speed to low speed. Therefore, if the rotational velocity is set at cor ^ 0, expression (9) can be obtained. [0047] [Mathematical 9]
[0048] Further, the second term on the right-hand side is a differential of high-frequency current, and the differential of high-frequency current is multiplied by an angular frequency coh of the high-frequency voltage.

Therefore, the second term on the right-hand side » the first term on the right-hand side is satisfied, so the first term on the right-hand side can be ignored. As a result, expression (10) can be obtained. [0049]
[0050] Here, if the high-frequency voltage vector is given as shown by expression (11), the high-frequency current vector (idh, iqh) is represented as expression (12) by substituting expression (11) into expression (10) and integrating both sides. [0051]

described- As shown in FIG. 10, the rotor magnetic flux vector $r is directed in the same direction as dm axis. Here, the rotor magnetic flux vector 3>r is projected to a direction parallel to the high-frequency voltage vector, to be <£drD in the d-axis direction, and a direction orthogonal to the high-frequency voltage vector, to be qrD in the q axis direction. Then, the 0drD and 0 qrD projected on d axis and q axis are set as the detected magnetic flux vector, and mathematical expressions thereof can be represented as expression (13). [0054]
[0055] In expression (13), 0f is the magnitude of the rotor magnetic flux vector <5r. In the case of permanent magnet synchronous machine, although f slightly varies depending on the temperature, $ f can be known by measuring 4> f in advance. In the case of induction machine or winding-field-type synchronous machine, $ f can be calculated from the magnitude of excited current or field current. Thus, £ is a known value.
Since f is a known value, it is only necessary to

know A 8, for calculating the detected magnetic flux vector from expression (13).
[0056] Here, a method for calculating A 6 will be discussed. Since A 0 is the difference between d axis and dm axis, A 0 can be calculated by obtaining the position of d axis and the position of dm axis. The position of d axis is
the estimated position 0 0 outputted from the adaptive observation section 8, and therefore is known. However, the position {hereinafter, referred to as 0 dm) of dm axis is not known, and therefore needs to be calculated by another method.
For example, as disclosed in International Publication No. WO2009-040965, an estimated position calculated by superimposing high-frequency voltage that allows suitable estimation of a position of the dm-axis direction even in a region of zero speed to low speed is used as 0 dm, whereby A0 can be calculated. However, in the case of calculating A 0 by the disclosed method, it is necessary to newly provide a position estimation section, for calculating 0dm, and therefore the computing amount greatly increases.
[0057] In the present embodiment 1, looking at expression (12), A 0 is included in the magnitude of the high-frequency current vector. Therefore, instead of performing position estimation, although a high-frequency voltage vector is superimposed, the magnitude of the high-frequency current

vector is extracted, whereby the detected magnetic flux vector is calculated.
In the present embodiment 1, the magnitude of the high-frequency current vector iqh on q axis which is orthogonal to the high-frequency voltage vector, is used.
From expression (12), a magnitude |iqh| of the high-frequency current vector iqh is represented by expression (14), and therefore A0 can be calculated as shown by expression (15). [0058]
[0060] It is noted that the magnitude |iqh| of the high-frequency current vector can be calculated by expression (16). Since the angular frequency coh and a high-frequency voltage amplitude Vh of the high-frequency voltage can be freely set in the high-frequency voltage vector generator 43, they are known values. L and 1 can be calculated from Ld and Lq as

shown by expression (8) . In addition, since Ld and Lq can be known by measuring them in advance, L and 1 are also known values. It is noted that T in expression (16) is the period of the high-frequency current vector (idh, iqh). [0061]
[Mathematical 16]
|iqh|=^(iqh2dt ." (16)
[0062] As described above, the detected magnetic flux vectors tf»drD and tfiqrD are calculated by expressions (13),
(15), and (16).
It is noted that since the estimated position 6 0 outputted from the adaptive observation section 8 is equal to or close to a direction of the rotor magnetic flux vector <5r, 2A0 ^ 0 is satisfied, and sin2A0 can be approximated as sin2 A B ^ 2A 8 . Therefore, A 8 may be calculated as shown by expression (17) . In this case, the detected magnetic flux
vector ( tf> drD, h of the high-frequency voltage vector in expression (18) is applied to a detected
current vector, whereby an angular frequency o>h component is eliminated from the detected current vector.
The adder/subtractor 912 subtracts output of the notch filter 911 from the detected current vector, thereby calculating a high-frequency current vector corresponding to the angular frequency coh component from the detected current vector. It is noted that s in expression (18) is Laplace operator and qx is the depth of notch. [0066]

[0067] In FIG. 8, in the orthogonal component extractor 92, the orthogonal component selector 921 multiplies the high-frequency current vector (idh, iqh) by a matrix (0, 1)T, thereby selecting only iqh which is a component orthogonal to d axis, of the high-frequency current vector.
The amplitude calculator 922 performs calculation by expression (16), to calculate IiqhI which is the magnitude (amplitude) of iqh, and outputs |iqh|.
[0068] In the detected magnetic flux vector calculator 93 in FIG. 9, the difference calculator 931 performs calculation by expression (15) or (17) using IiqhI outputted from the orthogonal component extractor, and thereby outputs A 9 .
The cosine calculator 932 performs cosine calculation using A 0 outputted from the difference calculator 931, and thereby outputs cos A 8 . The multiplier 934 multiplies cosA0 outputted from the cosine calculator 932 by the magnitude 0 f of the rotor magnetic flux, and thereby outputs 4>drD which is a detected magnetic flux vector component parallel to the high-frequency voltage vector.
The sine calculator 933 performs sine calculation

using A 8 outputted from the difference calculator 931, and thereby outputs sinA 0 . The multiplier 935 multiplies sinA 8 outputted from the sine calculator 933 by the magnitude $f of the rotor magnetic flux, and thereby outputs <£qrD which is. a detected magnetic flux vector component orthogonal to the high-frequency voltage vector.
The above is the description of the configuration, function, and operation of the magnetic flux vector detection section 19.
[0069] Next, a modification of the present embodiment 1 will be described.
In the deviation amplification section 7 in FIG. 3, the matrixes He and Hf are gain matrixes defined by expression (1). In expression (1), hll to h44 are amplification gains, and the values of hll to h44 can be freely set.
Here, for example, as described in FIG. 9 of Japanese Patent No. 4672236, for hll to h42 of the matrix He, the value of each amplification gain can be changed depending
on the estimated velocity corO. Also, for hl3 to h44 of the
gain matrix Hf, the value of each amplification gain can be
changed depending on the estimated velocity corO.
In this case, as shown by a system configuration diagram of the modification in FIG. 11, the adaptive observation section 8 also outputs the estimated velocity co

rO. In addition, as shown in FIG. 12, in a deviation amplification section 70, the estimated velocity o>r0 is inputted to gain matrix calculators 701 and 702.
In FIG. 12, the adder/subtractor 73 adds vectors
5 outputted from the gain matrix calculator 701 and the gain matrix calculator 702, and thereby outputs the amplified deviation vector (el, e2, e3, e4)T.
In addition, as described above, in the case where the absolute value of the estimated velocity corO is great,
) the values of hl3 to h44 are set at zero in the gain matrix calculator 7 02 so that output of the gain matrix calculator 702 in a high rotational speed region becomes zero. [0070] As described above, the control device 1 for AC rotary machine according to embodiment 1 calculates the
i magnetic flux deviation vector which is a deviation between the detected magnetic flux vector and the estimated magnetic flux vector, and the current deviation vector which is a deviation between the detected current vector and the estimated current vector, and uses the amplified deviation
l vector obtained by amplifying the magnetic flux deviation vector, thereby making it possible to drive an AC rotary machine by using only one kind of estimated velocity or estimated position outputted from the adaptive observation section over a range from zero speed to high-speed region.
i Therefore, it is not necessary to cause both the low-

frequency region phase generator and the high-frequency
region phase generator to perform computing processing
simultaneously, so the amount of computing processing can be
greatly reduced,
i Further, in calculation of the detected magnetic
flux vector (tf>drD, 4>qrD) outputted from the magnetic fiux vector detection section, a high-frequency current vector having the same frequency component as a high-frequency voltage vector is extracted from the detected current vector. Of the extracted high-frequency current vector, by using the magnitude |iqh| of iqh which is a component orthogonal to the high-frequency voltage vector, the difference A 8 , which is included in liqhf, between d axis on which the high-frequency voltage vector is applied and dm axis which is parallel to the rotor magnetic flux vector is calculated, and then only A0 and the magnitude f of the rotor magnetic flux vector are used. Therefore, it is not necessary to prepare the rotor angle in advance, for calculating the detected magnetic flux vector, and therefore an effect of reducing the computing amount is provided. In addition, since response of estimation of the rotor angle is not relevant to the calculation of the detected magnetic flux vector at all, an effect of maintaining response performance such as velocity control response or torque control response at a high level is provided.

[0071] Embodiment 2
In a control device for AC rotary machine according to embodiment 2, in calculation of the detected magnetic flux vector ($drD, qrD), the magnetic flux vector detection section 9 extracts, from a detected current vector, a high-frequency current vector having the same frequency component as the high-frequency voltage vector.. Of the extracted high-frequency current vector, by using the magnitude |idh| of idh which is a component parallel to the high-frequency voltage vector, the difference A0, which is included in lidh], between d axis on which the high-frequency voltage vector is applied and dm axis which is parallel to the rotor magnetic flux vector is calculated, and then only A 6 and the magnitude 0 f of the rotor magnetic flux vector are used. [0072] In the present embodiment 2, the configuration is the same as in embodiment 1 except for the magnetic flux vector detection section, so the description of a part other than the magnetic flux vector detection section is omitted.
Hereinafter, embodiment 2 of the present invention will be described based on FIGS. 13 to 15 showing a configuration diagram and an internal configuration diagram of the magnetic flux vector detection section 9 according to-the control device for AC rotary machine.
In FIGS. 13 to 15, components that are the same as or correspond to those in FIGS. 6 to 9 of embodiment 1 are

denoted by the same reference characters.
[0073] FIG. 13 is a configuration diagram of the magnetic flux vector detection section 9 of the present embodiment. In FIG. 13, parts different from embodiment 1 are only a parallel component extractor 94 and a detected magnetic flux vector calculator 95. The filter 91 is the same as in embodiment 1, so the description thereof is omitted.
FIG. 14 is a configuration diagram of the parallel component extractor 94. A parallel component selector 941 multiplies the high-frequency current vector (idh, iqh) by a matrix {1, 0)T, thereby selecting only idh which is a high-frequency current vector parallel to d axis. An amplitude calculator 942 calculates the magnitude |idh| of idh by expression (19), and outputs the magnitude |idh|. [0074]
[0075] FIG. 15 is a configuration diagram of the detected magnetic flux vector calculator 95. From expression (12), IidhI outputted from the parallel component extractor 94 is represented by expression (20). A difference calculator 951 performs calculation by expressions (20.) and (21) , and
thereby outputs A 6 . [0076]

I [0078] The cosine calculator 932 performs cosine calculation using A Q outputted from the difference calculator 951, and thereby outputs cosAfl. The multiplier 934 multiplies cos A 9 outputted from the cosine calculator 932 by the magnitude <£f of the rotor magnetic flux, and thereby outputs the resultant value as $drD which is a detected magnetic flux vector parallel to the high-frequency voltage vector.
The sine calculator 933 performs sine calculation using A B outputted from the difference calculator 951, and thereby outputs sinAO. The multiplier 935 multiplies sinA 6 outputted from the sine calculator 933 by the magnitude f of the rotor magnetic flux, and thereby outputs the resultant value as <£qrD which is a detected magnetic flux vector orthogonal to the high-frequency voltage vector.

[0079] As described above, in the control device for AC rotary machine according to einbodiment 2, in calculation of
the detected magnetic flux vector (#drD, qrD) outputted from the magnetic flux vector detection section 9, a high-frequency current vector having the same frequency component as a high-frequency voltage vector is extracted from the detected current vector. Of the extracted high-frequency current vector, by using the magnitude |idh| of idh which is a component parallel to the high-frequency voltage vector, the difference A 9, which is included in IidhI, between d axis on which the high-frequency voltage vector is applied and dm axis which is parallel to the rotor magnetic flux vector is calculated, and then only A 0 and the magnitude f of the rotor magnetic flux vector are used. Therefore, as in embodiment 1, an effect of reducing the computing amount and an effect of maintaining response performance such as velocity control response or torque control response at a high level are provided. [0080] Embodiment 3
In embodiment 1, the detected current vector outputted from the current vector detection section is a detected current vector on d-q axes which are an orthogonal coordinate system rotating in synchronization with the rotor of the AC rotary machine, and the control section, the deviation vector calculation section, the deviation

amplification section, the adaptive observation section, and the magnetic flux vector detection section are also configured on d-g axes.
On the other hand, in the present embodiment 3, the detected current vector outputted from the current vector detection section is a detected current vector (ia, ij3) on a-& axes which are known as an orthogonal coordinate system . at rest, and the control section, the deviation vector calculation section, the deviation amplification section, the adaptive observation section, and the magnetic flux vector detection section are also configured on a-$ axes. [0081] Hereinafter, embodiment 3 of the present invention will be described based on FIG. 16 showing a system configuration diagram of a control device 10 for AC rotary machine, FIG. 17 showing a configuration diagram of a control section, FIG. 18 showing an internal configuration diagram of the control section, FIG. 19 showing a configuration diagram of an adaptive observation section, FIG. 20 showing an internal configuration diagram of the adaptive observation section, FIG. 21 showing a configuration diagram of a magnetic flux vector detection section, FIGS. 22 to 24 showing internal configuration diagrams of the magnetic flux vector detection section, and FIG. 25 showing an explanation diagram of a rotor magnetic flux vector. [0082] First, the entire system configuration including

the control device 10 for AC rotary machine according to embodiment 3 of the present invention will be described.
In FIG. 16, the control device 10 for AC rotary machine is composed of a current vector detection section 13, a control section 14, the voltage application section 5, the deviation vector calculation section 6, the deviation amplification section 7, an adaptive observation section 18, and a magnetic flux vector detection section 19, in order to control the AC rotary machine 2.
In FIG. 16, components that are the same as or correspond to those in FIG. 1 are denoted by the same reference characters.
[0083] The control device 10 for AC rotary machine of embodiment 3 is different from the control device 1 for AC rotary machine of embodiment 1 only in the configurations of the current vector detection section 13, the control section 14, the adaptive observation section 18, and the magnetic flux vector detection section 19. Therefore, hereinafter, the configuration, function, and operation of each constituent part of the control device 10 for AC rotary machine will be described in order, focusing on the difference.
[0084] The current vector detection section 13 detects three-phase current flowing in the AC rotary machine 2, and outputs, as a detected current vector (ia, i/3), a-axis

current ia and #-axis current i£ which are respectively obtained by a coordinate converter 131 performing coordinate conversion of the three-phase current into an a-axis direction and a #-axis direction on a-j9 axes which are an
> orthogonal coordinate system at rest of the rotor of the AC rotary machine 2.
[0085] Next, based on FIG. 17, .the configuration, function, and operation of the control section 14 will be described.
The control section 14 is composed of a coordinate
i converter 141, an adder/subtractor 142, a current controller 143, a coordinate converter 144, a high-frequency voltage vector generator 145, and an adder/subtractor 146.
It is noted that computing by the control section 14 is performed on d-q axes. The coordinate converter 141 converts the detected current vector (ia, ij3) into a vector on d-q axes by using an estimated position 8 0 outputted from the adaptive observation section 18 described later. The adder/subtractor 142 subtracts a detected current vector (ids, iqs) outputted from the coordinate converter 141, from a current instruction vector (id_ref, iq_ref) given from the outside. The current controller 143 performs proportional integral control so that the deviation between the current instruction vector and the detected current vector, outputted ■ from the adder/subtractor 142, becomes zero, and outputs a fundamental wave voltage vector (vdf, vqf). The coordinate

converter 144 converts the fundamental wave voltage vector (vdf, vqf) into a voltage instruction vector (vuf, vvf, vwf) in a coordinate system at rest from d-q axes, by using the
estimated position 0 0, and outputs the voltage instruction
J vector (vuf, wf, vwf) .
In the high-frequency voltage vector generator 145, as shown in the configuration diagram of the high-frequency voltage vector generator 145 in FIG. 18, a high-frequency voltage instruction vector (van, v/Jh) represented by
I expression (22) on a-j3 axes which are an orthogonal
coordinate system at rest is given, and then converted into a high-frequency voltage vector instruction {vuh, vvh, vwh) in a coordinate system at rest by a coordinate converter 401, and then the high-frequency voltage vector instruction (vuh,
> wh, vwh) is outputted.
In the present embodiment 3, since the high-frequency voltage vector instruction is thus given, a rotational-voltage high-frequency voltage vector instruction can be given unlike alternating voltage of embodiment 1. [0086]
[0087] The adder/subtractor 146 adds the voltage

instruction vector (vuf, vvf, vwf) outputted from the coordinate converter 144 and the high-frequency voltage vector instruction (vuh, wh, vwh}, and thereby outputs the voltage instruction vector (vu, vv, vw).
[0088] Next, the deviation vector calculation section 6 will be described.
The deviation vector calculation section 6, as in embodiment 1, although its input and output are changed from d-q axes to a-j3 axes, outputs a current deviation vector (ei a, ei |3 ) which is a difference between an estimated current vector (iaO, i/3 0) and the detected current vector (ia, i j3 ) , and a magnetic flux deviation vector (e<£ar, e$#r) which is a difference between an estimated magnetic flux vector (tf> a rO, 0 j8r0) and a detected magnetic flux vector (0 arD, ]3 rD).
[0089] Next, the deviation amplification section 7 will be described.
Also the deviation amplification section 7, as in embodiment 1, although its input and output are changed from
d-q axes to a-jS axes, outputs an amplified deviation vector (el, e2, e3, e4) obtained by adding the current deviation
vector (eia, ei 0 ) multiplied by the matrix He of expression (1) and the magnetic flux deviation vector (e<£ar, e$j3r)
multiplied by the matrix Hf of expression (1). [0090] Next, the adaptive observation section 18 will be

described. Also for the adaptive observation section 18, the basic configuration is the same as that of the adaptive observation section 8 of embodiment 1, but the matrix A and the matrix CI therein are different.
FIG. 19 is a configuration diagram of the adaptive observation section 18, and FIG. 20 is a configuration diagram of a status observer 182 which is a main constituent element of the adaptive observation section 18.
In FIG. 19, the adaptive observation section 18 is composed of a coordinate converter 181/ the status observer 182, and an integrator 183.
In FIG. 20, the status observer 182 is composed of gain matrix calculators 831, 833, 835, and 836, an adder/subtractor 832, an integrator 834, and a velocity estimator 837.
[0091] Next, the entire operation of the adaptive observation section 18 will be described.
In the adaptive observation section 18, the following expressions (23) to (25) respectively correspond to calculations by expressions (2) to (4) in the adaptive observation section 8 of embodiment 1, and the following expressions (26) and (27) respectively correspond to calculations by expressions (6) and (7). [0092]

[0097] Next, the function and operation of each main constituent element of the adaptive observation section 18 will be described.
In FIG. 19, the coordinate converter 181 converts a three-phase AC voltage instruction vector outputted from the control section 14 into a voltage instruction vector (vets, v $ s) on a-j8 axes which are an orthogonal coordinate system at rest, and outputs the voltage instruction vector (vas, v £s) to the status observer 182. In the status observer 182 in FIG. 20, the gain matrix calculator 831 multiplies the
voltage instruction vector {vas, vas)T outputted from the coordinate converter 181, by the matrix B of expression (23), and outputs the result. The adder/subtractor 832 outputs a vector obtained by addition and subtraction among output of the gain matrix calculator 831, output of the gain matrix calculator 833, and the amplified deviation vector (el, e2, e3, e4)T. The integrator 834 integrates each element of the vector outputted from the adder/subtractor 832, and outputs the resultant value as a vector (<£as0, <£#s0, 4> arO, (f> (i r0)T, The above corresponds to the right-hand side of

expression (24). The left-hand side of expression (24) corresponds to input to the integrator 834.
The gain matrix calculator 835 multiplies the
vector ( 0 asO, 0 0 sO, 0 ctrO, 0 # rO)T by the matrix CI of expression (23) and thereby outputs the estimated current
vector (0asO, 0j3sO)T. This corresponds to expression (26).
Here, the vector (0asO, 0#sO, arO, 0 0rO)T is a stator/rotor estimated magnetic flux vector.
The gain matrix calculator 836 multiplies the vector ( 0 asO, 0 #s0, 0 a rO, 0 $ rO)T by the matrix C2 and thereby outputs the estimated magnetic flux vector (0arO, 0 0rO)T. This corresponds to expression (27).
The gain matrix calculator 833 receives the
estimated velocity corO outputted from the velocity estimator 837, applies the matrix A to the vector (0 asO, 0 £ sO, 0 arO, 0j3rO)T outputted from the integrator 834, and outputs the result to the adder/subtractor 832.
The integrator 183 in FIG. 19, as in embodiment 1, integrates the estimated velocity corO outputted from the status observer 182 by expression (5), thereby calculating
the estimated position 0 0.
The above is the description of the configuration, function, and operation of the adaptive observation section 18. [0098] Next, the configuration, function, and operation of

the magnetic flux vector detection section 19 will be described based on FIGS. 21 to 25.
First, the configuration of the magnetic flux vector detection section 19 will be described.
In FIG. 21 showing a configuration diagram of the magnetic flux vector detection section 19, the magnetic flux vector detection section 19 is composed of the filter 91, an a-axis component extractor 192, a j3 -axis component extractor 193, and a detected magnetic flux vector calculator 194.
As shown in FIG. 22, the a-axis component extractor 192 is composed of an a-axis component selector 961 and an amplitude calculator 962. As shown in FIG. 23, the #-axis component extractor 193 is composed of a j3 -axis component selector 963 and an amplitude calculator 964. As shown in FIG. 24, the detected magnetic flux vector calculator 194 is composed of a difference calculator 965, a cosine calculator 966, a sine calculator 967, and multipliers 968 and 969.
[0099] Next, the function and operation of the magnetic flux vector detection section 19 will be described. First, the entire operation of the magnetic flux vector detection section 19 will be described, and then the function and operation of each constituent element will be described. [0100] A mathematical expression on a-Q axes of the AC

rotary machine 2 when the high-frequency voltage vector (van, v/3h) is applied thereto by the high-frequency voltage vector generator 145 in FIG. 17 can be represented as expression (28) . Here, p in the expression is a differential operator. [0101]
[0102] As in embodiment 1, the high-frequency voltage vector is applied only in a region of zero speed to low speed. Therefore, if the rotational velocity is set at cor ^ 0, expression (29) can be obtained.
[0103]

first term on the right-hand side is satisfied, so the first term on the right-hand side can be ignored. As a result, expression (30) can be obtained. [0105]
[0106] Here, since the high-frequency voltage vector (van, Vj8h) is given by expression (22), the high-frequency current vector (iah, iah) is represented as expression (31) by substituting expression (22) into expression (30) and integrating both sides. [0107]
[0108] Next, the detected magnetic flux vector will be described. As shown in FIG. 25, at a given moment, the rotor magnetic flux vector $r and a axis have a relationship shown in FIG. 25, and the angle therebetween is set at 0dm. Since the rotor position is set in a direction parallel to the rotor magnetic flux vector, the rotor position as seen from a axis is 0dm. As described above, in the present

embodiment 3, since the deviation vector calculation section 6 calculates a magnetic flux deviation vector on a-j3 axes, the detected magnetic flux vector outputted from the magnetic
flux vector detection section 19 is on a-& axes. In FIG. 25, of the rotor magnetic flux vector, a projected component in
the a-axis direction is set at $ arD and a projected component in the 0 -axis direction is set at tf>|3rD. Then, $ arD and (j> j8 rD are calculated by expression (32), and the detected magnetic flux vector (# arD, tf> )3rD) is outputted from the magnetic flux vector detection section 19. [0109]
[0110] In expression (32), ^ f is the magnitude of the rotor magnetic flux vector Or, and therefore is a known value. It is only necessary to know 0 dm, for calculating the components arD and 4> #rD of the detected magnetic flux vector from expression (32).
Looking at expression (31), since 0dm is included in the magnitude of the high-frequency current vector (iah, i0h), 0 dm can be calculated from the magnitude of the high-frequency current vector (iah, i#h). When the magnitude of

the high-frequency current iah on a axis is set at |iah( and the magnitude of the high-frequency current ij3h on fi axis is set at |i#h|, |iah| and Iij9h| are represented by expression (33). [0111]
[0112] In addition, the magnitudes licthi and |i,Sh| can be calculated by expression (34), using the high-frequency current vector (iah, i|3h). Here, T in expression (34) is the period of the high-frequency current vector (iah, i/3h). [0113]
[0114] From expression (33), a result of subtracting the reciprocal of |iahj from the reciprocal of |i£hj is represented by expression (35). From expression (35), 0dm can be calculated as shown by expression (36). [0115]

[0117] Thus, the detected magnetic flux vector ( <£> a rD, 0 #rD) can be calculated by expressions (32) and (36).
The above is the entire operation of the magnetic flux vector detection section 19.
[0118] Next, the function and operation of each constituent element of the magnetic flux vector detection section 19 will be described.
The filter 91 in FIG. 21 is the same as in embodiment 1 except that its input and output are changed to a detected current vector and a high-frequency current vector on a-jS axes. The filter 91 outputs the high-frequency current vector (iah, i|3h), based on the detected current vector (ias, i#s) .
As shown in FIG. 22, in the a—axis component extractor 192, the a-axis component selector 961 multiplies the high-frequency current vector (iah, "ij3h) by a matrix (1, 0)T, thereby selecting only iah which is a component in the

a-axis direction, of the high-frequency current vector. The amplitude calculator 962 performs calculation by expression
(34), to calculate the magnitude |iah| of iah, and outputs the magnitude |iah|.
As shown in FIG. 23, in the $ -axis component extractor 193, the j3-axis component selector 963 multiplies the high-frequency current vector (iah/ ij3h) by a matrix {0, 1)T, thereby selecting only i/Jh which is a component in the
)3-axis direction, of the high-frequency current vector. The amplitude calculator 964 performs calculation by expression
(34), to calculate the magnitude |i£h| of i^h, and outputs the magnitude | i j3 h | .
[0119] The detected magnetic flux vector calculator 194 will be described. In FIG. 24, the difference calculator 965 performs calculation by expression (36) using |iah| outputted from the a-axis component extractor 192 and Iij3h| outputted from the £-axis component extractor 193, and thereby outputs 0dm. The cosine calculator 966 performs cosine calculation using 0dm outputted from the difference calculator 965, and thereby outputs cos 0 dm. The multiplier 968 multiplies cos 0 dm outputted from the cosine calculator 966 by the magnitude $f of the rotor magnetic flux, and thereby outputs arD which is a detected magnetic flux vector component in the a-axis direction. The sine calculator 967 performs sine calculation using 0dm outputted

from the difference calculator 965, and thereby outputs sin0 dm. The multiplier 969 multiplies sin 0 dm outputted from the sine calculator 967 by the magnitude 0 f of the rotor magnetic flux, and thereby outputs fS rD which is a detected magnetic flux vector component in the j3-axis direction.
The above is the description of the configuration, function, and operation of the magnetic flux vector detection section 19.
[0120] As described above, the control device for AC rotary machine according to embodiment 3 calculates the magnetic flux deviation vector which is a deviation between the detected magnetic flux vector and the estimated magnetic flux vector, and the current deviation vector which is a deviation between the detected current vector and the estimated current vector, and uses the amplified deviation vector obtained by amplifying the magnetic flux deviation vector, thereby making it possible to drive an AC rotary machine by only one kind of estimated velocity or estimated position outputted from the adaptive observation section over a range from zero speed to high-speed region. Therefore, it is not necessary to cause both the low-frequency region phase generator and the high-frequency region phase generator to perform computing processing simultaneously, so the amount of computing processing can be greatly reduced.
In addition, in calculation of the detected

magnetic flux vector (