Control device for rotating electric machine, control method for rotating electric machine, and control program for rotating electric machine

Abstract

A control system (100) comprises: a rotating electric machine (40) having multi-phase windings (51U, 51V, 51W); a first inverter (20) connected to first ends of the windings; and a second inverter (30) connected to second ends of the windings. A control device (60) of the rotating electric machine applied to the control system (100) comprises: a storage unit (60b) for storing a trained model (70) from which control information for controlling the operations of the first inverter and second inverter is output when a condition parameter including a control amount of the rotating electric machine is input; and a control unit (60a) for controlling the operations of the first inverter and second inverter on the basis of the control information output from the trained model. The trained model is generated by machine learning so that a zero-phase current (I0r) of the windings approaches zero.

Description

Rotating electric machine control device, rotating electric machine control method, and rotating electric machine control program CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application is based on Japanese Application No. 2023-214544, filed on December 20, 2023, the contents of which are incorporated herein by reference.

[0002] The present disclosure relates to a control device for a rotating electric machine, a control method for a rotating electric machine, and a control program for a rotating electric machine.

[0003] A control system including a rotating electric machine having an open winding is known. In this control system, a first inverter is connected to a first end of each phase winding of the rotating electric machine, and a second inverter is connected to a second end of each phase winding. The high-potential side of the first inverter is connected to the high-potential side of the second inverter by a high-potential side connecting wire, and the low-potential side of the first inverter is connected to the low-potential side of the second inverter by a low-potential side connecting wire. The control system described above can implement H-bridge drive, which performs PWM drive on the upper and lower arm switches of each inverter, when driving the rotating electric machine.

[0004] In this control system, the zero-sequence current, which is the sum of the fundamental currents flowing through the windings of each phase in the rotating electric machine, may not be zero. If the zero-sequence current is not zero, a zero-sequence voltage including a 3×(2n-1)th harmonic voltage is generated in the rotating electric machine. The zero-sequence voltage is a disturbance voltage for the rotating electric machine, and the zero-sequence voltage increases the driving loss of the rotating electric machine.

[0005] Therefore, a control device has been proposed that stores map information (correspondence information) that associates the amplitude and phase of harmonic voltages with parameters that are correlated with the fundamental current flowing through the windings of each phase. This control device suppresses the zero-phase current by controlling each inverter based on the map information and the acquired parameters. An example of such a control device is described in Patent Document 1.

[0006] Japanese Patent Application Laid-Open No. 2020-188590

[0007] However, since the zero-phase current may change depending on the specifications and condition (temperature, deterioration state, etc.) of the rotating electric machine, it is extremely time-consuming to create map information that can handle all of these conditions. Even if such map information could be created, the amount of storage required for such map information would be enormous, putting a strain on the memory capacity of the control device.

[0008] The present disclosure has been made in consideration of the above, and its purpose is to provide a control device for a rotating electric machine, a control method for a rotating electric machine, and a control program for a rotating electric machine that can reduce the burden on memory capacity.

[0009] A first means for solving the above problem is a control device for a rotating electric machine that is applied to a control system that includes a rotating electric machine having multi-phase windings, a first inverter that is connected to a DC power source and connected to a first end of the windings of each phase, and a second inverter that is connected to a second end of the windings of each phase, and that includes a memory unit that stores a learning model that outputs control information for controlling the operation of the first inverter and the second inverter when condition parameters including at least a control amount of the rotating electric machine are input, and a control unit that acquires the condition parameters, inputs the acquired condition parameters into the learning model, and controls the operation of the first inverter and the second inverter based on the control information output from the learning model, and the learning model is generated by machine learning so that the zero-phase current of the windings approaches zero.

[0010] According to the above configuration, by inputting the condition parameters into the learning model, it is possible to obtain control information for controlling the operation of the first inverter and the second inverter, which reduces the amount of storage required compared to when a huge amount of map information is generated and stored.

[0011] A second means for solving the above problem is a control method for a rotating electric machine implemented by a rotating electric machine control device applied to a control system including a rotating electric machine having multi-phase windings, a first inverter connected to a DC power source and connected to a first end of the windings of each phase, and a second inverter connected to a second end of the windings of each phase, the control method including the steps of acquiring condition parameters including at least a control amount of the rotating electric machine, inputting the condition parameters to a learning model stored in a memory unit, and acquiring control information for controlling the operation of the first inverter and the second inverter output from the learning model, and controlling the operation of the first inverter and the second inverter based on the control information output from the learning model, wherein the learning model is generated by machine learning so that the zero-phase current of the windings approaches zero.

[0012] According to the above configuration, by inputting the condition parameters into the learning model, it is possible to obtain control information for controlling the operation of the first inverter and the second inverter, which reduces the amount of storage required compared to when a huge amount of map information is generated and stored.

[0013] A third means for solving the above problem is a control program for a rotating electric machine implemented by a rotating electric machine control device applied to a control system including a rotating electric machine having multi-phase windings, a first inverter connected to a DC power source and connected to a first end of the windings of each phase, and a second inverter connected to a second end of the windings of each phase, the program comprising the steps of acquiring condition parameters including at least a control amount of the rotating electric machine, inputting the condition parameters into a learning model stored in a memory unit, and acquiring control information for controlling the operation of the first inverter and the second inverter output from the learning model, and controlling the operation of the first inverter and the second inverter based on the control information output from the learning model, wherein the learning model is generated by machine learning so that the zero-phase current of the windings approaches zero.

[0014] According to the above configuration, by inputting the condition parameters into the learning model, it is possible to obtain control information for controlling the operation of the first inverter and the second inverter, which reduces the amount of storage required compared to when a huge amount of map information is generated and stored.

[0015] The above and other objects, features, and advantages of the present disclosure will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which Fig. 1 is a diagram showing the configuration of a control system, Fig. 2 is a functional block diagram showing the function of a control device, Fig. 3 is a diagram showing the relationship between the amplitude of a fundamental current and a harmonic voltage, Fig. 4 is a diagram showing the relationship between the phase of a fundamental current and a harmonic voltage, Fig. 5 is an explanatory diagram of current amplitude and current phase, Fig. 6 is an explanatory diagram of a learning model, Fig. 7 is a block diagram showing a control device in a second embodiment, Fig. 8 is a diagram showing a network system in a third embodiment, Fig. 9 is a functional block diagram showing the function of a control device in a modified example, and Fig. 10 is a diagram showing machine learning in the modified example.

[0016] Several embodiments will be described with reference to the drawings. In several embodiments, functionally and / or structurally corresponding and / or associated parts may be designated by the same reference numerals. For corresponding and / or associated parts, reference may be made to the descriptions of other embodiments.

[0017] A first embodiment of a control device for a rotating electric machine according to the present disclosure will now be described with reference to the drawings. A control device 60 for a rotating electric machine 40 according to the present embodiment is mounted on an electric vehicle such as an electric car or a hybrid car, and is applied to a control system 100 for the rotating electric machine 40 (hereinafter simply referred to as the control system 100).

[0018] As shown in FIG. 1, the control system 100 of this embodiment includes a battery 10, a first inverter 20, a second inverter 30, a rotating electric machine 40, and a control device 60 for the rotating electric machine (hereinafter, in the first embodiment, simply referred to as the control device 60).

[0019] The battery 10 is, for example, a battery pack including a series connection of unit cells, and is a DC power supply in this embodiment. The unit cell is a single battery cell or a series connection of multiple battery cells. The battery cell is, for example, a secondary battery such as a lithium-ion battery.

[0020] The first inverter 20 and the second inverter 30 are power conversion circuits that convert DC power supplied from the battery 10 into three-phase AC power and supply it to the rotating electric machine 40 .

[0021] The first inverter 20 includes a series connection of a U-phase first upper arm switch SUHa and a U-phase first lower arm switch SULa, a V-phase first upper arm switch SVHa and a V-phase first lower arm switch SVLa, and a W-phase first upper arm switch SWHa and a W-phase first lower arm switch SWLa. Hereinafter, these will be collectively referred to as switches SUHa to SWLa.

[0022] Similarly, the second inverter 30 includes a series connection of a U-phase second upper arm switch SUHb and a U-phase second lower arm switch SULb, a series connection of a V-phase second upper arm switch SVHb and a V-phase second lower arm switch SVLb, and a series connection of a W-phase second upper arm switch SWHb and a W-phase second lower arm switch SWLb. Hereinafter, these will be collectively referred to as switches SUHb to SWLb.

[0023] In this embodiment, voltage-controlled semiconductor switching elements, more specifically, IGBTs, are used as the switches SUHa to SWLa and SUHb to SWLb. In this embodiment, the high-potential terminal of each switch is the collector, and the low-potential terminal is the emitter. Freewheeling diodes DUHa, DVHa, DWHa, DULa, DVLa, DWLa, DUHb, DVHb, DWHb, DULb, DVLb, and DWLb are connected in anti-parallel to the switches SUHa, SVHa, SWHa, SULa, SVLa, SWLa, SUHb, SVHb, SWHb, SULb, SVLb, and SWLb, respectively.

[0024] The collectors of the first upper arm switches SUHa, SVHa, SWHa of each phase and the collectors of the second upper arm switches SUHb, SVHb, SWHb of each phase are electrically connected by a positive bus 11 serving as a high-potential side connecting line such as a bus bar. The emitters of the first lower arm switches SULa, SVLa, SWLa of each phase and the emitters of the second lower arm switches SULb, SVLb, SWLb of each phase are electrically connected by a negative bus 12 serving as a low-potential side connecting line such as a bus bar.

[0025] The positive terminal of the battery 10 is electrically connected to the positive bus 11, and the negative terminal of the battery 10 is electrically connected to the negative bus 12. The battery 10 is electrically connected to the buses 11, 12 on the opposite side of the first inverter 20 from the second inverter 30.

[0026] The rotating electric machine 40 is an on-vehicle main motor. A rotor 41 of the rotating electric machine 40 is capable of transmitting power to drive wheels (not shown) of the vehicle. In this embodiment, the rotating electric machine 40 is a permanent magnet field type synchronous machine. The rotor 41 includes permanent magnets 42 (e.g., neodymium magnets) as field poles.

[0027] The rotating electric machine 40 includes a stator 50. The stator 50 includes armature windings, namely, a U-phase winding 51U, a V-phase winding 51V, and a W-phase winding 51W. The phase windings 51U, 51V, and 51W are arranged with an electrical angle of 120°. The phase windings 51U, 51V, and 51W are open-connected, and both ends of each phase winding 51U, 51V, and 51W are electrically connected to the first inverter 20 or the second inverter 30.

[0028] Specifically, in each phase, first ends 51Ua, 51Va, 51Wa of the windings 51U, 51V, 51W are electrically connected to first upper switches SUHa, SVHa, SWHa and first lower switches SULa, SVLa, SWLa of the corresponding phase. Also, second ends 51Ub, 51Vb, 51Wb of the windings 51U, 51V, 51W are electrically connected to second upper switches SUHb, SVHb, SWHb and second lower switches SULb, SVLb, SWLb of the corresponding phase.

[0029] The control system 100 includes a changeover switch 13. The changeover switch 13 is provided on the positive electrode side bus 11 between the first inverter 20 and the second inverter 30. When the changeover switch 13 is turned on, it electrically connects the first inverter 20 and the second inverter 30, and when turned off, it electrically disconnects the first inverter 20 and the second inverter 30. The changeover switch 13 is controlled by the control device 60. The changeover switch 13 is provided to switch the driving state of the control system 100, as will be described later.

[0030] For example, the changeover switch 13 is a semiconductor switching element such as an IGBT or a relay. When an IGBT is used as the changeover switch 13, a freewheel diode is connected in parallel to the changeover switch 13. In this case, the anode of the freewheel diode is electrically connected to the second inverter 30 side, and the cathode is electrically connected to the first inverter 20 side.

[0031] The control system 100 includes a power switch 14 and a capacitor 15. The power switch 14 is, for example, a semiconductor switching element or a relay. The power switch 14 is provided on the positive bus 11 between the positive terminal of the battery 10 and the first inverter 20. When the power switch 14 is turned on, it electrically connects the battery 10 and the rotating electric machine 40, and when it is turned off, it electrically disconnects the battery 10 and the rotating electric machine 40. The power switch 14 is driven by the control device 60.

[0032] A first end of the capacitor 15 is electrically connected to the positive bus 11 between the power switch 14 and the first inverter 20. A second end of the capacitor 15 is electrically connected to the negative bus 12 between the negative terminal of the battery 10 and the first inverter 20.

[0033] The control system 100 includes a current sensor 16 and a rotation angle sensor 17. The current sensor 16 detects phase currents Iu, Iv, and Iw flowing through the respective phase windings 51U, 51V, and 51W. In this embodiment, the current sensor 16 is provided at one of the ends of each phase winding 51U, 51V, and 51W closer to the first inverter 20. The sign of the detected value of the current sensor 16 is positive when the current flows from the first terminal 51Ua, 51Va, and 51Wa of each phase winding 51U, 51V, and 51W to the second terminal 51Ub, 51Vb, and 51Wb, and negative when the current flows from the second terminal 51Ub, 51Vb, and 51Wb to the first terminal 51Ua, 51Va, and 51Wa. Note that the current sensor 16 may also be provided at one of the ends of each phase winding 51U, 51V, and 51W closer to the second inverter 30.

[0034] The rotation angle sensor 17 is, for example, a resolver, and detects the electrical angle θ of the rotor 41. The phase currents Iur, Ivr, and Iwr detected by the current sensor 16 and the electrical angle θr detected by the rotation angle sensor 17 are input to the control device 60.

[0035] The control device 60 is primarily composed of a microcomputer including a processing unit 60a such as a CPU and a storage unit 60b such as various types of memory. The functions provided by the microcomputer can be provided by software stored in a physical memory device and a computer executing the software, software alone, hardware alone, or a combination thereof. For example, when the microcomputer is provided by electronic circuits, which are hardware, the functions can be provided by digital circuits including numerous logic circuits or analog circuits. For example, the processing unit 60a of the microcomputer executes programs stored in a non-transitory tangible storage medium (non-transitory tangible storage medium) that serves as the storage unit 60b. The programs include, for example, programs that realize the functions shown in FIG. 2 . Execution of the programs results in the execution of methods corresponding to the programs. The storage unit 60b is, for example, a non-volatile memory. The programs stored in the storage unit 60b can be downloaded and updated via a communication network such as the Internet, for example, via OTA (Over the Air).

[0036] As shown in FIG. 2 , the control device 60 functions as a mode switching unit 90 that switches the drive state of the control system 100 between a Y drive state and an H drive state by controlling the selector switch 13. The mode switching unit 90 switches the control system 100 to the Y drive state (first mode) by turning off the selector switch 13, turning on the second upper arm switches SUHb, SVHb, and SWHb of each phase, and turning off the lower arm switches SULb, SVLb, and SWLb of each phase. In the Y drive state, the phase windings 51U, 51V, and 51W are Y-connected via the second inverter 30. On the other hand, the mode switching unit 90 switches the control system 100 to the H drive state (second mode) by turning on the selector switch 13.

[0037] In the Y drive state, the control device 60 controls the switching of the switches SUHa to SWLa in the first inverter 20. In addition, in the H drive state, the control device 60 controls the switching of both the switches SUHa to SWLa in the first inverter 20 and the switches SUHb to SWLb in the second inverter 30. By appropriately switching between the Y drive state and the H drive state and executing switching control, the control system 100 can be made to have a high output and high efficiency.

[0038] When the control system 100 is in the H drive state and switching control is performed, a current flows between the first inverter 20 and the second inverter 30 via the positive bus 11 and the negative bus 12. In this case, a zero-phase current, which is the sum of the currents flowing through the phase windings 51U, 51V, and 51W, may be generated in the control system 100. The zero-phase current is one of the factors that can cause malfunctions in the control system 100 and deterioration of power consumption. Therefore, a technology for suppressing the generation of the zero-phase current is desired.

[0039] The configuration for suppressing the occurrence of zero-phase current will be described below. Fig. 2 is a functional block diagram for explaining various functions realized by the control device 60. Note that the various functions of the control device 60 are realized, for example, by the arithmetic processing device 60a executing a program stored in the storage unit 60b.

[0040] 2, the control device 60 functions as a setting unit 61. A torque request value Trq* is input to the setting unit 61 from a control device that is higher in level than the control device 60. The setting unit 61 calculates a d-axis current command value Id* and a q-axis current command value Iq* in a dq coordinate system based on the input torque request value Trq*. Hereinafter, the d-axis current command value Id* and the q-axis current command value Iq* may be collectively referred to as the d- and q-axis current command values ​​Id* and Iq*.

[0041] The control device 60 functions as a dq converter 62. The dq converter 62 receives the phase currents Iur, Ivr, and Iwr detected by the current sensor 16 and the electrical angle θr detected by the rotation angle sensor 17. The dq converter 62 calculates a d-axis current value Idr and a q-axis current value Iqr based on the detected phase currents Iur, Ivr, and Iwr and the electrical angle θr. The d-axis current value Idr and the q-axis current value Iqr may be collectively referred to as the d- and q-axis current values ​​Idr and Iqr.

[0042] The control device 60 functions as a current feedback unit 63. The current feedback unit 63 receives d- and q-axis current command values ​​Id* and Iq* and d- and q-axis current values ​​Idr and Iqr. The current feedback unit 63 calculates a d-axis current deviation, which is the difference between the d-axis current command value Id* and the d-axis current value Idr, and calculates a d-axis voltage command value Vd* as a manipulated variable for feedback-controlling the calculated d-axis current deviation to zero. The current feedback unit 63 also calculates a q-axis current deviation, which is the difference between the q-axis current command value Iq* and the q-axis current value Iqr, and calculates a q-axis voltage command value Vq* as a manipulated variable for feedback-controlling the calculated q-axis current deviation to zero. The feedback control is, for example, proportional-plus-integral control.

[0043] The control device 60 functions as a UVW converter 64. The d- and q-axis voltage command values ​​Vd* and Vq* and the detected electrical angle θr are input to the UVW converter 64. The UVW converter 64 calculates a U-phase voltage command value Vu*, a V-phase voltage command value Vv*, and a W-phase voltage command value Vw* based on the input d- and q-axis voltage command values ​​Vd* and Vq* and the electrical angle θr. Hereinafter, these may be collectively referred to as the respective phase voltage command values ​​Vu*, Vv*, and Vw*.

[0044] The phase voltage command values ​​Vu*, Vv*, and Vw* are command values ​​for the phase voltages Vu, Vv, and Vw. In this embodiment, the signs of the phase voltages Vu, Vv, and Vw are positive when the potentials of the first terminals 51Ua, 51Va, and 51Wa of the windings 51U, 51V, and 51W are higher than the potentials of the second terminals 51Ub, 51Vb, and 51Wb of the windings 51U, 51V, and 51W, respectively, and negative when the potentials of the second terminals 51Ub, 51Vb, and 51Wb of the windings 51U, 51V, and 51W are higher than the potentials of the first terminals 51Ua, 51Va, and 51Wa of the windings 51U, 51V, and 51W.

[0045] The control device 60 functions as a current calculation unit 65, an angular velocity calculation unit 66, and a correction unit 67. The current calculation unit 65 calculates a zero-phase current detection value Ior by summing the detected phase currents Iur, Ivr, and Iwr. FIG. 2 shows a first adder 65a and a second adder 65b as an example of the current calculation unit 65. The first adder 65a adds the detected W-phase current Iwr to the detected V-phase current Ivr. The second adder 65b adds the detected U-phase current Iur to the sum of the V- and W-phase currents Ivr and Iwr calculated by the first adder 65a. In this case, the output value of the second adder 65b is the zero-phase current detection value Ior.

[0046] The angular velocity calculation unit 66 calculates the electrical angular velocity ω as a time differential value of the electrical angle θr detected by the rotation angle sensor 17 .

[0047] The correction unit 67 receives a drive state signal S1 indicating the current drive state of the control system 100 from the mode switching unit 90. If the drive state signal S1 determines that the drive state of the control system 100 is the H drive state, the correction process described below is performed by the estimation unit 67a and U-, V-, and W-phase calculation units 68U, 68V, and 68V. Note that if the drive state of the control system 100 is determined to be the Y drive state, the correction process described below does not need to be performed.

[0048] The corrector 67 functions as an estimator 67a. The estimator 67a receives (acquires) the d- and q-axis current command values ​​Id* and Iq*, the zero-phase current detection value Ior calculated by the current calculator 65, and the electrical angular velocity ω and electrical angle θr calculated by the angular velocity calculator 66. The estimator 67a then inputs the acquired d- and q-axis current command values ​​Id* and Iq*, the zero-phase current detection value Ior, the electrical angle θr, and the electrical angular velocity ω to a learning model 70 stored in the storage unit 60b, thereby calculating (estimating) a correction amount Vcr for each phase voltage command value Vu*, Vv*, and Vw* as control information. In this embodiment, the correction amount Vcr is common to all phases and corresponds to control information for the first inverter 20 and the second inverter 30.

[0049] The learning model 70 is a trained model generated in advance through machine learning so that the zero-phase current detection value I0r approaches zero, and is stored in the storage unit 60b. The machine learning is, for example, deep learning. The learning model 70 will be described later.

[0050] The correction unit 67 corrects the phase voltage command values ​​Vu*, Vv*, and Vw* based on the calculated correction amount Vcr for the phase voltage command values ​​Vu*, Vv*, and Vw*. Specifically, the correction unit 67 includes U-, V-, and W-phase calculation units 68U, 68V, and 68W. The U-phase calculation unit 68U subtracts the correction amount Vcr for the U-phase voltage command value from the U-phase voltage command value Vu*. The V-phase calculation unit 68V subtracts the correction amount Vcr for the V-phase voltage command value from the V-phase voltage command value Vv*. The W-phase calculation unit 68W subtracts the correction amount Vcr for the W-phase voltage command value from the W-phase voltage command value Vw*. In this case, for example, when the phase currents Iu, Iv, and Iw are flowing in the positive direction, the phase voltage command values ​​Vu*, Vv*, and Vw* are corrected so that the magnitudes of the positive phase voltages Vu, Vv, and Vw are reduced. Furthermore, for example, when the phase currents Iu, Iv, and Iw flow in the negative direction, the phase voltage command values ​​Vu*, Vv*, and Vw* are corrected so that the magnitudes of the negative phase voltages Vu, Vv, and Vw are reduced. Furthermore, when the correction amount Vcr is subtracted, the phase is shifted before subtraction so as to correspond to the phase voltage command values ​​Vu*, Vv*, and Vw*.

[0051] The control device 60 functions as a modulator 69. The modulator 69 receives corrected phase voltage command values ​​Vu*, Vv*, and Vw*, which are output values ​​from the phase calculation units 68U, 68V, and 68W. The modulator 69 generates operation signals SUHa to SWLa of the first inverter 20 and operation signals SUHb to SWLb of the second inverter 30 based on the corrected phase voltage command values ​​Vu*, Vv*, and Vw* and a carrier signal. The carrier signal is, for example, a triangular wave signal. Switching control of the first inverter 20 and the second inverter 30 is performed based on the generated operation signals.

[0052] Next, the learning model 70 will be described. First, the parameters input to the learning model 70 will be described. The phase currents Iur, Ivr, and Iwr (fundamental currents) flowing through the windings 51U, 51V, and 51W of each phase are correlated with the zero-phase voltage. The relationship between the harmonic voltage Vz and the fundamental current will be described below using the third-order harmonic voltage Vz, which is the main component of the zero-phase voltage. FIG. 3 shows the relationship between the amplitude Ie and the current phase β of the fundamental current with respect to the amplitude Ψz of the harmonic voltage Vz. FIG. 4 shows the relationship between the amplitude Ie and the current phase β of the fundamental current with respect to the phase θz of the harmonic voltage Vz. FIG. 5 shows the relationship between the d-axis current Id, the q-axis current Iq, the current amplitude Ie, and the current phase β. The current amplitude Ie and the current phase β are expressed as in Equation 1 and Equation 2 using the d-axis and q-axis currents Id and Iq, which are converted from the phase currents Iu, Iv, and Iw.

[0053] Ie = √(Id^2 + Iq^2) (Equation 1) β = arctan(Id / Iq) (Equation 2) As shown in Fig. 3, the amplitude Ψz of the harmonic voltage Vz has a correlation in which it increases as the current phase β or the current amplitude Ie increases. Also, the phase θz of the harmonic voltage Vz has a correlation in which it decreases as the current phase β or the current amplitude Ie decreases. Therefore, the d- and q-axis current command values ​​Id* and Iq* that specify the d- and q-axis currents Id and Iq are considered to be correlated with the amplitude Ψz and phase θz of the harmonic voltage Vz.

[0054] If the amplitude Ψz and phase θz of the harmonic voltage Vz can be estimated, a correction amount R* for reducing the harmonic voltage Vz can be calculated. Specifically, the correction amount R* is expressed as in (Equation 3) using the amplitude Ψz, phase θz, electrical angle θr, and electrical angular velocity ω. The correction amount R* is common to all three phases.

[0055] R*=3ω×Ψz×sin(3θr+θz) (Equation 3) Considering the above, it is considered that if machine learning is performed using the d- and q-axis current command values ​​Id* and Iq*, the electrical angle θr, and the electrical angular velocity ω as input parameters, it will be possible to estimate the correction amount Vcr that can effectively suppress the zero-phase current. Note that in this embodiment, the d- and q-axis current command values ​​Id* and Iq*, the electrical angle θr, and the electrical angular velocity ω correspond to condition parameters, the d- and q-axis current command values ​​Id* and Iq* correspond to control amounts of the rotating electric machine 40, and the electrical angle θr and the electrical angular velocity ω correspond to state parameters that indicate the operating state of the rotating electric machine.

[0056] The zero-phase current is also considered to be affected by the state before the d-axis and q-axis current command values ​​Id* and Iq* are input, i.e., the current phase currents Iu, Iv, and Iw. The zero-phase current is also considered to change depending on the state (temperature and deterioration state) of the rotating electric machine 40. For this reason, if the zero-phase current detection value I0r is added as an input parameter and machine learning is performed, it is considered possible to estimate the correction amount Vcr according to the current zero-phase current and the state of the rotating electric machine 40. In this embodiment, the zero-phase current detection value I0r corresponds to an evaluation parameter.

[0057] Next, an example of a learning model will be described. As shown in FIG. 6, the learning model 70 is a model using a neural network with a multilayer perceptron structure, and has an input layer, a hidden layer, and an output layer. Note that each layer may be further divided into multiple layers. Each layer is composed of multiple perceptrons 71. The perceptron 71 in a given layer weights and biases the input value from the perceptron 71 in the previous layer (external in the case of the input layer), inputs it, performs a predetermined process, and outputs an output value to the perceptron 71 in the next layer (external in the case of the output layer).

[0058] Next, an example of machine learning will be described. Machine learning in this embodiment is performed by a dedicated electronic computer such as a workstation before shipping the control system 100 or the control device 60 (during manufacturing, design, etc.). The electronic computer has the same functions as the estimator 67a. The estimator 67a of the electronic computer inputs the various parameters (Id*, Iq*, ω, θr, Ior) described above to the learning model 70 before learning (before training) and causes the learning model 70 to estimate the correction amount Vcr. The electronic computer corrects the phase voltage command values ​​Vu*, Vv*, and Vw* based on this correction amount Vcr. Note that the method for calculating the pre-correction phase voltage command values ​​Vu*, Vv*, and Vw* from the d- and q-axis current command values ​​Id* and Iq* and the method for calculating the corrected phase voltage command values ​​Vu*, Vv*, and Vw* are as described above (see FIG. 2 ), and therefore will not be described here.

[0059] The electronic computer then controls the operation of the first inverter 20 and the second inverter 30 based on the corrected phase voltage command values ​​Vu*, Vv*, and Vw*, and detects the zero-phase current detection value I0r. The electronic computer then updates at least one of the weighting and bias of the perceptron constituting the learning model 70 so that the zero-phase current detection value I0r falls within an allowable range including zero, i.e., so that the pulsation of the zero-phase current detection value I0r approaches zero. The updating method is, for example, an error backpropagation method. The computer repeatedly executes these learning processes to generate the final learning model 70. The final learning model 70 is stored in the storage unit 60b before shipping (e.g., during design or manufacturing).

[0060] The effects of the above embodiment will be described.

[0061] The control device 60 inputs the condition parameters and evaluation parameters to a learning model 70 and controls the operations of the first inverter 20 and the second inverter 30 based on the control information output from the learning model 70. More specifically, an estimator 67a in a corrector 67 of the control device 60 inputs the d- and q-axis current command values ​​Id* and Iq*, the zero-phase current detection value I0r, the electrical angle θr, and the electrical angular velocity ω to the learning model 70, thereby calculating a correction amount Vcr for the phase voltage command values ​​Vu*, Vv*, and Vw* as control information. Then, U-, V-, and W-phase calculators 68U, 68V, and 68V correct the phase voltage command values ​​Vu*, Vv*, and Vw* by subtracting the correction amount Vcr from the phase voltage command values ​​Vu*, Vv*, and Vw*, respectively. Then, the control device 60 generates operation signals SUHa to SWLa of the first inverter 20 and operation signals SUHb to SWLb of the second inverter 30 based on the corrected phase voltage command values ​​Vu*, Vv*, Vw*, and executes switching control of the first inverter 20 and the second inverter 30 based on the generated operation signals. This eliminates the need to generate map information and also eliminates the need to store a huge amount of map information.

[0062] The condition parameters are the d-axis and q-axis current command values ​​Id* and Iq*, the electrical angle θr, and the electrical angular velocity ω. As described above, these parameters have a stronger correlation with the harmonic voltage Vz of the zero-phase voltage than other parameters. Therefore, by inputting the d-axis and q-axis current command values ​​Id* and Iq*, the electrical angle θr, and the electrical angular velocity ω into the learning model 70 as parameters, an appropriate correction amount that effectively suppresses the zero-phase current can be estimated. This is also expected to reduce the control load when calculating the learning amount and the correction amount Vcr. Furthermore, when the operating point of the rotating electric machine 40 transiently changes, the zero-phase voltage command can be quickly updated based on the change in motor current accompanying the change in the operating point, which is effective when used during frequent acceleration and deceleration.

[0063] The zero-phase current is affected by the state of the rotating electric machine 40. Therefore, the zero-phase current detection value I0r is input as an evaluation parameter. This makes it possible to estimate the correction amount Vcr that reflects the state of the rotating electric machine 40. It is also expected that the control load when calculating the learning amount and the correction amount Vcr can be reduced.

[0064] The learning model 70 is generated through machine learning so that the detected zero-phase current value I0r approaches zero. More specifically, various parameters are input to the pre-learning learning model 70 to estimate the correction amount Vcr. Then, corrected phase voltage command values ​​Vu*, Vv*, and Vw* are calculated based on the correction amount Vcr and other parameters. The operation of the first inverter 20 and the second inverter 30 is controlled based on the corrected phase voltage command values ​​Vu*, Vv*, and Vw* to detect the detected zero-phase current value I0r. At least one of the weighting and bias of the perceptron constituting the learning model 70 is updated so that the detected zero-phase current value I0r falls within an allowable range that includes zero. This learning process is then repeated to generate the final learning model 70. Because the learning model 70 is trained in this manner, the detected zero-phase current value I0r can be suppressed.

[0065] The learning model 70 is used to calculate the correction amount Vcr of each phase voltage command value Vu*, Vv*, Vw* as control information. This allows the functions used when the control system 100 is in the Y drive state, such as the setting unit 61, current feedback unit 63, UVW conversion unit 64, and modulation unit 69, to be used in the H drive state as well. This reduces the amount of design work.

[0066] The correction unit 67 suppresses the zero-phase current when the drive state of the control system 100 is the H drive state. Therefore, the learning model 70 is not necessary when the drive state is the Y drive state, and the amount of learning of the learning model 70 can be reduced.

[0067] Second Embodiment The configuration of the control system 100 described in the first embodiment may be partially modified. A second embodiment in which the configuration of the control system 100 described in the first embodiment is partially modified will be described below.

[0068] The learning model 70 of the second embodiment is configured to be updated by the control device 60 even after being installed in the vehicle. To explain in detail, as shown in Fig. 7, the control device 60 has a function as an update unit 80. The update unit 80 is realized, for example, by the arithmetic processing device 60a executing a program stored in the storage unit 60b.

[0069] The control device 60 acquires the various parameters input into the learning model 70 and the zero-phase current detection value I0r detected as a result of correction by the correction unit 67 based on the various parameters, and associates them and stores them in the memory unit 60b as historical data 81.

[0070] The update unit 80 of the control device 60 then refers to the history data 81 at a predetermined timing and updates at least one of the weighting and bias of the perceptron constituting the learning model 70 so that the zero-phase current detection value I0r falls within an allowable range including zero. The updating method is the same as in the first embodiment. The update unit 80 then stores the updated learning model 70 in the storage unit 60b. The predetermined timing may be while the vehicle is running or while it is stopped.

[0071] According to the second embodiment, the following effects are achieved.

[0072] By providing the update unit 80 that updates the learning model 70, the learning model 70 can be continuously updated in accordance with the state of the rotating electric machine 40. As a result, even if the rotating electric machine 40 deteriorates and its state changes, affecting the zero-phase current, the correction amount Vcr of each phase voltage command value Vu*, Vv*, Vw* can be calculated using the learning model 70 in accordance with the state of the rotating electric machine 40. Therefore, even if the state of the rotating electric machine 40 changes, the zero-phase current can be suitably suppressed.

[0073] Third Embodiment The configuration of the control system 100 described in the first embodiment may be partially modified. A third embodiment in which the configuration of the control system 100 described in the first embodiment is partially modified will be described below.

[0074] The learning model 70 of the third embodiment is configured to be updated by an external device via a communication network. FIG. 8 shows the overall configuration of a communication network system 300 of the third embodiment. The communication network system 300 includes a large number of vehicles 310 and an external server. The external server is a device operated by a service provider that provides a service. The external server of this embodiment is a cloud server 320. The vehicles 310 and the cloud server 320 are connected to each other so as to be able to communicate with each other via a communication network 330 (e.g., the Internet). The communication network 330 is at least one of a wired network and a wireless network.

[0075] The vehicle 310 includes a control system 100. The configuration of the control system 100 is substantially the same as that of the first embodiment. In the third embodiment, only the differences from the first embodiment will be described, and the same components as those in the first embodiment will be assigned the same reference numerals as those in the first embodiment, and detailed description thereof will be omitted.

[0076] The control device 60 of the control system 100 includes a communication unit 340. The communication unit 340 functions as an interface for transmitting and receiving data to and from external communication devices. In the control device 60, the communication unit 340 is connected to the arithmetic processing unit 60a and the storage unit 60b via a communication bus or the like.

[0077] The communication unit 340 transmits and receives data via the communication network 330. The communication unit 340 transmits data to other devices in accordance with instructions from the arithmetic processing unit 60a. The communication unit 340 also receives data transmitted from other devices and transmits the data to the arithmetic processing unit 60a or the storage unit 60b.

[0078] The cloud server 320 includes a control unit 321, a communication unit 322, and a storage unit 323. In the cloud server 320, the control unit 321, the communication unit 322, and the storage unit 323 are connected to one another via a communication bus 324.

[0079] The storage unit 323 includes a memory and a storage. The control unit 321 includes a processor, and controls the operation of the cloud server 320 by reading and executing programs stored in the storage unit 323. The communication unit 322 transmits and receives data via the communication network 330. The communication unit 322 transmits data to other devices in accordance with instructions from the control unit 321. The communication unit 322 receives data transmitted from other devices and transmits the data to at least one of the control unit 321 and the storage unit 323.

[0080] The learning model 70 of the third embodiment is configured to be updated by the cloud server 320 even after being installed in the vehicle.

[0081] To explain in more detail, the control device 60 acquires the various parameters input into the learning model 70 and the zero-phase current detection value I0r detected as a result of correction by the correction unit 67 based on the various parameters, and associates them and stores them in the memory unit 60b as historical data.

[0082] Then, the communication unit 340 of the control device 60 transmits the history data and the learning model 70 to the cloud server 320 via the communication network 330 at a predetermined timing. The predetermined timing may be while the vehicle 310 is traveling or while the vehicle 310 is stopped. Alternatively, for example, the predetermined timing may be when the battery 10 of the vehicle 310 is being charged by an external charger.

[0083] The communication unit 322 of the cloud server 320 transmits the received history data and learning model 70 to at least one of the control unit 321 and the memory unit 323.

[0084] The control unit 321 of the cloud server 320 refers to the received historical data and updates at least one of the weighting and bias of the perceptron that constitutes the learning model 70 so that the zero-phase current detection value I0r falls within an allowable range that includes zero. The updating method is the same as in the first embodiment. Then, the communication unit 322 of the cloud server 320 transmits the updated learning model 70 to the communication unit 340 of the vehicle 310 via the communication node 330. Upon receiving the updated learning model 70, the communication unit 340 of the vehicle 310 stores it in the storage unit 60b.

[0085] According to the third embodiment, the following effects are achieved.

[0086] Communication with the cloud server 320 is performed via the communication network 330, and the learning model 70 can be updated by the cloud server 320. This allows the learning model 70 to be continuously updated according to the state of the rotating electric machine 40. Therefore, even if the rotating electric machine 40 deteriorates and its state changes, affecting the zero-phase current, the correction amount Vcr of the phase voltage command values ​​Vu*, Vv*, and Vw* can be calculated using the learning model 70 according to the state of the rotating electric machine 40. Therefore, even if the state of the rotating electric machine 40 changes, the zero-phase current can be suitably suppressed.

[0087] Furthermore, since the learning model 70 is updated by the cloud server 320, there is no need to provide the vehicle 310 with a high-performance control device 60 capable of performing machine learning.

[0088] (Modifications) Hereinafter, modifications in which the configuration of the above embodiment is partially changed will be described.

[0089] In the above embodiment, the correction unit 67 receives the drive status signal S1 and performs correction processing only when the drive status is H. However, the correction unit 67 may perform correction processing regardless of the drive status without receiving the drive status signal S1. Note that the H drive status can also be determined from the zero-phase current detection value I0r, etc., even without receiving the drive status signal S1.

[0090] In the above embodiment, the phase currents Iu, Iv, and Iw may be disturbed and a zero-phase current may occur during a predetermined period of time from immediately after the drive state is changed. Therefore, the drive state signal S1 may be used as a parameter to be input to the learning model 70, and the learning model 70 may be trained to reflect changes in the drive state signal S1. This is expected to suppress the zero-phase current immediately after the drive state is changed.

[0091] In the above embodiment, when the drive state is changed, this is reflected in the d-axis and q-axis current values ​​Idr and Iqr, the zero-phase current detection value I0r, and the phase currents Iur, Ivr, and Iwr. Therefore, these detection values ​​may be used instead of the drive state signal S1.

[0092] In the above embodiment, environmental parameters related to the operating environment of the rotating electric machine 40, such as the temperature of the rotating electric machine 40 or the set atmospheric pressure, may be used as the condition parameters. If there is a change in the operating environment of the rotating electric machine 40, there is a possibility that the zero-phase current will change. Therefore, by using these environmental parameters and performing machine learning, it is possible to generate a learning model 70 that takes into account the influence of the operating environment of the rotating electric machine 40. It is expected that the zero-phase current will be suppressed even if the operating environment of the rotating electric machine 40 changes.

[0093] In the above embodiment, the correction amount Vcr is common to all phases. However, the learning model 70 may be constructed so that the correction amount Vcr differs for each of the U phase, V phase, and W phase.

[0094] In the above embodiment, the zero-phase current detection value I0r is used as the evaluation parameter, but a parameter other than the zero-phase current detection value I0r may be used as long as it is a parameter correlated with the zero-phase current detection value I0r detected as a result of controlling the rotary electric machine 40. For example, instead of the zero-phase current detection value I0r, the respective phase currents Iur, Ivr, and Iwr may be used, or the d-axis current value Idr may be used.

[0095] In the above embodiment, the d-axis and q-axis current command values ​​Id* and Iq*, the electrical angle θr, and the electrical angular velocity ω are used as the condition parameters. However, the type and number of parameters may be changed as long as they are parameters for instructing the operation of the rotary electric machine 40 and are correlated with the zero-phase current. For example, instead of the d-axis and q-axis current command values ​​Id* and Iq*, a torque request value Trq* indicating a control amount of the rotary electric machine 40 may be used. Also, one or both of the electrical angle θr and the electrical angular velocity ω may be omitted from the condition parameters. Also, although the correction amount Vcr is output (estimated) as control information, the corrected phase voltage command values ​​Vu*, Vv*, and Vw* may be output (estimated).

[0096] Below, a modified example will be described with reference to FIG. 9 , in which a learning model 170 is employed that employs the torque request value Trq*, the q-axis current value Iqr, the electrical angle θr, and the electrical angular velocity ω as condition parameters and outputs the phase voltage command values ​​Vu*, Vv*, and Vw* as control information.

[0097] 9 , the control device 60 has a function as a voltage estimator 400 instead of the setting unit 61, current feedback unit 63, UVW conversion unit 64, correction unit 67, estimator 67a, and U-, V-, and W-phase calculation units 68U, 68V, and 68W. The voltage estimator 400 acquires the torque request value Trq*, d- and q-axis current values ​​Idr and Iqr, the zero-phase current detection value I0r, the electrical angle θr, and the electrical angular velocity ω, and inputs them to a learning model 70.

[0098] In this modified example, the learning model 70 is trained by inputting the torque request value Trq* instead of the d-axis and q-axis current command values ​​Id* and Iq*. Furthermore, the learning model 70 in this modified example is configured to estimate (calculate) the (corrected) phase voltage command values ​​Vu*, Vv*, and Vw* instead of the correction value Vcr as control information, and performs machine learning. As in the above embodiment, the learning method (update method) is performed so that the zero-phase current detection value I0r falls within an allowable range including zero, i.e., so that the pulsation of the zero-phase current detection value I0r approaches zero.

[0099] According to this modified example, the setting unit 61, current feedback unit 63, UVW conversion unit 64, correction unit 67, estimation unit 67a, and U, V, and W phase calculation units 68U, 68V, and 68W can be omitted, which is expected to simplify the configuration of the control device 60.

[0100] In the above embodiment, the evaluation parameter may be omitted.

[0101] In the above embodiment, the machine learning method, i.e., the learning method of the learning model 70, may be changed as desired. For example, in the above embodiment, machine learning is performed so that the zero-phase current detection value I0r approaches zero. However, machine learning may be performed so that the harmonic currents included in the phase currents Iu, Iv, and Iw approach zero. Furthermore, machine learning may be performed so that the harmonic currents or the zero-phase currents are calculated or estimated from the d- and q-axis current values ​​Idr and Iqr, and so that these currents approach zero.

[0102] In the above embodiment, the following machine learning may be performed. As in the above embodiment, the subject of this machine learning may be any of the electronic computer, the update unit 80, and the cloud server 320. Here, the description will be focused on the electronic computer. As shown in FIG. 10 , the electronic computer has the same function as the estimation unit 67a. The estimation unit 67a of the electronic computer inputs the various parameters (Id*, Iq*, ω, θr, Ior) described above to a learning model 70 before learning (before training) and causes the learning model 70 to estimate a correction amount Vcr. The electronic computer inputs this correction amount Vcr to a zero-phase current calculation unit 91 included in the electronic computer, which calculates a zero-phase current calculation value Ic. More specifically, the zero-phase current calculation unit 91 corrects the phase voltage command values ​​Vu*, Vv*, and Vw* based on the correction amount Vcr. The method for calculating the pre-correction phase voltage command values ​​Vu*, Vv*, Vw* from the d- and q-axis current command values ​​Id*, Iq* and the d- and q-axis currents Id, Iq, etc., and the method for calculating the post-correction phase voltage command values ​​Vu*, Vv*, Vw* from the correction amount Vcr and the pre-correction phase voltage command values ​​Vu*, Vv*, Vw* are as described above, and therefore will not be described here.

[0103] A zero-phase current calculation unit 91 of the computer calculates a zero-phase current calculation value Ic from the corrected phase voltage command values ​​Vu*, Vv*, and Vw*. An update unit 92 of the electronic computer calculates the error (such as a least square error) between the zero-phase current detection value I0r and the zero-phase current calculation value Ic, and updates at least one of the weighting and bias of the perceptron constituting the learning model 70 using a method such as the error backpropagation method so as to reduce this error.

[0104] In the above embodiment, a learning model 70 may be employed that has undergone machine learning to output the amplitude Ψz and phase θz of the harmonic voltage Vz as control information, and the amplitude Ψz and phase θz may be estimated by the learning model 70. The control device 60 may then calculate a correction amount (R*) for the voltage command value of each phase to be output to each inverter 20, 30, based on the amplitude Ψz and phase θz of the harmonic voltage Vz, so that the zero-sequence current of the winding falls within a tolerance range that includes zero. More specifically, the correction unit 67 included in the control device 60 calculates the correction amount R* for the voltage command value from Equation 3 described in the first embodiment, using the amplitude Ψz, phase θz, electrical angle θr, and electrical angular velocity ω of the harmonic voltage Vz. Then, the U, V, W phase calculation units 68U, 68V, 68W perform correction by subtracting the correction amount R* from each phase voltage command value Vu*, Vv*, Vw*, and control each inverter 20, 30 based on the corrected each phase voltage command value Vu*, Vv*, Vw*.

[0105] When generating the learning model 70 that estimates the amplitude Ψz and phase θz of the harmonic voltage Vz, the input parameters (condition parameters, evaluation parameters) are the same as those in the above-described embodiment and modified example. During machine learning, the third-order harmonic voltage Vz, which is the main component of the zero-phase current, is identified from the detected phase currents Iur, Ivr, Iwr, etc., and learning is performed so that the amplitude Ψz and phase θz of the harmonic voltage Vz match the amplitude Ψz and phase θz estimated by the learning model 70 (so that there is no error).

[0106] In the above embodiment, when performing machine learning, data obtained by experiment (zero-phase current detection value) or data obtained by simulation may be used. Also, a surrogate model may be used as the learning model 70.

[0107] In the above embodiment, if the control system 100 is not set to the Y drive state, the changeover switch 13 may be omitted.

[0108] The above-described embodiments and their modifications can be combined within the scope of possible combinations.

[0109] The controller and methods described herein may be implemented by a special-purpose computer configured with a processor and memory programmed to perform one or more functions embodied in a computer program. Alternatively, the controller and methods described herein may be implemented by a special-purpose computer configured with a processor configured with one or more dedicated hardware logic circuits. Alternatively, the controller and methods described herein may be implemented by one or more special-purpose computers configured with a processor and memory programmed to perform one or more functions in combination with a processor configured with one or more hardware logic circuits. Furthermore, the computer program may be stored as instructions executed by a computer on a computer-readable non-transitory storage medium.

[0110] The following describes characteristic configurations extracted from the above-described embodiments.

[0111] [Configuration 1] A control device (60) for a rotating electric machine applied to a control system (100) comprising: a rotating electric machine (40) having multi-phase windings (51U, 51V, 51W); a first inverter (20) connected to a DC power source (10) and connected to a first end of both ends of the windings of each phase; and a second inverter (30) connected to a second end of both ends of the windings of each phase, the control device comprising: a memory unit (60b) that stores a learning model (70) that outputs control information for controlling the operation of the first inverter and the second inverter when condition parameters including at least a control amount of the rotating electric machine are input; and a control unit (60a) that acquires the condition parameters, inputs the acquired condition parameters into the learning model, and controls the operation of the first inverter and the second inverter based on the control information output from the learning model.

[0112] [Configuration 2] The control device for a rotating electric machine according to Configuration 1, wherein the learning model is generated by machine learning so that the zero-phase sequence current (I0r) of the winding approaches zero.

[0113] [Configuration 3] The control device for a rotating electric machine according to Configuration 1 or 2, wherein the control unit acquires an evaluation parameter correlated with a zero-phase current of the winding, and inputs the evaluation parameter to the learning model in addition to the condition parameter.

[0114] [Configuration 4] The control device for a rotating electric machine according to any one of Configurations 1 to 3, wherein the control unit acquires a d-axis current detection value (Idr), or each phase current (Iu, Iv, Iw), or a zero-phase current detection value (I0r) as an evaluation parameter correlated with the zero-phase current of the winding, and inputs the evaluation parameter in addition to the condition parameter to the learning model.

[0115] [Configuration 5] A control device for a rotating electric machine according to any one of configurations 1 to 4, wherein the learning model is a model using a neural network with a multilayer perceptron structure, and the learning model is generated by updating at least one of the weighting and bias of each neuron constituting the learning model so as to reduce an error between a zero-phase current detection value identified from an evaluation parameter correlated with the zero-phase current flowing in the winding as a result of controlling the operation of the first inverter and the second inverter based on the control information, and a zero-phase current calculation value calculated based on the control information.

[0116] [Configuration 6] The control device for a rotating electric machine according to any one of configurations 1 to 5, wherein the control information includes a voltage command value or a correction amount (Vcr) of the voltage command value for each phase to be output to each of the inverters.

[0117] [Configuration 7] The control device for a rotary electric machine according to any one of Configurations 1 to 6, wherein the control information includes an amplitude (Ψz) and a phase (θz) of a harmonic voltage (Vz) generated in the rotary electric machine, and the control unit calculates a correction amount (R*) of a voltage command value for each phase to be output to each of the inverters based on the amplitude and phase included in the control information so that a zero-phase current of the winding falls within a tolerance range including zero, corrects the voltage command value for each phase based on the correction amount, and controls each of the inverters based on the corrected voltage command value for each phase.

[0118] [Configuration 8] The control device for a rotating electric machine according to any one of Configurations 1 to 7, wherein the condition parameters include a torque request value (Tr*) which is a control amount of the rotating electric machine, a q-axis current detection value (Iqr), an electrical angle (θr) of the rotating electric machine, and an electrical angular velocity (ω) of the rotating electric machine.

[0119] [Configuration 9] The control device for a rotating electric machine according to any one of Configurations 1 to 8, wherein the condition parameters include a d-axis current command value (Id*), a q-axis current command value (Iq*), an electrical angle (θr) of the rotating electric machine, and an electrical angular velocity (ω) of the rotating electric machine.

[0120] [Configuration 10] The control device for a rotating electric machine according to any one of configurations 1 to 9, wherein the condition parameters include a temperature or an air pressure of the rotating electric machine.

[0121] [Configuration 11] The control device for a rotating electric machine according to any one of Configurations 1 to 10, further comprising an update unit (80) that acquires an evaluation parameter correlated with a zero-phase current of the winding and updates the learning model based on the evaluation parameter.

[0122] [Configuration 12] A control device for a rotating electric machine according to any one of configurations 1 to 11, comprising a communication unit (340) that transmits an evaluation parameter correlated with the zero-phase current of the winding to an external device (320) via a communication network (330), receives the learning model updated by the external device, and updates the learning model stored in a memory unit.

[0123] [Configuration 13] The control device for a rotating electric machine according to Configuration 12, wherein the rotating electric machine is mounted on a vehicle, and the communication unit connects to the communication network while the vehicle is stopped and updates the learning model.

[0124] [Configuration 14] The first inverter has upper switches (SUHa, SVHa, SWHa) and lower switches (SULa, SVLa, SWLa) connected in series for each phase, and a connection point between the upper switches and the lower switches is connected to the first end of the winding for each phase; the second inverter has upper switches (SUHb, SVHb, SWHb) and lower switches (SULb, SVLb, SWLb) connected in series for each phase, and a connection point between the upper switches and the lower switches is connected to the second end of the winding for each phase; a high potential side connection line (11) connecting the high potential side of the first inverter and the high potential side of the second inverter; and a low potential side connection line (12) connecting the low potential side of the first inverter and the low potential side of the second inverter. A control device for a rotating electric machine according to any of configurations 1 to 13, further comprising: a switch (13) provided on at least one of the high potential side connecting line and the low potential side connecting line; and a mode switching unit (90) configured to switch between a first mode in which the switch is in an open state and performs switching drive of the upper and lower arm switches in one of the first and second inverters, and performs neutral point driving in which at least one of the upper and lower arm switches in the other inverter is maintained in an on state, and a second mode in which the switch is in a closed state and performs switching drive of the upper and lower arm switches in each of the inverters, when the control unit is switched to the second mode by the mode switching unit, controls the operation of the first inverter and the second inverter based on control information output from the learning model.

[0125] [Configuration 15] A control method for a rotating electric machine implemented by a control device (60) for a rotating electric machine applied to a control system (100) including: a rotating electric machine (40) having multi-phase windings (51U, 51V, 51W); a first inverter (20) connected to a DC power source (10) and connected to a first end of both ends of the windings of each phase; and a second inverter (30) connected to a second end of both ends of the windings of each phase, the control method comprising the steps of: acquiring condition parameters including at least a control amount of the rotating electric machine; inputting the condition parameters into a learning model (70) stored in a memory unit (60b); and acquiring control information for controlling the operation of the first inverter and the second inverter output from the learning model; and controlling the operation of the first inverter and the second inverter based on the control information output from the learning model.

[0126] [Configuration 16] A control program for a rotating electric machine implemented by a control device (60) for a rotating electric machine applied to a control system (100) including: a rotating electric machine (40) having multi-phase windings (51U, 51V, 51W); a first inverter (20) connected to a DC power source (10) and connected to a first end of both ends of the windings of each phase; and a second inverter (30) connected to a second end of both ends of the windings of each phase, the control program for a rotating electric machine executing the steps of: acquiring condition parameters including at least control variables of the rotating electric machine, inputting the condition parameters to a learning model (70) stored in a memory unit (60b), and acquiring control information for controlling the operation of the first inverter and the second inverter output from the learning model; and controlling the operation of the first inverter and the second inverter based on the control information output from the learning model.

[0127] Although the present disclosure has been described with reference to the embodiments, it is understood that the present disclosure is not limited to the embodiments or structures. The present disclosure also encompasses various modifications and equivalent modifications. In addition, various combinations and forms, including only one element, more than one element, or less than one element, are also within the scope and spirit of the present disclosure.

Claims

1. A control device (60) for a rotating electric machine applied to a control system (100) including: a rotating electric machine (40) having a multi-phase winding (51U, 51V, 51W); a first inverter (20) connected to a DC power source (10) and connected to a first end of the winding of each phase; and a second inverter (30) connected to a second end of the winding of each phase, the control device comprising: a memory unit (60b) that stores a learning model (70) that outputs control information for controlling the operation of the first inverter and the second inverter when a condition parameter including at least a control amount of the rotating electric machine is input; and a control unit (60a) that acquires the condition parameters, inputs the acquired condition parameters to the learning model, and controls the operation of the first inverter and the second inverter based on the control information output from the learning model, the learning model being generated by machine learning so that the zero-phase current (I0r) of the winding approaches zero.

2. The control device for a rotating electric machine according to claim 1, wherein the control unit acquires an evaluation parameter correlated with the zero-phase current of the winding, and inputs the evaluation parameter to the learning model in addition to the condition parameter.

3. The control unit of the rotating electric machine control device according to claim 1, wherein the control unit acquires a d-axis current detection value (Idr), or each phase current (Iur, Ivr, Iwr), or a zero-phase current detection value (I0r) as an evaluation parameter correlated with the zero-phase current of the winding, and inputs the evaluation parameter in addition to the condition parameter to the learning model.

4. A control device for a rotating electric machine as described in claim 3, wherein the learning model is a model using a neural network with a multi-layer perceptron structure, and the learning model is generated by updating at least one of the weighting and bias of each neuron constituting the learning model so as to reduce an error between a zero-phase current detection value identified from an evaluation parameter correlated with the zero-phase current flowing in the winding as a result of controlling the operation of the first inverter and the second inverter based on the control information, and a zero-phase current calculation value calculated based on the control information.

5. A control device for a rotating electric machine according to any one of claims 1 to 4, wherein the control information includes a voltage command value or a correction amount (Vcr) of the voltage command value for each phase to be output to each of the inverters.

6. A control device for a rotating electric machine as described in any one of claims 1 to 4, wherein the control information includes an amplitude (Ψz) and a phase (θz) of a harmonic voltage (Vz) generated in the rotating electric machine, and the control unit calculates a correction amount (R*) of a voltage command value of each phase to be output to each of the inverters based on the amplitude and phase included in the control information so that the zero-sequence current of the winding is within a tolerance range including zero, corrects the voltage command value of each phase based on the correction amount, and controls each of the inverters based on the corrected voltage command value of each phase.

7. A control device for a rotating electric machine as described in any one of claims 1 to 4, wherein the condition parameters include a torque request value (Tr*), which is a control variable of the rotating electric machine, a q-axis current detection value (Iqr), an electrical angle (θr) of the rotating electric machine, and an electrical angular speed (ω) of the rotating electric machine.

8. A control device for a rotating electric machine as described in any one of claims 1 to 4, wherein the condition parameters include a d-axis current command value (Id*), a q-axis current command value (Iq*), an electrical angle (θr) of the rotating electric machine, and an electrical angular velocity (ω) of the rotating electric machine.

9. The control device for a rotating electric machine according to any one of claims 1 to 4, wherein the condition parameters include a temperature or an air pressure of the rotating electric machine.

10. A control device for a rotating electric machine as described in any one of claims 1 to 4, comprising an update unit (80) that acquires an evaluation parameter correlated with the zero-phase current of the winding and updates the learning model based on the evaluation parameter.

11. A control device for a rotating electric machine as described in any one of claims 1 to 4, comprising a communication unit (340) that transmits evaluation parameters correlated with the zero-phase current of the winding to an external device (320) via a communication network (330), receives the learning model updated by the external device, and updates the learning model stored in the memory unit.

12. The control device for a rotating electric machine as described in claim 11, wherein the rotating electric machine is mounted on a vehicle, and the communication unit connects to the communication network while the vehicle is stopped and updates the learning model.

13. The first inverter has upper arm switches (SUHa, SVHa, SWHa) and lower arm switches (SULa, SVLa, SWLa) connected in series for each phase, and a connection point between the upper arm switches and the lower arm switches is connected to the first end of the winding for each phase; the second inverter has upper arm switches (SUHb, SVHb, SWHb) and lower arm switches (SULb, SVLb, SWLb) connected in series for each phase, and a connection point between the upper arm switches and the lower arm switches is connected to the second end of the winding for each phase; a high potential side connection line (11) connecting the high potential side of the first inverter and the high potential side of the second inverter; and a low potential side connection line (12) connecting the low potential side of the first inverter and the low potential side of the second inverter. A control device for a rotating electric machine according to any one of claims 1 to 4, further comprising: a switch (13) provided on at least one of the high potential side connecting line and the low potential side connecting line; and the control unit comprises a mode switching unit (90) that switches between a first mode in which the switch is in an open state, and neutral point driving is performed in which switching drive of upper and lower arm switches in one of the first inverter and the second inverter is performed and at least one of the upper and lower arm switches in the other inverter is maintained in an on state, and a second mode in which the switch is in a closed state and switching drive of the upper and lower arm switches in each of the inverters is performed; and when the control unit is switched to the second mode by the mode switching unit, the control unit controls the operation of the first inverter and the second inverter based on control information output from the learning model.

14. A method for controlling a rotating electric machine implemented by a rotating electric machine control device (60) applied to a control system (100) including: a rotating electric machine (40) having multi-phase windings (51U, 51V, 51W); a first inverter (20) connected to a DC power source (10) and connected to a first end of both ends of the windings of each phase; and a second inverter (30) connected to a second end of both ends of the windings of each phase, the method including the steps of: acquiring condition parameters including at least a control amount of the rotating electric machine, inputting the condition parameters to a learning model (70) stored in a memory unit (60b), and acquiring control information for controlling the operation of the first inverter and the second inverter, the control information being output from the learning model; and controlling the operation of the first inverter and the second inverter based on the control information output from the learning model, wherein the learning model is generated by machine learning so that a zero-phase current (I0r) of the winding approaches zero.

15. A control program for a rotating electric machine implemented by a control device (60) for a rotating electric machine applied to a control system (100) including a rotating electric machine (40) having a multi-phase winding (51U, 51V, 51W), a first inverter (20) connected to a DC power source (10) and connected to a first end of the winding of each phase, and a second inverter (30) connected to a second end of the winding of each phase, the control program comprising the steps of: acquiring condition parameters including at least a control amount of the rotating electric machine, inputting the condition parameters to a learning model (70) stored in a memory unit (60b), and acquiring control information for controlling the operation of the first inverter and the second inverter output from the learning model; and controlling the operation of the first inverter and the second inverter based on the control information output from the learning model, the learning model being generated by machine learning so that a zero-phase current (I0r) of the winding approaches zero.

Images