Drive control method and system for interior permanent magnet synchronous motor

Abstract

The application discloses a driving control method and system of an interior permanent magnet synchronous motor, adopts a three-level neutral-point-clamped inverter to drive the interior permanent magnet synchronous motor, and comprises the following steps: generating a reference current according to a motor speed requirement; based on a limited set, using an extended state observer to respectively perform two-step current prediction on stator currents of the interior permanent magnet synchronous motor under different inverter switch states; based on a low-pass filter, separating a fundamental component and a harmonic component of a measured stator current of the interior permanent magnet synchronous motor; based on a preset cost function, calculating a cost value corresponding to each switch state, and selecting a switch state with the minimum cost value as an optimal switch state; based on the optimal switch state, generating a control instruction, and controlling the three-level neutral-point-clamped inverter to generate a three-phase voltage according to the control instruction, so as to drive the interior permanent magnet synchronous motor. The method can adaptively suppress harmonics and balance the neutral-point voltage, and improves the current waveform quality and system robustness.

Description

Technical Field

[0001] This application relates to the field of motor control, and in particular to a drive control method and system for a built-in permanent magnet synchronous motor. Background Technology

[0002] Built-in permanent magnet synchronous motors (IPMSMs) are widely used in new energy vehicles, industrial servo systems and other fields due to their advantages such as high power density and wide speed range.

[0003] Unlike surface-mount motors, IPMSMs exhibit salient pole effects (unequal d-axis and q-axis inductances), making the cross-coupling terms in their mathematical model more complex, and parameter variations significantly impact control performance. Three-level neutral-point clamped (NPC) inverters are widely used in medium- and high-power motor drive systems due to their lower device voltage stress, superior output voltage waveform quality, and lower total harmonic distortion (THD). However, the application of NPC inverters in embedded permanent magnet synchronous motors introduces a DC bus neutral-point voltage balancing control problem. Imbalance in the upper and lower DC bus capacitor voltages leads to an asymmetrical output voltage waveform and exacerbates current harmonic distortion. Furthermore, in systems combining NPC inverters with embedded permanent magnet synchronous motors, the coupling relationship between switching frequency penalty and harmonic suppression becomes more complex, making it difficult to achieve a balance between steady-state harmonics and dynamic response over a wide operating range. Summary of the Invention

[0004] In a first aspect, this application provides a drive control method for an embedded permanent magnet synchronous motor, which employs a three-level neutral point clamping inverter to drive the embedded permanent magnet synchronous motor. The drive control method includes: A reference current is generated based on the motor speed requirement; Based on a finite set, an extended state observer is used to perform two-step current prediction on the stator current of the built-in permanent magnet synchronous motor under different inverter switching states, and obtain the corresponding two-step predicted current values ​​under different switching states. Based on low-pass filtering, the fundamental and harmonic components of the measured quantum current of the built-in permanent magnet synchronous motor are separated to obtain the harmonic components. The cost value corresponding to each switching state is calculated based on a preset cost function. The switching state that minimizes the cost value is selected as the optimal switching state. The preset cost function includes a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term. The two-step current prediction error term is configured to track the error between the two-step predicted current value and the reference current. The adaptive harmonic suppression term is configured to adjust the current tracking accuracy using adaptive weights. The adaptive weights are negatively correlated with the harmonic components. The midpoint voltage balance term is configured to penalize the midpoint potential deviation of each candidate switching state. Based on the optimal switching state, control commands are generated. According to the control commands, the three-level neutral point clamping inverter is controlled to generate three-phase voltage to drive the built-in permanent magnet synchronous motor.

[0005] In one embodiment, the cost function further includes a linearly superimposed adaptive switching frequency reduction term, which is configured to consist of the product of the number of switching actions and the adaptive switching penalty coefficient. The number of switching actions represents the total number of transitions in the switching states of each phase of the inverter within the current control cycle relative to the previous moment. The adaptive switching penalty coefficient is modulated by a harmonic current suppression factor, which is configured as follows: When the amplitude of the harmonic component exceeds the preset harmonic current threshold, the harmonic current suppression factor is less than 1 to reduce the value of the adaptive switching penalty coefficient, thereby allowing for increased switching action to prioritize the suppression of harmonics. When the amplitude of the harmonic component does not exceed the preset harmonic current threshold, the harmonic current suppression factor remains at 1, so that the adaptive switching penalty coefficient is maintained at the reference level determined by the mechanical angular velocity, so as to reduce switching losses first.

[0006] In one embodiment, the adaptive switching penalty coefficient is configured to be determined by the product of the base switching penalty weight, the frequency factor related to the rotational speed, the absolute value of the motor mechanical angular velocity, and the harmonic current suppression factor.

[0007] In one embodiment, the cost function further includes a linearly superimposed current smoothing penalty term, which is configured to calculate the current change acceleration based on the second-order difference of the two-step predicted current values, and use the current change acceleration as a penalty index to suppress high-frequency oscillating harmonics.

[0008] In one embodiment, the adaptive weights are represented by the following formula: ; In the formula, express Axis current tracking error weighting, express The reference current value of the shaft. This indicates that the state obtained through the extended state observer The predicted current value of the axis at time k+1. express Axial harmonic component amplitude, , This represents a very small positive number used to prevent the denominator from being zero.

[0009] In one embodiment, the extended state observer is represented by the following formula: ; In the formula, Sampling time, Let k be the predicted current at time k. Let k be the estimated value of the lumped disturbance at time k. Let K be the measured quantum current at time k. ， , To ensure uniform motor angular velocity tuning, This indicates that it is based on the nominal inductance. Gain, This represents the voltage vector at time k.

[0010] In one embodiment, based on the predicted current at time k+1 generated by the extended state observer, a two-step current prediction is performed using the following formula to obtain the two-step current prediction value: ; In the formula, This represents the two-step current prediction value. This represents the predicted current at time k+1 generated by the extended state observer. This represents the lumped perturbation at time k+1 generated by the extended state observer. Indicates based on nominal inductance Gain, This represents the voltage vector corresponding to the candidate switch state.

[0011] Secondly, this application also provides a drive control system for a built-in permanent magnet synchronous motor. The drive control system uses a three-level neutral point clamping inverter to drive the built-in permanent magnet synchronous motor, including: The speed outer loop controller is configured to generate a reference current based on the motor speed requirement; The current inner loop predictive controller is configured to perform two-step current prediction on the stator current of the built-in permanent magnet synchronous motor under different inverter switching states based on a finite set and using an extended state observer to obtain the corresponding two-step predicted current values ​​under different switching states. The low-pass filter harmonic extraction module is configured to separate the fundamental component and harmonic component of the measured quantum current of the built-in permanent magnet synchronous motor based on low-pass filtering to obtain the harmonic component. The multi-objective optimization decision module is configured to calculate the cost value corresponding to each switching state based on a preset cost function, and select the switching state that minimizes the cost value as the optimal switching state. The preset cost function includes a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term. The two-step current prediction error term is configured to track the error between the two-step predicted current value and the reference current. The adaptive harmonic suppression term is configured to adjust the current tracking accuracy using adaptive weights, and the adaptive weights are negatively correlated with the harmonic components. The midpoint voltage balance term is configured to penalize the midpoint potential deviation of each candidate switching state. The drive execution module generates control commands based on the optimal switching state, and controls the three-level neutral point clamping inverter to generate three-phase voltage according to the control commands to drive the built-in permanent magnet synchronous motor.

[0012] In one embodiment, the cost function further includes a linearly superimposed adaptive switching frequency reduction term, which is configured to consist of the product of the number of switching actions and the adaptive switching penalty coefficient. The number of switching actions represents the total number of transitions in the switching states of each phase of the inverter within the current control cycle relative to the previous moment. The adaptive switching penalty coefficient is determined by the product of the basic switching penalty weight, the speed-related frequency factor, the absolute value of the motor mechanical angular velocity, and the harmonic current suppression factor. The harmonic current suppression factor is configured as follows: When the amplitude of the harmonic component exceeds the preset harmonic current threshold, the harmonic current suppression factor is less than 1 to reduce the value of the adaptive switching penalty coefficient, thereby allowing for increased switching action to prioritize the suppression of harmonics. When the amplitude of the harmonic component does not exceed the preset harmonic current threshold, the harmonic current suppression factor remains at 1, so that the adaptive switching penalty coefficient is maintained at the reference level determined by the mechanical angular velocity, so as to reduce switching losses first.

[0013] The aforementioned drive control method for the built-in permanent magnet synchronous motor generates a reference current based on speed requirements; employs an extended state observer to perform two-step current prediction for all candidate switching states, obtaining two-step current predictions under different switching states; and separates harmonic components from the measured current using low-pass filtering. A preset cost function is constructed to calculate the cost function value for each candidate switching state, and the candidate switching state with the smallest cost function value is selected as the optimal switching state output. This cost function includes at least a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term, which are used to ensure current tracking accuracy, dynamically adjust control weights according to harmonic magnitude to suppress harmonics, and penalize the impact of switching actions on the DC side midpoint potential, respectively. This method fully considers the special characteristics of combining a three-level NPC with an IPMSM system. It achieves robust current prediction independent of system parameters based on an extended state observer, avoiding the influence of the d- and q-axis salient pole effects of the IPMSM system on the current prediction error term in the subsequent cost function. In addition, the method provided in this application also considers the midpoint voltage balance target in the cost function, enabling the three-level NPC to adapt to the IPMSM system. At the same time, an adaptive harmonic suppression term is introduced to dynamically adjust the harmonic state. By setting an adaptive weight that is negatively correlated with the harmonic components, the method automatically strengthens harmonic suppression when harmonic distortion is severe and relaxes harmonic constraints to reduce switching losses when the harmonics meet the requirements. Thus, it is possible to simultaneously optimize current quality and switching losses over a wide load and speed range. Attached Figure Description

[0014] Figure 1 This is a flowchart of a drive control method for a built-in permanent magnet synchronous motor in one embodiment; Figure 2 This is a topology diagram of a three-level midpoint clamped inverter in one embodiment; Figure 3 This is a schematic diagram of 19 different voltage vectors in one embodiment; Figure 4 This is a framework diagram of the drive control system for a built-in permanent magnet synchronous motor in one embodiment. Figure 5 This is a framework diagram of the drive control system for a built-in permanent magnet synchronous motor in another embodiment. Detailed Implementation

[0015] The present application will be described in detail below with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present application. Any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the protection scope of the present application.

[0016] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0017] In one embodiment, such as Figure 1 As shown, a drive control method for an embedded permanent magnet synchronous motor is provided, which uses a three-level neutral point clamping inverter to drive the embedded permanent magnet synchronous motor. The method includes the following steps: Step 101: Generate a reference current based on the motor speed requirement; Specifically, the speed outer-loop PI controller receives the actual speed of the built-in permanent magnet synchronous motor and compares it with the given speed command to generate a speed error signal. The speed outer-loop PI controller then generates a q-axis reference current based on this speed error signal. And set the d-axis reference current This is to achieve maximum torque-to-current ratio control.

[0018] Step 102: Based on the finite set, the extended state observer is used to perform two-step current prediction on the stator current of the built-in permanent magnet synchronous motor under different inverter switching states, and obtain the corresponding two-step predicted current values ​​under different switching states. Specifically, at each sampling time k, the Extended State Observer (ESO) uses the motor stator current and the applied voltage vector actually sampled at time k to estimate the current prediction and lumped disturbance estimate at time k+1 online.

[0019] Based on the current prediction and lumped disturbance estimate at time k+1 estimated by the extended state observer, a traversal prediction is performed for all 27 candidate switching states of the three-level neutral-point clamp (NPC) inverter. For the voltage vector corresponding to each candidate state, the voltage vector corresponding to each candidate state, along with the current prediction and lumped disturbance estimate at time k+1 estimated by the extended state observer, are substituted into the discrete prediction model to calculate one control cycle backward, thereby predicting the d-axis and q-axis stator current predictions of the motor at time k+2 under different switching states, and thus obtaining the two-step predicted current values ​​of the motor under different switching states.

[0020] Step 103: Based on low-pass filtering, separate the fundamental component and harmonic component of the measured quantum current of the built-in permanent magnet synchronous motor to obtain the harmonic component. Specifically, the actual measured stator currents of the motor along the d-axis and q-axis are collected, and low-pass filtering is applied to both. The low-pass filter effectively attenuates high-frequency harmonic components in the stator current signal, thus preserving and outputting the fundamental components of the current, namely the d-axis fundamental current and the q-axis fundamental current.

[0021] Furthermore, the original measured quantum current values ​​of the d-axis and q-axis collected at the same time are subtracted from the fundamental current of the d-axis and the fundamental current of the q-axis extracted by the low-pass filter, respectively. The resulting difference is the d-axis harmonic component and the q-axis harmonic component.

[0022] Step 104: Calculate the cost value corresponding to each switching state based on the preset cost function, and select the switching state that minimizes the cost value as the optimal switching state. The preset cost function includes a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term. The two-step current prediction error term is configured to track the error between the two-step predicted current value and the reference current. The adaptive harmonic suppression term is configured to adjust the current tracking accuracy using adaptive weights. The adaptive weights are negatively correlated with the harmonic components. The midpoint voltage balance term is configured to penalize the midpoint potential deviation of each candidate switching state. Specifically, for each candidate switch state, the corresponding two-step predicted current value is obtained, i.e., the two-step predicted current values ​​along the d-axis and q-axis obtained after two-step current prediction based on the extended state observer. The error values ​​between the two-step predicted current value along the d-axis and the reference current along the d-axis, and the error values ​​between the two-step predicted current value along the q-axis and the reference current along the q-axis are calculated separately. The cost function of the two-step current prediction error term is the sum of the squares of the d-axis current error value and the squares of the q-axis current error value.

[0023] The d-axis and q-axis harmonic components are separated in real time from the measured currents along the d and q axes using a low-pass filter. Adaptive weights are calculated, which are inversely proportional to the amplitude of the corresponding harmonic component. When the harmonic component of a certain axis is large, the adaptive weight for that axis decreases, causing the optimizer to prioritize switching actions that effectively suppress that harmonic component. Conversely, when the harmonic component of a certain axis is small, the adaptive weight for that axis increases to restore the required current tracking accuracy. The cost function of the adaptive harmonic suppression term consists of the sum of squared current errors weighted by the adaptive weights.

[0024] For each candidate switching state, the impact of applying that switching state on the DC-side midpoint potential of the three-level NPC inverter is predicted, and the midpoint potential deviation corresponding to each candidate switching state is calculated. The cost function of the midpoint voltage balance term is the absolute value of this midpoint potential deviation. When evaluating candidate switching states, the midpoint voltage balance is incorporated into a unified cost function optimization, eliminating the need for a separate midpoint balance controller and avoiding parameter coordination issues between two controllers.

[0025] In each control cycle, for all 27 candidate switch states, the values ​​of the three cost functions mentioned above are calculated, and the three cost function values ​​are added together to obtain the total cost function value. The candidate switch state with the smallest total cost function value is selected as the optimal switch state for the current cycle.

[0026] Step 105: Generate control commands based on the optimal switching state, and control the three-level neutral point clamping inverter to generate three-phase voltage according to the control commands to drive the built-in permanent magnet synchronous motor.

[0027] Specifically, after evaluating the preset cost functions of all 27 candidate switching states, the candidate switching state that minimizes the cost function value is compared and selected, and this candidate switching state is determined as the optimal switching state for the current control cycle. The optimal switching state This is a combination containing information from the three-phase switches. This optimal switch state vector is output to the gate drive circuit of the three-level neutral-point clamp (NPC) inverter.

[0028] The inverter's gate drive circuit determines the optimal switching state received. At the start of the next sampling cycle, the corresponding power switching transistor (such as IGBT) is turned on and off, thereby generating the corresponding three-phase AC voltage at the inverter output.

[0029] In this embodiment, the method generates a reference current based on the rotational speed requirement; it uses an extended state observer to perform two-step current prediction on all candidate switching states, obtaining two-step current predictions for different switching states; and it separates harmonic components from the measured current using low-pass filtering. A preset cost function is constructed to calculate the cost function value for each candidate switching state, and the candidate switching state with the smallest cost function value is selected as the optimal switching state output. This cost function includes at least a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term, which are used to ensure current tracking accuracy, dynamically adjust control weights according to harmonic magnitude to suppress harmonics, and penalize the impact of switching actions on the DC side midpoint potential, respectively. This method can achieve parameter-independent robust prediction based on the extended state observer, and through the preset cost function, dynamically adjust adaptive weights according to the real-time harmonic state to coordinate waveform quality and current tracking accuracy, comprehensively optimizing current quality and harmonic suppression effects.

[0030] In one embodiment, the topology of a three-level neutral-point clamp (NPC) inverter is as follows: Figure 2 As shown, each phase arm of the bridge can output three level states, which are defined respectively. ( The DC side is powered by power supply V. dc The power supply is filtered and voltage divided by two series-connected supporting capacitors C1 and C2, with the midpoint denoted as n. The inverter section contains three identical bridge arms (bridge arms 1, 2, and 3), each phase bridge arm consisting of four power switching transistors (S) with anti-parallel diodes. a1 -S a4 It is composed of three phases and can output three switching states: P, O, and N; the three-phase output voltage is V. a v b v c and the three-phase output current i a i b i c Input to the built-in permanent magnet synchronous motor (IPMSM).

[0031] For a three-level neutral-point clamp (NPC) inverter, there are a total of 3³ = 27 candidate states for the switching state combinations. These candidate states correspond to the following in the space vector plane: Figure 3 The diagram shows 19 different voltage vectors (including zero vector, small vector, medium vector, and large vector). Figure 3As shown, each three-phase label composed of the letters P, O, and N (such as "POO" and "ONN") represents a specific combination of switching states, where P, O, and N represent the positive, zero, and negative output levels of each phase bridge arm, respectively. These vectors are symmetrically distributed radially around the origin (the zero vector corresponding to the OOO state) and can be classified into zero vectors, small vectors (each corresponding to two switching states, such as POO and ONN), medium vectors, and large vectors based on their amplitude and position.

[0032] The midpoint current is: .

[0033] No. The midpoint voltage deviation under each candidate state is: ; in, , These are the voltages of the upper and lower DC bus capacitors, respectively. For DC bus capacitors, Sampling time.

[0034] In one embodiment, an integrated permanent magnet synchronous motor (IPMSM) is used. The dynamic equation for shaft current can be rewritten in lumped disturbance form: .

[0035] in, Based on nominal inductance Gain, This is the lumped disturbance term, which includes all parameter uncertainties. , , , The terms include (variation), cross-coupling terms, and back electromotive force terms. It should be noted that precise knowledge of each specific parameter is not required during the calculation.

[0036] against axis, A second-order linear extended state observer is designed for each axis, and the predicted current at time k+1 is obtained after discretization. and lumped disturbance estimate The expression is as follows: ; In the formula, Sampling time, Let k be the predicted current at time k. Let k be the estimated value of the lumped disturbance at time k. Let K be the measured quantum current at time k. ， , To ensure uniform motor angular velocity tuning, This indicates that it is based on the nominal inductance. Gain, This represents the voltage vector at time k.

[0037] In this embodiment, the extended state observer only needs the actual sampled motor stator current and the applied voltage vector, without any motor parameters, and has inherent robustness to the time-varying parameters caused by the salient pole effect of the built-in permanent magnet synchronous motor.

[0038] In one embodiment, the predicted current at time k+1 is based on the extended state observer. and lumped disturbance estimate , for the Voltage vector corresponding to each candidate switch state The two-step current prediction value is obtained by performing the following formula: ; In the formula, This represents the two-step current prediction value. This represents the predicted current at time k+1 generated by the extended state observer. This represents the lumped perturbation at time k+1 generated by the extended state observer. Indicates based on nominal inductance Gain, This represents the voltage vector corresponding to the candidate switch state.

[0039] It should be noted that this two-step prediction mechanism compensates for a one-cycle calculation delay, thereby ensuring that the selected optimal switching vector minimizes the current tracking error in the next cycle after actual application. Simultaneously, since the d-axis and q-axis predictions are performed independently based on estimates from the extended state observer, this process includes compensation for the salient pole effect of the built-in permanent magnet synchronous motor.

[0040] In one embodiment, the preset cost function includes a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term; the two-step current prediction error term is configured to track the error between the two-step predicted current value and the reference current.

[0041] Specifically, the two-step predicted current values ​​along the d-axis are calculated separately. With d-axis reference current The error value between them, and the two-step predicted current value along the q-axis. With q-axis reference current The error value between the two-step current prediction terms. The cost of the two-step current prediction error term is the sum of the squares of the d-axis current error value and the squares of the q-axis current error value, and thus the two-step current prediction error term can be obtained. The cost function is calculated using the following formula: .

[0042] In one embodiment, the adaptive harmonic suppression term is configured to adjust the current tracking accuracy with adaptive weights that are negatively correlated with the harmonic components.

[0043] Specifically, the fundamental component is extracted from the measured d-axis and q-axis current values ​​using a first-order low-pass filter. : .

[0044] Harmonic components are The adaptive weights are represented by the following formula: ; In the formula, express Axis current tracking error weighting, express The reference current value of the shaft. This indicates that the state obtained through the extended state observer The predicted current value of the axis at time k+1. express Axial harmonic component amplitude, , Based on the weighting coefficient, This represents a very small positive number used to prevent the denominator from being zero.

[0045] When harmonic components When the weight is large, adaptive weight The optimizer automatically reduces harmonic suppression by prioritizing switching states that effectively mitigate harmonics; when harmonics are low, their weights are automatically increased to ensure current tracking accuracy. This adaptive mechanism enables real-time adjustment of harmonic suppression without manual intervention. , The adaptive weights are calculated separately for each axis, overcoming the inadequacy of the uniform weights in the traditional scheme for the salient pole effect.

[0046] The cost of the adaptive harmonic suppression term is composed of the sum of squared current errors after adaptive weighting. The cost function is calculated using the following formula: .

[0047] In one embodiment, the midpoint voltage balance term is configured to penalize midpoint potential deviations in each candidate switching state.

[0048] Specifically, for each candidate switching state, the impact of applying that switching state on the DC-side neutral point potential of the three-level NPC inverter is predicted, and the neutral point potential deviation corresponding to each candidate switching state is calculated. Midpoint voltage balance term The cost function value is the predicted midpoint potential deviation. The absolute value of the midpoint voltage balance term The cost function is calculated using the following formula: .

[0049] in, It is the preset midpoint voltage balance weight.

[0050] In one embodiment, the cost function further includes a linearly superimposed adaptive switching frequency reduction term, which is configured to consist of the product of the number of switching actions and the adaptive switching penalty coefficient. The number of switching actions represents the total number of transitions in the switching states of each phase of the inverter within the current control cycle relative to the previous moment. The adaptive switching penalty coefficient is modulated by a harmonic current suppression factor, which is configured as follows: When the amplitude of the harmonic component exceeds the preset harmonic current threshold, the harmonic current suppression factor is less than 1 to reduce the value of the adaptive switching penalty coefficient, thereby allowing for increased switching action to prioritize the suppression of harmonics. When the amplitude of the harmonic component does not exceed the preset harmonic current threshold, the harmonic current suppression factor remains at 1, so that the adaptive switching penalty coefficient is maintained at the reference level determined by the mechanical angular velocity, so as to reduce switching losses first.

[0051] Specifically, the switching state (P, O, N) of each phase (a, b, c) in the candidate state is compared with the actual switching state of that phase at the previous moment. If the states are different, it is counted as one transition. The transition counts of all three phases are summed, and the total number of transition counts is the number of switching actions corresponding to that candidate state.

[0052] Furthermore, the adaptive switching penalty coefficient is calculated. Adaptive switching penalty coefficient It is a weighting coefficient that changes dynamically due to the modulation of the harmonic current suppression factor, and an adaptive switching penalty coefficient. The calculation formula is: .

[0053] in, It is the preset basic switch penalty weight; It is a frequency factor related to rotational speed; It is the absolute value of the mechanical angular velocity of the motor. It is a harmonic current suppression factor.

[0054] The working mechanism is as follows: It is a preset harmonic current threshold. It is the amplitude of the harmonic current at the current moment.

[0055] When the harmonic current amplitude Exceeding the threshold Harmonic current suppression factor The value is less than 1, which leads to an adaptive switching penalty coefficient. The value is reduced. This causes the optimizer to automatically relax the penalty for the number of switching operations during evaluation, tending to select states that are more effective in suppressing harmonics (more frequent switching operations) and prioritizing the quality of the current waveform.

[0056] When the harmonic current amplitude Not exceeding the threshold Harmonic current suppression factor The value remains at 1. At this point, the switching penalty coefficient... Mainly composed of basic switch penalty weights Frequency factor related to rotational speed The absolute value of the mechanical angular velocity of the motor and harmonic current suppression factor The decision at this point is to prioritize reducing switching losses. This is achieved by sensing the amplitude of harmonic currents in real time. With harmonic current threshold The magnitude of the signal is dynamically adjusted to regulate the switching frequency intensity, thereby achieving automatic coordination between harmonic suppression and switching losses.

[0057] The number of switching actions calculated above is compared with the adaptive switching penalty coefficient. Multiplying these terms yields the adaptive switching frequency reduction term for the i-th candidate switching state. The formula for calculating the cost function is as follows: .

[0058] In one embodiment, the cost function further includes a linearly superimposed current smoothing penalty term, which is configured to calculate the current change acceleration based on the second-order difference of the two-step predicted current values, and use the current change acceleration as a penalty index to suppress high-frequency oscillating harmonics.

[0059] Specifically, for the i-th candidate switch state, the two-step predicted current value corresponding to the i-th candidate switch state is used. (where g represents the d-axis or q-axis), combined with the current value estimated at time k+1 by the extended state observer. and the measured current value at time k. The second-order differences of the d-axis and q-axis currents are calculated separately to obtain the current-current accelerations along the d-axis and q-axis. The calculation formulas are as follows: .

[0060] The sum of the squares of the calculated second-order difference values ​​of the d-axis and q-axis currents is then multiplied by a preset current smoothing weighting coefficient. The current smoothness penalty term is obtained. The cost function is calculated using the following formula: .

[0061] Current smoothness penalty term By penalizing abrupt changes in the rate of change of current, high-frequency oscillating harmonics caused by switching transitions are suppressed from a time-domain perspective, complementing the adaptive harmonic suppression term.

[0062] In one embodiment, the cost function further includes a linearly superimposed voltage tracking term, which is configured to evaluate the error between the voltage vector corresponding to the candidate switch state and the reference voltage. The reference voltage is calculated based on the two-step predicted current value obtained through the extended state observer, combined with the reference current, according to the deadbeat principle.

[0063] Specifically, based on the current prediction value of the extended state observer at time k+1 With lumped disturbance estimate Given reference current The reference voltage is calculated using the deadbeat principle. : .

[0064] The reference voltage This represents the optimal voltage required for the current to accurately track the reference value at the next moment, and will serve as the benchmark for the voltage tracking term in the cost function. Because the reference voltage... The system generates parameters from extended state observer estimates rather than relying on precise motor parameters, resulting in significantly better robustness than traditional deadbeat control.

[0065] Furthermore, for the i-th candidate switch state, the corresponding voltage vector is: Voltage tracking term The cost function is the voltage vector corresponding to the candidate switching state. With reference voltage The square of the difference is calculated using the following formula: .

[0066] In this embodiment, the reference voltage is quickly selected from all 27 candidate voltage vectors. The closest voltage vector. Due to the reference voltage. Generated by an extended state observer that is insensitive to parameter variations, it does not depend on precise motor model parameters, and therefore uses a reference voltage. The voltage tracking term evaluation based on this is more robust.

[0067] The optimal choice is the preset cost function. The value is the above six sub-cost voltage tracking items. Two-step current prediction error term Adaptive harmonic suppression term Current smoothness penalty term Adaptive switching frequency reduction term and the midpoint voltage balance term The sum of linear superpositions: .

[0068] This cost function integrates six objectives—voltage tracking, two-step current prediction, adaptive harmonic suppression, current smoothing, adaptive switching reduction, and midpoint voltage balance—into the FCS-PCC cost function framework, making it suitable for three-level NPC inverters driving built-in permanent magnet synchronous motors.

[0069] In one embodiment, a comparative experiment was conducted using a 5.8kW built-in permanent magnet synchronous motor experimental platform under a rated load condition of 300r / min. The experimental results show that, compared with traditional predictive control schemes that only include basic control objectives, the method provided in this embodiment can reduce the total harmonic distortion (THD) of the output phase current from 23.27% to 15.16%, a relative reduction of up to 34.8%, demonstrating the effectiveness of the preset cost function in improving the current waveform quality.

[0070] It should be noted that the performance improvement is partly attributed to the third-order Lagrange extrapolation delay compensation technology used in the embodiments of this application. Compared with the traditional linear extrapolation method, this method can achieve higher accuracy rotor position angle prediction under high-speed and acceleration / deceleration dynamic operation conditions of the built-in permanent magnet synchronous motor, thereby effectively reducing the degradation of current prediction quality caused by coordinate transformation errors and further suppressing current harmonic distortion in the dynamic process.

[0071] Based on the same concept, such as Figure 4 As shown, this application also provides a drive control system for a built-in permanent magnet synchronous motor. The drive control system uses a three-level neutral point clamping inverter to drive the built-in permanent magnet synchronous motor. The system includes: The speed outer loop controller 401 is configured to generate a reference current based on the motor speed requirement; The current inner loop predictive controller 402 is configured to perform two-step current prediction on the stator current of the built-in permanent magnet synchronous motor under different inverter switching states based on a finite set and using an extended state observer to obtain the corresponding two-step predicted current values ​​under different switching states. The low-pass filter harmonic extraction module 403 is configured to separate the fundamental component and harmonic component of the measured quantum current of the built-in permanent magnet synchronous motor based on low-pass filtering to obtain the harmonic component. The multi-objective optimization decision module 404 is configured to calculate the cost value corresponding to each switching state based on a preset cost function, and select the switching state that minimizes the cost value as the optimal switching state. The preset cost function includes a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term. The two-step current prediction error term is configured to track the error between the two-step predicted current value and the reference current. The adaptive harmonic suppression term is configured to adjust the current tracking accuracy using adaptive weights. The adaptive weights are negatively correlated with the harmonic components. The midpoint voltage balance term is configured to penalize the midpoint potential deviation of each candidate switching state. The drive execution module 405 generates control commands based on the optimal switching state, and controls the three-level neutral point clamping inverter to generate three-phase voltage according to the control commands to drive the built-in permanent magnet synchronous motor.

[0072] Preferably, in another embodiment, the drive control system framework diagram of the built-in permanent magnet synchronous motor is as follows: Figure 5 As shown.

[0073] The drive control system for the embedded permanent magnet synchronous motor (IPMSM) uses a three-level neutral point clamp (NPC) inverter as the actuator. The outer-loop speed PI controller generates a q-axis reference current based on the error between the speed command ω* and the actual speed ω. And set the d-axis reference current Maximum torque-to-current ratio control is implemented. q-axis reference current. With d-axis reference current Constitutes the electronic reference current .

[0074] At each sampling time k, the measured three-phase stator current The dq-axis current is obtained through Clarke and Park coordinate transformation. and To achieve parameter robustness, the current prediction module based on the extended state observer (ESO) utilizes... and Based on the voltage applied in the previous cycle, the current value at time k+1 can be predicted online. , And the estimated lumped disturbance value. Meanwhile, the measured rotor position is processed by a one-step delay compensation module (the transfer function includes z). -1 (This represents storing and retrieving data from the previous moment) is used for third-order Lagrange extrapolation to obtain the predicted position for inverse coordinate transformation.

[0075] The multi-objective optimization cost function module receives the predicted current, reference current, and dynamic weighting coefficients calculated from the harmonic components extracted by low-pass filtering from the ESO. With harmonic state quantities Furthermore, the switch status recording module uses a delay unit z -1 Provides the switching state S(k) from the previous moment. The optimizer considers all candidate switching states S(k) of the three-level inverter. q (For different voltage vectors) iterate and evaluate, for each S q Calculate the value of the preset cost function, which integrates objectives such as voltage tracking, current error, harmonic suppression, switching frequency, and midpoint balance.

[0076] The optimizer compares all cost function values ​​and selects the candidate state that minimizes the total cost as the optimal switching state S. q Optimal switching state S q After S(k) is updated by the switch signal recording module, the specific three-phase drive signal S is generated by the pulse distribution logic. a ,S b ,S c At time k+1, a voltage is applied to the three-level NPC inverter, thereby driving the IPMSM to operate. (V in the diagram...) dc This is the DC bus voltage.

[0077] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0078] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The above are merely preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the scope of protection of this application.

Claims

1. A drive control method for an embedded permanent magnet synchronous motor, comprising using a three-level neutral point clamping inverter to drive the embedded permanent magnet synchronous motor, characterized in that, The drive control method includes: A reference current is generated based on the motor speed requirement; Based on a finite set, an extended state observer is used to perform two-step current prediction on the stator current of the built-in permanent magnet synchronous motor under different inverter switching states, so as to obtain the corresponding two-step predicted current values ​​under different switching states. Based on low-pass filtering, the fundamental and harmonic components of the measured quantum current of the built-in permanent magnet synchronous motor are separated to obtain the harmonic components. The cost value corresponding to each switching state is calculated based on a preset cost function. The switching state that minimizes the cost value is selected as the optimal switching state. The preset cost function includes a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term. The two-step current prediction error term is configured to track the error between the two-step predicted current value and the reference current. The adaptive harmonic suppression term is configured to adjust the current tracking accuracy using adaptive weights, which are negatively correlated with the harmonic components. The midpoint voltage balance term is configured to penalize the midpoint potential deviation of each candidate switching state. The adaptive weights are expressed by the following formula: ； In the formula, express Axis current tracking error weighting, express The reference current value of the shaft. This indicates that the state obtained through the extended state observer The predicted current value of the axis at time k+1. express Axial harmonic component amplitude, , This represents a very small positive number used to prevent the denominator from being zero. The cost of the adaptive harmonic suppression term is composed of the sum of squared current errors after adaptive weighting. The cost function is calculated using the following formula: ， In the formula, This represents the weight of the second-order difference penalty term for the d-axis current. This represents the two-step predicted current value along the d-axis. Indicates the d-axis reference current. This represents the weight of the second-order difference penalty term for the q-axis current. This represents the two-step predicted current value along the q-axis. Indicates the q-axis reference current; Based on the optimal switching state, a control command is generated, and the three-level neutral point clamping inverter is controlled to generate a three-phase voltage according to the control command to drive the built-in permanent magnet synchronous motor. Based on the predicted current at time k+1 generated by the extended state observer, a two-step current prediction is performed using the following formula to obtain the two-step current prediction value: ； In the formula, This represents the two-step current prediction value. This represents the predicted current at time k+1 generated by the extended state observer. This represents the lumped perturbation at time k+1 generated by the extended state observer. Indicates based on nominal inductance Gain, This represents the voltage vector corresponding to the candidate switch state.

2. The drive control method for the built-in permanent magnet synchronous motor according to claim 1, characterized in that, The cost function also includes a linearly superimposed adaptive switching frequency reduction term, which is configured to consist of the product of the number of switching actions and the adaptive switching penalty coefficient. The number of switching actions represents the total number of transitions in the switching state of each phase of the inverter within the current control cycle relative to the previous moment. The adaptive switching penalty coefficient is modulated by a harmonic current suppression factor, which is configured as follows: When the amplitude of the harmonic component exceeds the preset harmonic current threshold, the harmonic current suppression factor is less than 1, so as to reduce the value of the adaptive switching penalty coefficient, thereby allowing increased switching action to prioritize the suppression of harmonics. When the amplitude of the harmonic component does not exceed the preset harmonic current threshold, the harmonic current suppression factor remains at 1, so that the adaptive switching penalty coefficient is maintained at the reference level determined by the mechanical angular velocity, so as to reduce switching losses first.

3. The drive control method for the built-in permanent magnet synchronous motor according to claim 2, characterized in that, The adaptive switching penalty coefficient is configured to be determined by the product of the basic switching penalty weight, the frequency factor related to the rotational speed, the absolute value of the motor's mechanical angular velocity, and the harmonic current suppression factor.

4. The drive control method for the built-in permanent magnet synchronous motor according to claim 1, characterized in that, The cost function also includes a linearly superimposed current smoothness penalty term, which is configured to calculate the current change acceleration based on the second-order difference of the two-step predicted current values, and use the current change acceleration as a penalty index to suppress high-frequency oscillating harmonics.

5. The drive control method for the built-in permanent magnet synchronous motor according to claim 1, characterized in that, The cost function also includes a linearly superimposed voltage tracking term, which is configured to evaluate the error between the voltage vector corresponding to the candidate switch state and the reference voltage. The reference voltage is calculated based on the two-step predicted current value obtained through the extended state observer, combined with the reference current, according to the deadbeat principle.

6. The drive control method for the built-in permanent magnet synchronous motor according to claim 1, characterized in that, The extended state observer is represented by the following formula: ； In the formula, Sampling time, Let k be the predicted current at time k. Let be the estimated value of the lumped disturbance at time k. Let K be the measured quantum current at time k. ， , To ensure uniform motor angular velocity tuning, This indicates that it is based on the nominal inductance. Gain, This represents the voltage vector at time k.

7. A drive control system for a built-in permanent magnet synchronous motor, wherein the drive control system uses a three-level neutral point clamping inverter to drive the built-in permanent magnet synchronous motor, characterized in that, The drive control system employs the drive control method for the built-in permanent magnet synchronous motor as described in any one of claims 1 to 6, and the drive control system comprises: The speed outer loop controller is configured to generate a reference current based on the motor speed requirement; The inner current prediction controller is configured to perform two-step current prediction on the stator current of the built-in permanent magnet synchronous motor under different inverter switching states based on a finite set and using an extended state observer to obtain the corresponding two-step predicted current values ​​under different switching states. The low-pass filter harmonic extraction module is configured to separate the fundamental component and harmonic component of the measured quantum current of the built-in permanent magnet synchronous motor based on low-pass filtering to obtain the harmonic component. The multi-objective optimization decision module is configured to calculate the cost value corresponding to each switching state based on a preset cost function, and select the switching state that minimizes the cost value as the optimal switching state. The preset cost function includes a linearly superimposed two-step current prediction error term, an adaptive harmonic suppression term, and a midpoint voltage balance term. The two-step current prediction error term is configured to track the error between the two-step predicted current value and the reference current. The adaptive harmonic suppression term is configured to adjust the current tracking accuracy using adaptive weights, and the adaptive weights are negatively correlated with the harmonic components. The midpoint voltage balance term is configured to penalize the midpoint potential deviation of each candidate switching state. The drive execution module generates control commands based on the optimal switching state, and controls the three-level neutral point clamping inverter to generate three-phase voltage according to the control commands, so as to drive the built-in permanent magnet synchronous motor.

8. The drive control system for the built-in permanent magnet synchronous motor according to claim 7, characterized in that, The cost function also includes a linearly superimposed adaptive switching frequency reduction term, which is configured to consist of the product of the number of switching actions and the adaptive switching penalty coefficient. The number of switching actions represents the total number of transitions in the switching state of each phase of the inverter within the current control cycle relative to the previous moment. The adaptive switching penalty coefficient is determined by the product of the basic switching penalty weight, the frequency factor related to the rotational speed, the absolute value of the motor's mechanical angular velocity, and the harmonic current suppression factor. The harmonic current suppression factor is configured as follows: When the amplitude of the harmonic component exceeds the preset harmonic current threshold, the harmonic current suppression factor is less than 1, so as to reduce the value of the adaptive switching penalty coefficient, thereby allowing increased switching action to prioritize the suppression of harmonics. When the amplitude of the harmonic component does not exceed the preset harmonic current threshold, the harmonic current suppression factor remains at 1, so that the adaptive switching penalty coefficient is maintained at the reference level determined by the mechanical angular velocity, so as to reduce switching losses first.

Images