A multi-three-phase permanent magnet motor vibration harmonic suppression optimization system and method

Abstract

The application relates to the technical field of industrial control systems and discloses a multi-three-phase permanent magnet motor vibration harmonic suppression optimization system and method, which comprises the following steps: a state quantity acquisition module acquires motor phase current and bus voltage sequences to generate a joint data source; a high-dimensional feature space-time decoupling unit separates and compensates for sampling time delay by using a supercalibration operator, and separates a disturbance feature matrix by using a decoupling transformation matrix; an adaptive control strategy optimization unit updates adaptive control gain by using a discrete filter, and performs adaptive control law optimization and optimization and calculates a dynamic voltage compensation vector according to a gradient table when an amplitude is over the limit; and an adjustment instruction correction module adjusts weight according to an integral saturation coefficient, superimposes a vector on a reference instruction to output an adaptive adjustment instruction. The application eliminates current noise in a low-speed dead zone, realizes in-situ decoupling of a disturbance feature and dynamic feedforward hedging, widens an operation control bandwidth of an industrial control adaptive system and improves loop tracking precision.

Description

Technical Field

[0001] This invention relates to the field of industrial control system technology, and more specifically, to a vibration harmonic suppression optimization system and method for multi-phase three-phase permanent magnet motors. Background Technology

[0002] Currently, the regulation and control system of multi-winding power units is widely used in precision drive conditions. Conventional multi-channel architecture relies on independent regulators inside each winding to compare the given signal and the feedback signal to output a reference voltage command. Although this architecture can regulate the steady-state response of a single-channel winding, the loop tracking accuracy depends on the phase alignment between the feedback signal and the reference command. Due to the fixed transmission delay of 10μs inside the hardware circuit, the system generates feedback phase shift and introduces control response lag when the operating frequency increases.

[0003] In high-frequency operation with multiple circuits sharing a compact space, the voltage dead zone overlap distortion generated by switching actions exhibits nonlinear characteristics. Spatial electromagnetic cross-inductance between winding channels causes spatiotemporally coherent nonlinear distortion residuals in the multi-channel control flow. These distortion residuals are coupled across physical dimensions. Since the adjustment logic of each channel is isolated, the control closed loop cannot perceive the disturbance evolution process of adjacent channels in real time. Unmodeled overlap interference exacerbates the deterioration of system state tracking under high-power alternating conditions. In addition to the constraints of multi-channel physical layout and inherent hardware delays, harmonic suppression schemes at the control algorithm level also have shortcomings. For example, Chinese invention patent CN103441726B discloses a proportional resonant regulator-based method. The vector control method for dual three-phase permanent magnet motors introduces a proportional resonant regulator in the decoupled harmonic subplane to control the low-order current harmonics in a closed loop. However, under complex operating conditions of high-power alternating current and compact space with multi-channel nonlinear strong coherence, this type of control algorithm based on a continuous domain ideal model exposes its fundamental limitations: it implicitly relies on the absolute phase alignment of the feedback signal, does not consider the inherent fixed transmission delay of the hardware conditioning circuit, and causes phase misalignment or even positive feedback instability when running at high frequencies; the parallel operation of multiple proportional resonant regulators cascaded or multi-frequency arrays involves complex transcendental function calculations. Under the constraint of the extremely short main interrupt cycle of the microprocessor in the digital control system, the computational complexity of the algorithm increases sharply, leading to instantaneous overload of single-cycle computing power and main interrupt failure.

[0004] Therefore, the technical problem to be solved by this invention is how to extract the in-situ spatiotemporal features of multi-channel coherent nonlinear feedback distortion interference and accurately calculate the nonlinear feedforward dynamic voltage compensation vector within the constraint of a 50μs main interrupt cycle of a single microprocessor in the discrete control domain, thereby improving the instruction follow-up accuracy and loop convergence stability of the vibration harmonic suppression optimization system of multi-three-phase permanent magnet motors under the industrial control system architecture. Summary of the Invention

[0005] To address the problems in the background art, this invention provides a vibration harmonic suppression optimization system for multi-phase three-phase permanent magnet motors, the system comprising: The state quantity acquisition module is used to synchronously acquire the discrete sequence of phase current and the DC bus voltage pulsation sequence of the controlled motor through multiple parallel sampling channels, combine them to generate a multi-element joint data source and output it to the high-dimensional feature spatiotemporal decoupling unit. The high-dimensional feature spatiotemporal decoupling unit is used to use the phase advance calibration operator to shift the phase of the multi-element joint data source in the complex plane to compensate for the fixed transmission delay, and uses the spatiotemporal decoupling transformation matrix to separate the spatial magnetic field distortion components between multi-phase windings, and outputs the multi-dimensional harmonic disturbance feature matrix to the adaptive control strategy optimization unit. The adaptive control strategy optimization unit is used to update the adaptive filtering gain of the multidimensional harmonic disturbance characteristic matrix in real time at a set octave point using a variable structure discrete filter. When the amplitude of the multidimensional harmonic disturbance characteristic matrix exceeds the set sliding mode boundary, it retrieves and adjusts the feedforward control gain step by step according to the internally stored discretized step table. It converges and calculates the dynamic voltage compensation vector within a single 50μs interrupt cycle and outputs it to the adjustment command correction module. The adjustment command correction module is used to determine the dynamic adaptive weighting coefficient based on the current integral saturation coefficient, and to superimpose the dynamic voltage compensation vector onto the reference control voltage command in situ. The cascaded aliasing generates an adaptive adjustment command and outputs it to the power inverter circuit.

[0006] Preferably, the system also includes a cross-channel residual state exchange matrix and an unbalanced arbitration unit. Multiple sets of parallel control loops share a multi-dimensional harmonic disturbance characteristic matrix in real time through the cross-channel residual state exchange matrix. The unbalanced arbitration unit is used to monitor the harmonic residuals between each set of parallel control loops in real time, and when the mutual difference between the harmonic residual of any channel and the other channels exceeds 20%, the feedforward control gain is adjusted according to the mutual difference.

[0007] Preferably, the status acquisition module includes a current Hall sensor and a voltage sampling circuit; wherein, the current Hall sensor is used to synchronously acquire the multi-phase current discrete signals of the controlled motor at a sampling frequency of 20kHz, and extract the phase current discrete sequence after cascaded digital low-pass filtering; the voltage sampling circuit is used to capture the DC bus voltage of the power inverter circuit in real time, and calculate the AC / DC decoupling impedance component to isolate the bus voltage pulsation sequence.

[0008] Preferably, the high-dimensional feature spatiotemporal decoupling unit is internally configured with a discrete-time lead phase calibration unit; wherein, the discrete-time lead phase calibration unit is used to calculate the lead compensation phase angle based on the inherent transmission delay of the digital microprocessor, and to use the lead compensation phase angle to translate and align the coordinates to eliminate the phase misalignment of the discrete sequence of phase current; the high-dimensional feature spatiotemporal decoupling unit is also used to decouple the aligned multi-channel data stream through the spatiotemporal decoupling transformation matrix, and separate the spatial magnetic field distortion component to generate a multidimensional harmonic disturbance feature matrix.

[0009] Preferably, the adaptive control strategy optimization unit includes a sliding mode boundary monitor; wherein, the sliding mode boundary monitor is used to track the modulus change of the multidimensional harmonic disturbance characteristic matrix in real time, and when the modulus change exceeds the set safety margin threshold, it calls the variable structure discrete filter to update the adaptive filter gain at the third, fifth and seventh harmonics of the fundamental frequency.

[0010] Preferably, the adaptive control strategy optimization unit internally stores a discretized step-by-step mapping table; wherein, when the amplitude of the multidimensional harmonic disturbance characteristic matrix exceeds the set sliding mode boundary, the adaptive control strategy optimization unit is used to call the discretized step-by-step mapping table to retrieve the feedforward gain correction value step by step, and adjust the feedforward control gain according to the retrieved discrete gain step amount.

[0011] Preferably, the system further includes a space voltage vector modulation unit; wherein the space voltage vector modulation unit is used to receive adaptive adjustment commands, and to compensate the nonlinear dead-zone voltage deviation of the power inverter circuit in the field to generate multiple pulse width modulation signals, and output the pulse width modulation signals to the power inverter circuit.

[0012] Preferably, when the harmonic residual of any channel is detected to differ from the mutual difference of the other channels by more than 20%, the unbalanced arbitration unit generates circulating current suppression commands and load reduction constraint weights between channels based on the amplitude ratio of the mutual difference, and loads the load reduction constraint weights to the adaptive control strategy optimization units of each set of parallel control loops to correct the feedforward control gain of each channel.

[0013] Preferably, the system is deployed on a multi-core industrial microprocessor platform; wherein, the state quantity acquisition module, the high-dimensional feature spatiotemporal decoupling unit, the adaptive control strategy optimization unit, and the adjustment instruction correction module pipeline data between multiple hardware cores of the multi-core industrial microprocessor platform, and the hardware cores are synchronized and aligned through the master clock hardware-triggered interrupt signal.

[0014] A method for optimizing vibration harmonic suppression in a multi-phase three-phase permanent magnet motor, applied to a multi-phase three-phase permanent magnet motor vibration harmonic suppression optimization system, includes the following steps: Step S1: The state quantity acquisition module synchronously acquires the phase current discrete sequence and DC bus voltage pulsation sequence of the controlled motor through multiple parallel sampling channels, combines them to generate a multi-element joint data source and outputs it to the high-dimensional feature spatiotemporal decoupling unit. Step S2: The phase advance calibration operator inside the high-dimensional feature spatiotemporal decoupling unit is used to shift the phase of the multi-element joint data source in the complex plane to compensate for the fixed transmission delay. The spatiotemporal decoupling transformation matrix is ​​used to separate the spatial magnetic field distortion components between the multi-phase windings and output the multi-dimensional harmonic disturbance feature matrix to the adaptive control strategy optimization unit. Step S3: The adaptive control strategy optimization unit uses the variable structure discrete filter inside to update the adaptive filter gain of the multidimensional harmonic disturbance feature matrix in real time at the set octave point. When the amplitude of the multidimensional harmonic disturbance feature matrix exceeds the set sliding mode boundary, the feedforward control gain is retrieved and adjusted step by step according to the internally stored discretized step table. The dynamic voltage compensation vector is calculated and output to the adjustment command correction module within a single 50μs interrupt cycle. Step S4: The adjustment command correction module determines the dynamic adaptive weighting coefficient based on the current integral saturation coefficient, and the dynamic voltage compensation vector is superimposed on the reference control voltage command in situ. The cascaded superposition generates an adaptive adjustment command and outputs it to the power inverter circuit.

[0015] The present invention has at least the following beneficial effects: 1. In the optimization of vibration harmonic suppression of multi-phase three-phase permanent magnet motors, the state quantity acquisition module synchronously acquires the phase current discrete time sequence and DC bus voltage pulsation sequence of the controlled multi-phase three-phase power system. The two digital quantities form a joint data source and are sent to the high-dimensional characteristic spatiotemporal decoupling unit. When the system is under a specific low-speed condition, the fundamental component in the phase current discrete time sequence weakens and the switching noise amplitude increases. At this time, the high-dimensional characteristic spatiotemporal decoupling unit extracts the high-frequency power feedback residual component in the DC bus voltage pulsation sequence and generates real-time offset compensation for the distortion of the phase current sampling signal, so that the control loop maintains a stable control residual separation state globally and eliminates the problem of oscillation and divergence of adjustment commands caused by the sharp drop in the signal-to-noise ratio of a single source.

[0016] 2. The high-dimensional feature spatiotemporal decoupling unit calls the internally configured discrete-time series phase advance calibration operator, calculates the digital sampling phase compensation angle based on the current feedback fundamental frequency, performs time-domain advance translation on the phase current discrete-time series, and safely offsets the 10μs fixed transmission delay generated by the hardware conditioning circuit. The calibrated data, together with the DC bus voltage pulsation sequence, is imported into the heterogeneous spatiotemporal topology mapping matrix. Peak noise is eliminated by the discrete exclusion rule based on variance characteristics, and the spatial magnetomotive force distortion and the time-domain dead-zone voltage overlap distortion are separated in situ to synthesize a high-dimensional nonlinear harmonic disturbance tensor, avoiding the positive feedback instability problem induced by phase misalignment in traditional feedback control.

[0017] 3. The adaptive control strategy optimization unit receives the high-dimensional nonlinear harmonic disturbance tensor output by the high-dimensional feature spatiotemporal decoupling unit. Within the preset voltage utilization and current safety boundary constraints, it uses a variable structure discrete filter to perform gain updates at a specific octave of the fundamental frequency. The update process follows a discretized step-by-step rule. When the disturbance tensor amplitude exceeds the set sliding mode boundary, the system adjusts the feedforward control gain by step-by-step lookup according to the preset gradient table. The nonlinear feedforward dynamic voltage compensation vector is quickly converged within a single 50μs interrupt cycle inside the digital microprocessor, avoiding instantaneous overload of computing power and main interrupt failure caused by the calculation of the inversion of the transcendental function matrix. Attached Figure Description

[0018] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings, in which several embodiments of the invention are illustrated by way of example and not limitation, wherein: Figure 1 This is a flowchart of the steps in the optimization method for vibration harmonic suppression of multi-phase permanent magnet motors of the present invention; Figure 2 This is a structural diagram of the vibration harmonic suppression optimization system module for multi-phase three-phase permanent magnet motors of the present invention. Detailed Implementation

[0019] The principles and spirit of the present invention will now be described with reference to several exemplary embodiments in conjunction with the accompanying drawings. It should be understood that these embodiments are provided merely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0020] A vibration harmonic suppression optimization system for multi-phase three-phase permanent magnet motors, the system comprising: The state quantity acquisition module is used to synchronously acquire the discrete sequence of phase current and the DC bus voltage pulsation sequence of the controlled motor through multiple parallel sampling channels, combine them to generate a multi-element joint data source and output it to the high-dimensional feature spatiotemporal decoupling unit. The high-dimensional feature spatiotemporal decoupling unit is used to use the phase advance calibration operator to shift the phase of the multi-element joint data source in the complex plane to compensate for the fixed transmission delay, and uses the spatiotemporal decoupling transformation matrix to separate the spatial magnetic field distortion components between multi-phase windings, and outputs the multi-dimensional harmonic disturbance feature matrix to the adaptive control strategy optimization unit. The adaptive control strategy optimization unit is used to update the adaptive filtering gain of the multidimensional harmonic disturbance characteristic matrix in real time at a set octave point using a variable structure discrete filter. When the amplitude of the multidimensional harmonic disturbance characteristic matrix exceeds the set sliding mode boundary, it retrieves and adjusts the feedforward control gain step by step according to the internally stored discretized step table. It converges and calculates the dynamic voltage compensation vector within a single 50μs interrupt cycle and outputs it to the adjustment command correction module. The adjustment command correction module is used to determine the dynamic adaptive weighting coefficient based on the current integral saturation coefficient, and to superimpose the dynamic voltage compensation vector onto the reference control voltage command in situ. The cascaded aliasing generates an adaptive adjustment command and outputs it to the power inverter circuit.

[0021] Preferably, the system also includes a cross-channel residual state exchange matrix and an unbalanced arbitration unit. Multiple sets of parallel control loops share a multi-dimensional harmonic disturbance characteristic matrix in real time through the cross-channel residual state exchange matrix. The unbalanced arbitration unit is used to monitor the harmonic residuals between each set of parallel control loops in real time, and when the mutual difference between the harmonic residual of any channel and the other channels exceeds 20%, the feedforward control gain is adjusted according to the mutual difference.

[0022] Preferably, the status acquisition module includes a current Hall sensor and a voltage sampling circuit; wherein, the current Hall sensor is used to synchronously acquire the multi-phase current discrete signals of the controlled motor at a sampling frequency of 20kHz, and extract the phase current discrete sequence after cascaded digital low-pass filtering; the voltage sampling circuit is used to capture the DC bus voltage of the power inverter circuit in real time, and calculate the AC / DC decoupling impedance component to isolate the bus voltage pulsation sequence.

[0023] Preferably, the high-dimensional feature spatiotemporal decoupling unit is internally configured with a discrete-time lead phase calibration unit; wherein, the discrete-time lead phase calibration unit is used to calculate the lead compensation phase angle based on the inherent transmission delay of the digital microprocessor, and to use the lead compensation phase angle to translate and align the coordinates to eliminate the phase misalignment of the discrete sequence of phase current; the high-dimensional feature spatiotemporal decoupling unit is also used to decouple the aligned multi-channel data stream through the spatiotemporal decoupling transformation matrix, and separate the spatial magnetic field distortion component to generate a multidimensional harmonic disturbance feature matrix.

[0024] Preferably, the adaptive control strategy optimization unit includes a sliding mode boundary monitor; wherein, the sliding mode boundary monitor is used to track the modulus change of the multidimensional harmonic disturbance characteristic matrix in real time, and when the modulus change exceeds the set safety margin threshold, it calls the variable structure discrete filter to update the adaptive filter gain at the third, fifth and seventh harmonics of the fundamental frequency.

[0025] Preferably, the adaptive control strategy optimization unit internally stores a discretized step-by-step mapping table; wherein, when the amplitude of the multidimensional harmonic disturbance characteristic matrix exceeds the set sliding mode boundary, the adaptive control strategy optimization unit is used to call the discretized step-by-step mapping table to retrieve the feedforward gain correction value step by step, and adjust the feedforward control gain according to the retrieved discrete gain step amount.

[0026] Preferably, the system further includes a space voltage vector modulation unit; wherein the space voltage vector modulation unit is used to receive adaptive adjustment commands, and to compensate the nonlinear dead-zone voltage deviation of the power inverter circuit in the field to generate multiple pulse width modulation signals, and output the pulse width modulation signals to the power inverter circuit.

[0027] Preferably, when the harmonic residual of any channel is detected to differ from the mutual difference of the other channels by more than 20%, the unbalanced arbitration unit generates circulating current suppression commands and load reduction constraint weights between channels based on the amplitude ratio of the mutual difference, and loads the load reduction constraint weights to the adaptive control strategy optimization units of each set of parallel control loops to correct the feedforward control gain of each channel.

[0028] Preferably, the system is deployed on a multi-core industrial microprocessor platform; wherein, the state quantity acquisition module, the high-dimensional feature spatiotemporal decoupling unit, the adaptive control strategy optimization unit, and the adjustment instruction correction module pipeline data between multiple hardware cores of the multi-core industrial microprocessor platform, and the hardware cores are synchronized and aligned through the master clock hardware-triggered interrupt signal.

[0029] A method for optimizing vibration harmonic suppression in a multi-phase three-phase permanent magnet motor, applied to a multi-phase three-phase permanent magnet motor vibration harmonic suppression optimization system, includes the following steps: Step S1: The state quantity acquisition module synchronously acquires the phase current discrete sequence and DC bus voltage pulsation sequence of the controlled motor through multiple parallel sampling channels, combines them to generate a multi-element joint data source and outputs it to the high-dimensional feature spatiotemporal decoupling unit. Step S2: The phase advance calibration operator inside the high-dimensional feature spatiotemporal decoupling unit is used to shift the phase of the multi-element joint data source in the complex plane to compensate for the fixed transmission delay. The spatiotemporal decoupling transformation matrix is ​​used to separate the spatial magnetic field distortion components between the multi-phase windings and output the multi-dimensional harmonic disturbance feature matrix to the adaptive control strategy optimization unit. Step S3: The adaptive control strategy optimization unit uses the variable structure discrete filter inside to update the adaptive filter gain of the multidimensional harmonic disturbance feature matrix in real time at the set octave point. When the amplitude of the multidimensional harmonic disturbance feature matrix exceeds the set sliding mode boundary, the feedforward control gain is retrieved and adjusted step by step according to the internally stored discretized step table. The dynamic voltage compensation vector is calculated and output to the adjustment command correction module within a single 50μs interrupt cycle. Step S4: The adjustment command correction module determines the dynamic adaptive weighting coefficient based on the current integral saturation coefficient, and the dynamic voltage compensation vector is superimposed on the reference control voltage command in situ. The cascaded superposition generates an adaptive adjustment command and outputs it to the power inverter circuit.

[0030] Example 1: A vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor is applied to the drive field of high-power industrial fans. Under long-term continuous heavy-load operation, this system achieves adaptive offsetting of electromagnetic cross-inductance and dead-zone distortion residuals of multiple windings by real-time acquisition of the discrete-time sequence of phase current and the DC bus voltage pulsation sequence. The state quantity acquisition module adopts multiple parallel sampling channels to synchronously acquire the current of each phase winding at a sampling frequency of 20kHz, and the acquired raw current signal is processed by digital low-pass filtering to form the discrete sequence of phase current. The DC bus voltage sampling circuit captures the DC bus voltage of the power inverter circuit in real time, and isolates the DC bus voltage pulsation sequence through impedance decoupling. Specifically, the AC / DC decoupling impedance components are calculated by establishing a dynamic impedance mathematical model of the bus. The system is implemented by real-time acquisition of the instantaneous value of the DC bus voltage and the total input current of the inverter circuit. In the digital domain, a high-pass filter is used to filter out the DC steady-state component, extracting the AC pulsating voltage and current. The decoupling impedance component is calculated by dividing the root mean square value of the AC pulsating voltage by the root mean square value of the AC pulsating current, representing the dynamic equivalent impedance generated by the inverter bridge switching action. By subtracting the nonlinear voltage drop loss generated by this dynamic impedance component on the DC side from the measured instantaneous bus voltage value, the DC bus voltage pulsation sequence generated by the inverter dead zone overlap can be separated. The state quantity acquisition module outputs the above two digital quantities as a joint data source in parallel. The high-dimensional feature spatiotemporal decoupling unit receives the joint data source and uses a discrete-time series phase lead calibration operator based on the current fundamental frequency. The process involves performing a first-order Taylor series phase shift calculation to advance the discrete sequence of phase currents along the time axis, offsetting the 10μs fixed transmission delay caused by the hardware conditioning circuit. This achieves synchronous alignment of the current signal and the DC bus voltage pulsation sequence in the time domain. The aligned data is then imported into a heterogeneous spatiotemporal topology mapping matrix. Discrete variance feature analysis within a preset time window is used to remove spike noise and separate the first characteristic component representing spatial magnetomotive force distortion and the second characteristic component representing time-domain dead-zone voltage overlap distortion in situ. This results in the synthesis of a high-dimensional nonlinear harmonic perturbation tensor. To correlate and extract the characteristic quantities of spatially distributed electromagnetic field physical distortion and time-domain discrete switching pulse distortion at the digital signal level, this invention achieves this by constructing a physical intermediate medium and a spatiotemporal energy transfer chain. Specifically, the time-domain dead-zone voltage overlap distortion directly leads to instantaneous voltage loss in the inverter output. This loss induces a high-frequency distortion current in the stator winding, which is then transformed into spatial nonlinear magnetomotive force disturbance in the stator air gap. Therefore, the time-domain voltage distortion and spatial magnetomotive force distortion are physically coupled through the instantaneous fluctuation of the air gap magnetic field energy. At the data level, this invention uses a heterogeneous spatiotemporal topology mapping matrix as an intermediate medium to transform the variance of the time-domain bus voltage pulsation within a fixed sampling period into an equivalent spatial voltage vector deviation, which is then uniformly normalized to a high-dimensional spatial geometric dimension. This allows for the calculation and generation of a high-dimensional nonlinear harmonic disturbance tensor that can simultaneously characterize both time-domain and spatial nonlinear features.

[0031] Adaptive control strategy optimization unit receives Within the preset voltage utilization and current safety boundary constraints, gain updates are performed at discrete points of 5th, 7th, and 11th harmonics of the fundamental frequency using a variable structure discrete filter. When the magnitude exceeds the safe sliding mode boundary, the system calls the internally stored discretized step mapping table to retrieve the dynamic voltage compensation vector within a single 50μs interrupt cycle. In the actual control process, the convergence calculation within a single 50-microsecond interrupt cycle employs a direct lookup table without nested loops and a first-order linear interpolation algorithm. When the amplitude of the multidimensional harmonic disturbance characteristic matrix exceeds the boundary, the digital microprocessor directly uses the current amplitude error as an address pointer and steps through the internally stored discretized step table within a single main interrupt. This eliminates the need for complex transcendental function matrix inversion calculations. The system's iteration termination is directly limited by the discrete step size boundary. Only three direction determinations and gain correction value accumulations are performed within a single interrupt, ensuring the calculation duty cycle is controlled within 30%. This forces convergence and smoothly outputs the dynamic voltage compensation vector before the end of the single 50-microsecond interrupt cycle. The adjustment instruction correction module receives... With main closed-loop reference voltage command The dynamic adaptive weighting coefficient is determined based on the current integral saturation coefficient. Execute cascaded instruction corrections and calculate and generate adaptive adjustment instructions. , ,in, The dynamic gain stabilization factor of the inverter circuit is determined by the ratio of the turn-on voltage drop of the inverter bridge switch to the DC bus voltage. The specific mapping path for converting this ratio into a dimensionless dynamic gain stabilization factor is as follows: A one-dimensional linear mapping lookup table is pre-stored within the system. Its input is connected to the sampled bus voltage value and the nominal rated voltage drop of the switch. During actual operation, the microprocessor calculates the real-time ratio of these two values ​​and uses this ratio as an index into the lookup table, mapping out the corresponding stabilization factor value. When the ratio fluctuates between 0.01 and 0.05, the corresponding stabilization factor monotonically increases between 0.92 and 0.98. This transforms the rigid voltage loss at the hardware level into a gain correction coefficient at the control law level, and the dynamic gain stabilization factor is calculated in the adaptive adjustment instruction. Inverter voltage utilization conditioning constant Presenting a monotonic numerical correspondence, the parameter conversion path is determined by periodically acquiring the instantaneous value of the DC bus voltage of the power inverter circuit and the nominal rated on-state voltage drop of the inverter bridge switching transistors through a multi-core industrial microprocessor platform. The microprocessor calculates the real-time ratio of the on-state voltage drop to the instantaneous value of the DC bus voltage in each 50-microsecond main interrupt cycle, which serves as the index input item for a one-dimensional linear image lookup table. When the real-time ratio fluctuates between 0.01 and 0.05, the lookup table outputs a dimensionless scaling coefficient that monotonically increases between 0.92 and 0.98 and serves as the current quantization parameter. This coefficient is used to scale the dynamic voltage compensation vector output by the adaptive control strategy optimization unit. The adjustment command correction module superimposes the scaled dynamic voltage compensation vector onto the reference control voltage command of the main control closed loop in situ, converting the voltage loss at the hardware level into the gain correction coefficient at the control end.

[0032] Adaptive adjustment command The data is sent to the power inverter circuit to drive the multi-phase three-phase power system to perform actions. When the harmonic residual amplitude of any control channel in the system differs from that of the other channels by more than 20%, the cross-channel residual state exchange matrix triggers the fault-tolerant control flow. The unbalanced arbitration unit actively reduces the output weight of the damaged channel and simultaneously increases the output weight of the intact channel proportionally. The coefficient is used to maintain the steady-state convergence of the global system. The aforementioned 20% unbalance critical threshold is determined based on engineering limit tests of the boundaries of multi-channel magnetic circuit saturation and thermal instability. Experimental results show that when the mutual difference of electromagnetic cross-inductance residuals between channels is less than 20%, the system's own closed-loop integral regulator can still maintain stable operation through conventional dynamic balancing. However, once the difference reaches or exceeds 20%, the electromagnetic cross-inductance torque pulsation of the damaged channel will nonlinearly superimpose with that of the adjacent channel, causing the harmonic current of the control loop to amplify rapidly. This results in the local temperature rise of the switching transistor soaring to more than 30 degrees Celsius within 50 milliseconds, thereby triggering the inverter overcurrent protection shutdown. Therefore, setting the critical threshold to 20% can retain sufficient fault tolerance margin before the system experiences cascading instability. Under this condition, the control loop can complete gain reconstruction and update within 2ms, reducing the harmonic residual amplitude by more than 82% across the entire speed range, and widening the closed-loop control bandwidth to 2.5 times the original.

[0033] Example 2: This experiment verifies the engineering effectiveness of the harmonic suppression optimization system by building a fully digital closed-loop control environment on a multi-axis linkage permanent magnet motor simulated load platform. The experimental platform is equipped with a high-frequency torque pulsation analyzer with a sampling accuracy of 0.01%, and a multi-channel parallel high-speed ADC acquisition card to fully capture the original physical signals containing environmental noise. The test object is a three-phase permanent magnet synchronous power system with a rated power of 50kW, and a high-precision eddy current brake is connected to the load end to simulate the variable speed and nonlinear load fluctuation conditions in industrial settings. The experimental design establishes a multi-dimensional control system: the experimental group runs the complete system scheme, and performs phase alignment and disturbance feature separation of the sampled current through a high-dimensional feature spatiotemporal decoupling unit, and performs discrete gain adjustment using an adaptive control strategy optimization unit; the first control group uses a transmission... The system employs a unified PI dual-closed-loop control strategy without introducing phase calibration and harmonic reconstruction logic. The second control group uses a feedback scheme with a low-pass hardware filter to eliminate high-dimensional spatiotemporal decoupling units. The test process covers the fundamental frequency range of 10Hz to 150Hz, and steady-state operation data is recorded at every 5Hz step point to evaluate the harmonic suppression performance across the entire speed range. The sampling frequency is set based on a trade-off between data processing real-time performance and signal aliasing margin. The test results show that when the sampling frequency is below 15kHz, dead-zone distortion harmonics at low speeds alias towards the fundamental frequency side, leading to feedback control instability. When the sampling frequency exceeds 25kHz, the duty cycle of the single-cycle interrupt processing load reaches 85%, causing phase lag in the control command output. Based on this, the system sets the sampling frequency to 20kHz to balance the data processing load and signal acquisition bandwidth.

[0034] Key performance indicators measured in the experiment: Under standard operating conditions with a fundamental frequency of 50Hz and a load torque of 80Nm, the measured total harmonic distortion (THD) of the current in the first control group was 8.42%, including the 5th and 7th harmonic components; after introducing hard filtering in the second control group, the THD decreased to 4.15%, but the closed-loop tracking bandwidth decreased from the original 350Hz to 120Hz, resulting in an increase in dynamic tracking error under acceleration and deceleration conditions. The experimental group running this scheme measured a current THD of 1.48%, and the closed-loop tracking bandwidth remained at 875Hz, verifying the synergistic improvement of harmonic suppression effectiveness and dynamic response capability.

[0035] To verify the logic fault-tolerant performance, a white noise component with a signal-to-noise ratio of 20dB was actively superimposed at the phase current input. During the test, the high-dimensional feature spatiotemporal decoupling unit eliminated the noise component within a 50μs interruption period through discrete variance feature analysis, maintaining the stability of the harmonic disturbance feature matrix. When the analog control channel experienced performance drift, causing the difference in harmonic residual amplitude between channels to fluctuate from 5% to 22%, the unbalanced arbitration unit adjusted the dynamic adaptive weighting coefficient of the intact channels. Gain reconstruction is completed within 2ms, reducing the global THD to 1.62%, demonstrating the system's dynamic redundancy capability under boundary conditions. For the gain update frequency range defined in this invention, extreme boundary verification experiments were conducted. When the gain update frequency is below four times the fundamental frequency, the harmonic offset command and the dead-zone voltage residual overlap, increasing the system oscillation amplitude by 15%. When the update frequency exceeds 13 times the fundamental frequency, due to the physical switching delay of the power inverter circuit, the compensation command generated by the control loop and the actual current change undergo phase reversal. Based on this, the gain update is limited to discrete points at 5, 7, and 11 times the fundamental frequency, forming the optimal working window to ensure harmonic suppression effectiveness. This invention adaptively selects different combinations of frequency points under different industrial drive conditions, particularly in the low-speed or light-load conditions defined in this invention. Under standard operating conditions, due to the large proportion of inverter dead time in the control cycle, system harmonics are mainly concentrated in the lower frequency band. Therefore, the sliding mode boundary monitor calls the filter to update at the 3rd, 5th, and 7th harmonics of the fundamental frequency. However, under the aforementioned high-power, high-speed continuous heavy-load operation conditions, low-order harmonics are weakened by the large inductance, while the residuals of high-order spatial electromagnetic cross-inductance at the 5th, 7th, and 11th harmonics caused by switching actions become the main contradiction causing system oscillations. The system automatically and smoothly switches between these two frequency configuration schemes based on the real-time detected fundamental operating frequency, thereby ensuring the logical consistency and harmonic suppression efficiency of multi-dimensional gain updates under all operating conditions. The sliding mode boundary monitor inside the adaptive control strategy optimization unit tracks the magnitude change of the multi-dimensional harmonic disturbance characteristic matrix in real time, according to the operating fundamental frequency. The frequency doubling point of the variable structure discrete filter, which switches the amplitude of the electromagnetic residual caused by load torque variation, is at the operating fundamental frequency of the controlled motor. In low-speed operating conditions (below 50Hz and with load torque below 50% of rated torque), the inverter dead time accounts for a proportion exceeding the set threshold in the control cycle, and system harmonic energy is concentrated in the low-frequency band. The sliding mode boundary monitor calls the variable structure discrete filter to update the 3rd, 5th, and 7th harmonic points of the fundamental frequency. In high-power continuous heavy-load operating conditions (where the controlled motor's fundamental frequency exceeds 50Hz or the load torque exceeds 50% of rated torque) and the characteristic matrix modulus exceeds the sliding mode boundary value of 0.12, the inverter circuit switching action generates 5th, 7th, and 11th harmonic high-order spatial electromagnetic cross-inductance residual disturbance components. The control strategy optimization unit determines the appropriate components based on the detected operating fundamental frequency. The filter frequency is switched to discrete points of 5th, 7th and 11th harmonics, and the feedforward control gain is optimized by iteratively searching the internally stored discrete step-by-step mapping table within a single 50-microsecond interrupt cycle, and a dynamic voltage compensation vector is output.

[0036] Example 3: The current vibration harmonic suppression optimization system for multi-phase three-phase permanent magnet motors is applied to industrial high-voltage variable frequency drive control scenarios. In this scenario, due to the interaction between the inverter switching frequency and the distributed inductance of the motor windings, nonlinear current distortion is generated. The system constructs a closed-loop regulation mechanism to suppress vibration harmonics. An interference observer is set in the control system to receive the instantaneous value of the phase current. With DC bus voltage The fundamental frequency component is extracted using a current vector decomposition algorithm, and the remaining high-frequency current components are defined as harmonic residual signals. The core of harmonic suppression optimization logic lies in establishing a dynamic mapping between the phase current fundamental frequency and the high-frequency residual. The system is based on... Frequency response settings for a multi-bandpass filter array, with the center frequency of each filter... Align the 5th, 7th, and 11th harmonic components of the motor respectively to isolate the harmonic disturbance characteristic matrix.

[0037] The system processes harmonic residual signals. At that time, a time constant-based approach is introduced. A first-order low-pass filter operator is used for signal smoothing, where the time constant is... With sampling period The relationship is set as By suppressing high-frequency switching noise during current sampling, this invention ensures that the residual signal entering the adjustment logic has stable phase characteristics. To address the control phase lag caused by setting the time constant to 5 times the sampling period, this invention incorporates a dynamic phase angle compensation flow in the control loop. Specifically, although the low-pass filter operator introduces a deterministic phase delay while suppressing high-frequency switching noise, since the sampling period is fixed at 50 microseconds, this delay manifests as a fixed electrical angle lag at a specific operating frequency. The phase advance calibration operator configured in the high-dimensional feature spatiotemporal decoupling unit of this invention linearly accumulates the fixed delay time caused by the low-pass filter with the inherent 10 microsecond transmission delay of the hardware circuit when calculating the advance compensation phase angle, and uniformly performs first-order Taylor series phase advance shift compensation. Thus, the phase lag caused by subsequent filtering is rigidly offset by the pre-advance advance calibration, eliminating time-domain physical logic conflicts and avoiding oscillations and divergences caused by the phase reversal between the feedforward command and the actual current. The control strategy optimization unit receives the filtered residual component and, based on the set control gain... Perform feedforward offset calculations to generate dynamic voltage compensation commands. When a current tracking error is detected Exceeding the set security boundary threshold At that time, the system automatically triggers the gain update stream to... As a unit compensation gain, iterative optimization is performed. Specifically, the initial value of the unit compensation gain is determined based on the ratio of the total inductance of the loop to the sampling period under the calibration condition. Its step size is set to 0.05% of the fixed physical gain. When performing iterative optimization in the gain update flow, the system increases the feedforward control gain step by step with this step size and evaluates the current tracking error after each increase. If the tracking error decreases by less than 0.01% after three consecutive iterations or the total gain reaches the preset maximum safety limit, the system forcibly terminates the iteration and locks the current gain value. This serves as a limiting mechanism to ensure single-cycle convergence.

[0038] During system operation, when the harmonic components of the damaged winding channel exhibit abnormal amplitude shifts, the arbitration logic unit triggers a state exchange matrix when the difference between the harmonic residuals of the currents in each winding exceeds 20%. The system logic then automatically adjusts the weighting coefficients of the failed winding. The gain was adjusted to 0.5, and the missing compensation gain was proportionally distributed to the gain register of the intact winding to maintain the stability of the global output torque. Under this condition, the closed-loop control bandwidth remained at 800Hz, and the total harmonic distortion (THD) was controlled within 1.5%, ensuring the continuous controllability of the harmonic suppression logic during local winding faults. Test data recorded: Under the condition of a fundamental frequency of 50Hz and a load of 80Nm, the current THD was 8.5% without the optimization system of this invention enabled, and this index dropped to 1.5% after harmonic offsetting; when the load torque increased to 120Nm, the harmonic suppression module... The dynamic weight matching of the coefficients suppresses the amplitude of the 5th harmonic to below 0.8% of the fundamental amplitude. This adjustment logic eliminates the control oscillation caused by current sampling distortion and realizes in-situ offsetting and dynamic stable tracking of motor vibration harmonics.

[0039] Example 4: Before deploying a current multi-phase three-phase permanent magnet motor system under continuous frequency conversion control conditions, a standardized offline calibration and benchmark construction process is performed to eliminate initial inconsistencies in the electrical characteristics between each phase winding. The first stage of the calibration process is the flux linkage observation under the static state of the windings. The system injects a sinusoidal excitation signal with a frequency of 100Hz and an amplitude of 5% of the rated current into each phase winding, records the instantaneous voltage drop and current response across the windings, and calculates the equivalent mutual inductance coefficient matrix of each winding using the least squares method. ,in Indicates the first Phase and Di The coupling inductance between phase windings.

[0040] The second stage of the calibration process is dynamic rotor position reference calibration. The system drive motor runs at a constant low speed of 200 rpm, and the electrical angle feedback from the encoder is collected. The phase difference compensation function is fitted to the zero-crossing time of the induced electromotive force of each phase winding. ,function Defined as the offset deviation between the physical installation angle and the electrical phase, its calculation formula is as follows: ,in, To measure the phase angle corresponding to the zero-crossing point of the measured potential, the system will... The data is discretized into a lookup table containing 360 data points and stored in the control unit's memory. This table serves as the initial phase reference for all subsequent harmonic offsetting algorithms, ensuring precise alignment between control commands and the electromagnetic space field inside the motor.

[0041] To address component thermal drift during long-term operation, the system embeds an online reference reconfiguration mechanism in the main control loop. This mechanism detects temperature sensor readings from the inverter power module. Comparison with initial calibration temperature When the deviation exceeds 20℃, the system is based on the temperature coefficient of resistance. Automatic correction of stator resistance estimation value The corrected logic is as follows: ,in, The system is based on the corrected winding static resistance at the calibration temperature. The voltage drop compensation value is updated in real time to offset the interference of impedance changes caused by winding heating on the accuracy of the harmonic offset algorithm, ensuring that the control command response under continuous frequency conversion is kept within the preset error limit.

[0042] Example 5: In a multi-phase three-phase permanent magnet synchronous motor drive system, to eliminate current harmonics caused by inverter dead-zone voltage, the system executes standardized pre-calibration and online compensation procedures. At the beginning of system deployment or after motor rotor replacement, the drive control unit, through the current sampling module, operates smoothly for 500ms in a zero-load current command state at a sampling frequency of 20kHz to obtain the discrete sequence of current in each phase on the stator side of the motor. and bus voltage sequence The high-dimensional feature spatiotemporal decoupling unit performs discrete phase alignment on the acquired multi-data. Based on the inherent 10μs transmission delay of the system's hardware filtering circuit, the microprocessor determines the lead compensation phase angle. ,in, Using the current operating fundamental frequency of the motor, the system utilizes this phase angle to shift the coordinate system, eliminates the phase misalignment of the discrete sequence of phase currents, and separates the spatial magnetic field distortion component through a spatiotemporal decoupling transformation matrix. Specifically, the spatiotemporal decoupling transformation matrix adopts a cascaded composite structure of an extended Clarke transformation matrix and a multiphase decoupling space vector topology matrix. In actual operation, the algebraic elements of this matrix are composed of the cosine and sine values ​​of the spatial electrical angle difference between multiple sets of symmetrical three-phase windings. This matrix is ​​used to project the aligned multi-channel phase current discrete sequence from the stator stationary coordinate system to a multidimensional orthogonal subspace. The row elements of the core transformation matrix correspond to the fundamental plane, the third harmonic spatial component, and the higher-order cross-inductance spurious components. The aliased multi-channel current signals are stripped away through discrete matrix multiplication, directly filtering out the fundamental wave and constant DC bias, and thus outputting a high-dimensional data structure containing only the spatial magnetic field distortion components between each physical channel. The modulus value exceeds the preset sliding boundary. The system enters the parameter adaptive adjustment state. The preset sliding mode boundary value of 0.12 is determined by combining offline calibration experiments with the system safety margin. In the calibration process, the drive control unit operates without feedforward compensation, and the relative amplitude of the stator current fundamental distortion component caused by the dead zone is measured to be 0.05. Considering the nonlinear load fluctuations and external high-frequency interference under high-power alternating conditions, the system introduces a safety control coefficient of 2.4 times. The steady-state distortion benchmark is multiplied by the safety coefficient, thereby deriving the sliding mode boundary value of 0.12. When the magnitude of the high-dimensional nonlinear harmonic disturbance tensor is lower than 0.12, it indicates that the system is in the safe steady-state control domain and there is no need to start feedforward optimization. When it is higher than 0.12, the adaptive adjustment is precisely triggered, avoiding the instantaneous overload of computing power caused by the microprocessor frequently performing table lookup optimization.

[0043] The system retrieves the optimal feedforward control gain in real time based on a preset discretized step mapping table. The mapping table was preset by the offline calibration experiment. By traversing the load fluctuation range at the rated speed of 3000 rpm, the gain values ​​corresponding to different current distortion rates were recorded. Within a 50μs control interruption period, the system, based on the dynamic voltage compensation vector... Execution instruction modification, modification logic follows ,in, The final output is the adaptive adjustment instruction. Reference voltage command, The set adaptive weight coefficients, The dynamic gain stabilization factor in the adaptive adjustment command calculation is related to the inverter voltage utilization conditioning constant. The system exhibits a monotonic numerical correspondence. When the impedance of phase A winding deviates due to local temperature rise, the unbalanced arbitration unit performs cross-channel residual state exchange. When the monitored residual of phase A differs from that of the other channels by 22%, the system automatically adjusts the residual of phase A. Adjust to 0.35, and connect phase B and phase C. The synchronous adjustment is set to 1.15. By adjusting the compensation gain of each intact channel, the drive control unit achieves the offsetting balance of the overall system harmonic disturbance, ensuring that the current residual of the drive link remains within 0.21A under unsteady conditions. It should be clarified that the aforementioned unbalance arbitration and weight adjustment logic has a hierarchical nested correspondence between multiple parallel control loops and between phases within a single loop. Specifically, the controlled motor is a multi-three-phase permanent magnet motor, containing multiple independent three-phase winding channels. In this embodiment, when the impedance deviation of the first phase winding in the first winding channel deviates due to local temperature rise, it will directly cause the entire first winding channel to deviate. The spatial magnetic field characteristics of the output are distorted, which causes the overall harmonic residual amplitude of the channel to differ from that of the adjacent second and third winding channels by more than 22%. After detecting the imbalance between the channels, the unbalance arbitration unit triggers cross-channel residual state exchange. It not only lowers the overall control command output weight of the first damaged channel at the channel level, but also goes deeper into the channel to directionally reduce the adaptive weight coefficient of the first phase that has failed to 0.35. The compensation intensity is then shared synchronously by the other two phases in the same channel and the adjacent intact channels, thus constructing a complete topology fault-tolerant closed loop both within and between channels.

[0044] The above description is only a few preferred embodiments of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, technical solutions formed by replacing the above-mentioned features with the technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A vibration harmonic suppression optimization system for multi-phase three-phase permanent magnet motors, characterized in that the system... include: The state quantity acquisition module is used to synchronously acquire the discrete sequence of phase current and the DC bus voltage pulsation sequence of the controlled motor through multiple parallel sampling channels, combine them to generate a multi-element joint data source and output it to the high-dimensional feature spatiotemporal decoupling unit. The high-dimensional feature spatiotemporal decoupling unit is used to use the phase advance calibration operator to shift the phase of the multi-element joint data source in the complex plane to compensate for the fixed transmission delay, and uses the spatiotemporal decoupling transformation matrix to separate the spatial magnetic field distortion components between multi-phase windings, and outputs the multi-dimensional harmonic disturbance feature matrix to the adaptive control strategy optimization unit. The adaptive control strategy optimization unit is used to update the adaptive filtering gain of the multidimensional harmonic disturbance characteristic matrix in real time at a set octave point using a variable structure discrete filter. When the amplitude of the multidimensional harmonic disturbance characteristic matrix exceeds the set sliding mode boundary, it retrieves and adjusts the feedforward control gain step by step according to the internally stored discretized step table. It converges and calculates the dynamic voltage compensation vector within a single 50μs interrupt cycle and outputs it to the adjustment command correction module. The adjustment command correction module is used to determine the dynamic adaptive weighting coefficient based on the current integral saturation coefficient, and to superimpose the dynamic voltage compensation vector onto the reference control voltage command in situ. The cascaded aliasing generates an adaptive adjustment command and outputs it to the power inverter circuit.

2. The vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor according to claim 1, characterized in that, The system also includes a cross-channel residual state exchange matrix and an unbalanced arbitration unit. Multiple sets of parallel control loops share the multidimensional harmonic disturbance characteristic matrix in real time through the cross-channel residual state exchange matrix. The unbalanced arbitration unit is used to monitor the harmonic residuals between each set of parallel control loops in real time, and when the mutual difference between the harmonic residual of any channel and the other channels exceeds 20%, the feedforward control gain is adjusted according to the mutual difference.

3. The vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor according to claim 1, characterized in that, The status acquisition module includes a current Hall sensor and a voltage sampling circuit. The current Hall sensor is used to synchronously acquire the discrete signals of multiple phase currents of the controlled motor at a sampling frequency of 20kHz, and extract the discrete sequence of phase currents after cascaded digital low-pass filtering. The voltage sampling circuit is used to capture the DC bus voltage of the power inverter circuit in real time and calculate the AC / DC decoupling impedance components to isolate the bus voltage pulsation sequence.

4. The vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor according to claim 1, characterized in that, The high-dimensional feature spatiotemporal decoupling unit is internally configured with a discrete-time lead phase calibration unit. The discrete-time lead phase calibration unit is used to calculate the lead compensation phase angle based on the inherent transmission delay of the digital microprocessor, and to use the lead compensation phase angle to translate and align the coordinates to eliminate the phase misalignment of the discrete sequence of phase current. The high-dimensional feature spatiotemporal decoupling unit is also used to decouple the aligned multi-channel data stream through the spatiotemporal decoupling transformation matrix, and to separate the spatial magnetic field distortion component to generate a multidimensional harmonic disturbance feature matrix.

5. The vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor according to claim 1, characterized in that, The adaptive control strategy optimization unit includes a sliding mode boundary monitor; wherein, the sliding mode boundary monitor is used to track the modulus change of the multidimensional harmonic disturbance characteristic matrix in real time, and when the modulus change exceeds the set safety margin threshold, it calls the variable structure discrete filter to update the adaptive filter gain at the third, fifth and seventh harmonics of the fundamental frequency.

6. The vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor according to claim 1, characterized in that, The adaptive control strategy optimization unit internally stores a discretized step-by-step mapping table. When the amplitude of the multidimensional harmonic disturbance characteristic matrix exceeds the set sliding mode boundary, the adaptive control strategy optimization unit calls the discretized step-by-step mapping table to retrieve the feedforward gain correction value step by step, and adjusts the feedforward control gain according to the retrieved discrete gain step amount.

7. The vibration harmonic suppression optimization system for multi-phase three-phase permanent magnet motors according to claim 1, characterized in that, The system also includes a space voltage vector modulation unit; wherein, the space voltage vector modulation unit is used to receive adaptive adjustment commands and compensate the nonlinear dead-zone voltage deviation of the power inverter circuit in the field to generate multiple pulse width modulation signals, and output the pulse width modulation signals to the power inverter circuit.

8. The vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor according to claim 2, characterized in that, The unbalanced arbitration unit is used to generate circulating current suppression commands and load reduction constraint weights between channels based on the amplitude ratio of the mutual difference when the mutual difference between the harmonic residual of any channel and the other channels exceeds 20%. The load reduction constraint weights are then distributed and loaded into the adaptive control strategy optimization unit of each set of parallel control loops to correct the feedforward control gain of each channel.

9. The vibration harmonic suppression optimization system for a multi-phase three-phase permanent magnet motor according to claim 1, characterized in that, The system is deployed on a multi-core industrial microprocessor platform. The state quantity acquisition module, the high-dimensional feature spatiotemporal decoupling unit, the adaptive control strategy optimization unit, and the adjustment instruction correction module pipeline data between multiple hardware cores of the multi-core industrial microprocessor platform, and the hardware cores are synchronized and aligned through hardware-triggered interrupt signals triggered by the master clock.

10. A method for optimizing vibration harmonic suppression of a multi-phase three-phase permanent magnet motor, applied to the multi-phase three-phase permanent magnet motor vibration harmonic suppression optimization system according to any one of claims 1-9, characterized in that, Includes the following steps: Step S1: The state quantity acquisition module synchronously acquires the phase current discrete sequence and DC bus voltage pulsation sequence of the controlled motor through multiple parallel sampling channels, combines them to generate a multi-element joint data source and outputs it to the high-dimensional feature spatiotemporal decoupling unit. Step S2: The phase advance calibration operator inside the high-dimensional feature spatiotemporal decoupling unit is used to shift the phase of the multi-element joint data source in the complex plane to compensate for the fixed transmission delay. The spatiotemporal decoupling transformation matrix is ​​used to separate the spatial magnetic field distortion components between the multi-phase windings and output the multi-dimensional harmonic disturbance feature matrix to the adaptive control strategy optimization unit. Step S3: The adaptive control strategy optimization unit uses the variable structure discrete filter inside to update the adaptive filter gain of the multidimensional harmonic disturbance feature matrix in real time at the set octave point. When the amplitude of the multidimensional harmonic disturbance feature matrix exceeds the set sliding mode boundary, the feedforward control gain is retrieved and adjusted step by step according to the internally stored discretized step table. The dynamic voltage compensation vector is calculated and output to the adjustment command correction module within a single 50μs interrupt cycle. Step S4: The adjustment command correction module determines the dynamic adaptive weighting coefficient based on the current integral saturation coefficient, and the dynamic voltage compensation vector is superimposed on the reference control voltage command in situ. The cascaded superposition generates an adaptive adjustment command and outputs it to the power inverter circuit.

Images