Motor harmonic suppression adaptive system and method based on active disturbance rejection control

Abstract

The invention relates to the technical field of motor control, and discloses a motor harmonic suppression adaptive system and method based on active disturbance rejection control, and the system comprises a harmonic separation module which is used for separating multi-band harmonic components in motor current in real time; the multi-mode resonance extended state observer (MRESO) comprises a plurality of independent sub-modules which are respectively corresponding to disturbance observation of different harmonic frequency bands; the fuzzy RBF neural network parameter setting device is used for dynamically adjusting the observer bandwidth and the nonlinear feedback gain of the MRESO, and the dynamic feed-forward compensator is used for generating a backward voltage signal based on a harmonic impedance model. According to the invention, harmonic waves can be accurately separated and compensated, and the total harmonic distortion rate is reduced; harmonic frequency band modes are distinguished, and the observation precision is improved; parameters can be dynamically adjusted according to various variables to adapt to wide-range working condition changes; intelligent decision is made according to a fuzzy rule base, and intelligent control is achieved; a harmonic impedance matrix is identified on line through a recursive least square method, and the dynamic feed-forward compensation effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control, and more particularly to a motor harmonic suppression adaptive system and method based on active disturbance rejection control. Background Art

[0002] In the field of motor control, certain progress has been made in harmonic suppression technology, but there are still many problems to be solved in existing methods. The following conducts a comparative analysis of the deficiencies of existing technologies through relevant patented technologies.

[0003] Limitations of Traditional PI Control and Harmonic Injection Method

[0004] When dealing with high-frequency harmonics, traditional PI controllers are difficult to balance phase lag and steady-state error. For example, in the patent "Permanent Magnet Synchronous Generator Current Harmonic Suppression Method Based on Active Disturbance Rejection Repetitive Control" with the publication number CN114123900A, although an active disturbance rejection controller is introduced, its core relies on a linear extended state observer, and its ability to separate multi-band harmonics is poor, making it impossible to achieve precise harmonic dynamic compensation.

[0005] The harmonic injection method depends on an accurate motor model. Once the parameters are mismatched or the load changes suddenly, the control performance will be greatly reduced. At the same time, the periodic delay characteristic of the repetitive controller also limits its dynamic response speed.

[0006] Insufficient Adaptability of Conventional Active Disturbance Rejection Control (ADRC)

[0007] Existing active disturbance rejection technologies mostly adopt a single observer structure. Taking the patent "A Speed Control System for Sensorless Permanent Magnet Synchronous Motor Based on Active Disturbance Rejection Control" with the publication number CN118826562A as an example, this patent improves the anti-disturbance ability through a Luenberger observer or PI control, but its observer design does not distinguish the modes in the harmonic frequency band and is insufficient in capturing periodic harmonic characteristics. Simulations show that in the case of the superposition of the 5th and 7th harmonics, the total harmonic distortion (THD) is still higher than 2%.

[0008] In addition, the fixed-parameter ADRC adopted in the patent "A Composite Active Disturbance Rejection Controller for Asynchronous Motors" with the publication number CN119030384A is difficult to adapt to wide-range operating conditions. When the load changes suddenly, manual experience is often required to adjust the parameters, and the degree of intelligence is low.

[0009] Technical Bottlenecks in Harmonic Separation and Parameter Adaptation

[0010] Existing harmonic detection technologies often use FFT or phase-locked loops. For example, in the patent "A Motor Control Method Based on an Adaptive Fuzzy Active Disturbance Rejection Algorithm" with the publication number CN118311873A, although fuzzy inference is introduced to optimize parameters, a real-time harmonic separation module is not integrated, resulting in a limited control bandwidth and the inability to independently compensate for multi-band harmonics.

[0011] Another example is the patent "A Design Method of a Position Controller for a Permanent Magnet Synchronous Motor Based on Active Disturbance Rejection" with the publication number CN119298736A. The parameter tuning of its active disturbance rejection control scheme depends on offline optimization (such as the particle swarm algorithm) and lacks an online dynamic adjustment mechanism, making it difficult to handle time-varying disturbances during motor operation.

[0012] In view of the above problems, the present invention proposes a motor harmonic suppression adaptive system and method based on active disturbance rejection control, aiming to effectively solve the above technical problems. Summary of the Invention

[0013] To overcome the above defects of the existing technology, the present invention provides a motor harmonic suppression adaptive system and method based on active disturbance rejection control to solve the problems in the above background technology.

[0014] The present invention provides the following technical solutions: A motor harmonic suppression adaptive system and method based on active disturbance rejection control, including:

[0015] A harmonic separation module for real-time separating multi-band harmonic components in the motor current;

[0016] A multi-modal resonant extended state observer (MRESO), which includes multiple independent sub-modules, respectively corresponding to the disturbance observation of different harmonic frequency bands;

[0017] A fuzzy RBF neural network parameter tuner for dynamically adjusting the observer bandwidth and non-linear feedback gain of the MRESO;

[0018] A dynamic feedforward compensator for generating a reverse voltage signal based on a harmonic impedance model and synthesizing a final control quantity with the output of the active disturbance rejection controller.

[0019] Furthermore, each sub-module of the multi-modal MRESO includes:

[0020] A complex coefficient resonant controller for generating orthogonal harmonic reference signals;

[0021] A band-pass filter bank, whose cut-off frequency is dynamically adjusted according to the harmonic order k and the frequency offset Δω, and the expression is:

[0022] f cut = kω e ±Δω, where ω e is the electrical angular frequency of the motor.

[0023] Furthermore, the transfer function of the complex coefficient resonant controller is as follows:

[0024]

[0025] where k r is the resonant gain, and k is the harmonic order (5th / 7th / 11th order).

[0026] Furthermore, the harmonic separation module includes:

[0027] A synchronous rotating coordinate transformation unit that converts three-phase current to the harmonic synchronous rotating coordinate system;

[0028] A complex coefficient harmonic extraction network that extracts the amplitudes and phases of each harmonic through a multi-path parallel resonator group to generate a harmonic spectrum feature vector.

[0029] Furthermore, the input layer variables of the fuzzy RBF neural network parameter tuner include:

[0030] The change in current harmonic distortion rate ΔTHD;

[0031] The variance of rotational speed fluctuation

[0032] The harmonic frequency offset Δf;

[0033] The output layer parameters include the ESO bandwidth correction coefficient β and the nonlinear feedback gain α.

[0034] Furthermore, the fuzzy rule base of the parameter tuner includes:

[0035] Rule 1: If ΔTHD > 2% and then increase the ESO bandwidth by 20% and increase the feedback gain α to 1.5 times;

[0036] Rule 2: If Δf > 10 Hz continuously for more than 3 control cycles, then trigger the online update of the neural network weights.

[0037] Furthermore, the harmonic voltage generation formula of the dynamic feedforward compensator is:

[0038]

[0039] where Z dq (jω h ) is the harmonic impedance matrix identified online through the recursive least squares method, is the reference value of the harmonic current.

[0040] Furthermore, the identification method of the harmonic impedance matrix is:

[0041] Inject a harmonic voltage test signal with an amplitude of 5% of the rated current into the motor;

[0042] Based on the voltage-current response data, update the real and imaginary parts of Z every 5 ms. dq

[0043] Furthermore, an adaptive method for motor harmonic suppression based on active disturbance rejection control includes the following steps:

[0044] S1. Collect the three-phase current signals of the motor and decompose them into harmonic components in the synchronous rotating coordinate system through coordinate transformation;

[0045] S2. Extract the amplitudes and phases of each harmonic through a complex coefficient resonator bank and construct a harmonic spectrum feature vector;

[0046] S3. Multimodal MRESO performs independent disturbance observation on the harmonics in each frequency band and outputs the total disturbance estimation value;

[0047] S4. Dynamically allocate the weight coefficients of each ESO sub-module according to the harmonic energy ratio;

[0048] S5. The fuzzy neural network adjusts the ESO bandwidth and feedback gain in real time according to the load mutation gradient and harmonic frequency deviation;

[0049] S6. The dynamic feedforward compensator generates a reverse harmonic voltage, which is superimposed with the ADRC output to generate the final PWM control signal.

[0050] Furthermore, the weight coefficient allocation formula in S4 is:

[0051]

[0052] Where is the effective value of the k-th harmonic current.

[0053] The technical effects and advantages of the present invention:

[0054] The present invention can accurately separate and compensate harmonics, reduce the total harmonic distortion rate; distinguish the harmonic frequency band modes to improve the observation accuracy; can dynamically adjust parameters according to various variables to adapt to wide-range working condition changes; make intelligent decisions based on the fuzzy rule base to achieve intelligent control; and identify the harmonic impedance matrix online through the recursive least squares method to improve the dynamic feedforward compensation effect.

[0055] Specifically:

[0056] High-efficiency and precise harmonic suppression ability: By the collaborative work of the harmonic separation module, the multi-modal resonant extended state observer (MRESO), and the dynamic feedforward compensator, the present invention realizes the precise separation, accurate observation, and effective compensation of multi-band harmonics in the motor current. The harmonic separation module separates harmonic components in real time. MRESO independently observes disturbances for different harmonic frequency bands, and the dynamic feedforward compensator generates a reverse voltage to cancel the harmonic influence. Compared with the traditional PI control and harmonic injection method, it overcomes the defects of phase lag, difficulty in balancing steady-state error, and dependence on an accurate motor model. Under the condition of the superposition of the 5th and 7th harmonics, it can significantly reduce the total harmonic distortion rate (THD), make the motor current waveform closer to a sine wave, improve the power quality, reduce the motor loss and heating, and extend the service life of the motor.

[0057] Powerful operating condition adaptability: The fuzzy RBF neural network parameter tuner dynamically adjusts the observer bandwidth and non-linear feedback gain of MRESO according to multiple variables such as the change in current harmonic distortion rate, the variance of speed fluctuation, and the deviation of harmonic frequency. In the face of complex operating conditions such as load mutation, speed fluctuation, and harmonic frequency drift during motor operation, the system can quickly and adaptively adjust the control parameters. Compared with the conventional active disturbance rejection control technology with fixed parameters, it can automatically adapt to wide-range operating condition changes without manual intervention, maintain the stable and efficient operation of the motor, and improve the reliability and stability of the system.

[0058] Highly intelligent control level: The fuzzy rule base of the parameter tuner sets rules such as "if ΔTHD > 2% and then increase the ESO bandwidth by 20% and increase the feedback gain α to 1.5 times" "if Δf > 10Hz continuously exceeds 3 control cycles, then trigger the online update of the neural network weights", etc. The system can automatically trigger corresponding control strategies based on real-time monitoring data to achieve intelligent decision-making and dynamic optimization. This feature solves the problem in the prior art that parameter tuning depends on offline optimization and lacks an online dynamic adjustment mechanism, making the motor control more intelligent and efficient, reducing labor costs, and improving the overall operating efficiency of the motor system.

[0059] Accurate identification of the harmonic impedance matrix: The dynamic feedforward compensator online identifies the harmonic impedance matrix by the recursive least squares method, injects a harmonic voltage test signal of 5% of the rated current into the motor, and updates the real and imaginary parts of the impedance matrix every 5ms. This accurate online identification method ensures the accuracy of harmonic voltage generation, makes the dynamic feedforward compensation more precise, effectively improves the harmonic suppression effect, and enhances the system's dynamic tracking and compensation ability for motor harmonics. Description of the Drawings

[0060] Figure 1 is the flowchart of the method in the present invention;

[0061] Figure 2 is the system architecture diagram of the present invention. Detailed implementation manners

[0062] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0063] It can be understood that the terms "first", "second", etc. used in the present application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish the first element from another element.

[0064] In order to enable those skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present application.

[0065] Embodiment:

[0066] The present invention provides a motor harmonic suppression adaptive system and method based on active disturbance rejection control, including:

[0067] A harmonic separation module for separating multi-band harmonic components in the motor current in real time;

[0068] It is implemented by using a digital signal processor (DSP) or a field programmable gate array (FPGA). In the hardware circuit, first, the three-phase current signals of the motor are collected through current sensors and converted into voltage signals suitable for chip processing through conditioning circuits. For the synchronous rotating coordinate transformation unit, according to the electrical angular frequency of the motor, the three-phase current is converted to the harmonic synchronous rotating coordinate system by using trigonometric function calculation. The complex coefficient harmonic extraction network extracts the amplitudes and phases of each harmonic by writing an algorithm of a multi-path parallel resonator group in the chip and using its frequency selection characteristics for specific frequencies, and finally generates a harmonic spectrum feature vector;

[0069] The specific extraction measures for harmonic extraction are as follows:

[0070] Hardware circuit basis: The three-phase current signals of the motor are collected by using current sensors (such as LEM Hall current sensors), and the collected analog signals are converted into digital signals suitable for chip processing through an A / D converter, providing a data basis for subsequent harmonic extraction.

[0071] Synchronous Rotating Coordinate Transformation: This transformation is implemented by leveraging the rich logic resources of the FPGA. A trigonometric function lookup table is programmed inside the FPGA, and based on the electrical angular frequency of the motor, the sine and cosine values required for the coordinate transformation are calculated in real time. The collected three-phase current signals are multiplied by the corresponding sine and cosine values respectively, and then through the operations of adders and subtracters, the three-phase current is converted to the harmonic synchronous rotating coordinate system, and the harmonic components in this coordinate system are obtained. This hardware implementation method can process data quickly and in parallel, meeting the real-time requirements of harmonic extraction.

[0072] Complex Coefficient Harmonic Extraction Network: This network is implemented by programming in a Digital Signal Processor (DSP). According to the frequencies of each harmonic, a multi-path parallel digital resonator is designed, and each resonator adopts a structure based on a second-order Infinite Impulse Response (IIR) filter. By adjusting the coefficients of the filter, each resonator can resonate at a specific harmonic frequency. The current signal after synchronous rotating coordinate transformation is input into the multi-path parallel resonator group. Utilizing the frequency selection characteristics of the resonator for specific frequency signals, after filtering, the amplitude and phase information of each harmonic are extracted. The DSP has the ability of fast multiplication and accumulation operations, which can efficiently complete the harmonic extraction task, and organize the extracted amplitude and phase information into a harmonic spectrum feature vector, which is stored in the memory of the DSP for subsequent use.

[0073] Specific Algorithm Implementation: In the DSP, specific coefficients are set for each resonator so that it can effectively filter specific harmonic frequencies. For example, for the 5th harmonic, according to its frequency, the coefficients of the resonator are calculated and set so that the resonator can accurately extract the amplitude and phase of the 5th harmonic. In the software code, through loop and conditional judgment statements, the multi-path parallel resonator group is managed and controlled to ensure that each resonator can work properly, and the extracted harmonic information is integrated to generate a harmonic spectrum feature vector;

[0074] Multi-modal Resonant Extended State Observer (MRESO), which contains multiple independent sub-modules, corresponding to the disturbance observations of different harmonic frequency bands respectively;

[0075] Based on a microcontroller (such as the STM32 series) for development. The complex coefficient resonant controller of each sub-module, according to the given transfer function is transformed into the form of a difference equation through discretization processing, and in the microcontroller, an orthogonal harmonic reference signal is calculated and generated by means of loop iteration. The band-pass filter bank can be implemented using an Infinite Impulse Response (IIR) filter or a Finite Impulse Response (FIR) filter. According to the cut-off frequency expression f cut = kω e ±Δω, the coefficients of the filter are dynamically configured in the software to achieve the filtering process of signals in a specific harmonic frequency band;

[0076] A fuzzy RBF neural network parameter tuner is used to dynamically adjust the observer bandwidth and non - linear feedback gain of MRESO;

[0077] Build a model on a computer with the help of the Python language and related machine learning libraries (such as TensorFlow or PyTorch). Take the change in current harmonic distortion rate ΔTHD, the variance of rotational speed fluctuation The harmonic frequency offset Δf as the input - layer variables, and collect these data in real - time through a data acquisition card and input them into the model. The fuzzy rule base writes logical judgment codes according to the set rules (such as Rule 1: If ΔTHD>2% and then increase the ESO bandwidth by 20% and increase the feedback gain α to 1.5 times. For example, Rule 2: If Δf>10Hz continuously exceeds 3 control cycles, then trigger the online update of the neural network weights). Optimize the network through training so that it outputs the ESO bandwidth correction coefficient β and the non - linear feedback gain α, and then transmit these parameters to MRESO.

[0078] A dynamic feed - forward compensator generates a reverse voltage signal based on the harmonic impedance model and synthesizes the final control quantity with the output of the active disturbance rejection controller.

[0079] With a dedicated power - electronic control chip (such as TMS320F28335 of TI) as the core. According to the harmonic voltage generation formula First, online identify the harmonic impedance matrix Z dq (jω h ) by the recursive least - squares method. In hardware, inject a harmonic voltage test signal with an amplitude of 5% of the rated current into the motor, collect voltage - current response data using voltage sensors and current sensors, and run the recursive least - squares algorithm in the chip every 5ms to update the real and imaginary parts of Z dq After calculating the reverse harmonic voltage synthesize the final control quantity with the output of the active disturbance rejection controller through addition operation in the chip to drive the power inverter of the motor.

[0080] Each sub - module of the multi - modal MRESO includes:

[0081] A complex - coefficient resonant controller is used to generate orthogonal harmonic reference signals;

[0082] At the software level, use MATLAB / Simulink as the development tool for algorithm design and verification. According to the transfer function of the complex - coefficient resonant controller Obtain the time - domain expression by using the inverse Laplace transform, and then convert it into a difference equation of a discrete - time system through a discretization method (such as the bilinear transformation method). Build the corresponding module in Simulink, input the electrical angular frequency ω of the motor e, the resonance gain k r Parameters such as the harmonic order k, etc., can generate orthogonal harmonic reference signals. Then, this algorithm is transplanted to run on an actual control chip (such as a DSP).

[0083] A band - pass filter bank, whose cut - off frequencies are dynamically adjusted according to the harmonic order k and the frequency offset Δω, and the expression is:

[0084] f cut = kω e ±Δω, where ω e is the electrical angular frequency of the motor.

[0085] Adopt a combination of analog circuits and digital signal processing. For the analog part, an integrated analog band - pass filter chip (such as MAX274, etc.) can be selected, and the center frequency and bandwidth range of the filter are initially set through the configuration of external resistors and capacitors. The digital part uses a microprocessor (such as the ARM series) for precise frequency adjustment. According to the formula f cut = kω e ±Δω, the cut - off frequency is calculated in real - time, and then a control voltage is output through a digital - to - analog converter (DAC) to finely adjust the center frequency of the analog filter, realizing precise filtering of signals in different harmonic frequency bands;

[0086] The transfer function of the complex - coefficient resonance controller is:

[0087]

[0088] where k r is the resonance gain, and k is the harmonic order (5 / 7 / 11 times).

[0089] In hardware implementation, an operational amplifier is used to build an analog circuit to approximately implement this transfer function. Utilize the integration and differentiation characteristics of the operational amplifier, combined with the feedback network composed of resistors and capacitors, to construct a second - order band - pass filter structure. According to the parameters k r , k and ω e , calculate the specific values of the resistors and capacitors. In software implementation, digital signal processing algorithms are used. For example, in a DSP, direct digital frequency synthesis (DDS) technology is adopted to generate the corresponding frequency response signal according to the transfer function. First, the transfer function is discretized, and then the iterative calculation of the discrete system is realized through programming to generate the required signal.

[0090] The harmonic separation module includes:

[0091] A synchronous rotation coordinate transformation unit that converts three - phase currents to the harmonic synchronous rotation coordinate system;

[0092] Implemented in FPGA. Utilizing the rich logic resources of FPGA, a trigonometric function lookup table is programmed, and according to the electrical angular frequency ω of the motor e The sine and cosine values required for coordinate transformation are calculated in real time. The collected three-phase current signals are multiplied by the corresponding sine and cosine values respectively, and then through the operations of adders and subtracters, the transformation of the three-phase current to the harmonic synchronous rotating coordinate system is realized. This hardware implementation method can process data quickly and in parallel, meeting the real-time requirements.

[0093] The complex coefficient harmonic extraction network extracts the amplitudes and phases of each harmonic through a multi-path parallel resonator group, generating a harmonic spectrum feature vector.

[0094] Implemented by programming in a digital signal processor (DSP). According to the frequencies of each harmonic, a multi-path parallel digital resonator is designed. Each resonator adopts the structure based on a second-order infinite impulse response (IIR) filter, and by adjusting the coefficients of the filter, it resonates at a specific harmonic frequency. The current signal after synchronous rotating coordinate transformation is input into the multi-path parallel resonator group, and after filtering, the amplitudes and phase information of each harmonic are extracted. Utilizing the fast multiplication and accumulation operation ability of the DSP, the harmonic extraction task is efficiently completed, and the extracted information is organized into a harmonic spectrum feature vector.

[0095] The input layer variables of the described fuzzy RBF neural network parameter tuner include:

[0096] The change amount of current harmonic distortion rate ΔTHD;

[0097] The variance of rotational speed fluctuation ;

[0098] The harmonic frequency offset Δf;

[0099] The output layer parameters include the ESO bandwidth correction coefficient β and the nonlinear feedback gain α.

[0100] Using the Python language combined with the Keras library to build a fuzzy RBF neural network model. The three-phase current and rotational speed data during the operation of the motor are collected in real time through a data acquisition device (such as an NI data acquisition card). In the software, the change amount of current harmonic distortion rate ΔTHD is calculated based on the collected current data, and the variance of rotational speed fluctuation is calculated through statistical analysis of the rotational speed data The harmonic frequency offset Δf is obtained by using a frequency detection algorithm. After normalizing these data, they are input into the neural network model. After the training and inference of the network, the ESO bandwidth correction coefficient β and the nonlinear feedback gain α are output, and these parameters are transmitted to the multi-modal resonant extended state observer (MRESO).

[0101] The fuzzy rule base of the described parameter tuner includes:

[0102] Rule 1: If ΔTHD > 2% and then increase the ESO bandwidth by 20% and increase the feedback gain α to 1.5 times;

[0103] Rule 2: If Δf > 10Hz continuously exceeds 3 control cycles, then trigger the online update of the neural network weights.

[0104] Implement in the code logic of Python. Use conditional judgment statements (such as if-else statements) to judge the change in the current harmonic distortion rate ΔTHD, the variance of the rotational speed fluctuation and the harmonic frequency offset Δf collected. When the conditions of Rule 1 are met, modify the parameter value of the ESO bandwidth in MRESO through the code to increase it by 20%, and at the same time update the value of the feedback gain a to 1.5 times the original. For Rule 2, set a counter. When it is detected that Δf > 10Hz, the counter is incremented by 1. If the counter is greater than 0 for 3 consecutive control cycles, call the online update function of the neural network to adjust the weights of the network to optimize the performance of the fuzzy RBF neural network.

[0105] The harmonic voltage generation formula of the described dynamic feedforward compensator is:

[0106]

[0107] where Z dq (jω h ) is the harmonic impedance matrix identified online by the recursive least squares method, is the harmonic current reference value.

[0108] Implement on a power electronics control board, taking the TI's TMS320F28069 as an example. First, use the on-board ADC module to collect the voltage and current signals of the motor to obtain the voltage-current response data. Write the algorithm of the recursive least squares method in the software, and update the real and imaginary parts of the harmonic impedance matrix Z dq (jω h ) every 5ms according to the collected data. Substitute the known harmonic current reference value and the updated harmonic impedance matrix into the formula to calculate the reverse harmonic voltage

[0109] The identification method of the described harmonic impedance matrix is:

[0110] Inject a harmonic voltage test signal with an amplitude of 5% of the rated current into the motor;

[0111] Update Z once every 5ms based on the voltage-current response datadq The real part and the imaginary part.

[0112] On the hardware circuit, a harmonic voltage test signal with an amplitude of 5% of the rated current is generated by a signal generator (such as the Agilent 33500B series function generator), amplified by a power amplifier and then injected into the motor. The voltage sensor and current sensor on the motor drive board are used to collect the voltage and current response data of the motor. In terms of software, the recursive least squares algorithm is written in embedded C language in a microcontroller (such as the STM32F4 series). The collected voltage-current data is stored in the memory of the microcontroller in a certain format, and the recursive least squares function is called every 5 ms to update the real part and the imaginary part of the harmonic impedance matrix Zdq, realizing the online identification of the harmonic impedance matrix.

[0113] An adaptive method for motor harmonic suppression based on active disturbance rejection control includes the following steps:

[0114] S1. Collect the three-phase current signals of the motor, and decompose them into harmonic components in the synchronous rotating coordinate system through coordinate transformation;

[0115] Use a current sensor (such as a LEM Hall current sensor) to collect the three-phase current signals of the motor, and convert the collected analog signals into digital signals through an A / D converter. Write a coordinate transformation algorithm in a microcontroller (such as the STM32F7 series). According to the electrical angular frequency ω of the motor e , use trigonometric functions to calculate and convert the three-phase current to the synchronous rotating coordinate system to obtain harmonic components;

[0116] S2. Extract the amplitudes and phases of each harmonic through a complex coefficient resonator bank, and construct a harmonic spectrum feature vector;

[0117] It is implemented in a digital signal processor (DSP). According to the frequencies of each harmonic, design a complex coefficient resonator bank, and each resonator is tuned for a specific harmonic frequency. Input the current signal after coordinate transformation into the complex coefficient resonator bank, and extract the amplitude and phase information of each harmonic through filtering and calculation. Organize this information into a harmonic spectrum feature vector and store it in the memory of the DSP for subsequent use;

[0118] S3. Multimodal MRESO performs independent disturbance observation on the harmonics in each frequency band and outputs the total disturbance estimation value;

[0119] Implementing multi-modal MRESO based on FPGA. Utilize the parallel processing ability of FPGA to build each sub-module (complex coefficient resonant controller and band-pass filter bank) in parallel. Each sub-module corresponds to a different harmonic frequency band, processes the input harmonic signal, and realizes independent disturbance observation. Add the outputs of each sub-module through an adder to obtain the total disturbance estimation value, which is output to the subsequent control link.

[0120] S4. Dynamically allocate the weight coefficients of each ESO sub-module according to the harmonic energy ratio;

[0121] Implement in the software algorithm. According to the formula

[0122]

[0123] Use the effective values of the harmonic currents collected to calculate the weight coefficients of each ESO sub-module in the microcontroller. Recalculate the weight coefficients at regular time intervals (such as 10 ms) to adapt to the change of harmonic energy during the operation of the motor.

[0124] S5. The fuzzy neural network adjusts the ESO bandwidth and feedback gain in real time according to the load mutation gradient and harmonic frequency deviation;

[0125] Build a fuzzy neural network model using Python and TensorFlow. Collect the load torque and speed signals of the motor through sensors to calculate the load mutation gradient. At the same time, obtain the harmonic frequency deviation data. Input these data into the trained fuzzy neural network model, and after the inference calculation of the network, output the ESO bandwidth correction coefficient and feedback gain adjustment value. Transmit these adjustment values to the multi-modal MRESO through a communication interface (such as SPI or CAN) to achieve real-time adjustment.

[0126] S6. The dynamic feed-forward compensator generates a reverse harmonic voltage, which is superimposed on the output of the ADRC to generate the final PWM control signal.

[0127] Implement in a power electronics control chip (such as TI's TMS320F28377D). Calculate the reverse harmonic voltage according to the harmonic voltage generation formula of the dynamic feed-forward compensator, and use the internal arithmetic unit of the chip to superimpose the reverse harmonic voltage on the output of the active disturbance rejection controller (ADRC). Input the superimposed signal into the PWM generator module, and generate the final PWM control signal by configuring the parameters of the PWM generator (such as frequency, duty cycle, etc.) to drive the power inverter of the motor.

[0128] The weight coefficient allocation formula described in S4 is:

[0129]

[0130] wherein is the effective value of the k-th harmonic current.

[0131] In software programming, taking C language as an example. In the code of the microcontroller, first define an array to store the effective values of the 5th, 7th, and 11th harmonic currents After obtaining these effective values through the harmonic detection algorithm, use loop statements and multiplication and addition operations to calculate the denominator according to the formula Then calculate the weight coefficient w corresponding to each k value respectively k . Store the calculated weight coefficients in the memory for adjusting the weights of each ESO sub-module in the multi-modal MRESO.

[0132] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combinations of these technical features do not conflict, they should all be considered as the scope described in this specification.

[0133] The above embodiments only represent several implementation manners of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent should be subject to the appended claims.

[0134] The above is only the preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should all be included in the protection scope of the present invention.

Claims

1. An adaptive system for motor harmonic suppression based on active disturbance rejection control, characterized in that, Comprising: A harmonic separation module, which is used to separate multi - band harmonic components in the motor current in real time; A multi - modal resonant extended state observer (MRESO), which includes multiple independent sub - modules, corresponding to the disturbance observation of different harmonic frequency bands respectively; A fuzzy RBF neural network parameter tuner, which is used to dynamically adjust the observer bandwidth and non - linear feedback gain of MRESO; A dynamic feed - forward compensator, which generates a reverse voltage signal based on the harmonic impedance model and synthesizes the final control quantity with the output of the active disturbance rejection controller.

2. The adaptive system for suppressing motor harmonics based on active disturbance rejection control according to claim 1, characterized in that: Each sub - module of the multi - modal MRESO includes: A complex - coefficient resonant controller, which is used to generate orthogonal harmonic reference signals; A band - pass filter bank, whose cut - off frequency is dynamically adjusted according to the harmonic order k and the frequency offset Δω, and the expression is: f cut = kω e ±Δω, where ω e is the electrical angular frequency of the motor.

3. An adaptive system for suppressing motor harmonics based on active disturbance rejection control according to claim 2, characterized in that: The transfer function of the complex - coefficient resonant controller is: where k r is the resonance gain, and k is the harmonic order (5th / 7th / 11th).

4. An adaptive system for motor harmonic suppression based on active disturbance rejection control according to claim 1, characterized in that: The harmonic separation module includes: A synchronous rotating coordinate transformation unit, which converts three - phase current to the harmonic synchronous rotating coordinate system; A complex - coefficient harmonic extraction network, which extracts the amplitude and phase of each harmonic through a multi - path parallel resonator group and generates a harmonic spectrum feature vector.

5. An adaptive system for motor harmonic suppression based on active disturbance rejection control according to claim 1, characterized in that: The input - layer variables of the fuzzy RBF neural network parameter tuner include: The change in current harmonic distortion rate ΔTHD; Variance of rotational speed fluctuation The harmonic frequency offset Δf; The output - layer parameters include the ESO bandwidth correction coefficient β and the non - linear feedback gain α.

6. The adaptive system for motor harmonic suppression based on active disturbance rejection control according to claim 5, wherein: The fuzzy rule base of the parameter tuner contains: Rule 1: If ΔTHD > 2% and then increase the ESO bandwidth by 20% and increase the feedback gain Δ to 1.5 times; Rule 2: If Δf>10Hz lasts for more than 3 control cycles, then trigger the online update of the neural network weights.

7. An adaptive system for suppressing motor harmonics based on active disturbance rejection control according to claim 1, wherein: The harmonic voltage generation formula of the dynamic feed - forward compensator is: where Z dq (jω h ) is the harmonic impedance matrix identified online by the recursive least squares method, is the reference value of the harmonic current.

8. An adaptive system for suppressing motor harmonics based on active disturbance rejection control according to claim 7, characterized in that: The identification method of the harmonic impedance matrix is: Inject a harmonic voltage test signal with an amplitude of 5% of the rated current into the motor; Update the real and imaginary parts of Z every 5 ms based on the voltage-current response data dq respectively.

9. An adaptive method for motor harmonic suppression based on active disturbance rejection control, characterized in that, Including the following steps: S1. Collect the three - phase current signals of the motor, and decompose them into harmonic components in the synchronous rotating coordinate system through coordinate transformation; S2. Extract the amplitude and phase of each harmonic through a complex - coefficient resonator group, and construct a harmonic spectrum feature vector; S3. The multi - modal MRESO conducts independent disturbance observation on each frequency - band harmonic and outputs the total disturbance estimation value; S4. Dynamically allocate the weight coefficients of each ESO sub - module according to the harmonic energy ratio; S5. The fuzzy neural network adjusts the ESO bandwidth and feedback gain in real time according to the load mutation gradient and the harmonic frequency deviation; S6. The dynamic feed - forward compensator generates a reverse harmonic voltage, which is superimposed with the ADRC output to generate the final PWM control signal.

10. An adaptive method for motor harmonic suppression based on active disturbance rejection control according to claim 9, characterized in that: The weight - coefficient allocation formula in S4 is: wherein is the effective value of the k-th harmonic current.

Images