Bus voltage auxiliary method for suppressing position error of electrolytic capacitor-free permanent magnet motor

Abstract

The application discloses a bus voltage auxiliary non-electrolytic capacitor permanent magnet motor position error suppression method, belongs to the technical field of motor control, and aims at solving the position estimation error fluctuation problem caused by the direct current bus voltage ripple. The application comprises the following steps: collecting the direct current bus voltage and the stator current, obtaining the extended back electromotive force observation value and the preliminary estimated electric angle of the last control period; extracting the twice frequency ripple component of the direct current bus voltage and constructing the synchronous quadrature input base; transforming the extended back electromotive force to the estimated synchronous rotating coordinate system to obtain the d-axis extended back electromotive force component; adopting the ripple synchronous adaptive linear neuron to fit and suppress the d-axis extended back electromotive force component to obtain the d-axis extended back electromotive force component after the ripple suppression; obtaining the preliminary estimated electric angle of the current control period through the preliminary position phase-locked loop; constructing the fundamental component adaptive linear neuron input base to reconstruct the alpha-beta axis extended back electromotive force fundamental component; and obtaining the final estimated electric angle and electric angular velocity through the phase-locked loop, which are used for the position sensorless closed-loop control.

Description

Technical Field

[0001] This invention belongs to the field of motor control technology, specifically relating to a method for suppressing position errors in a permanent magnet motor without electrolytic capacitors and assisted by bus voltage. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) offer advantages such as high power density, high efficiency, fast dynamic response, and reliable operation, making them widely used in electric drive systems for electric transportation, industrial servos, home appliances, and aerospace. Traditional motor drives typically employ large-capacity electrolytic capacitors to stabilize the DC bus voltage. However, electrolytic capacitors suffer from drawbacks such as large size, short lifespan, and temperature sensitivity, which can limit system power density and long-term reliability. Electrolytic capacitor-free PMSM drive systems, which replace electrolytic capacitors with small-capacity film capacitors, offer improved system lifespan and reduced size, thus attracting widespread attention.

[0003] In electrolytic capacitor-free drive systems, the DC bus capacitor value is significantly reduced, resulting in a noticeable ripple at twice the grid frequency and its harmonics on the DC bus voltage. This bus voltage ripple further couples to the inverter output voltage, motor stator current, and extended back EMF observation circuitry, generating sideband harmonics in the extended back EMF. For sensorless control methods based on extended back EMF, the extended back EMF phase information is directly used for rotor position estimation; therefore, its sideband harmonics cause fluctuations in position estimation errors, affecting current decoupling effectiveness and system stability.

[0004] To suppress extended back EMF harmonics and the resulting position estimation errors, existing methods typically incorporate a filtering stage between the extended back EMF observer and the phase-locked loop (PLL), or achieve specific harmonic suppression by reconstructing the observer's frequency characteristics. For example, complex filtering, generalized integral filtering, and master-slave observer structures can reduce extended back EMF sideband harmonics to some extent; however, these methods usually rely on preset center frequencies and filtering bandwidths. When changes in operating frequency, fluctuations in estimated rotational speed, or discretization errors cause center frequency shifts, the filtering performance and dynamic response are affected.

[0005] However, traditional filtering methods involve trade-offs between sideband harmonic suppression, fundamental phase accuracy, and transient convergence speed. Reducing the filter bandwidth enhances harmonic suppression but easily introduces fundamental phase lag, leading to DC bias in position estimation errors. Increasing the filter bandwidth improves phase characteristics and dynamic response but reduces the suppression of sideband harmonics. Therefore, how to fully utilize the amplitude and phase information of the DC bus voltage ripple to suppress extended back EMF sideband harmonics while maintaining fundamental phase accuracy and improving the transient convergence speed of position estimation errors has become a pressing technical problem to be solved in the sensorless control of electrolytic capacitor-free permanent magnet synchronous motors. Summary of the Invention

[0006] To address the issues of DC bus voltage ripple causing extended back EMF sideband harmonics in permanent magnet synchronous motor (PMSM) drive systems without electrolytic capacitors, leading to position estimation error fluctuations, and the problems of fundamental phase lag, DC bias in position estimation error, and slow transient convergence speed in traditional filtering methods, this invention provides a bus voltage-assisted position error suppression method for PSMs without electrolytic capacitors.

[0007] The present invention discloses a method for suppressing position errors in a permanent magnet motor without electrolytic capacitors assisted by bus voltage. The method includes the following steps:

[0008] S1. Collect the DC bus voltage and stator current of the permanent magnet synchronous motor electrolytic capacitor-free drive system, and obtain the extended back electromotive force observation value and the preliminary estimated electrical angle of the previous control cycle.

[0009] S2. Extract the second harmonic ripple component of the DC bus voltage from the DC bus voltage, and construct an orthogonal input base that is synchronized with the second harmonic ripple component of the DC bus voltage;

[0010] S3. Based on the preliminary estimated electrical angle of the previous control cycle, transform the observed value of the extended back electromotive force to the estimated synchronous rotating coordinate system to obtain the d-axis extended back electromotive force component.

[0011] S4. Based on the orthogonal input basis of the DC bus voltage second harmonic ripple component synchronization, the ripple synchronization adaptive linear neuron is used to fit and suppress the d-axis extended back EMF ripple component to obtain the ripple-suppressed d-axis extended back EMF component.

[0012] S5. Based on the d-axis extended back EMF component after ripple suppression, obtain the preliminary estimated electrical angle of the current control cycle through the preliminary position phase-locked loop.

[0013] S6. Based on the preliminary estimated electrical angle of the current control cycle, construct the input basis of the fundamental component adaptive linear neuron, and reconstruct the fundamental components of the α-axis extended back EMF and the β-axis extended back EMF.

[0014] S7. Input the reconstructed α-axis extended back EMF fundamental component and β-axis extended back EMF fundamental component into the final position phase-locked loop to obtain the final estimated electrical angle and the final estimated electrical angular velocity, and use them for sensorless closed-loop control.

[0015] Preferably, in step S1, the DC bus voltage includes a DC component and an AC ripple component with twice the grid frequency as the base frequency; the extended back EMF observation value is obtained by an extended back EMF observer, and the extended back EMF is determined based on the motor's d-axis inductance, q-axis inductance, d-axis current, q-axis current, electric angular velocity, and permanent magnet flux linkage.

[0016] Preferably, in step S2, the second harmonic ripple component of the DC bus voltage is the ripple component at twice the grid frequency in the DC bus voltage, which is extracted from the DC bus voltage using a bandpass filter; an in-phase component and a quadrature component that are synchronized with the second harmonic ripple component of the DC bus voltage are obtained through a quadrature signal generator, and the in-phase component and the quadrature component constitute the quadrature input base.

[0017] Preferably, steps S1 to S5 constitute a DC bus voltage harmonic-assisted position pre-estimation stage. This stage uses the ripple synchronous adaptive linear neuron to fit and suppress the ripple component in the d-axis extended back electromotive force, and then obtains the preliminary estimated electrical angle of the current control cycle through the preliminary position phase-locked loop.

[0018] Preferably, in step S4, the ripple synchronous adaptive linear neuron uses the orthogonal input basis as input, calculates the fitted d-axis extended back EMF ripple component through weight coefficients, subtracts the fitted d-axis extended back EMF ripple component from the d-axis extended back EMF component, and obtains the ripple-suppressed d-axis extended back EMF component; the weight coefficients are updated using the least mean square algorithm.

[0019] Preferably, in step S5, the preliminary position phase-locked loop includes a proportional-integral (PI) regulator. With zero as a given value, the ripple-suppressed d-axis extended back EMF component is fed into the PI regulator to obtain the preliminary estimated electrical angular velocity of the current control cycle. The preliminary estimated electrical angular velocity of the current control cycle is integrated to obtain the preliminary estimated electrical angle of the current control cycle. The PI regulator is configured with independent proportional coefficients and integral coefficients.

[0020] Preferably, steps S6 to S7 constitute an extended back EMF reconstruction stage based on fundamental component adaptive linear neurons. This stage constructs an input basis using the cosine and sine values ​​of the preliminary estimated electrical angle of the current control cycle. The α-axis and β-axis extended back EMF fundamental components are reconstructed using α-axis and β-axis fundamental component adaptive linear neurons, respectively. The reconstructed fundamental components are input into the final position phase-locked loop to obtain the final estimated electrical angle and the final estimated electrical angular velocity.

[0021] Preferably, the α-axis fundamental component adaptive linear neuron is configured with independent α-axis weight coefficients and updated using the least mean square algorithm, and the β-axis fundamental component adaptive linear neuron is configured with independent β-axis weight coefficients and updated using the least mean square algorithm; the update error of the α-axis weight coefficients is the difference between the observed α-axis extended back EMF and the reconstructed α-axis extended back EMF fundamental component, and the update error of the β-axis weight coefficients is the difference between the observed β-axis extended back EMF and the reconstructed β-axis extended back EMF fundamental component; the fundamental component adaptive linear neuron is configured with a dedicated update step size.

[0022] Preferably, in step S7, the final estimated electrical angle is used for the current loop Park transform and inverse Park transform, and the final estimated electrical angular velocity is used for the extended back EMF observer.

[0023] The beneficial effects of this invention are:

[0024] 1. Enhanced suppression of sideband harmonics in the extended back EMF, improving position estimation accuracy. Compared with traditional complex filtering methods, this invention uses the second harmonic ripple component of the DC bus voltage and its orthogonal components as the input basis of the ripple synchronization adaptive linear neuron. Therefore, the ripple synchronization adaptive linear neuron can directly suppress the ripple component in the d-axis extended back EMF, weakening the sideband harmonics of the extended back EMF at its source and reducing position estimation error fluctuations.

[0025] 2. This invention avoids fundamental phase lag and DC bias in position estimation errors, improving steady-state control performance. Traditional filtering methods typically rely on filter bandwidth to suppress sideband harmonics. When the bandwidth is small, it easily introduces fundamental phase lag in the extended back EMF, leading to DC bias in position estimation errors. This invention uses a ripple-synchronous adaptive linear neuron to suppress d-axis extended back EMF ripple and reconstructs the α-β axis extended back EMF fundamental components, maintaining fundamental phase accuracy while suppressing harmonics.

[0026] 3. Improve transient convergence speed and enhance adaptability to different operating conditions. The ripple synchronization input base constructed in this invention includes the amplitude and phase information of the DC bus voltage ripple, enabling the equivalent update step size of the ripple synchronization adaptive linear neuron to adjust with changes in the DC bus voltage ripple amplitude. Compared to adaptive methods using preset sine and cosine input bases, this invention reduces the learning burden of weight coefficients, which is beneficial for improving convergence speed under conditions of varying ripple amplitude and operating frequency.

[0027] In summary, this invention can reconstruct extended back EMF using DC bus voltage ripple information without changing the hardware structure of the permanent magnet synchronous motor electrolytic capacitor-free drive system. It has the advantages of strong sideband harmonic suppression capability, accurate fundamental phase, small DC bias for position estimation error, fast transient convergence speed, and ease of engineering implementation. Attached Figure Description

[0028] Figure 1 This is an overall control block diagram of the method described in this invention;

[0029] Figure 2 This is a schematic diagram of the signal processing principle of the strategy proposed in this invention;

[0030] Figure 3 This is a block diagram of the traditional position estimation error suppression control based on cross-decoupling complex filtering;

[0031] Figure 4 This is a flowchart illustrating the method described in this invention;

[0032] Figure 5 This is an equivalent structural block diagram of the ripple synchronous adaptive linear neuron in the embodiments of the present invention;

[0033] Figure 6 This is a schematic diagram of the equivalent mapping from DC bus voltage ripple to extended back electromotive force ripple and the weight decomposition relationship under different input bases in an embodiment of the present invention.

[0034] Figure 7 These are overall waveforms of DC bus voltage, extended back EMF, and position estimation error corresponding to different control methods at the rated operating frequency in specific embodiments. Figure 7 (a) is the overall waveform without the position estimation error suppression algorithm applied. Figure 7 (b) shows the overall waveform when using the traditional complex filtering method. Figure 7 (c) is the overall waveform diagram when the method of the present invention is used;

[0035] Figure 8 This is a schematic diagram of the extended back electromotive force component harmonic FFT analysis results corresponding to different control methods in specific embodiments, wherein... Figure 8 (a) is a schematic diagram of the analysis results without applying the position estimation error suppression algorithm. Figure 8 (b) is a schematic diagram of the analysis results when using the traditional complex filtering method. Figure 8 (c) is a schematic diagram of the analysis results when the method of the present invention is used;

[0036] Figure 9 This is a comparison diagram of the transient convergence process of DC bus voltage, extended back EMF component, and position estimation error when using the traditional complex filtering method and the method of this invention in a specific embodiment. Figure 9 (a) is a diagram of the transient convergence process when using the traditional complex filtering method. Figure 9 (b) is a diagram of the transient convergence process when using the method of the present invention;

[0037] Figure 10The specific embodiments show position estimation error fluctuation components and DC bias statistics at different operating frequencies, using the traditional complex filtering method and the method of this invention, without applying the algorithm. Figure 10 (a) is a statistical chart of position estimation error fluctuation components at different operating frequencies. Figure 10 (b) A statistical chart of DC bias for position estimation error at different operating frequencies. Detailed Implementation

[0038] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0039] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0040] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; and the term "one embodiment" means "at least one embodiment". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0041] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0042] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0043] Electrolytic capacitor-free permanent magnet synchronous motor drive systems use small-value film capacitors instead of traditional large-capacity electrolytic capacitors. Because the DC bus capacitor value is significantly reduced, the DC bus voltage will exhibit a noticeable ripple at twice the grid frequency and its harmonics, depending on the input grid voltage. For example... Figure 1As shown, the bus voltage ripple is modulated and coupled into the motor stator voltage and stator current through the inverter, generating sideband harmonic components in the stator current that are superimposed on the fundamental frequency. This, in turn, causes sideband harmonics in the extended back EMF observation. For sensorless control methods based on extended back EMF, the extended back EMF phase information is directly used for rotor position estimation. Its sideband harmonics cause fluctuations in position estimation errors, affecting the current decoupling effect and the stable operation performance of the system.

[0044] Traditional methods typically involve adding complex filtering or generalized integral filtering between the extended back EMF observer and the phase-locked loop to suppress sideband harmonics. Figure 3 This is a block diagram for suppressing position estimation errors based on traditional cross-decoupling complex filtering. The traditional position estimation error suppression method based on cross-decoupling complex filtering uses current feedback... and αβ axis voltage commands in a two-phase stationary coordinate system Voltage command , As input to the extended back EMF observer, the α-axis extended back EMF observation value is obtained. and β-axis extended back electromotive force observations Subsequently, the cross-decoupling stage utilizes the positive and negative sideband components of the αβ axis. , , , Cross-compensation was performed on the extended back EMF observations, and the fundamental components of the αβ-axis extended back EMF were extracted using complex filtering. , And sideband components. Extracted αβ-axis fundamental wave components. and Input phase-locked loop to obtain estimated electrical angle And estimating electric angular velocity This method can reduce position estimation error fluctuations to some extent, but the filtering effect depends on the filter's center frequency and bandwidth. When the bandwidth is small, although the sideband harmonic suppression capability is enhanced, it is easy to introduce fundamental phase lag, which in turn leads to DC bias in the position estimation error; when the bandwidth is large, the fundamental phase characteristics are improved, but the sideband harmonic suppression capability decreases.

[0045] This invention addresses the disturbance source by incorporating DC bus voltage ripple information into the adaptive filtering process. Figure 1 As shown in the overall system control diagram, the strategy proposed in this invention obtains the α-axis extended back EMF observation value in the two-phase stationary coordinate system through an extended back EMF observer. and β-axis extended back electromotive force observations Simultaneously, the DC bus voltage is collected. And extract its ripple component at twice the grid frequency.

[0046] See Figure 1 The position and connection relationship of the strategy proposed in this invention in the overall control block diagram are as follows: Figure 1 The system shown is a permanent magnet synchronous motor electrolytic capacitor-free drive system, including a single-phase AC power supply, a diode rectifier, and a DC-side inductor. DC bus thin film capacitor The system employs a three-phase voltage-source inverter (SVPWM) and a permanent magnet synchronous motor (PMSM). It utilizes a dual closed-loop control structure with an outer speed loop and an inner current loop, and the motor's electrical angular velocity setpoint is... With the final estimated electric angular velocity The q-axis current setpoint is obtained after subtraction via the speed loop proportional-integral controller (the first PI). d-axis current setpoint Set to zero. q-axis current setpoint. With q-axis current feedback value After subtraction, the q-axis voltage command is obtained through the q-axis current loop proportional-integral regulator (the second PI). d-axis current given With d-axis current feedback value After subtraction, the d-axis voltage command is obtained through the d-axis current loop proportional-integral regulator. q-axis voltage command and d-axis voltage command The α-axis voltage command is obtained through inverse Park transformation (dq / αβ). and β-axis voltage command The inverter drive signal is then generated via a space vector pulse width modulation module. The three-phase stator current of the permanent magnet synchronous motor... , , After sampling, the data is input into the extended back EMF observer, which outputs the α-axis extended back EMF observation value. and β-axis extended back electromotive force observations To the strategy module proposed in this invention; simultaneously, the DC bus voltage sampling value The strategy module proposed in this invention is also input. The strategy module proposed in this invention outputs the final estimated electrical angle. The current loop reverse Park transform module and the Park transform module (in the DC bus voltage harmonic auxiliary position pre-estimation module) output the final estimated electric angular velocity. This leads to the proportional-integral (PI) regulator of the speed loop. The strategy proposed in this invention introduces the second harmonic ripple component of the DC bus voltage and its orthogonal components into the position estimation stage. The input is the α-axis extended back EMF observation value output by the extended back EMF observer. and β-axis extended back electromotive force observations and DC bus voltage sampling value The output is the reconstructed α-axis extended back EMF fundamental component. and the fundamental component of the β-axis extended back electromotive force The reconstructed fundamental component is input into the final position phase-locked loop to obtain the final estimated electrical angle. and the final estimated electric angular velocity Finally, the electrical angle was estimated. and the final estimated electric angular velocity The output of the strategy proposed in this invention is used for sensorless closed-loop control.

[0047] Combination Figures 1 to 3 This invention provides a method for suppressing the position error of a permanent magnet motor without electrolytic capacitors, assisted by bus voltage. The method includes the following steps:

[0048] S1. Collect the DC bus voltage and stator current of the permanent magnet synchronous motor electrolytic capacitor-free drive system, and obtain the extended back electromotive force observation value and the preliminary estimated electrical angle of the previous control cycle.

[0049] Specifically, the controller acquires the DC bus voltage and the stator current of the permanent magnet synchronous motor in real time, and obtains the extended back EMF observation values ​​in a two-phase stationary coordinate system based on the extended back EMF observer. Simultaneously, the controller acquires the preliminary estimated electrical angle from the previous control cycle, which is used for coordinate transformation in subsequent steps. The extended back EMF observation values ​​contain sideband harmonic components caused by the DC bus voltage ripple, which are the targets for subsequent suppression. Acquiring these signals provides the data foundation for subsequent ripple extraction and adaptive suppression.

[0050] S2. Extract the second harmonic ripple component of the DC bus voltage from the DC bus voltage, and construct an orthogonal input base that is synchronized with the second harmonic ripple component of the DC bus voltage;

[0051] Specifically, since DC bus voltage ripple is a disturbance source causing sideband harmonics in the extended back EMF, this invention extracts the ripple component at twice the grid frequency from the DC bus voltage and constructs in-phase and quadrature components synchronized with this ripple component using a quadrature signal generator. The in-phase and quadrature components constitute the orthogonal input basis of the ripple-synchronized adaptive linear neuron. This input basis contains the amplitude and phase information of the disturbance source and will subsequently serve as the reference input for the ripple-synchronized adaptive linear neuron to fit the ripple component in the d-axis extended back EMF. Because the input basis is directly related to the disturbance source, the adaptive neuron does not need to additionally learn the characteristics of the disturbance source, thereby reducing the weight learning burden.

[0052] S3. Based on the preliminary estimated electrical angle of the previous control cycle, transform the observed value of the extended back electromotive force to the estimated synchronous rotating coordinate system to obtain the d-axis extended back electromotive force component.

[0053] Specifically, based on the preliminary estimated cosine and sine values ​​of the electrical angle from the previous control cycle, a coordinate transformation is performed on the observed values ​​of the α-axis extended back EMF and β-axis extended back EMF to obtain the d-axis and q-axis extended back EMF components in the estimated synchronous rotating coordinate system. In the estimated synchronous rotating coordinate system, the extended back EMF disturbance affected by the DC bus voltage ripple mainly manifests as the ripple component in the d-axis extended back EMF. Therefore, subsequent ripple fitting and suppression can be performed on the d-axis extended back EMF component.

[0054] S4. Based on the orthogonal input basis of the DC bus voltage second harmonic ripple component synchronization, the ripple synchronization adaptive linear neuron is used to fit and suppress the d-axis extended back EMF ripple component to obtain the ripple-suppressed d-axis extended back EMF component.

[0055] Specifically, the ripple-synchronous adaptive linear neuron uses the orthogonal input basis constructed in step S2 as input. It calculates the fitted d-axis extended back EMF ripple component by weighted summation of the input basis using weight coefficients. Subtracting the fitted d-axis extended back EMF ripple component from the d-axis extended back EMF component yields the ripple-suppressed d-axis extended back EMF component. The weight coefficients are updated using the least mean square algorithm, with the update error equal to the ripple-suppressed d-axis extended back EMF component. Since the input basis is synchronized with the DC bus voltage ripple, the ripple-synchronous adaptive linear neuron can directly adaptively fit and cancel the ripple component in the d-axis extended back EMF, thereby weakening the extended back EMF sideband harmonics at the source. Unlike traditional fixed-bandwidth filtering, this adaptive mechanism can automatically adjust the weights when operating conditions change, maintaining the ripple suppression effect.

[0056] S5. Based on the d-axis extended back EMF component after ripple suppression, obtain the preliminary estimated electrical angle of the current control cycle through the preliminary position phase-locked loop.

[0057] Specifically, the ripple-suppressed d-axis extended back EMF component is fed into the preliminary position phase-locked loop (PLL). The preliminary position PLL includes a proportional-integral (PI) controller. With zero as a setpoint, the ripple-suppressed d-axis extended back EMF component is fed as feedback into the PLL to obtain the preliminary estimated electrical angular velocity for the current control cycle. Integrating this preliminary estimated electrical angular velocity yields the preliminary estimated electrical angle for the current control cycle. The output of the preliminary position PLL is the position estimate corresponding to the ripple-suppressed extended back EMF.

[0058] S6. Based on the preliminary estimated electrical angle of the current control cycle, construct the input basis of the fundamental component adaptive linear neuron, and reconstruct the fundamental components of the α-axis extended back EMF and the β-axis extended back EMF.

[0059] Specifically, the input vector of the fundamental component adaptive linear neuron is constructed using the cosine and sine values ​​of the preliminary estimated electrical angle of the current control cycle. The α-axis and β-axis fundamental components of the extended back EMF are reconstructed using the α-axis and β-axis fundamental component adaptive linear neurons, respectively. The α-axis fundamental component adaptive linear neuron is configured with independent α-axis weight coefficients and updated using the least mean square algorithm; similarly, the β-axis fundamental component adaptive linear neuron is configured with independent β-axis weight coefficients and updated using the least mean square algorithm. Through the reconstruction by the fundamental component adaptive linear neuron, residual harmonic components are further eliminated, a pure fundamental component is extracted, and the fundamental phase lag caused by traditional filtering is avoided.

[0060] S7. Input the reconstructed α-axis extended back EMF fundamental component and β-axis extended back EMF fundamental component into the final position phase-locked loop to obtain the final estimated electrical angle and the final estimated electrical angular velocity, and use them for sensorless closed-loop control.

[0061] Specifically, the fundamental components of the α-axis extended back EMF and the β-axis extended back EMF obtained from step S6 are input into the final position phase-locked loop to obtain the final estimated electrical angle and the final estimated electrical angular velocity. The final estimated electrical angle is used for Park transformation and inverse Park transformation in the current loop, enabling the stator current and voltage commands to be converted between the estimated synchronous rotating coordinate system and the two-phase stationary coordinate system; the final estimated electrical angular velocity is used for the extended back EMF observer. Thus, closed-loop control of the permanent magnet synchronous motor electrolytic capacitor-free drive system is achieved without adding a position sensor, and position estimation error fluctuations caused by DC bus voltage ripple are suppressed.

[0062] like Figure 4 As shown, the method of the present invention is executed in the following order: extraction of DC bus voltage second harmonic ripple, construction of ripple synchronous orthogonal input basis, coordinate transformation of extended back EMF, suppression of d-axis extended back EMF ripple, preliminary estimation of electrical angle generation, reconstruction of αβ-axis extended back EMF fundamental wave, and final position estimation by phase-locked loop. This process corresponds to steps S1 to S7, and can combine the DC bus voltage second harmonic ripple information with the extended back EMF reconstruction process.

[0063] Furthermore, steps S1 to S5 constitute a DC bus voltage harmonic-assisted position pre-estimation stage. This stage uses the ripple synchronous adaptive linear neuron to fit and suppress the ripple component in the d-axis extended back electromotive force, and then obtains the preliminary estimated electrical angle of the current control cycle through the preliminary position phase-locked loop.

[0064] Specifically, such as Figure 2As shown, the DC bus voltage harmonic-assisted position prediction stage consists of a bandpass filter (BPF), an orthogonal signal generator, a fundamental component input base constructor, a ripple synchronization adaptive linear neuron, a Park transform module, and a preliminary position phase-locked loop. DC bus voltage First, the signal is fed into a bandpass filter to extract the ripple component at twice the grid frequency. The center angular frequency of this bandpass filter is set to twice the grid angular frequency. The extracted ripple component... The signal is fed into an orthogonal signal generator to construct an in-phase component synchronized with the second harmonic ripple component of the DC bus voltage. and orthogonal components ,Depend on and This constitutes the orthogonal input basis of the ripple-synchronous adaptive linear neuron. Simultaneously, the α-axis extended back EMF observation value output by the extended back EMF observer is... and β-axis extended back electromotive force observations The Park transformation module converts the data into the d-axis extended back EMF component in the synchronous rotating coordinate system. The electrical angle required for the Park transformation is the preliminary estimated electrical angle from the previous control cycle. d-axis extended back EMF component Output of adaptive linear neurons synchronized with ripple By subtraction, the d-axis extended back EMF component after ripple suppression is obtained. Ripple-synchronized adaptive linear neurons use in-phase components. and orthogonal components Using the input basis as the basis, the weighted sum is obtained by weighting the input basis with weight coefficients. The weighting coefficients are calculated using the least mean square algorithm. Update the error signal. The d-axis extended back EMF component after ripple suppression. The initial position phase-locked loop (PLL) is fed in. The PLL includes a proportional-integral (PI) controller, which takes zero as its setpoint and... The feedback signal is fed into the proportional-integral controller to obtain a preliminary estimate of the electrical angular velocity for the current control cycle. ,right Integrating the components yields a preliminary estimate of the electrical angle for the current control cycle. The fundamental component is input to the base constructor based on the preliminary estimated electrical angle. Constructing input basis , This step starts with the disturbance source and directly introduces the DC bus voltage ripple information into the adaptive filtering process, so that the ripple suppression effect is not affected by changes in the motor operating frequency.

[0065] Further, steps S6 to S7 constitute an extended back EMF reconstruction stage based on fundamental component adaptive linear neurons. This stage constructs an input basis using the cosine and sine values ​​of the preliminary estimated electrical angle of the current control cycle. The α-axis and β-axis extended back EMF fundamental components are reconstructed using α-axis and β-axis fundamental component adaptive linear neurons, respectively. The reconstructed fundamental components are input into the final position phase-locked loop to obtain the final estimated electrical angle and the final estimated electrical angular velocity.

[0066] Specifically, such as Figure 2 As shown, the extended back EMF reconstruction stage based on fundamental component adaptive linear neurons consists of α-axis fundamental component adaptive linear neurons and β-axis fundamental component adaptive linear neurons. The α-axis fundamental component adaptive linear neurons... , As input, reconstruct the fundamental component of the α-axis extended back EMF. ;β-axis fundamental wave component adaptive linear neuron with , As input, reconstruct the fundamental component of the β-axis extended back EMF. The final position phase-locked loop is based on and Obtain the final estimated electrical angle and the final estimated electric angular velocity This step does not rely on a fixed-bandwidth filter. Instead, it uses adaptive neurons to adjust the weight coefficients online to approximate the fundamental component. Therefore, it does not introduce fundamental phase lag and avoids the DC bias error in position estimation caused by traditional filtering methods.

[0067] Further, in step S1, the DC bus voltage includes a DC component and an AC ripple component with twice the grid frequency as the fundamental frequency; the extended back EMF observation value is obtained by an extended back EMF observer, and the extended back EMF is determined based on the motor's d-axis inductance, q-axis inductance, d-axis current, q-axis current, electric angular velocity, and permanent magnet flux linkage.

[0068] Specifically, in electrolytic capacitor-free drive systems, because the DC bus uses a small-capacity film capacitor, the DC bus voltage will generate ripple at twice the grid frequency and its harmonics, depending on the input grid voltage. This ripple is the direct cause of sideband harmonics in the extended back electromotive force (EMF). The amplitude of the extended back EMF is determined by the motor parameters and operating conditions, and the observed value of the extended back EMF is obtained in real time through an extended back EMF observer.

[0069] Combination Figure 1 The electrolytic capacitor-free drive system for permanent magnet synchronous motors includes a single-phase AC power supply, a diode rectifier, a small-capacity film capacitor, a three-phase voltage-source inverter, and a permanent magnet synchronous motor. DC bus voltage. It can be represented as:

[0070] ;

[0071] In the formula, This is the DC bus voltage. This represents the DC component of the DC bus voltage. and The DC bus voltage is respectively The amplitude and phase of the second harmonic components, This is the angular frequency of the power grid.

[0072] The controller collects DC bus voltage in real time. The stator current of the permanent magnet synchronous motor was measured, and the extended back EMF observation values ​​in the two-phase stationary coordinate system were obtained using the extended back EMF observer. and The extended back electromotive force can be expressed as:

[0073] ;

[0074] In the formula, To expand the back electromotive force amplitude, , and These are the d-axis inductance and q-axis inductance of a permanent magnet synchronous motor, respectively. and These are the d-axis current and the q-axis current, respectively. The electric angular velocity of the motor. For differential operators, It is a permanent magnet flux linkage.

[0075] At the same time, the controller obtains a preliminary estimate of the electrical angle. The preliminary estimated electrical angle can be the estimated electrical angle output from the previous control cycle.

[0076] Further, in step S2, the second harmonic ripple component of the DC bus voltage is the ripple component at twice the grid frequency in the DC bus voltage, which is extracted from the DC bus voltage using a bandpass filter; and an in-phase component and a quadrature component that are synchronized with the second harmonic ripple component of the DC bus voltage are obtained through a quadrature signal generator, and the in-phase component and the quadrature component constitute the quadrature input base.

[0077] Specifically, a bandpass filter (BPF) is used to extract the ripple component at twice the grid frequency from the DC bus voltage. The center angular frequency of the BPF is set to twice the grid angular frequency. The extracted ripple component is then used to obtain in-phase and quadrature components synchronized with it through a quadrature signal generator. These in-phase and quadrature components form a quadrature input basis, which serves as the input to the ripple-synchronized adaptive linear neuron. This input basis is synchronized with the disturbance source and contains the amplitude and phase information of the DC bus voltage ripple, enabling the adaptive neuron to directly utilize the disturbance source information for ripple fitting, thereby reducing the weight learning burden and improving the convergence speed.

[0078] In a preferred embodiment, combined with Figure 1 and Figure 2 The transfer function of the bandpass filter is:

[0079] ;

[0080] In the formula, The transfer function of the bandpass filter. The bandwidth coefficient of the bandpass filter. This is the center angular frequency of the bandpass filter. This is the second harmonic ripple component of the DC bus voltage. and DC bus voltage and its second harmonic ripple component The image function in the complex frequency domain.

[0081] To synchronize the input basis of the adaptive linear neuron with the second harmonic ripple component of the DC bus voltage, an orthogonal signal generator is then used to obtain the signal... Synchronous orthogonal input bases, wherein the orthogonal signal generator satisfies:

[0082] ;

[0083] In the formula, This is an in-phase component synchronized with the second harmonic ripple component of the DC bus voltage. This is an orthogonal component synchronized with the second harmonic ripple component of the DC bus voltage. The bandwidth coefficient of the quadrature signal generator. , They are respectively , The first derivative of . From and Constructing the input vector for a ripple-synchronous adaptive linear neuron:

[0084] ;

[0085] In the formula, This is the input vector for the ripple-synchronous adaptive linear neuron. This indicates the transpose. Since this input vector is directly constructed from the DC bus voltage ripple, it contains the amplitude and phase information of the disturbance source.

[0086] Further, in step S3, based on the preliminary estimated electrical angle of the previous control cycle, the extended back EMF observation value is transformed to the estimated synchronous rotating coordinate system to obtain the d-axis extended back EMF component.

[0087] In a preferred embodiment, combined with Figure 1 The coordinate transformation expression is:

[0088] ;

[0089] In the formula, and The d-axis and q-axis extended back EMF components in the synchronous rotating coordinate system are estimated respectively. In the synchronous rotating coordinate system, the extended back EMF disturbance affected by the DC bus voltage ripple mainly manifests as the ripple component in the d-axis extended back EMF. Therefore, it is possible to target... Perform ripple fitting and suppression.

[0090] Further, in step S4, the ripple synchronous adaptive linear neuron uses the orthogonal input basis as input, calculates the fitted d-axis extended back EMF ripple component through weight coefficients, subtracts the fitted d-axis extended back EMF ripple component from the d-axis extended back EMF component, and obtains the ripple-suppressed d-axis extended back EMF component; the weight coefficients are updated using the least mean square algorithm.

[0091] Specifically, such as Figure 5 As shown, the ripple synchronous adaptive linear neuron expands the back electromotive force observation along the d-axis. As the signal to be suppressed, the ripple component is twice the DC bus voltage and the grid frequency. and its orthogonal components As the input basis, where The amplitude of the ripple component at twice the grid frequency of the DC bus voltage. The angular frequency of the power grid. The phase of the ripple component at twice the grid frequency of the DC bus voltage is given. The ripple synchronous adaptive linear neuron obtains the fitted d-axis extended back EMF ripple component by multiplying the weight coefficients with the input basis and summing the results. Then from After subtraction, the d-axis extended back EMF component after ripple suppression is obtained. .at the same time, With setting update step size After multiplying, respectively with and Multiply by each other to form the weight update amount; the weight update amount is then multiplied by the weight update amount after a unit delay. The weight coefficients of the previous control cycle are summed to obtain the weight coefficients of the current control cycle. The weight coefficients are updated online using the least mean square algorithm, eliminating the need for offline training and enabling them to adapt to changes in operating conditions.

[0092] In a preferred embodiment, combined with Figure 5 The expression for the output and weight update of the ripple-synchronous adaptive linear neuron is as follows:

[0093] ;

[0094] In the formula, The d-axis extended back EMF ripple component is obtained by fitting the ripple synchronous adaptive linear neuron. The d-axis extended back EMF component after ripple suppression. Let be the weight coefficient vector of the ripple-synchronized adaptive linear neuron, where For the in-phase component weighting coefficients of the ripple-synchronized adaptive linear neuron, These are the orthogonal component weighting coefficients of the ripple-synchronous adaptive linear neuron.

[0095] The weight coefficients are updated using the least mean square algorithm, and the update expression is as follows:

[0096] ;

[0097] In the formula, For the first The ripple synchronization adaptive linear neuron weight coefficient vector for each control cycle. For the first The ripple synchronization adaptive linear neuron weight coefficient vector for each control cycle. To set the update step size, For the first The d-axis extended back EMF component after ripple suppression in each control cycle For the first Ripple synchronization input vector in each control cycle.

[0098] because Composed of the DC bus voltage ripple synchronization component and its orthogonal components, the equivalent update step size of the ripple synchronization adaptive linear neuron satisfies:

[0099] ;

[0100] In the formula, To achieve an equivalent update step size, The ripple component is twice the DC bus voltage and the grid frequency. The amplitude of the bus voltage ripple is thus increased. Therefore, as the amplitude of the bus voltage ripple increases, the proposed strategy can improve the fitting and suppression speed of the corresponding ripple component.

[0101] The following is combined with Figure 6 This invention explains the advantages of the DC bus voltage frequency-doubled ripple synchronous input base from the perspective of weight decomposition principles. Traditional frequency-doubled quadrature input bases include in-phase inputs. and orthogonal input ,in To control the cycle number, To control the cycle; and These represent the first and second weighting coefficients, respectively, when using a traditional quadrature input base. The DC bus voltage ripple vector at twice the grid frequency can be expressed as... Its amplitude is Phase is Its orthogonal vector can be represented as There is an equivalent complex mapping relationship between the DC bus voltage ripple and the extended back electromotive force ripple, which can be represented by an amplitude coefficient. and phase shift Therefore, the extended back EMF ripple vector can be represented as: When using the DC bus voltage ripple synchronous input base described in this invention, and They are respectively And the ideal weight coefficients corresponding to their orthogonal vectors. Figure 6 It is evident that the core difference between the traditional fixed-frequency input base and the synchronous input base of this invention lies in the different weight decomposition methods. The synchronous input base used in this invention already contains the amplitude and phase information of the disturbance source. Therefore, the adaptive weights only need to learn the amplitude and phase characteristics of the equivalent channel, and the weight learning burden is lighter than that of the traditional fixed-frequency input base. Figure 5 This demonstrates the specific implementation structure of the ripple synchronous adaptive linear neuron used in this invention—using the second harmonic ripple component of the DC bus voltage and its orthogonal components as the input basis, and updating the weight coefficients online through the least mean square algorithm, it can quickly track changes in the disturbance source. The two work together to explain the theoretical basis for improving the transient convergence speed of the method in this invention from both the weight decomposition principle and algorithm implementation perspectives.

[0102] Further, in step S5, the preliminary position phase-locked loop includes a proportional-integral (PI) regulator. With zero as a given value, the ripple-suppressed d-axis extended back EMF component is fed into the PI regulator to obtain the preliminary estimated electrical angular velocity of the current control cycle. The preliminary estimated electrical angular velocity of the current control cycle is integrated to obtain the preliminary estimated electrical angle of the current control cycle. The PI regulator is configured with independent proportional coefficients and integral coefficients.

[0103] Specifically, the initial position phase-locked loop uses the ripple-suppressed d-axis extended back EMF component as an error signal. This error is adjusted to zero by a proportional-integral (PI) controller, thus obtaining a preliminary estimated electrical angular velocity. Integrating this electrical angular velocity yields a preliminary estimated electrical angle. The proportional and integral coefficients of the PLC are independently configured and can be tuned separately according to the system's dynamic response and steady-state accuracy requirements.

[0104] In a preferred embodiment, combined with Figure 2 The setpoint of the proportional-integral controller =0, the preliminary estimate of the electric angular velocity is obtained by the following formula:

[0105] ;

[0106] Further integration of the preliminary estimated electric angular velocity yields the preliminary estimated electric angle:

[0107] ;

[0108] In the formula, To make a preliminary estimate of the electric angular velocity, To make a preliminary estimate of the electrical angle, and These are the proportional coefficient and integral coefficient of the proportional-integral controller, respectively. It is a frequency domain operator.

[0109] Furthermore, the α-axis fundamental component adaptive linear neuron is configured with independent α-axis weight coefficients and updated using the least mean square algorithm, and the β-axis fundamental component adaptive linear neuron is configured with independent β-axis weight coefficients and updated using the least mean square algorithm; the update error of the α-axis weight coefficients is the difference between the observed α-axis extended back EMF and the reconstructed α-axis extended back EMF fundamental component, and the update error of the β-axis weight coefficients is the difference between the observed β-axis extended back EMF and the reconstructed β-axis extended back EMF fundamental component; the fundamental component adaptive linear neuron is configured with a dedicated update step size.

[0110] Specifically, such as Figure 2As shown, the α-axis fundamental component adaptive linear neuron and the β-axis fundamental component adaptive linear neuron have the same structure, both being two-input, single-output types. The two inputs of the α-axis fundamental component adaptive linear neuron are in-phase input quantities based on a preliminary estimated electrical angle. and orthogonal input The two inputs are multiplied by independent weighting coefficients. and Summing the results yields the reconstructed α-axis extended back electromotive force fundamental component. The two inputs of the β-axis fundamental wave component adaptive linear neuron are also... and Multiply by independent weighting coefficients respectively and Summing the results yields the reconstructed fundamental component of the β-axis extended back electromotive force. .

[0111] The error signal of the adaptive linear neuron for the α-axis fundamental component is the observed value of the α-axis extended back electromotive force. The fundamental component of the α-axis extended back electromotive force obtained from the reconstruction The difference, that is The error signal of the adaptive linear neuron for the β-axis fundamental component is the observed value of the β-axis extended back electromotive force. The fundamental component of the β-axis extended back electromotive force obtained from the reconstruction The difference, that is The weight coefficients of both axes are updated using the least mean square algorithm with their respective error signals, and the fundamental component adaptive linear neuron is configured with a dedicated update step size. The weight coefficients for each axis are independent of each other, updated separately, and do not affect each other.

[0112] By reconstructing the fundamental component using an adaptive linear neuron, the extended back electromotive force fundamental component can be separated from observations containing sideband harmonics. Since the adaptive neuron uses the cosine and sine values ​​of the initially estimated electrical angle as input bases, the phase of the input bases matches the fundamental component during reconstruction, without introducing additional phase lag. This avoids the fundamental phase lag and DC bias of position estimation errors caused by bandwidth limitations in traditional complex filtering methods.

[0113] In a preferred embodiment, combined with Figure 2 The input vector and reconstruction expression of the fundamental component adaptive linear neuron are as follows:

[0114] ;

[0115] In the formula, For the first The fundamental component of the adaptive linear neuron input vector for each control cycle. , and These are the in-phase and quadrature inputs, respectively, based on a preliminary estimated electrical angle. For the first Preliminary estimate of electrical angle for each control cycle.

[0116] Based on the input vector The fundamental components of the extended back electromotive force along the α-axis and β-axis are reconstructed respectively:

[0117]

[0118] In the formula, and The first The α-axis and β-axis extended back EMF fundamental components obtained by reconstructing the data over one control cycle and The first The adaptive linear neuron weight coefficient vector for the α-axis and β-axis fundamental wave components of each control cycle.

[0119] The weighting coefficients are updated using the least mean square algorithm:

[0120]

[0121] In the formula, Configure a dedicated update step size for the fundamental component adaptive linear neuron. and The first Extended back EMF observations for the α-axis and β-axis over each control cycle. For the first The fundamental component of the α-axis extended back EMF obtained from the reconstruction of each control cycle. For the first The fundamental component of the β-axis extended back EMF obtained from the reconstruction of each control cycle. and The first The adaptive linear neuron weight coefficient vector for the α-axis and β-axis fundamental wave components of each control cycle.

[0122] Furthermore, in step S7, the final estimated electrical angle is used for the current loop Park transformation and inverse Park transformation, and the final estimated electrical angular velocity is used for the extended back EMF observer.

[0123] Specifically, the final estimated electrical angle is used for Park and inverse Park transformations in the current loop, enabling the stator current and voltage commands to be converted between the estimated synchronous rotating coordinate system and the two-phase stationary coordinate system. The final estimated electrical angular velocity is used for the proportional-integral regulators in the speed and current loops, serving as the basis for speed feedback values ​​and current decoupling calculations. Thus, the method of this invention achieves closed-loop control of a permanent magnet synchronous motor electrolytic capacitor-free drive system without adding position sensors.

[0124] In a preferred embodiment, combined with Figure 2 The reconstructed α-axis extended back EMF fundamental component and the fundamental component of the β-axis extended back electromotive force Input the final position phase-locked loop to obtain the final estimated electrical angle. and the final estimated electric angular velocity The final estimated electrical angle is used for the Park transform and inverse Park transform of the current loop.

[0125] The final estimated electrical angle Used in the current loop inverse Park transformation module and the Park transformation module (in the DC bus voltage harmonic auxiliary position pre-estimation module), it enables the stator current and voltage commands to be converted between the estimated synchronous rotating coordinate system and the two-phase stationary coordinate system; the final estimated electric angular velocity This is used to extend the back EMF observer, current loop controller, or fundamental component input base constructor. Thus, this embodiment achieves closed-loop control of a permanent magnet synchronous motor electrolytic capacitor-free drive system without adding a position sensor, and suppresses position estimation error fluctuations caused by DC bus voltage ripple.

[0126] The beneficial effects of this invention are verified through specific experiments below. This embodiment is implemented on a capacitor-free permanent magnet synchronous motor drive system. The motor parameters are as follows: rated power 2.2kW, rated voltage 380V, rated current 5A, rated speed 1500r / min, stator resistance 2.5Ω, d-axis inductance 50mH, q-axis inductance 65mH, permanent magnet flux linkage 0.3Wb, and number of pole pairs 4. A small-capacity film capacitor with a capacitance of 20μF is used for the DC bus.

[0127] Figure 7 The diagram shows a comparison of experimental waveforms at the rated operating frequency without algorithm application, using the traditional complex filtering method, and using the method of this invention. Figure 7 This includes waveforms such as DC bus voltage, αβ-axis extended back EMF components, and position estimation error. Figure 7 (a) It can be seen that without the algorithm applied, the extended back EMF component exhibits significant fluctuations, and the position estimation error contains a large periodic fluctuation component. Figure 7(b) It can be seen that after adopting the traditional complex filtering method, the sideband harmonics of the extended back EMF are suppressed to a certain extent, and the fluctuation of the position estimation error is reduced. However, due to the introduction of fundamental phase lag by complex filtering, a significant DC bias is generated in the position estimation error. Figure 7 (c) It can be seen that after adopting the method of the present invention, the fluctuation of the extended back EMF is further reduced, the fluctuation of the position estimation error is effectively suppressed, and no obvious additional DC bias is introduced, which verifies the advantages of the present invention in maintaining the accuracy of the fundamental phase.

[0128] Figure 8 The results are shown in the harmonic FFT analysis of the extended back electromotive force component, using the traditional complex filtering method and the method of this invention, without applying the algorithm. Figure 8 (a) It can be seen that there are obvious sideband harmonics in the extended back electromotive force without the algorithm applied. Figure 8 (b) It can be seen that traditional complex filtering methods have a suppressive effect on some sideband harmonics, but their ability to suppress low-frequency sideband harmonics is limited. Figure 8 (c) It can be seen that the method of the present invention can simultaneously weaken the positive and negative sideband harmonics generated by the coupling of twice the power grid frequency and the motor frequency, verifying the advantages of the method of the present invention in terms of sideband harmonic suppression capability.

[0129] Figure 9 This is a comparison of the transient convergence process before and after using the traditional complex filtering method and the method of this invention. Figure 9 (a) It can be seen that the traditional complex filtering method requires a relatively long time to reach steady state. Figure 9 (b) It can be seen that the method of the present invention reduces the learning burden of adaptive weights on the amplitude and phase changes of ripple by synchronously inputting the DC bus voltage ripple, enabling the d-axis extended back EMF ripple component to be fitted and suppressed more quickly, thereby improving the transient convergence speed of the position estimation error. In summary, the method of the present invention can reduce the sideband harmonics of the extended back EMF and the fluctuation of the position estimation error without changing the hardware structure of the drive system, avoid the DC bias of the estimation error introduced by traditional complex filtering, and improve the transient performance of the system.

[0130] Figure 10 The graphs show the position estimation error fluctuation components and DC bias statistics at different operating frequencies: without algorithm application, using traditional complex filtering, and using the method of this invention. Figure 10 (a) It can be seen that the method of the present invention can effectively reduce the fluctuation of position estimation error at different operating frequencies. Figure 10 (b) It can be seen that the traditional complex filtering method introduces a significant DC bias at different frequencies, while the DC bias introduced by the method of the present invention is significantly reduced, which verifies the adaptability and robustness of the present invention at different operating frequencies.

[0131] In summary, this invention introduces a DC bus voltage ripple synchronous input base, utilizes ripple synchronous adaptive linear neurons to suppress the d-axis extended back EMF ripple component, and combines this with fundamental component adaptive linear neurons to reconstruct the αβ-axis extended back EMF fundamental component. Compared with traditional filtering methods, this invention can simultaneously address sideband harmonic suppression, fundamental phase accuracy, and transient convergence speed, effectively improving the position estimation accuracy of a permanent magnet synchronous motor electrolytic capacitor-free drive system.

[0132] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for suppressing position error in a permanent magnet motor without electrolytic capacitors and assisted by bus voltage, characterized in that, The method includes the following steps: S1. Collect the DC bus voltage and stator current of the permanent magnet synchronous motor electrolytic capacitor-free drive system, and obtain the extended back electromotive force observation value and the preliminary estimated electrical angle of the previous control cycle. S2. Extract the second harmonic ripple component of the DC bus voltage from the DC bus voltage, and construct an orthogonal input base that is synchronized with the second harmonic ripple component of the DC bus voltage; S3. Based on the preliminary estimated electrical angle of the previous control cycle, transform the observed value of the extended back electromotive force to the estimated synchronous rotating coordinate system to obtain the d-axis extended back electromotive force component. S4. Based on the orthogonal input basis of the DC bus voltage second harmonic ripple component synchronization, the ripple synchronization adaptive linear neuron is used to fit and suppress the d-axis extended back EMF ripple component to obtain the ripple-suppressed d-axis extended back EMF component. S5. Based on the d-axis extended back EMF component after ripple suppression, obtain the preliminary estimated electrical angle of the current control cycle through the preliminary position phase-locked loop. S6. Based on the preliminary estimated electrical angle of the current control cycle, construct the input basis of the fundamental component adaptive linear neuron, and reconstruct the fundamental components of the α-axis extended back EMF and the β-axis extended back EMF. S7. Input the reconstructed α-axis extended back EMF fundamental component and β-axis extended back EMF fundamental component into the final position phase-locked loop to obtain the final estimated electrical angle and the final estimated electrical angular velocity, and use them for sensorless closed-loop control.

2. The method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 1, characterized in that, In step S1, the DC bus voltage includes a DC component and an AC ripple component with twice the grid frequency as the fundamental frequency; the extended back EMF observation value is obtained by an extended back EMF observer, and the extended back EMF is determined based on the motor's d-axis inductance, q-axis inductance, d-axis current, q-axis current, electric angular velocity, and permanent magnet flux linkage.

3. The method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 2, characterized in that, In step S2, the DC bus voltage second harmonic ripple component is the ripple component at twice the grid frequency in the DC bus voltage, which is extracted from the DC bus voltage using a bandpass filter; the in-phase component and quadrature component synchronized with the DC bus voltage second harmonic ripple component are obtained through a quadrature signal generator, and the in-phase component and quadrature component constitute the quadrature input base.

4. The method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 1, characterized in that, Steps S1 to S5 constitute the DC bus voltage harmonic-assisted position pre-estimation stage. This stage uses the ripple synchronous adaptive linear neuron to fit and suppress the ripple component in the d-axis extended back electromotive force, and then obtains the preliminary estimated electrical angle of the current control cycle through the preliminary position phase-locked loop.

5. The method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 4, characterized in that, In step S4, the ripple synchronous adaptive linear neuron uses the orthogonal input basis as input, calculates the fitted d-axis extended back EMF ripple component through weight coefficients, subtracts the fitted d-axis extended back EMF ripple component from the d-axis extended back EMF component, and obtains the ripple-suppressed d-axis extended back EMF component; the weight coefficients are updated using the least mean square algorithm.

6. The method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 4, characterized in that, In step S5, the preliminary position phase-locked loop includes a proportional-integral (PI) regulator. With zero as a given value, the ripple-suppressed d-axis extended back EMF component is fed into the PI regulator to obtain the preliminary estimated electrical angular velocity of the current control cycle. The preliminary estimated electrical angular velocity of the current control cycle is integrated to obtain the preliminary estimated electrical angle of the current control cycle. The PI regulator is configured with independent proportional coefficients and integral coefficients.

7. The method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 1, characterized in that, Steps S6 to S7 constitute an extended back EMF reconstruction stage based on fundamental component adaptive linear neurons. This stage constructs an input basis using the cosine and sine values ​​of the preliminary estimated electrical angle of the current control cycle. The α-axis and β-axis extended back EMF fundamental components are reconstructed using α-axis and β-axis fundamental component adaptive linear neurons, respectively. The reconstructed fundamental components are input into the final position phase-locked loop to obtain the final estimated electrical angle and the final estimated electrical angular velocity.

8. The method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 7, characterized in that, The α-axis fundamental component adaptive linear neuron is configured with independent α-axis weight coefficients and updated using the least mean square algorithm. The β-axis fundamental component adaptive linear neuron is configured with independent β-axis weight coefficients and updated using the least mean square algorithm. The update error of the α-axis weight coefficients is the difference between the observed α-axis extended back EMF and the reconstructed α-axis extended back EMF fundamental component. The update error of the β-axis weight coefficients is the difference between the observed β-axis extended back EMF and the reconstructed β-axis extended back EMF fundamental component. The fundamental component adaptive linear neuron is configured with a dedicated update step size.

9. A method for suppressing position error of a permanent magnet motor without electrolytic capacitor assisted by bus voltage according to claim 7, characterized in that, In step S7, the final estimated electrical angle is used for the current loop Park transformation and inverse Park transformation, and the final estimated electrical angular velocity is used for the extended back EMF observer.

Images