Permanent magnet synchronous motor off-line inductance identification method and system
Patent Information
- Application Number
- CN202611245706.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-17
- Publication Date
- 2026-10-20
AI Technical Summary
然而，当前的这些离线电感辨识方法通常需要设计特定激励过程、依赖较准确的逆变器非线性模型或先验补偿参数，在参数未知、硬件成本受限和寄生阻抗变化的场景下仍然存在电感辨识误差较大的问题
[0008]总体而言，通过本发明所构思的以上技术方案与现有技术相比，本发明主要具有以下有益效果。
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of motor parameter identification, and more specifically, relates to an offline inductance identification method and system for permanent magnet synchronous motors. Background Technology
[0002] Permanent magnet synchronous motors (PMSMs) are widely used in electric vehicles, industrial servos, robot actuators, and precision transmissions due to their high power density, fast dynamic response, and high efficiency. The d-axis and q-axis inductances are crucial parameters in the PMSM control model, and their accuracy directly affects control performance, including current loop tuning, torque estimation, field weakening control, rotor position estimation, and condition monitoring. Traditional PMSM inductance parameters are typically obtained from manufacturer-specified values, offline measurements, or online identification algorithms. Offline PMSM inductance identification methods allow for the injection of test signals when the motor is stationary or at low speeds, and the inductance parameters can be calculated by measuring the voltage and current responses. Since the offline identification process does not rely on the motor's normal load operation, it reduces the impact of operating conditions on the identification results. However, this method requires measuring voltage and current response. Low-cost drive systems generally do not have additional terminal voltage sensors. In this case, the actual terminal voltage cannot be obtained directly. The terminal voltage can only be reconstructed from the duty cycle, DC bus voltage, and control commands. However, due to the influence of inverter dead zone, power device conduction voltage drop, bus voltage sampling error, cable resistance, winding resistance variation, and current sampling noise, there will be a deviation between the reconstructed voltage and the actual terminal voltage, which will cause inductor identification error.
[0003] To address the issue of offline inductance identification errors in permanent magnet synchronous motors (PMSMs), current main methods involve injecting specific test signals and establishing compensation models. The paper "Offline Inductance Identification of IPMSM With Sequence-Pulse Injection" proposes an embedded offline inductance identification method for PMSMs based on sequence pulse injection, extracting d-axis and q-axis inductance information through pulse voltage and current responses. The paper "Offline Inductance Identification of PMSM With Adaptive Inverter Nonlinearity Compensation" proposes an offline inductance identification method for PMSMs with adaptive inverter nonlinearity compensation, improving identification accuracy by compensating for inverter nonlinearity. However, these current offline inductance identification methods typically require designing specific excitation processes, relying on relatively accurate inverter nonlinearity models or prior compensation parameters, and still suffer from significant inductance identification errors in scenarios with unknown parameters, limited hardware costs, and varying parasitic impedance.
[0004] Therefore, how to accurately identify the inductance of a permanent magnet synchronous motor without setting a terminal voltage sensor and without knowing the model parameters is a technical problem that urgently needs to be solved. Summary of the Invention
[0005] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides an offline inductance identification method and system for permanent magnet synchronous motors, the purpose of which is to achieve accurate identification of the inductance of permanent magnet synchronous motors without setting up a terminal voltage sensor and without knowing the model parameters.
[0006] According to a first aspect of the present invention, an offline inductance identification method for a permanent magnet synchronous motor is provided, comprising: In the offline state of a permanent magnet synchronous motor with its rotor locked, voltage excitation signals under different operating conditions are sequentially injected into the motor, and the current response signals of the motor under different excitations are obtained; The voltage excitation signal and current response signal are used within different time windows. For each time window, a corresponding candidate response set is constructed to obtain... There are several candidate response sets. The different candidate responses in each set are the complex frequency domain impedances calculated by extracting the complex frequency domain current and voltage within the window at different test orientation angles and target frequency angles. For each candidate response set, the corresponding candidate inductance identification result is calculated. Candidate inductor identification results k=1,2,3,…… ; The number is a preset quantity, and k is the sequence number of the candidate response set; the voltage excitation signals under different operating conditions have different amplitudes, phases, frequencies, and excitation methods. Will The candidate inductor identification results are mapped to the inductor parameter space to obtain Candidate points ; For each candidate point According to time stability Spectral Dispersion Real part divergence of impedance and cross-conditional consistency A comprehensive evaluation was conducted to determine the strength of the principal evidence. , and , , Anticorrelation, with Positive correlation; Used to measure the difference between candidate points obtained based on the first half window and candidate points obtained based on the second half window in the k-th time window. Used to measure the dispersion of the complex frequency domain impedance spectrum amplitude in the k-th candidate response set. Used to measure the dispersion of the real part of the complex frequency domain impedance in the set of k-th candidate responses. It is used to measure how close the k-th candidate point is to other candidate points obtained under different operating conditions in the inductance parameter space; Perform kernel density estimation algorithm on inductance parameter space and use the main evidence strength For candidate points The kernel function is used to weight the results to obtain the main evidence weighted kernel density field, and the maximum peak point is found as the main peak point. ; Strength of primary evidence By applying nonlinear constraint, a conservative strength of evidence is obtained. ; Perform kernel density estimation algorithm on inductance parameter space and with conservative evidence strength For candidate points We weight the kernel functions to obtain the conservative evidence-weighted kernel density field and find the maximum peak point as the conservative peak point. ; For the main peak point and conservative peak point Weighted fusion is performed to obtain the final parameter points. ; Final parameter points The inductance parameter space is reverse-mapped to the d-axis inductance and q-axis inductance as the final identification result.
[0007] According to a second aspect of the present invention, an offline inductance identification system for a permanent magnet synchronous motor is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the preceding claims.
[0008] Overall, compared with the prior art, the technical solutions conceived in this invention have the following beneficial effects.
[0009] This invention enables the identification of d-axis and q-axis inductors in a permanent magnet synchronous motor (PMSM) under offline conditions with a locked rotor. By applying different excitation voltages to generate a two-dimensional composite voltage trajectory and acquiring the corresponding response current, candidate inductance values can be obtained under multiple frequencies, angles, trajectory segments, and excitation modes. However, due to the limited dimensionality of response information in each window, the candidate inductance values are not accurate. Subsequently, all candidate inductors are uniformly mapped to an inductance parameter space as spatial candidate points for analysis. The inconsistency evidence strength of each candidate point in the inductance parameter space is calculated. This evidence strength describes the candidate response quality of the corresponding point. Furthermore, based on the evidence kernel density peak finding, the main peak and conservative peak are obtained from the candidate field. The d-axis and q-axis inductances are output through dual-peak fusion. This method achieves accurate identification of PMSM inductors without the need for terminal voltage sensors and without knowing the model parameters. Attached Figure Description
[0010] Figure 1 This is a flowchart of the steps of an offline inductance identification method for a permanent magnet synchronous motor according to an embodiment of the present invention.
[0011] Figure 2 This is a schematic diagram of the motor system in one embodiment of the present invention.
[0012] Figure 3 This is a schematic diagram of a two-dimensional composite voltage trajectory in one embodiment of the present invention.
[0013] Figure 4 This is a curve showing the relationship between injection voltage and current response in one embodiment of the present invention; where (a) is the injection voltage command and (b) is the current response.
[0014] Figure 5 This is a statistical chart of inductor identification consistency in one embodiment of the present invention.
[0015] Figure 6 This is the statistical analysis of the maximum error in inductor identification per combination in one embodiment of the present invention.
[0016] Figure 7 This is a statistical distribution of inductor identification error in one embodiment of the present invention. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.
[0018] Example 1 like Figure 1 The diagram shows a flowchart of the offline inductance identification method for a permanent magnet synchronous motor according to an embodiment of the present invention, which mainly includes the following steps.
[0019] S1: In the offline state of the permanent magnet synchronous motor with the rotor locked, voltage excitation signals under different operating conditions are sequentially injected into the motor, and the current response signals of the motor under different excitations are obtained. Candidate inductor identification results .
[0020] Specifically, to obtain The voltage excitation signal and current response signal are used within different time windows. For each time window, a corresponding candidate response set is constructed to obtain... There are several candidate response sets. The different candidate responses in each set are the complex frequency domain impedances calculated by extracting the complex frequency domain current and voltage within the window at different test orientation angles and target frequency angles. For each candidate response set, the corresponding candidate inductance identification result is calculated. Candidate inductor identification results k=1,2,3,…… ; The preset quantity is k, which is the sequence number of the candidate response set. The voltage excitation signals under different operating conditions have different amplitudes, phases, frequencies, and excitation methods.
[0021] This step involves applying different excitations to the motor and acquiring the corresponding responses, thereby obtaining voltage excitation signals and current response signals under different operating conditions. These signals can then be obtained through a time window. Different sets of (voltage excitation signals, current response signals) cover different operating conditions. Inductance identification is performed based on the excitation and current response signals within each time window, which can yield... The different candidate inductor identification results need to be based on coverage of different operating conditions. The accurate inductance parameters are calculated from the results of identifying different candidate inductors.
[0022] In one embodiment, an αβ injection voltage in the αβ coordinate system can be injected into the permanent magnet synchronous motor in an offline state with the rotor locked. By adjusting the amplitude, phase, frequency, and excitation mode of the αβ injection voltage, a two-dimensional composite voltage trajectory with multi-directional coverage is formed in the αβ plane, and the αβ current response of the motor under the excitation of the αβ injection voltage is obtained; wherein, the αβ injection voltage is a voltage excitation signal, and the αβ current response is a current response signal. It is understood that other excitation methods can also be used, for example, the αβ injection voltage command generated internally by the controller can be used directly, or it can be reconstructed from the three-phase PWM duty cycle and the DC bus voltage. The goal is to cover as many different operating conditions as possible, making the candidate inductor identification results as rich as possible.
[0023] like Figure 2 The diagram shown is a schematic representation of a motor system according to an embodiment of the present invention. The voltage excitation signal is expressed as a two-dimensional voltage command u in the αβ stationary coordinate system. α u β The voltage is injected in the form of αβ, which is then converted into a three-phase switch control signal S by a space vector pulse width modulation (SVPWM) module. abc The drive inverter converts the DC bus voltage u dc The voltage is modulated into a three-phase voltage and applied to the three-phase windings of a permanent magnet synchronous motor (PMSM); the motor generates a three-phase current response i under this voltage excitation. a i b i c The signal is then fed into the Clarke transformation module and transformed into the current response i in the αβ stationary coordinate system. α i β Record the above excitations and responses for subsequent inductor parameter identification.
[0024] When the motor rotor is locked offline, speed-related terms and mechanical motion effects are suppressed, and the αβ injected voltage... It can be represented as: ; ; In the formula, and Inject voltage components along the α-axis and β-axis, respectively. and These represent the amplitudes of the injected voltages along the α and β axes, respectively. and The frequencies of the injected voltages along the α and β axes are respectively. and These represent the phases of the injected voltages along the α and β axes, respectively.
[0025] like Figure 3The diagram shown is a schematic representation of a two-dimensional composite voltage trajectory according to an embodiment of the present invention. By adjusting the amplitude, frequency, phase, and excitation mode of the injected voltages along the α and β axes, a trajectory such as... can be formed in the αβ plane. Figure 3 The diagram shows a two-dimensional composite voltage trajectory with multi-directional coverage. Compared to traditional unidirectional high-frequency injection, the two-dimensional composite high-frequency trajectory can cover multiple spatial directions and phase combinations within a single test cycle. Taking the Lissajous or Rosette trajectory as an example, the controller generates high-frequency voltage commands with different frequencies and adjustable phases along the α and β axes, respectively. When a non-integer or irrational frequency ratio approximation is used between the two frequencies, the injection trajectory exhibits richer spatial sweep characteristics within a finite time, thus allowing the current response to contain more frequency domain information related to the inductance differences along the d and q axes.
[0026] In one embodiment, frequency You can choose 400Hz, 700Hz, and 1100Hz, the frequency ratio is... For non-integers, such as It is 2.4. Phase and The angles can be 30°, 60°, 90°, 120°, and 150°, and the αβ injection voltage... (in) , The combinations of (30°, 120°), (60°, 150°), (30°, 150°), (60°, 90°), (90°, 150°), or (30°, 90°) can form two-dimensional composite voltage trajectories with directional diversity. Excitation modes include irrational frequency ratio (INR) Lissajous / Rosette trajectories, mirror-balanced INR Lissajous / Rosette trajectories, and amplitude perturbation INR Lissajous / Rosette trajectories. The injected amplitude can be 16 V or 24 V.
[0027] Under the αβ injected voltage excitation, the three-phase stator current is collected. , , The αβ current response can be obtained through Clarke transformation. : ; In the formula, The current response in the αβ coordinate system. This is the Clarke transformation matrix.
[0028] like Figure 4The figure shows the correspondence between the injected voltage and the current response in one embodiment of the present invention, where (a) is the injected voltage command and (b) is the current response. In the same test sample, the two-dimensional composite trajectory injection command and the Clarke current response have a corresponding relationship. The voltage in the figure is the controller injection command or equivalent injected voltage, not the measured value from the external terminal voltage sensor; the current is the αβ current obtained by Clarke transformation of the three-phase sampled current. By simultaneously observing the injection command and the current response, a data foundation can be provided for subsequent frequency domain projection, candidate response generation, and evidence calculation.
[0029] Subsequently, obtain The αβ injection voltage and αβ current responses within different time windows are used to independently construct corresponding candidate response sets based on the αβ injection voltage and αβ current responses within each time window, and inductance identification is then performed. Candidate inductor identification results , and These are the d-axis inductance candidate value and the q-axis inductance candidate value in the k-th candidate inductance identification result, respectively; This is the preset quantity.
[0030] Specifically, it can be intercepted The excitation and current response signals are obtained within different time windows. For each time window, existing offline identification methods can be used for inductance identification to obtain... Candidate inductor identification results .
[0031] In one embodiment, the frequency domain impedance method can be used to identify the inductance. The process of constructing a corresponding candidate response set for each time window and calculating the candidate inductance identification result is as follows.
[0032] (1) Select multiple test direction angles and target angular frequency m = 1, 2, ..., M, where m is the candidate response number in the candidate response set and M is the number of candidate responses in the candidate response set.
[0033] (2) Inject the αβ voltage within the corresponding time window and αβ current response Angles along the m-th test direction respectively Directional unit vector By performing frequency domain projection, the voltage scalar is obtained. and current scalar t is the sampling time.
[0034] Wherein, the direction unit vector for: .
[0035] voltage scalar and current scalar for: ; .
[0036] (3) At a given target angular frequency Extract the voltage scalar respectively and current scalar The complex frequency domain components are used to obtain the complex frequency domain voltage. and complex frequency domain current .
[0037] Among them, complex frequency domain voltage and complex frequency domain current It can be represented as: ; ; In the formula, The target angular frequency is represented Frequency domain extraction operator, Specifically, frequency domain extraction operators are used. Time window The calculation method for frequency domain projection of the component x(t) to be projected within the range is as follows: ; In the formula, This represents the start time of the k-th time window. The time window represents the end time of the k-th time window, T is the length of the time window, ω is the target angular frequency, and w(t) is the window function. For example, a rectangular window can be used, or a Hann window can be used if it is necessary to reduce spectral leakage.
[0038] (4) Calculate the test direction angle and target angular frequency The impedance in the complex frequency domain is as follows: .
[0039] (5) Calculate the corresponding candidate values of the inductance response: .
[0040] (6) Combine candidate inductance response values under different test directions to construct d-axis and q-axis candidate inductance values, and obtain the k-th candidate inductance identification result.
[0041] S2, will The candidate inductor identification results are mapped to the inductor parameter space to obtain Candidate points .
[0042] Specifically, will Candidate inductor identification results Mapping to the inductor parameter space, we get There are several candidate points, which can be specifically subjected to a logarithmic transformation: ; In the formula, For the first 1 candidate point and These are the candidate inductance values for the d-axis and q-axis corresponding to the candidate point, respectively.
[0043] It should be noted that when it is necessary to map the candidate points back to the inductor later, the inverse function of the mapping function can be used.
[0044] S3, For each candidate point According to time stability Spectral Dispersion Real part divergence of impedance and cross-conditional consistency A comprehensive evaluation was conducted to determine the strength of the principal evidence. .
[0045] and , , Anticorrelation, with Positive correlation.
[0046] Specifically, the strength of primary evidence It can be represented as: ; In the formula, This is a multi-indicator fusion function that satisfies... and , , Anticorrelation, with Positive correlation.
[0047] Response time stability measures whether the response of the same candidate is consistent across different time windows; spectral dispersion measures the degree of dispersion of the same candidate in adjacent frequencies or harmonic components; impedance real dispersion reflects the degree of interference of resistive or non-inductive components on the candidate response; cross-condition consistency measures whether the candidate can obtain similar responses under different voltage amplitudes, frequencies, angles, or excitation modes. These indicators are used to obtain the candidate evidence strength through an evidence mapping function. The higher the evidence strength, the more representative the candidate is of the consistency center for multi-condition responses.
[0048] Time stability This is used to measure the difference between the first and second halves of the response in the k-th time window. Specifically, the original data in the k-th time window is divided into a first half and a second half by time, and two temporary candidate points are formed under the same conditions for each half. and , This is the distance between the two temporary candidate points, calculated using the following formula: ; In the formula, and These are temporary candidate points obtained based on data from the time periods before and after the k-th time window.
[0049] Spectral Dispersion Used to measure the dispersion of the complex frequency domain impedance spectrum amplitude in the k-th candidate response set. The calculation formula is: ; In the formula, This is the normality consistency correction coefficient, which can be taken as 1.4826. Let m be the spectral amplitude of the complex frequency domain impedance in the k-th candidate response set. This indicates the search for the median spectral amplitude from the set of candidate responses for the kth rank. This indicates the computation of the k-th candidate response set. spectral amplitude of a complex frequency domain impedance The absolute difference between the amplitude and the median spectral value. This indicates the search for the absolute difference between the medians.
[0050] Real divergence of impedance The dispersion of the real part of the complex frequency domain impedance in the k-th candidate response set. The calculation formula is: ; In the formula, This is the normality consistency correction coefficient, which can be taken as 1.4826. For the m-th complex impedance in the k-th candidate response set, This indicates the search for the real part of the median complex impedance from the set of candidate responses of the kth generation. This indicates the computation of the k-th candidate response set. Real part of the complex impedance The absolute difference between the real part of the median complex impedance and the median. This indicates the search for the absolute difference between the medians.
[0051] Cross-conditional consistency The formula used to measure the similarity between the k-th candidate point and candidate points obtained under other operating conditions in the inductance parameter space is as follows: ; In the formula, For the set of working conditions different from candidate point k, Let f be the set of candidate points generated under operating condition f. The set cross-condition consistency kernel width.
[0052] Furthermore, time stability can be assessed separately. Spectral Dispersion Real part divergence of impedance and cross-conditional consistency Normalization is performed to obtain the normalized time stability. Spectral Dispersion Real part divergence of impedance and cross-conditional consistency .
[0053] Subsequently, the inconsistency quantity is calculated. : ; In the formula, All are preset non-negative weights, satisfying .
[0054] Subsequently, the strength of the primary evidence was calculated. : ; The smaller the value, the more consistent the candidate point is in terms of time, spectrum, real part of impedance, and transconditional characteristics. The larger.
[0055] S4. Perform kernel density estimation algorithm on the inductance parameter space and use the main evidence strength For candidate points The kernel function is used to weight the results to obtain the main evidence weighted kernel density field, and the maximum peak point is found as the main peak point. .
[0056] Kernel density estimation (KDE) is a nonparametric density estimation method. It sets a kernel function (commonly Gaussian kernel) and bandwidth parameter, places a smooth kernel shape centered on each candidate point, and then superimposes and normalizes all kernels point by point in space to obtain the density estimate for any position, thus forming a "kernel density field". This transforms discrete candidate points into a continuous probability density surface, thereby revealing the clustering structure and main modes of data in the feature space, and then the peak point can be identified as the optimal parameter point.
[0057] Specifically, for any candidate point The formula for calculating its density estimate in the master evidence weighted kernel density field is as follows: ; In the formula, Candidate points in the main evidence weighted kernel density field The density estimate, For kernel function, For the first The kernel width of each candidate point This represents the total number of candidate points.
[0058] Main peak It can be represented as: .
[0059] S5. Strength of primary evidence By applying nonlinear constraint, a conservative strength of evidence is obtained. .
[0060] Specifically, regarding the strength of primary evidence Applying nonlinear constraint can enhance the consistency of evidence and yield a conservative strength of evidence. .
[0061] In one embodiment, conservative evidence strength The calculation formula is: ; In the formula, the limiting parameter γ is set to be greater than 1.
[0062] S6. Perform kernel density estimation algorithm on the inductance parameter space and conservatively assess the strength of evidence. For candidate points We weight the kernel functions to obtain the conservative evidence-weighted kernel density field and find the maximum peak point as the conservative peak point. .
[0063] Specifically, a conservative evidence-weighted kernel density field is constructed in the same manner as in step S4.
[0064] Specifically, for any candidate point The formula for calculating its density estimate in the conservative evidence-weighted kernel density field is as follows: ; In the formula, Candidate points in a conservative evidence-weighted kernel density field The density estimate, To preserve the strength of evidence, For kernel function, For the first The conservative kernel width of each candidate point.
[0065] Among them, the conservative kernel width Greater than or equal to the kernel width The calculation formula can be: ; In the formula, the coefficient η is set to ≥ 0.
[0066] Taking the selection of a Gaussian kernel as an example: .
[0067] Conservative peak point It can be represented as: ; The above adopts a conservative strength of evidence approach. This allows for further reduction of the weight of candidate points with weaker evidence, thus conserving the kernel width. This involves expanding the kernel width so that the conservative peaks reflect the stable clustering positions that the candidate field can still maintain under stronger consistency constraints. In other words, the conservative peaks are used to characterize the stable candidate clustering positions that can still be maintained under stronger consistency constraints.
[0068] S7, for the main peak point and conservative peak point Weighted fusion is performed to obtain the final parameter points. .
[0069] Specifically, the final parameter point It can be represented as: ; In the formula, λ is the fusion weight.
[0070] In one embodiment, the fusion weight λ can be adaptively adjusted as follows.
[0071] The formula for calculating the fusion weight λ is: ; In the formula, These represent the support of the primary evidence weighted kernel density field and the conservative evidence weighted kernel density field, respectively.
[0072] For any kernel density field, its support The calculation formula is: ; In the formula, These represent the peak density of the current kernel density field, the number of valid votes, and the average evidence strength near the peak, respectively. The formulas for calculating the number of valid votes N and the average strength of evidence E near the peak are: ; ; ; In the formula, These represent candidate points in the current kernel density field. Strength of evidence and nuclear width This represents the peak point found in the current kernel density field. Candidate points Towards The voting weight.
[0073] Based on the above calculation formulas, the support of the main evidence weighted kernel density field and the conservative evidence weighted kernel density field can be calculated separately, and then the adaptive value of the fusion weight λ can be calculated for the main peak point. and conservative peak point Weighted fusion is performed to obtain the final parameter points. .
[0074] S8, Set the final parameter points The inductance parameter space is reverse-mapped to the d-axis inductance and q-axis inductance as the final identification result.
[0075] Specifically, if step S2 uses a logarithmic transformation mapping method to map the inductor to candidate points in the inductor parameter space, then step S8 will finalize the parameter points. The formula for the inverse mapping from the inductance parameter space to the d-axis inductance and q-axis inductance is: ; in, and These represent the final parameter points respectively. The two components corresponding to the d-axis inductance and the q-axis inductance are shown in the figure.
[0076] Example 2 The present invention also relates to an offline inductance identification system for a permanent magnet synchronous motor, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.
[0077] This offline inductance identification system for permanent magnet synchronous motors can be installed on computing devices such as desktop computers, laptops, handheld computers, and cloud servers. The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The memory can be used to store computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and by accessing data stored in the memory.
[0078] Example 3 The present invention also relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.
[0079] Specifically, the memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.
[0080] Example 4 This invention provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of the method described in the above embodiments of this invention.
[0081] The present invention will be verified through experiments below.
[0082] In the specific experiment, a permanent magnet synchronous motor simulation platform was used for verification. The motor was kept in a stalled or locked rotor state, and the controller applied a two-dimensional composite high-frequency voltage trajectory. The test data included 12 different d-axis and q-axis inductance combinations. Under each inductance combination, multiple excitation modes, voltage amplitudes, injection frequencies, and injection angles were set, resulting in a total of 1080 sets of two-dimensional composite high-frequency excitation simulation data. During data processing, the controller's internal injection voltage command was used as the equivalent injection voltage. The three-phase current was collected and converted into αβ current, followed by frequency domain projection, candidate field construction, evidence calculation, and bimodal fusion.
[0083] The identification results for different inductor combinations are shown in the table below:
[0084] like Figure 5 The figure shown is a statistical chart of inductor identification consistency in one embodiment of the present invention. Under 12 different d-axis and q-axis inductor combinations, the identified values and the true values are generally distributed near the ideal consistency line and are located within a narrow error band, indicating that the present invention can obtain relatively stable identification results under different inductor combinations.
[0085] like Figure 6 The figure shows the maximum error statistics for inductor identification by combination in one embodiment of the present invention. The maximum absolute error corresponding to each inductor combination remains at a low level. Experimental data shows that the median maximum absolute error of the 12 inductor combinations is about 1.157%, the average maximum absolute error is about 1.161%, and the worst maximum absolute error is about 1.753%.
[0086] like Figure 7 The figure shows the statistical distribution of inductor identification errors in one embodiment of the present invention. The distributions of Ld error, Lq error, and maximum error are relatively concentrated, without significant dispersion. This result demonstrates that the present invention can maintain relatively stable offline inductor identification accuracy under multiple sets of different inductor parameters.
[0087] The above experimental results demonstrate that the present invention can complete inductor identification for multiple Ld / Lq combinations under stalled offline conditions, and the results show consistency across combinations. Its engineering significance mainly includes: first, it can be used for automatic parameter initialization of the driver after factory release or repair / replacement, reducing manual offline measurement steps; second, it can complete inductor identification using the driver's existing PWM, bus voltage, and current sampling information without adding an additional terminal voltage sensor.
[0088] Finally, it should be noted that the above embodiments are not the only implementation methods. In other implementation methods, the two-dimensional composite high-frequency trajectory can be a dual-frequency orthogonal trajectory, a piecewise rotating trajectory, a frequency-modulated trajectory, a random-phase multi-frequency trajectory, or other composite trajectories that can cover the αβ plane; the frequency domain projection can be synchronous integration, fast Fourier transform, sliding window spectrum analysis, or windowed least squares frequency estimation; the kernel function can be a Gaussian kernel, an Epanechnikov kernel, or other radial kernels; the fusion weight can be determined by one or more of peak density, effective vote count, average evidence strength, and candidate field concentration; the system can be deployed in a host computer, driver embedded processor, DSP, MCU, edge computing module, or production line calibration equipment.
[0089] The technical features of the embodiments described above can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. It should be noted that the terms "in one embodiment," "for example," and "again" in this invention are intended to illustrate the invention and are not intended to limit the invention.
[0090] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.
Claims
1. A method for offline inductance identification of a permanent magnet synchronous motor, characterized in that, include: In the offline state of the permanent magnet synchronous motor with the rotor locked, voltage excitation signals under different operating conditions are injected into the motor in sequence, and the current response signals of the motor under different excitations are obtained. get The voltage excitation signal and current response signal are used within different time windows. For each time window, a corresponding candidate response set is constructed to obtain... There are several candidate response sets. The different candidate responses in each set are the complex frequency domain impedances calculated by extracting the complex frequency domain current and voltage within the window at different test orientation angles and target frequency angles. For each candidate response set, the corresponding candidate inductance identification result is calculated. Candidate inductor identification results k=1,2,3,…… ; The number is a preset quantity, and k is the sequence number of the candidate response set; the voltage excitation signals under different operating conditions have different amplitudes, phases, frequencies, and excitation methods. Will The candidate inductor identification results are mapped to the inductor parameter space to obtain Candidate points ; For each candidate point According to time stability Spectral Dispersion Real part divergence of impedance and cross-conditional consistency A comprehensive evaluation was conducted to determine the strength of the principal evidence. , and , , Anticorrelation, with Positive correlation; Used to measure the difference between candidate points obtained based on the first half window and candidate points obtained based on the second half window in the k-th time window. Used to measure the dispersion of the complex frequency domain impedance spectrum amplitude in the k-th candidate response set. Used to measure the dispersion of the real part of the complex frequency domain impedance in the set of k-th candidate responses. It is used to measure how close the k-th candidate point is to other candidate points obtained under different operating conditions in the inductance parameter space; Perform kernel density estimation algorithm on inductance parameter space and use the main evidence strength For candidate points The kernel function is used to weight the results to obtain the main evidence weighted kernel density field, and the maximum peak point is found as the main peak point. ; Strength of primary evidence By applying nonlinear constraint, a conservative strength of evidence is obtained. ; Perform kernel density estimation algorithm on inductance parameter space and with conservative evidence strength For candidate points We weight the kernel functions to obtain the conservative evidence-weighted kernel density field and find the maximum peak point as the conservative peak point. ; For the main peak point and conservative peak point Weighted fusion is performed to obtain the final parameter points. ; Final parameter points The inductance parameter space is reverse-mapped to the d-axis inductance and q-axis inductance as the final identification result.
2. The offline inductance identification method for permanent magnet synchronous motors as described in claim 1, characterized in that, The process involves sequentially injecting voltage excitation signals under different operating conditions into the motor and acquiring the motor's current response signals under different excitations, including: An αβ injection voltage in the αβ coordinate system is injected into the motor. By adjusting the amplitude, phase, frequency, and excitation mode of the αβ injection voltage, a two-dimensional composite voltage trajectory with multi-directional coverage is formed in the αβ plane, and the αβ current response of the motor under the excitation of the αβ injection voltage is obtained. Here, the αβ injection voltage is a voltage excitation signal, and the αβ current response is a current response signal.
3. The offline inductance identification method for permanent magnet synchronous motors as described in claim 2, characterized in that, The process of constructing a corresponding candidate response set for each time window and calculating its candidate inductance identification results includes: Select multiple test direction angles and target angular frequency m = 1, 2, ..., M, where m is the candidate response index in the candidate response set and M is the number of candidate responses in the candidate response set; The αβ injection voltage within the corresponding time window and αβ current response Angles along the m-th test direction respectively Directional unit vector By performing frequency domain projection, the voltage scalar is obtained. and current scalar t is the sampling time; At a given target angular frequency Extract the voltage scalar respectively and current scalar The complex frequency domain components are used to obtain the complex frequency domain voltage. and complex frequency domain current ; Calculate the test direction angle and target angular frequency Complex frequency domain impedance ; Calculate the corresponding candidate values for the inductor response; ; By fusing candidate inductance response values from different test directions, candidate inductance values for the d-axis and q-axis are constructed, and the k-th candidate inductance identification result is obtained.
4. The offline inductance identification method for permanent magnet synchronous motors as described in claim 1, characterized in that, Will The mapping method for the candidate inductor identification results to the inductor parameter space is as follows: ； In the formula, These are the d-axis inductance identification value and the q-axis inductance identification value in the k-th candidate inductance identification result, respectively.
5. The offline inductance identification method for permanent magnet synchronous motors as described in claim 1, characterized in that, The formula for calculating the strength of conservative evidence is: ； In the formula, the strength of the main evidence Set the limiting parameter γ>
1.
6. The offline inductance identification method for permanent magnet synchronous motors as described in claim 5, characterized in that, For any candidate point Its density estimation in the main evidence weighted kernel density field The calculation formula is: ； For any candidate point Its density estimation in a conservative evidence-weighted kernel density field The calculation formula is: ； In the formula, For kernel function, For the first The kernel width of each candidate point in the main evidence-weighted kernel density field For the first The conservative kernel width of each candidate point in the conservative evidence-weighted kernel density field; Conservative kernel width With kernel width satisfy: ； In the formula, the coefficient η is set to ≥ 0.
7. The offline inductance identification method for permanent magnet synchronous motors as described in claim 1, characterized in that, Final parameter points The calculation formula is: ； In the formula, λ is the fusion weight; The formula for calculating the fusion weight λ is as follows: ； In the formula, These are the support values for the weighted kernel density field of the primary evidence and the weighted kernel density field of the conservative evidence, respectively. For any kernel density field, its support The calculation formula is: ； In the formula, These represent the peak density of the current kernel density field, the number of valid votes, and the average evidence strength near the peak, respectively. The formulas for calculating the number of valid votes N and the average strength of evidence E near the peak are: ； ； ； In the formula, They represent candidate points in the current kernel density field. Strength of evidence and nuclear width This represents the peak point found in the current kernel density field. Candidate points Towards The voting weight.
8. An offline inductance identification system for a permanent magnet synchronous motor, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.
10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 7.