An off-line identification method for permanent magnet flux linkage and direct and quadrature axis inductance of permanent magnet synchronous motor

Abstract

The application discloses an off-line identification method for permanent magnet flux linkage and direct and quadrature axis inductance of a permanent magnet synchronous motor and relates to the technical field of motor control. The method comprises a permanent magnet flux linkage identification stage and a direct and quadrature axis inductance identification stage. The permanent magnet flux linkage identification stage comprises the following steps: A1, obtaining a stator resistance, driving the motor to a first rotating speed with a q-axis current, setting the d-axis current to zero, setting the q-axis current to zero after the first rotating speed is reached, and allowing the motor to freely reduce the rotating speed. In the prior art, the motor is in a "loaded" state when the flux linkage is identified, and the resistance voltage drop must be deducted. The application adopts the process design of "firstly increasing the rotating speed, then setting to zero, and finally reducing the rotating speed", so that the motor is in a pure no-load coasting state at the sampling moment, the resistance voltage drop is zero, and the transmission influence of the resistance error on the flux linkage identification is fundamentally eliminated. This process design is not a simple superposition of each step, but a system-level innovation with a synergistic effect.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, specifically to an offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) are widely used in industrial speed regulation, servo drives, new energy vehicles, and aerospace due to their advantages such as small size, high efficiency, high power density, and high torque-to-inertia ratio. Vector control is the mainstream control scheme for PMSMs. Its core is to decompose the three-phase stator current into direct-axis (d-axis) current and quadrature-axis (q-axis) current through coordinate transformation, thereby achieving decoupled control of excitation components and torque components. The performance of vector control is highly dependent on the accuracy of the motor's electromagnetic parameters. The direct-axis inductance L_d and quadrature-axis inductance L_q determine the decoupling effect of the current loop and the tuning parameters of the PI regulator, while the permanent magnet flux linkage Ψ_f directly affects the torque control accuracy, the field weakening control boundary, and the effectiveness of the maximum torque-current ratio control strategy. However, the parameters provided by motor manufacturers are often inaccurate or not provided at all. In actual operation, the electromagnetic parameters of the motor will deviate from the design values ​​due to factors such as magnetic saturation, cross-coupling, and temperature changes. Therefore, offline parameter identification before the motor is put into operation to obtain accurate L_d, L_q, and Ψ_f is of great significance for ensuring the control performance of the vector control system. Currently, various technical solutions have been proposed in the field of offline parameter identification for permanent magnet synchronous motors. A common solution is to identify the stator resistance using the DC volt-ampere method, identify the d-axis inductance using the voltage step method or high-frequency injection method, identify the q-axis inductance using the model reference adaptive method, and finally identify the permanent magnet flux linkage by running the motor under no-load to the rated speed. For example, CN103825524A discloses an offline identification method for basic electrical parameters of a permanent magnet synchronous motor. This method employs the DC volt-ampere method to measure resistance, the voltage step method to measure d-axis inductance, the model reference adaptive method to identify q-axis inductance, and the method of identifying rotor flux linkage from no-load operation to rated speed. However, this method has significant shortcomings: First, the voltage step method for identifying d-axis inductance relies on the accurate measurement of the time it takes for the d-axis current to rise to 0.632 times the steady-state current, making it susceptible to sampling noise and ADC quantization errors. Second, the model reference adaptive method for identifying q-axis inductance requires establishing a current observation model and introducing a PI controller for iterative convergence, resulting in a large computational load and convergence speed significantly affected by initial values. Third, the permanent magnet flux linkage identification relies solely on single-point sampling, exhibiting weak anti-interference capabilities and difficulty in guaranteeing identification accuracy. Another approach is to use a high-frequency signal injection method for offline parameter identification. For example, CN103560736A discloses a static offline parameter identification method for a permanent magnet synchronous motor, which identifies direct-axis inductance and quadrature-axis inductance by injecting a high-frequency rotating voltage into the stator winding. Although this method can identify inductance in a static state, it requires an additional high-frequency signal injection stage and a complex signal extraction algorithm (such as discrete Fourier transform). The algorithm is complex to implement, and high-frequency injection may cause motor noise and additional losses.

[0003] Another approach utilizes intelligent optimization algorithms for parameter identification. For example, CN118739933A discloses an offline parameter identification method for adaptive permanent magnet synchronous motors based on a ramp-up algorithm. This method achieves parameter identification by dynamically changing the control parameters of the current loop and the control circuit. While this type of method has a certain degree of robustness, the algorithm is complex, computationally intensive, and requires high computing power from the main control chip. Parameter identification methods based on genetic algorithms also suffer from high computational complexity and slow convergence speed. Furthermore, existing technologies also employ recursive least squares methods for parameter identification, but this method relies on the appropriate selection of the forgetting factor and is susceptible to data saturation issues. In summary, existing offline parameter identification technologies for permanent magnet synchronous motors mainly suffer from the following technical problems: 1. The computational load is large and the algorithm is complex. Existing solutions mostly adopt model reference adaptation, intelligent optimization algorithm or recursive least squares method, which require repeated iterative convergence or complex signal processing. The computational burden is heavy and it is not easy to implement on microcontrollers or DSPs with limited resources.

[0004] 2. Identification accuracy is greatly affected by sampling noise. Most existing solutions use single or small amounts of sampling data for parameter calculation, resulting in weak anti-interference capabilities. Sampling noise and operating condition disturbances can easily cause the identification results to deviate from the true values. Although some solutions mention taking the average multiple times, they lack a systematic steady-state determination and data validity verification mechanism.

[0005] 3. Lack of coordination between parameter identification stages. Existing solutions often separate the identification of stator resistance, d-axis inductance, q-axis inductance and permanent magnet flux into multiple independent stages, failing to fully utilize the physical constraints between parameters to improve overall identification accuracy.

[0006] To address the shortcomings of the existing technologies, this invention proposes an offline identification method for permanent magnet flux linkage and quadrature-axis inductance of permanent magnet synchronous motors. Through a two-stage step-by-step identification strategy, a free-speed-decreasing steady-state sampling mechanism, multi-dimensional composite steady-state judgment and data pre-validation, and multi-set data mean filtering, high-precision, low-computation offline parameter identification is achieved at the pure software level. No additional hardware equipment is required, and it has good engineering practicality and mass production calibration adaptability. Summary of the Invention

[0007] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor, thereby solving the aforementioned problems in the prior art.

[0008] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: an offline identification method for permanent magnet flux linkage and quadrature-direct axis inductance of a permanent magnet synchronous motor, comprising a permanent magnet flux linkage identification stage and a quadrature-direct axis inductance identification stage; The permanent magnet flux identification stage includes: Step A1: Obtain the stator resistance, drive the motor to speed up to the first speed with the q-axis current, and the d-axis current is zero. After reaching the first speed, set the q-axis current to zero and allow the motor to decelerate freely. Step A2: When the motor freely decelerates to the second speed and meets the steady-state determination condition, it is determined that the motor has entered the steady-state operating condition, and the second speed is lower than the first speed. Step A3: Collect N sets of data under steady-state conditions. Each set of data must include at least the rotational speed, d-axis voltage, and q-axis voltage. During the data collection process, the validity of each set of data is verified according to preset rules. Step A4: Calculate the electric angular velocity and back electromotive force amplitude based on the verified data of each group, then calculate the corresponding permanent magnet flux identification value for each group, and take the average of the verified permanent magnet flux identification values ​​to obtain the final identification result of the permanent magnet flux. The quadrature-axis inductor identification stage includes: Step B1: Inject a current of a preset amplitude into the d-axis; Step B2: Control the motor to run at the third speed using a speed closed-loop method, and wait for the motor to enter a steady-state condition; Step B3: Collect M sets of data under steady-state conditions. Each set of data must include at least d-axis voltage, q-axis voltage, d-axis current, and q-axis current. During the data collection process, the validity of each set of data is verified according to preset rules. Step B4: Based on the verified data and the final identification results of the permanent magnet flux linkage, calculate the corresponding direct-axis inductance and quadrature-axis inductance identification values ​​for each group according to the steady-state voltage equation in the dq coordinate system. Then, average the verified identification values ​​for each group to obtain the final identification results of the direct-axis inductance and quadrature-axis inductance.

[0009] Preferably, the stator resistance is obtained in step A1 using one of the following methods: Method 1, DC Volt-Ampere Method: With the motor stationary and shaft locked, a controllable DC voltage is applied to any two phases of the stator windings. The steady-state voltage U_dc and steady-state current I_dc are collected, and then... Calculate the equivalent resistance of a single-phase winding, and take the average value after multiple sets of sampling calculations; Method 2: Use a multimeter to directly measure the resistance of the three-phase terminals of the motor and then convert it to obtain the single-phase resistance.

[0010] Preferably, the steady-state determination condition in steps A2 and B2 is: the rotational speed fluctuation does not exceed d-axis current fluctuation does not exceed Rated current and q-axis current fluctuations shall not exceed Furthermore, all three conditions must be met simultaneously for at least 200ms.

[0011] Preferably, in step A1, the first speed is 60% to 90% of the rated speed of the motor; in step A2, the second speed is 30% to 60% of the rated speed of the motor, and the difference between the first speed and the second speed is not less than 20% of the rated speed of the motor.

[0012] Preferably, the preset rules in steps A3 and B3 include: Speed ​​verification: If the speed value in the current data deviates from the preset target speed for this stage by more than [percentage missing], If so, discard that set of data; Amplitude verification: If the voltage or current value in the current data exceeds the value corresponding to the rated operating condition of the motor. If the range is too wide, then discard the data set. Only when a set of data passes both of the above verifications can that set of data participate in subsequent calculations.

[0013] Preferably, the sampling frequency of the data collected in steps A3 and B3 is 10kHz to 20kHz, the time interval between two adjacent sets of valid data is 5ms to 20ms, and the number of collection groups N and M are each independently selected from 10 to 50 groups.

[0014] Preferably, the current with preset amplitude in step B1 is 5% to 25% of the rated current of the motor; and the third speed in step B2 is 20% to 60% of the rated speed of the motor.

[0015] Preferably, after determining that the motor has entered a steady-state condition in steps A2 and B2, a continuous stabilization waiting time of 100ms to 300ms is added before starting data acquisition in step A3 or B3.

[0016] Preferably, in step A4, the electric angular velocity is calculated according to... Calculate the back electromotive force amplitude according to Calculation, permanent magnet flux according to Calculate, where n is the rotational speed, p is the number of pole pairs of the motor, and u_d and u_q are the d-axis voltage and q-axis voltage, respectively.

[0017] Preferably, in step B4, the direct-axis inductance and quadrature-axis inductance are calculated based on the steady-state voltage equation in the dq coordinate system, specifically according to the following formula: Where L_d is the direct-axis inductance and L_q is the quadrature-axis inductance. For stator resistance, Electric angular velocity, The final identification result of the permanent magnet flux obtained in step A4 of claim 1 is as follows. These represent the d-axis voltage, q-axis voltage, d-axis current, and q-axis current, respectively.

[0018] (III) Beneficial Effects This invention provides an offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance in a permanent magnet synchronous motor, which has the following beneficial effects: (1) In the prior art, the motor is in a "loaded" state when identifying the magnetic flux, and the resistance voltage drop must be deducted; the present invention, through the process design of "first speeding up, then setting to zero, and then speeding down", makes the motor in a pure no-load gliding state at the sampling time, and the resistance voltage drop is zero, which fundamentally eliminates the transmission influence of resistance error on magnetic flux identification. This process design is not a simple superposition of the steps, but a system-level innovation with synergistic effect. (2) Existing technologies use single-threshold steady-state determination and post-processing filtering; this invention introduces a steady-state confirmation mechanism of "rotation speed + dual-axis current" three-dimensional composite criterion + 200ms time window, as well as a feedforward data verification mechanism of "determination first, calculation later", which simultaneously ensures the reliability of the calculation input from the two dimensions of data source and data processing. (3) Existing technologies separate the identification of magnetic flux and inductance; the present invention constructs a serial dependency relationship of “magnetic flux identification output - inductance identification input”, making the whole scheme a closed-loop self-consistent system, and the error of any link will be found by the physical rationality test of the subsequent links. Attached Figure Description

[0019] Figure 1 This is a flowchart of the permanent magnet flux identification stage in the offline identification method for permanent magnet synchronous motors and quadrature-direct axis inductors of the present invention. Figure 2 This is a flowchart of the offline identification stage of the quadrature and direct axis inductance in a permanent magnet synchronous motor according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figures 1-2 The present invention provides an offline identification method for permanent magnet flux linkage and quadrature-direct axis inductance of permanent magnet synchronous motor, including a permanent magnet flux linkage identification stage and a quadrature-direct axis inductance identification stage; The permanent magnet flux identification stage includes: Step A1: Obtain the stator resistance, drive the motor to speed up to the first speed with the q-axis current, and the d-axis current is zero. After reaching the first speed, set the q-axis current to zero and allow the motor to decelerate freely. Step A2: When the motor freely decelerates to the second speed and meets the steady-state determination condition, it is determined that the motor has entered the steady-state operating condition, and the second speed is lower than the first speed. Step A3: Collect N sets of data under steady-state conditions. Each set of data must include at least the rotational speed, d-axis voltage, and q-axis voltage. During the data collection process, the validity of each set of data is verified according to preset rules. Step A4: Calculate the electric angular velocity and back electromotive force amplitude based on the verified data of each group, then calculate the corresponding permanent magnet flux identification value for each group, and take the average of the verified permanent magnet flux identification values ​​to obtain the final identification result of the permanent magnet flux. The quadrature-axis inductor identification stage includes: Step B1: Inject a current of a preset amplitude into the d-axis; Step B2: Control the motor to run at the third speed using a speed closed-loop method, and wait for the motor to enter a steady-state condition; Step B3: Collect M sets of data under steady-state conditions. Each set of data must include at least d-axis voltage, q-axis voltage, d-axis current, and q-axis current. During the data collection process, the validity of each set of data is verified according to preset rules. Step B4: Based on the verified data and the final identification results of the permanent magnet flux linkage, calculate the corresponding direct-axis inductance and quadrature-axis inductance identification values ​​for each group based on the steady-state voltage equation in the dq coordinate system. Then, average the verified identification values ​​for each group to obtain the final identification results of the direct-axis inductance and quadrature-axis inductance. In step A4, the electric angular velocity is as follows Calculate the back electromotive force amplitude according to Calculation, permanent magnet flux according to Calculate, where n is the rotational speed, p is the number of pole pairs of the motor, and u_d and u_q are the d-axis voltage and q-axis voltage, respectively; In step B4, the direct-axis inductance and quadrature-axis inductance are calculated based on the steady-state voltage equation in the dq coordinate system, specifically using the following formulas: Where L_d is the direct-axis inductance and L_q is the quadrature-axis inductance. For stator resistance, Electric angular velocity, The final identification result of the permanent magnet flux obtained in step A4 of claim 1 is as follows. These represent the d-axis voltage, q-axis voltage, d-axis current, and q-axis current, respectively.

[0022] In this embodiment, in step A1 of the permanent magnet flux identification stage: This solution employs an open-loop speed ramp-up method. Traditional methods typically run the motor directly to its rated speed for single-point sampling. However, this invention uses an open-loop speed mode and sets the first speed to 60%–90% of the rated speed, rather than requiring it to reach the rated speed. This design has several advantages. First, it avoids the safety risks associated with high-speed operation by setting the ramp-up target below the rated speed, reducing mechanical stress and safety hazards during high-speed motor rotation. This is particularly suitable for testing environments lacking speed feedback or with inadequate protection mechanisms. Second, it creates sufficient kinetic energy reserves for subsequent free deceleration. A sufficient speed difference (not less than 20% of the rated speed) is reserved between the first and subsequent second speeds, ensuring that the motor has sufficient inertial gliding distance after the q-axis current is zeroed, thereby collecting a sufficient number of effective data points during deceleration. Third, the open-loop speed mode avoids interference from the speed loop regulator on the current response, allowing the q-axis current to quickly return to zero after being zeroed. This eliminates the current tailing problem caused by speed loop integral saturation, ensuring the purity of the "no-load" operating condition.

[0023] In step A2 of the permanent magnet flux identification stage: This scheme employs free descent and fixed-point sampling, which offers the following advantages: First, the conventional method for identifying permanent magnet flux linkage in existing technologies is to directly sample and calculate the flux linkage after the motor has reached a certain speed. In this method, the motor is in a "loaded" driving state, the q-axis current is not zero, and there is a voltage drop across the stator resistance. The calculation of the back electromotive force must deduct the voltage drop term, introducing additional calculation errors. This invention creatively adopts a "speed-up, zero-current, and speed-down" process: after reaching the first speed, the q-axis current is completely zero, allowing the motor to enter a pure inertial coasting state. At this time, the motor is in a true no-load condition, the d-axis current is zero, the q-axis current is zero, the stator resistance voltage drop is zero, and the motor terminal voltage is the pure back electromotive force. This design simplifies flux linkage calculation. The direct division method eliminates the need to deduct the voltage drop across the resistance, fundamentally eliminating the transmission effect of resistance measurement error on the accuracy of magnetic flux identification. This is the most essential difference from existing technologies. Second, existing technologies typically sample when the motor is running stably at a certain speed, while this invention chooses to sample during the free-fall process. The underlying technical consideration is that the electromagnetic torque of the motor is zero during the free-fall process, the current loop is in the purest zero-current state, there is no dynamic adjustment disturbance of the current loop PI regulator, and the sampled voltage signal is a pure back EMF waveform with the highest signal-to-noise ratio. In contrast, there is an accelerating torque current in the acceleration phase and a torque current required to maintain the speed in the steady-state operation phase, neither of which can obtain a pure no-load back EMF. Third, existing technologies typically only monitor whether the rotational speed is stable (single threshold judgment), but "rotational speed stability" is not the same as "electrical quantity stability". The current inner loop may have slight oscillations, while the rotational speed feedback shows stability due to the filtering effect of mechanical inertia. This invention introduces a three-dimensional composite criterion of "rotational speed + d-axis current + q-axis current", and requires that it be met simultaneously for at least 200ms. This design effectively avoids sampling errors caused by misjudging the steady state due to the non-convergence of the current inner loop. Fourth, an additional 100ms to 300ms waiting time is added after determining the steady state to ensure that the system has sufficient time to eliminate transient disturbances that may exist at the moment of determination, further ensuring the absolute stability of the system state at the sampling time; Fifth, a hard constraint is set that the difference between the first speed and the second speed is not less than 20% of the rated speed. This constraint ensures that the motor has sufficient kinetic energy reserve and time window to complete the collection of N sets of data during the free deceleration process. If the difference is too small, the deceleration period is too short and there is not enough time to collect enough effective data. If the difference is too large (although it is theoretically feasible, it is difficult to achieve in actual operation), the acceleration process will take too long. The 20% difference lower limit is the optimal solution domain obtained by coupling analysis of the motor mechanical time constant and the sampling period, and is not a conventional design choice for those skilled in the art.

[0024] In step A3 of the permanent magnet flux identification stage: The conventional approach in existing technologies is: collect all data → substitute into the formula for calculation → remove outliers from the calculation results (post-processing filtering). This post-processing method has a fatal flaw: when a set of abnormal data (such as extreme voltage values ​​caused by sampling spikes) is substituted into the formula, it may cause the single calculation result to deviate significantly (especially when the denominator is a small value, small disturbances will be amplified). Even if the outlier is removed later, the outlier has already "contaminated" the calculation process (for example, it has already participated in the calculation of some intermediate variables before averaging). This invention employs a feedforward hard threshold rejection mechanism that prioritizes "judgment before calculation": data is verified before entering the calculation stage, and abnormal data does not participate in any calculation. This design effectively avoids "contamination" of the calculation process by outliers, fundamentally ensuring that all data involved in the mean calculation are reliable. At the same time, "real-time" judgment means that data collection and validity verification are carried out in parallel without adding extra processing time overhead. Speed ​​verification threshold The selection of (approximately 1% of the target rotational speed) and the amplitude verification threshold ±20% ensures sufficient data throughput (avoiding insufficient effective data due to overly strict thresholds) while effectively eliminating abnormal spikes introduced by sampling noise and ADC quantization errors.

[0025] In step A4 of the permanent magnet flux identification stage: Existing technologies for identifying permanent magnet flux linkage typically require first estimating the αβ-axis flux linkage components, then deriving the effective flux linkage, then fitting the permanent magnet flux linkage, and finally correlating the flux linkage with the inductance. This is a multi-step iterative calculation process, with each step introducing computational errors that accumulate progressively. In contrast, this invention directly utilizes a dq-axis steady-state model under no-load conditions. and The two formulas can calculate the flux linkage value in one step without any iteration or recursion, and the amount of computation is extremely small. The averaging of multiple data sets further suppressed the impact of residual sampling noise. The sampling size of N=30 sets ensured statistical validity (the sample size was sufficient to eliminate random errors) without excessively increasing the acquisition time (the total acquisition time was only 0.3 seconds). In step B1 of the quadrature-direct axis inductance identification stage: The amplitude of the injected current is not chosen arbitrarily. This invention limits the injected current to a narrow range of 5% to 25% of the rated current, which is based on two considerations: Firstly, to prevent the motor from overheating, excessive injection current will lead to increased copper losses. If the motor temperature rises too quickly, it may not only damage the motor, but also cause the resistance value to drift due to temperature changes, affecting the recognition accuracy. Secondly, to ensure the signal-to-noise ratio, if the injected current is too small, the signal amplitudes of u_d and i_d will be too low and easily submerged by sampling noise, resulting in an excessive error in the calculated L_d. A current amplitude of 5% to 25% achieves the optimal balance between "not causing significant heat generation" and "ensuring a sufficient signal-to-noise ratio". Furthermore, the choice of d-axis injection instead of q-axis injection also has deeper considerations: d-axis current only affects the saturation of the direct-axis magnetic circuit and has little effect on the q-axis inductance; while q-axis current will generate torque (causing speed fluctuations) and cross-coupling effect at the same time, which is not conducive to steady-state maintenance. D-axis injection can make the equation full rank while minimizing the disturbance to the steady state of the system.

[0026] In step B2 of the quadrature-direct axis inductance identification stage: This step selects a combination of operating conditions of "medium-low speed (20% to 60% of rated speed) + speed closed loop", which is significantly different from existing technologies: Existing technologies for identifying inductance often employ pulse voltage or high-frequency injection methods under static conditions. While these methods do not require rotating the motor, they have significant drawbacks: the pulse voltage method relies on the accurate measurement of the current rise time (0.632 times the steady-state current), which is susceptible to ADC sampling delay and quantization errors; the high-frequency injection method requires additional signal generation and extraction circuitry and may cause motor whine. This invention identifies inductors at low to medium rotational speeds, which has the following advantages: First, the back electromotive force term Substitute it into the formula as a known quantity (because) (The identification has been completed in the first stage), which allows L_d and L_q to be solved analytically directly from the steady-state voltage equation without any iteration or high-frequency signal processing; Secondly, the low to medium speed (20% to 60% of the rated speed) avoids the safety risks of high-speed operation, while ensuring sufficient back electromotive force amplitude, so that the u_d and u_q signals have sufficient signal-to-noise ratio.

[0027] Third, the speed closed-loop ensures precise controllability of the rotational speed, making... Being a precisely known quantity, rather than a measured value, further reduces sources of error. In step B3 of the quadrature-direct axis inductance identification stage: The validity verification rules in this step are consistent with those in the first stage, ensuring the uniformity and reproducibility of the data processing logic in both stages. It is worth noting that the amplitude verification in this stage covers four quantities (u_d, u_q, i_d, i_q) simultaneously, which is more stringent than the voltage verification in the first stage. This is because both voltage and current appear in the denominator of the inductance calculation formula, and any abnormality in any quantity may lead to a serious deviation in the calculation result.

[0028] In step B4 of the quadrature-axis inductor identification stage: This step has the following advantages: The creativity of this step is reflected in the following aspects: First, it directly solves analytically with zero iterations and zero recursion. Existing technologies for identifying L_d and L_q often use model reference adaptive methods, recursive least squares methods, or intelligent optimization algorithms. These methods all require repeated iterations to converge, resulting in large computational loads and convergence speeds that are significantly affected by initial values. In contrast, this invention directly starts from the steady-state voltage equation and obtains the closed-form analytical solutions of L_d and L_q through simple algebraic transformations without any iterative process. This difference has an order-of-magnitude advantage in computational efficiency. This invention can complete all inductor identification with only a few multiplication and division operations, while iterative methods may require hundreds or even thousands of iterations. Second, the two-stage serial strong dependency architecture ensures the physical consistency of parameters. The inductance calculation formula in this stage directly substitutes the values ​​obtained from the first stage identification. This means that if the magnetic flux identification in the first stage is inaccurate, the inductance identification in the second stage will inevitably be inaccurate. Conversely, if the finally identified L_d and L_q are physically reasonable, it also indirectly verifies... The accuracy of this serial architecture, where the output of the front stage serves as the forced input of the back stage, makes the entire identification scheme a closed-loop self-consistent system. Errors in any stage will be detected by the physical rationality verification of the subsequent stages. Existing technologies separate the identification of each parameter into independent modules, which cannot form such a self-consistent verification mechanism. Third, the mean processing of multiple sets of data further suppresses residual errors. Similar to the first stage, the sampling quantity of M=30 sets ensures statistical validity and eliminates residual sampling noise and random disturbances.

[0029] In step A1, the stator resistance is obtained using one of the following methods: Method 1, DC Volt-Ampere Method: With the motor stationary and shaft locked, a controllable DC voltage is applied to any two phases of the stator windings. The steady-state voltage U_dc and steady-state current I_dc are collected, and then... Calculate the equivalent resistance of a single-phase winding, and take the average value after multiple sets of sampling calculations; Method 2: Use a multimeter to directly measure the resistance of the three-phase terminals of the motor and then convert it to obtain the single-phase resistance.

[0030] In this embodiment, the present invention uses the stator resistor The acquisition method is limited to two parallel schemes: the DC volt-ampere method and the direct measurement method using a multimeter. This allows implementers to flexibly choose according to the on-site equipment conditions. When a programmable DC power supply is available, the volt-ampere method can obtain higher accuracy. When only basic electrical tools are available, the multimeter method can also meet the project requirements. The two methods serve as backups for each other, avoiding the predicament of the entire identification process being unable to be executed due to the limitation of a single acquisition method. At the same time, the volt-ampere method effectively eliminates the measurement errors introduced by contact resistance and temperature drift by averaging the results after sampling multiple sets of different voltages and currents. The multimeter method further suppresses single-phase measurement deviations through redundant measurement and averaging of the three-phase terminal resistances, thus providing a basis for the subsequent inductance identification formula. The parameters provide accurate and reliable input, fundamentally eliminating the transmission path from stator resistance error to inductance identification results.

[0031] The steady-state determination condition in steps A2 and B2 is: the speed fluctuation does not exceed d-axis current fluctuation does not exceed Rated current and q-axis current fluctuations shall not exceed Furthermore, all three conditions must be met simultaneously for at least 200ms.

[0032] In this embodiment, the present invention realizes a three-dimensional multi-dimensional joint determination of the motor steady state, overcoming the false steady state problem in the prior art that relies solely on a single threshold of rotational speed, where "the rotational speed appears stable due to mechanical inertia filtering, but the current inner loop has not actually converged." Rotational speed stability only represents mechanical dynamic balance, while current stability represents electromagnetic dynamic balance; both are indispensable. This solution fundamentally eliminates the sampling data failure caused by misjudging the steady state when there are still slight oscillations in the current inner loop through dual-domain joint verification of electrical quantities (d / q axis current) and mechanical quantities (rotational speed). At the same time, the 200ms duration requirement further eliminates the accidental steady state misjudgment triggered by instantaneous disturbances, ensuring the absolute reliability of the system state at the sampling moment.

[0033] In step A1, the first speed is 60% to 90% of the motor's rated speed; in step A2, the second speed is 30% to 60% of the motor's rated speed, and the difference between the first speed and the second speed is not less than 20% of the motor's rated speed.

[0034] In this embodiment, the present invention establishes a differentiated asymmetric mapping relationship between the acceleration target and the deceleration sampling point by limiting the preset first speed to 60%–90% of the rated speed and the preset second speed to 30%–60% of the rated speed, and by mandating that the difference between the two is not less than 20% of the rated speed. On the one hand, the range of the first speed ensures that the motor has sufficient kinetic energy reserves before free deceleration, so that the motor has sufficient inertial gliding distance to complete data acquisition after the q-axis current is set to zero. On the other hand, the range of the second speed ensures that the sampling point is located in the speed range with a moderate back EMF amplitude and a good signal-to-noise ratio, avoiding the back EMF signal being submerged by sampling noise when the speed is too low. Most importantly, the hard constraint that the difference between the two is not less than 20% of the rated speed ensures that the back EMF has a sufficient monotonic rate of change during the deceleration process, thereby having sufficient numerical differentiation between the data sets within the sampling window, and avoiding the loss of statistical validity due to information redundancy between the sampling points caused by the slow change of speed.

[0035] The preset rules in steps A3 and B3 include: Speed ​​verification: If the speed value in the current data deviates from the preset target speed for this stage by more than [percentage missing], If so, discard that set of data; Amplitude verification: If the voltage or current value in the current data exceeds the value corresponding to the rated operating condition of the motor. If the range is too wide, then discard the data set. Only when a set of data passes both of the above verifications can that set of data participate in subsequent calculations.

[0036] In this embodiment, the present invention constructs a feedforward data filtering mechanism of "judgment first, calculation later," which is fundamentally different from the post-processing filtering method of the prior art of "collecting all data → substituting into the formula for calculation → removing outliers in the calculation result." The fatal flaw of post-processing filtering is that abnormal data has already been substituted into the formula for calculation before it is discovered. When an outlier appears in the denominator, it may cause serious distortion of the single calculation result and contaminate the mean. However, the feedforward filtering of the present invention completes the judgment before the data enters the calculation stage. Abnormal data does not participate in any calculation at all, thus eliminating the contamination of the identification result by sampling spikes and ADC conversion anomalies from the source. At the same time, speed verification excludes valid data in which the actual speed deviates from the target value (reflecting operating condition drift), and amplitude verification excludes invalid data in which voltage and current surge instantaneously (reflecting sampling noise). The two have clear division of labor and complement each other, jointly ensuring that each set of data participating in the mean calculation is reliable in both the dimensions of operating condition validity and signal validity.

[0037] The sampling frequency of the data collected in steps A3 and B3 is 10kHz to 20kHz, the time interval between two adjacent valid data sets is 5ms to 20ms, and the number of data sets N and M are each independently selected from 10 to 50 sets.

[0038] In this embodiment, the present invention constructs a three-dimensional coordinated configuration relationship between sampling frequency, sampling density, and sample size. A sampling frequency of 10kHz to 20kHz is sufficient to capture all dynamic components of current and voltage signals without losing information, while avoiding unnecessary computational burden on the main control chip due to excessively high sampling frequencies. A sampling interval of 5ms to 20ms ensures that adjacent data sets are fully decoupled in time, avoiding the loss of statistical independence due to high correlation between samples caused by excessively dense sampling points. A sample size of 10 to 50 sets satisfies the statistical effectiveness requirements of mean filtering without causing the motor state (such as temperature and magnetic saturation) to drift due to excessively long acquisition time. The above three parameters match and work together to achieve an optimal balance between information integrity, sample independence, and state stability in the entire sampling process.

[0039] In step B1, the current with the preset amplitude is 5% to 25% of the motor's rated current; in step B2, the third speed is 20% to 60% of the motor's rated speed.

[0040] In this embodiment, the present invention constructs an optimal working condition combination of low-current excitation and medium-low speed rotation during the inductor identification stage. The 5% to 25% range of the d-axis injected current precisely balances the contradictory requirements of ensuring the full rank of the inductor identification equation and preventing motor overheating: below 5%, the i_d signal amplitude is too small and easily submerged by sampling noise. If the term is too small, the denominator in the L_d calculation formula approaches zero, amplifying the error. When it exceeds 25%, copper losses increase significantly and may cause the motor magnetic circuit to enter the saturation region, altering the inductance value to be identified. The preset third speed range of 20%–60% precisely balances the conflicting requirements of obtaining sufficient back EMF signal-to-noise ratio and ensuring operational safety: below 20%... If the inductance term is too small, the inductance term in the voltage equation will have a low proportion, which will reduce the identification sensitivity. If it is higher than 60%, high-speed rotation will bring safety hazards and may trigger the field weakening control zone to change the motor parameters. The two ranges work together to ensure that inductance identification is carried out under the condition that the signal is sufficient and the state is pure.

[0041] After determining that the motor has entered a steady-state condition in steps A2 and B2, a continuous stabilization waiting time of 100ms to 300ms is added before starting data acquisition in step A3 or B3.

[0042] In this embodiment, the present invention constructs a three-stage sampling sequence of steady-state determination → stable consolidation → data acquisition, effectively solving the problem of residual dynamic error that exists at the moment of determination when sampling occurs. At the moment the steady-state determination condition is met, although the system has met the threshold values ​​of each fluctuation quantity, it may still be in the final stage of the transition from dynamic to steady state (such as current fluctuations from...). convergence to (The stabilization process has not yet fully ended). Sampling should be started immediately at this time. The first few sets of data may still carry weak transient components. The waiting time of 100ms to 300ms provides sufficient rest period for the control system, allowing all integral terms of the regulators in the current loop and speed loop to fully converge and all oscillation modes to fully decay. This ensures that the first batch of collected data is in an absolute steady state, thereby eliminating the systematic deviation introduced by unreasonable sampling timing design and improving the repeatability and reproducibility of the identification results.

[0043] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0044] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0045] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor, characterized in that, This includes the permanent magnet flux identification stage and the quadrature-axis and direct-axis inductance identification stage; The permanent magnet flux identification stage includes: Step A1: Obtain the stator resistance, drive the motor to speed up to the first speed with the q-axis current, and the d-axis current is zero. After reaching the first speed, set the q-axis current to zero and allow the motor to decelerate freely. Step A2: When the motor freely decelerates to the second speed and meets the steady-state determination condition, it is determined that the motor has entered a steady-state operating condition, where the second speed is lower than the first speed. Step A3: Collect N sets of data under steady-state conditions. Each set of data must include at least the rotational speed, d-axis voltage, and q-axis voltage. During the data collection process, the validity of each set of data is verified according to preset rules. Step A4: Calculate the electric angular velocity and back electromotive force amplitude based on the verified data of each group, then calculate the corresponding permanent magnet flux identification value for each group, and take the average of the verified permanent magnet flux identification values ​​to obtain the final identification result of the permanent magnet flux. The quadrature-axis and direct-axis inductor identification stage includes: Step B1: Inject a current of a preset amplitude into the d-axis; Step B2: Control the motor to run at the third speed using a speed closed-loop method, and wait for the motor to enter a steady-state condition; Step B3: Collect M sets of data under steady-state conditions. Each set of data must include at least d-axis voltage, q-axis voltage, d-axis current, and q-axis current. During the data collection process, the validity of each set of data is verified according to the preset rules. Step B4: Based on the verified data of each group and the final identification result of the permanent magnet flux linkage, calculate the corresponding direct-axis inductance and quadrature-axis inductance identification values ​​of each group according to the steady-state voltage equation in the dq coordinate system, and average the verified identification values ​​of each group to obtain the final identification results of the direct-axis inductance and quadrature-axis inductance.

2. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: The stator resistance obtained in step A1 can be obtained using one of the following methods: Method 1, DC Volt-Ampere Method: With the motor stationary and shaft locked, a controllable DC voltage is applied to any two phases of the stator windings. The steady-state voltage U_dc and steady-state current I_dc are collected, and then... Calculate the equivalent resistance of a single-phase winding, and take the average value after multiple sets of sampling calculations; Method 2: Use a multimeter to directly measure the resistance of the three-phase terminals of the motor and then convert it to obtain the single-phase resistance.

3. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: The steady-state determination condition mentioned in steps A2 and B2 is: the rotational speed fluctuation does not exceed d-axis current fluctuation does not exceed Rated current and q-axis current fluctuations shall not exceed Furthermore, all three conditions must be met simultaneously for at least 200ms.

4. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: In step A1, the first speed is 60% to 90% of the rated speed of the motor; in step A2, the second speed is 30% to 60% of the rated speed of the motor, and the difference between the first speed and the second speed is not less than 20% of the rated speed of the motor.

5. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: The preset rules mentioned in steps A3 and B3 include: Speed ​​verification: If the speed value in the current data deviates from the preset target speed for this stage by more than [percentage missing], If so, discard that set of data; Amplitude verification: If the voltage or current value in the current data exceeds the value corresponding to the rated operating condition of the motor. If the range is too wide, then discard the data set. Only when a set of data passes both of the above verifications can that set of data participate in subsequent calculations.

6. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: The sampling frequency of the data collected in steps A3 and B3 is 10kHz to 20kHz, the time interval between two adjacent valid data sets is 5ms to 20ms, and the number of data sets N and M are each independently selected from 10 to 50 sets.

7. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: The current with the preset amplitude mentioned in step B1 is 5% to 25% of the rated current of the motor; the third speed mentioned in step B2 is 20% to 60% of the rated speed of the motor.

8. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: After determining that the motor has entered a steady-state condition in steps A2 and B2, a continuous stabilization waiting time of 100ms to 300ms is added before starting data acquisition in step A3 or B3.

9. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: The electric angular velocity mentioned in step A4 is as follows: The calculation is based on the back electromotive force amplitude. Calculations show that the permanent magnet linkage is based on... Calculate, where n is the rotational speed, p is the number of pole pairs of the motor, and u_d and u_q are the d-axis voltage and q-axis voltage, respectively.

10. The offline identification method for permanent magnet flux linkage and quadrature-axis and direct-axis inductance of a permanent magnet synchronous motor according to claim 1, characterized in that: The calculation of the direct-axis inductance and quadrature-axis inductance based on the steady-state voltage equation in the dq coordinate system described in step B4 is specifically calculated using the following formulas: Where L_d is the direct-axis inductance and L_q is the quadrature-axis inductance. For stator resistance, Electric angular velocity, The final identification result of the permanent magnet flux obtained in step A4 of claim 1 is as follows. These represent the d-axis voltage, q-axis voltage, d-axis current, and q-axis current, respectively.

Images