Unit cascade type motor speed control compensation method and system

Abstract

The application discloses a unit cascade type motor speed regulation control compensation method and system, relates to the motor control technical field, and solves the technical problems that parameter dynamic changes caused by temperature rise, magnetic saturation or load mutation during motor operation, real-time parameter identification is not performed, magnetic resistance torque compensation error is relatively large, and motor speed regulation control compensation accuracy is influenced; the difference value of magnetic resistance torque and Ld-Lq is strongly related, and the difference value of Ld-Lq is jointly influenced by motor design and working conditions. RLS can accurately quantify the nonlinear part of magnetic resistance torque by online identification of the difference value of Ld-Lq, and avoids the limitations of traditional table lookup method or offline calibration. Cross coupling exists in the d-q axis current loop, RLS can decouple the mutual influence between parameters by simultaneously identifying inductance parameters, and the current loop bandwidth is improved; meanwhile, the inductance value drift caused by motor temperature rise is adjusted by continuously monitoring the inductance change, and torque attenuation caused by temperature rise is avoided.

Description

Technical Field

[0001] This invention belongs to the field of motor control technology, specifically a method and system for speed control compensation of unit-cascaded motors. Background Technology

[0002] Traditional two-level frequency converters require series connection of components in high-voltage scenarios, leading to difficulties in voltage equalization and reduced efficiency. Cascaded structures, through modular design, directly connect low-voltage units in series to achieve high-voltage output, avoiding the complexity of component series connection. Motor systems account for over 60% of industrial electricity consumption; improving speed control accuracy and efficiency can significantly reduce energy consumption. Cascaded structures reduce harmonics, lower motor losses, and extend equipment life through multi-level output, making them widely applicable. A motor speed control compensation scheme based on a multi-level power unit series architecture is needed, achieving high-voltage, high-capacity output through modular design and combining dynamic compensation strategies to improve system performance.

[0003] Traditional methods typically use offline calibrated parameters such as inductance and flux linkage. However, when the motor is running, the temperature rises, the magnetic saturation occurs, or the load changes suddenly, causing the parameters to change dynamically. If the parameters are not identified in real time, the reluctance torque compensation error will be large, the motor speed control compensation will be inaccurate, and the d-axis current component needs to be increased to maintain torque output, which leads to an increase in the proportion of reactive power and a reduction in system efficiency. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a unit-cascaded motor speed control compensation method and system to solve the technical problem that the motor's operating temperature rise, magnetic saturation or sudden load change will cause dynamic changes in parameters, failure to identify parameters in real time, large magnetic reluctance torque compensation error, and affect the accuracy of motor speed control compensation.

[0005] To address the above problems, the first aspect of the present invention provides a method and system for speed control compensation of a unit-cascaded motor, comprising the following steps:

[0006] An edge node is established at each motor in the cascaded motor management area, forming an edge node cluster. This edge node cluster model is used to collect parameters and perform real-time compensation for the cascaded motor units, including:

[0007] An improved Prandtl-Ishlinskii model was adopted, and a temperature compensation term was introduced to correct the hysteresis operator parameters. A hysteresis model was constructed, and the model was identified by recursive least squares (RLS) to generate the dq-axis compensation voltage.

[0008] Based on the mathematical model of the motor's dq axis, the total torque is decomposed into permanent magnet torque Tpm and reluctance torque Tr;

[0009] In the current command generation stage, the recursive least squares method is used to identify the d-axis inductance and q-axis inductance in real time. Based on the difference between the d-axis inductance and q-axis inductance, the calculated value of reluctance torque is corrected and the reluctance torque compensation amount is calculated.

[0010] Edge nodes transmit the collected and compensation parameters to the cloud in real time for analysis, including:

[0011] Based on the thermal equivalent circuit model, the compensation values ​​of voltage and reluctance torque are corrected, and the flux linkage in the two-phase stationary coordinate system is compensated for flux linkage observation error through a limiting circuit. Control compensation signals are sent to the edge nodes according to the calculated correction parameters and error compensation parameters.

[0012] Optionally, in one example of the above aspects, an improved Prandtl-Ishlinskii model is used, in which a temperature compensation term is introduced to correct the hysteresis operator parameters, and a hysteresis model is constructed, including the following steps:

[0013] To address the effect of temperature on hysteresis characteristics, a temperature compensation term is introduced into the hysteresis operator:

[0014] ,

[0015] Where α is the temperature compensation coefficient, T(t) is the real-time temperature, and T0 is the reference temperature.

[0016] An improved PI model is constructed by combining two non-centrosymmetric hysteresis operators, and the hysteresis model is built as follows:

[0017] ,

[0018] Where H(x) is the hysteresis operator, wi and vi are the different weight coefficients of the i-th rising branch, x(t) is the input signal, and ri is the threshold of the i-th rising branch. , where n is the total number of rising branch hysteresis operators, and ht is the temperature compensation term.

[0019] Optionally, in one example of the above aspects, the dq-axis compensation voltage is generated by identifying the model using recursive least squares (RLS), including:

[0020] The recursive least squares (RLS) identification model is used to calculate the compensation voltage Vcomp for the d-axis and q-axis respectively:

[0021] ,

[0022] Where Vcomp is the compensation voltage, and H(x) is the compensation voltage. -1 This is a hysteresis inverse model, where Vref is the reference voltage, ωref is the reference angular frequency, ωn is the natural angular frequency, and k1 and k2 are the corresponding weighting coefficients.

[0023] Optionally, in one example of the above aspects, based on the mathematical model of the motor's dq axis, the total torque is decomposed into permanent magnet torque Tpm and reluctance torque Tr, including the following steps:

[0024] Based on the mathematical model of the motor's dq axis, the total torque Te is decomposed into permanent magnet torque Tpm and reluctance torque Tr:

[0025] ;

[0026] Where p is the number of pairs of rotor poles of the motor. Ld is the flux linkage of the permanent magnet, Lq is the d-axis inductance, Lq is the q-axis inductance, id is the d-axis current, and iq is the q-axis current.

[0027] Optionally, in one example of the above aspects, in the current command generation stage, the d-axis inductance and q-axis inductance are identified in real time using the recursive least squares method. Based on the difference between the d-axis inductance and q-axis inductance, the calculated reluctance torque value is corrected, and the reluctance torque compensation amount is calculated, including the following steps:

[0028] In the current command generation stage, a discretized and identifiable model is constructed using the dq-axis voltage equation:

[0029] ,

[0030] Where ud(k) is the d-axis voltage at discrete point k, uq(k) is the q-axis voltage at discrete point k, Ld is the d-axis inductance, Lq is the q-axis inductance, id(k) is the d-axis current at discrete point k, iq(k) is the q-axis current at discrete point k, Ts is the discretization time step, and ωe(k) is the angular frequency at discrete point k.

[0031] Organize into matrix form:

[0032] ,

[0033] in: , , ε(k) represents noise.

[0034] Initialize the identification parameter matrix θ(k) and covariance matrix P(k), and update the matrix θ(k) recursively using the recursive least squares method:

[0035] ,

[0036] ,

[0037] ,

[0038] Where K(k) is the gain matrix and λ is the forgetting factor (0 < λ ≤ 1), used to balance the influence of historical data and new data;

[0039] Based on the updated identification parameter matrix θ(k), the identified d-axis inductance is updated in real time. and q-axis inductance And transmit the data to the cloud in real time;

[0040] Based on the identified d-axis inductance and q-axis inductance Calculate the corrected reluctance torque:

[0041] ;

[0042] The corrected Trcomp is fed back to the current command generation stage to adjust the dq axis current command. and .

[0043] Optionally, in one example of the above aspects, the cloud-based system corrects the compensation values ​​for voltage and reluctance torque based on a thermal equivalent loop model, including the following steps:

[0044] A global parameter library is generated in the cloud to store motor model parameters and a temperature-magnetic saturation mapping table.

[0045] Based on the data uploaded by the edge nodes, the motor parameters are updated in batches for batch parameter identification. The electromagnetic field distribution of each cascaded motor at different temperatures is simulated through the finite element analysis model. The variation law of dq axis inductance Ld and Lq with temperature is extracted, and a temperature-inductance MAP is generated.

[0046] The winding temperature Tw is estimated in real time using a machine learning model based on a temperature-magnetic saturation mapping table, and the updated d-axis inductance and q-axis inductance are obtained by interpolation from the temperature-inductance MAP.

[0047] Calculate the reluctance torque for correction based on the updated d-axis and q-axis inductance;

[0048] Obtain the discretized identifiable model constructed from the edge nodes, and substitute the updated d-axis inductance and q-axis inductance into the corresponding model to obtain the compensation voltages for the d-axis and q-axis.

[0049] Optionally, in one example of the above aspects, flux linkage observation error compensation in the two-phase stationary coordinate system is performed by a limiting element, including the following steps:

[0050] Transform the magnetic flux linkage in the α-β coordinate system to the polar coordinate system:

[0051] ,

[0052] in, The stator flux linkage vector is given in the two-phase stationary coordinate system (α-β coordinate system). for The magnitude of the flux linkage vector, e, represents the strength of the flux linkage vector. jθ This represents the phase of the flux linkage vector. Let be the projection component of the flux linkage vector onto the α-axis. This is the projection component of the flux linkage vector onto the β-axis;

[0053] Independent amplitude and phase limiting are performed. The amplitude limiting is as follows: sat for flux linkage amplitude Perform a bandwidth limiting operation. ;

[0054] Discretize the phase and perform phase limiting correction:

[0055] Where θcomp(k) is the phase limiting after correction at the kth discrete point, θ(k) is the phase at the kth discrete point, θ(k−1) is the phase at the (k-1)th discrete point, Tsc is the discretization period, and Δθ is the limiting phase change amount;

[0056] The mean value of the corrected phase limit at each discrete point is calculated as the corrected phase limit θcomp. An inverse coordinate transformation is then performed to convert the compensated polar flux linkage back to the α-β coordinate system, yielding the corrected stator flux linkage vector. The specific formula is as follows:

[0057] ,

[0058] Among them, e jθcomp The phase representation of the flux linkage vector for the corrected phase-limited θcomp;

[0059] Compensation for flux linkage observation errors is achieved by using the phase representation of the flux linkage vector of the corrected phase-limited θcomp.

[0060] Optionally, in one example of the above aspects, sending a control compensation signal to the edge node based on the calculated correction parameters and error compensation parameters includes the following steps:

[0061] Obtain the dq-axis compensation voltage uploaded by the edge nodes, and update the identified d-axis inductance. and q-axis inductance The corrected reluctance torque;

[0062] Based on the updated d-axis and q-axis inductances, the corrected reluctance torque, and the compensation voltages for the d-axis and q-axis, the cloud compares the corresponding dq-axis inductances, the corrected reluctance torque, and the compensation voltages for the d-axis and q-axis. If the difference between the calculation results of the edge node and the calculation results of the cloud exceeds a preset threshold, the average value of the corresponding values ​​is used as the new correction parameter, and a correction parameter modification command is sent to the corresponding edge node.

[0063] The phase parameters of the flux linkage vector of the corrected phase-limited θcomp are sent to the edge nodes, and the edge nodes perform flux linkage observation error compensation based on the specific parameters.

[0064] According to another aspect of this disclosure, a unit-cascaded motor speed control compensation system is provided, which uses the unit-cascaded motor speed control compensation method described above to achieve unit-cascaded motor speed control compensation.

[0065] Compared with the prior art, the beneficial effects of the present invention are:

[0066] This invention demonstrates a strong correlation between reluctance torque and the difference between Ld and Lq, a difference influenced by both motor design and operating conditions. The Reluctance Scale (RLS) accurately quantifies the nonlinear component of the reluctance torque by identifying the Ld-Lq difference online, avoiding the limitations of traditional lookup table methods or offline calibration. When the motor load changes abruptly, transient changes in Ld and Lq cause torque fluctuations. The recursive nature of the RLS allows it to update inductance parameters with millisecond-level response speeds. Combined with a feedforward compensation strategy, this significantly shortens the torque recovery time.

[0067] This invention addresses the cross-coupling of the dq-axis current loop. By simultaneously identifying Ld and Lq, the RLS can decouple the mutual influence between parameters and improve the current loop bandwidth. At the same time, the motor temperature rise can cause the inductance value to drift. By continuously monitoring the inductance change, the RLS can automatically adjust the compensation amount to avoid torque attenuation caused by temperature rise.

[0068] This invention enables cloud-based analysis of coupling effects between units in a cascaded motor, generating coordinated compensation signals through a global optimization algorithm to avoid system conflicts caused by local compensation. When the deviation of the control compensation parameter of an edge node exceeds a threshold, the cloud can promptly adjust the compensation strategy to ensure system operation. Attached Figure Description

[0069] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0070] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0071] Figure 2 This is a schematic diagram of the system architecture of the present invention. Detailed Implementation

[0072] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0073] Please see Figures 1-2 The first aspect of this invention provides a method and system for speed control compensation of a cascaded motor, comprising the following steps:

[0074] An edge node is established at each motor in the cascaded motor management area, forming an edge node cluster. This edge node cluster model is used to collect parameters and perform real-time compensation for the cascaded motor units, including:

[0075] An improved Prandtl-Ishlinskii model was adopted, and a temperature compensation term was introduced to correct the hysteresis operator parameters. A hysteresis model was constructed, and the model was identified by recursive least squares (RLS) to generate the dq-axis compensation voltage.

[0076] Specifically, the traditional Prandtl-Ishlinskii model only describes the centrosymmetric hysteresis loop, while the hysteresis characteristics of motor drivers are typically asymmetric, such as those of piezoelectric ceramics. The improved model significantly improves modeling accuracy by introducing a non-centrosymmetric hysteresis operator to model the rising and falling branches separately.

[0077] In the cascaded motor structure, each unit can independently apply the improved model, and distributed compensation can be achieved through edge node clusters, reducing computational complexity and supporting system capacity expansion.

[0078] Temperature variations during motor operation can cause hysteresis characteristics to drift, such as the electrostriction effect of piezoelectric materials decaying with temperature. By introducing a temperature compensation term, the model can adjust the hysteresis operator parameters in real time to ensure that the compensation voltage matches the actual operating conditions. For example, under high-temperature conditions, the model can automatically increase the amplitude of the compensation voltage to counteract the effects of increased hysteresis nonlinearity.

[0079] The RLS algorithm achieves real-time identification of model parameters by recursively updating the covariance and gain matrices without the need for offline calibration. Combined with the low-latency computing capabilities of edge nodes, it can quickly generate dq-axis compensation voltages. It is robust to noise and parameter drift, and is especially suitable for operating conditions such as motor startup and sudden load changes.

[0080] By configuring an independent edge node for each motor unit, parameters such as current, voltage, and temperature can be collected locally, avoiding the communication bottleneck of centralized control.

[0081] Within the cluster, computing tasks can be dynamically allocated to nodes. When a node fails, the remaining nodes take over and compensate for the failure through redundancy design, ensuring continuous system operation. Furthermore, edge computing supports localized decision-making, reducing reliance on the cloud and improving system reliability.

[0082] By accurately offsetting the current distortion caused by hysteresis through dq-axis compensation voltage, the total harmonic distortion (THD) is reduced, the power factor is improved, and reactive power loss is reduced.

[0083] Based on the mathematical model of the motor's dq axis, the total torque is decomposed into permanent magnet torque Tpm and reluctance torque Tr;

[0084] In the current command generation stage, the recursive least squares method is used to identify the d-axis inductance and q-axis inductance in real time. Based on the difference between the d-axis inductance and q-axis inductance, the calculated value of reluctance torque is corrected and the reluctance torque compensation amount is calculated.

[0085] Specifically, in the speed control compensation method for unit-cascaded motors, the total torque is decomposed into permanent magnet torque (Tpm) and reluctance torque (Tr) based on the dq-axis mathematical model. The recursive least squares (RLS) method is used to identify the d-axis and q-axis inductances in real time, thereby correcting the calculated reluctance torque value and generating a compensation amount. During motor operation, Ld and Lq dynamically change due to temperature increases, magnetic saturation, or sudden load changes. Traditional methods using fixed inductance values ​​lead to deviations in reluctance torque calculations, while RLS, by identifying Ld and Lq in real time, can dynamically correct the calculated Tr value.

[0086] The reluctance torque is strongly correlated with the difference between Ld and Lq (ΔL = Lq - Ld), and ΔL is affected by both motor design (such as saliency ratio) and operating conditions. Reluctance torque resonator (RLS) can accurately quantify the nonlinear component of reluctance torque by identifying ΔL online, avoiding the limitations of traditional lookup table methods or offline calibration. When the motor load changes abruptly, the transient changes in Ld and Lq can cause torque fluctuations. The recursive nature of RLS allows it to update inductance parameters with a millisecond-level response speed. Combined with a feedforward compensation strategy, this can significantly shorten the torque recovery time.

[0087] The dq axis current loop has cross-coupling. By simultaneously identifying Ld and Lq, RLS can decouple the mutual influence between parameters and improve the current loop bandwidth. At the same time, the motor temperature rise will cause the inductance value to drift. By continuously monitoring the inductance change, RLS can automatically adjust the compensation amount to avoid torque attenuation caused by temperature rise.

[0088] The RLS algorithm has low complexity, making it suitable for deployment in edge nodes or low-power controllers. Compared to the EKF extended Kalman filter, RLS reduces computation while maintaining similar recognition accuracy. Precise magnetoresistive torque compensation reduces the d-axis current component (Id), thereby reducing copper and iron losses.

[0089] Edge nodes transmit the collected and compensation parameters to the cloud in real time for analysis, including:

[0090] Based on the thermal equivalent circuit model, the compensation values ​​of voltage and reluctance torque are corrected, and the flux linkage in the two-phase stationary coordinate system is compensated for flux linkage observation error through a limiting circuit. Control compensation signals are sent to the edge nodes according to the calculated correction parameters and error compensation parameters.

[0091] Specifically, parameters collected by edge nodes typically only reflect local states, while cloud-based thermal equivalent loop models can integrate electromagnetic, thermal, and mechanical characteristics of motors to construct a more comprehensive dynamic model. The cloud can store long-term operational data, and machine learning can be used to uncover implicit relationships between parameters and optimize compensation strategies.

[0092] Flux flux observations in a two-phase stationary coordinate system (α-β axis) are susceptible to parameter errors and noise interference, causing observed values ​​to deviate from the true values. Cloud-based systems constrain the amplitude and phase of the flux flux through amplitude limiting mechanisms, which can prevent system oscillations caused by the accumulation of observation errors.

[0093] In cascaded motors, the cloud can analyze the coupling effects between units and generate coordinated compensation signals through a global optimization algorithm to avoid system conflicts caused by local compensation. When the deviation of the control compensation parameter of an edge node exceeds a threshold, the cloud can adjust the compensation strategy in a timely manner to ensure system operation.

[0094] In one embodiment of the present invention, an improved Prandtl-Ishlinskii model is used, and a temperature compensation term is introduced to correct the hysteresis operator parameters to construct a hysteresis model, including the following steps:

[0095] To address the effect of temperature on hysteresis characteristics, the electrostriction coefficient of the piezoelectric ceramic in the motor varies with temperature. Therefore, a temperature compensation term is introduced into the hysteresis operator.

[0096] ,

[0097] Where α is the temperature compensation coefficient, T(t) is the real-time temperature, and T0 is the reference temperature. α is calibrated experimentally; for example, within the range of 25-80℃, the value of α ranges from 0.001 to 0.003 / ℃.

[0098] An improved PI model is constructed by combining two non-centrosymmetric hysteresis operators, and the hysteresis model is built as follows:

[0099] ,

[0100] Where H(x) is the hysteresis operator, wi and vi are the different weight coefficients of the i-th rising branch, x(t) is the input signal, and ri is the threshold of the i-th rising branch. , where n is the total number of rising branch hysteresis operators, and ht is the temperature compensation term.

[0101] In one embodiment of the present invention, the dq-axis compensation voltage is generated by identifying the model using the recursive least squares (RLS) method, including:

[0102] The recursive least squares (RLS) identification model is used to calculate the compensation voltage Vcomp for the d-axis and q-axis respectively:

[0103] ,

[0104] Where Vcomp is the compensation voltage, and H(x) is the compensation voltage. -1 This is a hysteresis inverse model, where Vref is the reference voltage, ωref is the reference angular frequency, ωn is the natural angular frequency, and k1 and k2 are the corresponding weighting coefficients. The settings of k1 and k2 can be calibrated through experiments. When the compensation voltage calculation error is less than the threshold, the mean values ​​of k1 and k2 are calculated and used as the corresponding weights. In this embodiment, k1 and k2 are set to 0.6 and 0.4, respectively.

[0105] In one embodiment of the present invention, based on the mathematical model of the motor's dq axis, the total torque is decomposed into permanent magnet torque Tpm and reluctance torque Tr, including the following steps:

[0106] Based on the mathematical model of the motor's dq axis, the total torque Te is decomposed into permanent magnet torque Tpm and reluctance torque Tr:

[0107] ;

[0108] Where p is the number of pairs of rotor poles of the motor. For permanent magnet flux linkage, Ld is the d-axis inductance, Lq is the q-axis inductance, id is the d-axis current, iq is the q-axis current, and Ld=Lq is the magnetic circuit asymmetry feature. Ld and Lq are obtained in real time through parameter identification.

[0109] In one embodiment of the present invention, in the current command generation stage, the d-axis inductance and q-axis inductance are identified in real time using the recursive least squares method. Based on the difference between the d-axis inductance and q-axis inductance, the calculated reluctance torque value is corrected, and the reluctance torque compensation amount is calculated, including the following steps:

[0110] In the current command generation stage, a discretized and identifiable model is constructed using the dq-axis voltage equation:

[0111] ,

[0112] Where ud(k) is the d-axis voltage at discrete point k, uq(k) is the q-axis voltage at discrete point k, Ld is the d-axis inductance, Lq is the q-axis inductance, id(k) is the d-axis current at discrete point k, iq(k) is the q-axis current at discrete point k, Ts is the discretization time step, and ωe(k) is the angular frequency at discrete point k.

[0113] Organize into matrix form:

[0114] ,

[0115] in: , , ε(k) represents noise.

[0116] Initialize the identification parameter matrix θ(k) and covariance matrix P(k), and update the matrix θ(k) recursively using the recursive least squares method:

[0117] ,

[0118] ,

[0119] ,

[0120] Where K(k) is the gain matrix and λ is the forgetting factor (0 < λ ≤ 1), used to balance the influence of historical data and new data;

[0121] Based on the updated identification parameter matrix θ(k), the identified d-axis inductance is updated in real time. and q-axis inductance And transmit the data to the cloud in real time;

[0122] Based on the identified d-axis inductance and q-axis inductance Calculate the corrected reluctance torque:

[0123] ;

[0124] The corrected Trcomp is fed back to the current command generation stage to adjust the dq axis current command. and .

[0125] For example, the reluctance torque compensation Δiq is calculated based on the difference between Ld and Lq, and the current command is corrected accordingly.

[0126] ,

[0127] in, The corrected current is Δiq, the correction parameter is kr, Ld is the d-axis inductance, Lq is the q-axis inductance, id is the d-axis current, and iq is the q-axis current.

[0128] In one embodiment of the present invention, the cloud platform corrects the compensation values ​​of voltage and reluctance torque based on a thermal equivalent circuit model, including the following steps:

[0129] A global parameter library is generated in the cloud to store motor model parameters and a temperature-magnetic saturation mapping table.

[0130] Based on the data uploaded by the edge nodes, the motor parameters are updated in batches for batch parameter identification. The electromagnetic field distribution of each cascaded motor at different temperatures is simulated through the finite element analysis model. The variation law of dq axis inductance Ld and Lq with temperature is extracted, and a temperature-inductance MAP is generated.

[0131] The winding temperature Tw is estimated in real time using a machine learning model based on a temperature-magnetic saturation mapping table, and the updated d-axis inductance and q-axis inductance are obtained by interpolation from the temperature-inductance MAP.

[0132] Calculate the reluctance torque for correction based on the updated d-axis and q-axis inductance;

[0133] Obtain the discretized identifiable model constructed from the edge nodes, and substitute the updated d-axis inductance and q-axis inductance into the corresponding model to obtain the compensation voltages for the d-axis and q-axis.

[0134] In one embodiment of the present invention, flux linkage observation error compensation is performed on the flux linkage in a two-phase stationary coordinate system through a limiting element, including the following steps:

[0135] Transform the magnetic flux linkage in the α-β coordinate system to the polar coordinate system:

[0136] ,

[0137] in, The stator flux linkage vector is given in the two-phase stationary coordinate system (α-β coordinate system). for The magnitude of the flux linkage vector, e, represents the strength of the flux linkage vector. jθThis represents the phase of the flux linkage vector. Let be the projection component of the flux linkage vector onto the α-axis. This is the projection component of the flux linkage vector onto the β-axis;

[0138] Independent amplitude and phase limiting are performed. The amplitude limiting is as follows: sat for flux linkage amplitude Perform a bandwidth limiting operation. ;

[0139] Discretize the phase and perform phase limiting correction:

[0140] ,

[0141] Where θcomp(k) is the phase limiting after correction at the kth discrete point, θ(k) is the phase at the kth discrete point, θ(k−1) is the phase at the (k-1)th discrete point, Tsc is the discretization period, and Δθ is the limiting phase change amount;

[0142] The mean value of the corrected phase limit at each discrete point is calculated as the corrected phase limit θcomp. An inverse coordinate transformation is then performed to convert the compensated polar flux linkage back to the α-β coordinate system, yielding the corrected stator flux linkage vector. The specific formula is as follows:

[0143] ,

[0144] Among them, e jθcomp The phase representation of the flux linkage vector for the corrected phase-limited θcomp;

[0145] Compensation for flux linkage observation errors is achieved by using the phase representation of the flux linkage vector of the corrected phase-limited θcomp.

[0146] In one embodiment of the present invention, sending a control compensation signal to the edge node based on the calculated correction parameters and error compensation parameters includes the following steps:

[0147] Obtain the dq-axis compensation voltage uploaded by the edge nodes, and update the identified d-axis inductance. and q-axis inductance The corrected reluctance torque;

[0148] Based on the updated d-axis and q-axis inductances, the corrected reluctance torque, and the compensation voltages for the d-axis and q-axis, the cloud compares the corresponding dq-axis inductances, the corrected reluctance torque, and the compensation voltages for the d-axis and q-axis. If the difference between the calculation results of the edge node and the calculation results of the cloud exceeds a preset threshold, the average value of the corresponding values ​​is used as the new correction parameter, and a correction parameter modification command is sent to the corresponding edge node.

[0149] The phase parameters of the flux linkage vector of the corrected phase-limited θcomp are sent to the edge nodes, and the edge nodes perform flux linkage observation error compensation based on the specific parameters.

[0150] In another embodiment of the present invention, a unit-cascaded motor speed control compensation system is used, which employs the unit-cascaded motor speed control compensation method described above to achieve unit-cascaded motor speed control compensation.

[0151] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A speed control compensation method for a unit-cascaded motor, characterized in that, Includes the following steps: An edge node is established at each motor in the cascaded motor management area, forming an edge node cluster. This edge node cluster model is used to collect parameters and perform real-time compensation for the cascaded motor units, including: An improved Prandtl-Ishlinskii model was adopted, and a temperature compensation term was introduced to correct the hysteresis operator parameters. A hysteresis model was constructed, and the model was identified by recursive least squares (RLS) to generate the dq-axis compensation voltage. Based on the mathematical model of the motor's dq axis, the total torque is decomposed into permanent magnet torque Tpm and reluctance torque Tr; In the current command generation stage, the recursive least squares method is used to identify the d-axis inductance and q-axis inductance in real time. Based on the difference between the d-axis inductance and q-axis inductance, the calculated value of reluctance torque is corrected and the reluctance torque compensation amount is calculated. Edge nodes transmit the collected and compensation parameters to the cloud in real time for analysis, including: Based on the thermal equivalent circuit model, the compensation values ​​of voltage and reluctance torque are corrected, and the flux linkage in the two-phase stationary coordinate system is compensated for flux linkage observation error through a limiting circuit. Control compensation signals are sent to the edge nodes according to the calculated correction parameters and error compensation parameters.

2. The speed control compensation method for a unit-cascaded motor according to claim 1, characterized in that, The improved Prandtl-Ishlinskii model is adopted, and a temperature compensation term is introduced to correct the hysteresis operator parameters. The hysteresis model is constructed by including the following steps: To address the effect of temperature on hysteresis characteristics, a temperature compensation term is introduced into the hysteresis operator: ， Where α is the temperature compensation coefficient, T(t) is the real-time temperature, and T0 is the reference temperature; An improved PI model is constructed by combining two non-centrosymmetric hysteresis operators, and the hysteresis model is built as follows: ， Where H(x) is the hysteresis operator, wi and vi are the different weight coefficients of the i-th rising branch, x(t) is the input signal, and ri is the threshold of the i-th rising branch. , where n is the total number of rising branch hysteresis operators, and ht is the temperature compensation term.

3. The speed control compensation method for a unit-cascaded motor according to claim 1, characterized in that, The model is identified using the recursive least squares (RLS) method, and the dq-axis compensation voltage is generated, including: The recursive least squares (RLS) identification model is used to calculate the compensation voltage Vcomp for the d-axis and q-axis respectively: ， Where Vcomp is the compensation voltage, and H(x) is the compensation voltage. -1 This is a hysteresis inverse model, where Vref is the reference voltage, ωref is the reference angular frequency, ωn is the natural angular frequency, and k1 and k2 are the corresponding weighting coefficients.

4. The speed control compensation method for a unit-cascaded motor according to claim 1, characterized in that, Based on the mathematical model of the motor's dq axis, the total torque is decomposed into permanent magnet torque Tpm and reluctance torque Tr, including the following steps: Based on the mathematical model of the motor's dq axis, the total torque Te is decomposed into permanent magnet torque Tpm and reluctance torque Tr: ； Where p is the number of pairs of rotor poles of the motor. Ld is the flux linkage of the permanent magnet, Lq is the d-axis inductance, Lq is the q-axis inductance, id is the d-axis current, and iq is the q-axis current.

5. The speed control compensation method for a unit-cascaded motor according to claim 1, characterized in that, In the current command generation stage, the recursive least squares method is used to identify the d-axis inductance and q-axis inductance in real time. Based on the difference between the d-axis inductance and q-axis inductance, the calculated reluctance torque value is corrected, and the reluctance torque compensation amount is calculated, including the following steps: In the current command generation stage, a discretized and identifiable model is constructed using the dq-axis voltage equation: ， Where ud(k) is the d-axis voltage at discrete point k, uq(k) is the q-axis voltage at discrete point k, Ld is the d-axis inductance, Lq is the q-axis inductance, id(k) is the d-axis current at discrete point k, iq(k) is the q-axis current at discrete point k, Ts is the discretization time step, and ωe(k) is the angular frequency at discrete point k. Organize into matrix form: ， in: , , ε(k) represents noise; Initialize the identification parameter matrix θ(k) and covariance matrix P(k), and update the matrix θ(k) recursively using the recursive least squares method: ， ， ， Where K(k) is the gain matrix and λ is the forgetting factor, which is used to balance the influence of historical data and new data; Based on the updated identification parameter matrix θ(k), the identified d-axis inductance is updated in real time. and q-axis inductance And transmit the data to the cloud in real time; Based on the identified d-axis inductance and q-axis inductance Calculate the corrected reluctance torque: ， The corrected Trcomp is fed back to the current command generation stage to adjust the dq axis current command. and .

6. The speed control compensation method for a unit-cascaded motor according to claim 1, characterized in that, The cloud-based system corrects the compensation values ​​for voltage and reluctance torque based on a thermal equivalent circuit model, including the following steps: A global parameter library is generated in the cloud to store motor model parameters and a temperature-magnetic saturation mapping table. Based on the data uploaded by the edge nodes, the motor parameters are updated in batches for batch parameter identification. The electromagnetic field distribution of each cascaded motor at different temperatures is simulated through the finite element analysis model. The variation law of dq axis inductance Ld and Lq with temperature is extracted, and a temperature-inductance MAP is generated. The winding temperature Tw is estimated in real time using a machine learning model based on a temperature-magnetic saturation mapping table, and the updated d-axis inductance and q-axis inductance are obtained by interpolation from the temperature-inductance MAP. Calculate the reluctance torque for correction based on the updated d-axis and q-axis inductance; Obtain the discretized identifiable model constructed from the edge nodes, and substitute the updated d-axis inductance and q-axis inductance into the corresponding model to obtain the compensation voltages for the d-axis and q-axis.

7. The speed control compensation method for a unit-cascaded motor according to claim 5, characterized in that, Compensation for flux linkage observation errors in a two-phase stationary coordinate system is achieved through a limiting process, including the following steps: Transform the magnetic flux linkage in the α-β coordinate system to the polar coordinate system: ， in, The stator flux linkage vector is given in the two-phase stationary coordinate system (α-β coordinate system). for The magnitude of the flux linkage vector, e, represents the strength of the flux linkage vector. jθ This represents the phase of the flux linkage vector. Let be the projection component of the flux linkage vector onto the α-axis. This is the projection component of the flux linkage vector onto the β-axis; Independent amplitude and phase limiting are performed. The amplitude limiting is as follows: sat for flux linkage amplitude Perform a bandwidth limiting operation. ; Discretize the phase and perform phase limiting correction: ， Where θcomp(k) is the phase limiting after correction at the kth discrete point, θ(k) is the phase at the kth discrete point, θ(k−1) is the phase at the (k-1)th discrete point, Tsc is the discretization period, Δθ is the limiting phase change amount, and Ts is the discretization time step. The mean value of the corrected phase limit at each discrete point is calculated as the corrected phase limit θcomp. An inverse coordinate transformation is then performed to convert the compensated polar flux linkage back to the α-β coordinate system, yielding the corrected stator flux linkage vector. The specific formula is as follows: ， Among them, e jθcomp The phase representation of the flux linkage vector for the corrected phase-limited θcomp; Compensation for flux linkage observation errors is achieved by using the phase representation of the flux linkage vector of the corrected phase-limited θcomp.

8. The speed control compensation method for a unit-cascaded motor according to claim 7, characterized in that, Based on the calculated correction parameters and error compensation parameters, control compensation signals are sent to the edge nodes, including the following steps: Obtain the dq-axis compensation voltage uploaded by the edge nodes, and update the identified d-axis inductance. and q-axis inductance The corrected reluctance torque; Based on the updated d-axis and q-axis inductances, the corrected reluctance torque, and the compensation voltages for the d-axis and q-axis, the cloud compares the corresponding dq-axis inductances, the corrected reluctance torque, and the compensation voltages for the d-axis and q-axis. If the difference between the calculation results of the edge node and the calculation results of the cloud exceeds a preset threshold, the average value of the corresponding values ​​is used as the new correction parameter, and a correction parameter modification command is sent to the corresponding edge node. The phase parameters of the flux linkage vector of the corrected phase-limited θcomp are sent to the edge nodes, and the edge nodes perform flux linkage observation error compensation based on the specific parameters.

9. A unit-cascaded motor speed control and compensation system, characterized in that, The system employs a speed control compensation method for unit-cascaded motors as described in any one of claims 1-8 to achieve speed control compensation for unit-cascaded motors.

Images