Permanent magnet synchronous motor fault-tolerant control method considering parameter variation

Abstract

This invention relates to the field of motor control technology, specifically to a phase-loss fault-tolerant control method for permanent magnet synchronous motors that considers parameter variations. The problem this invention addresses is how to promptly correct parameters in the prediction model when a phase loss causes time-varying drift in motor parameters, thereby improving the stability of fault-tolerant control. To solve this problem, this invention provides a control method comprising: real-time monitoring of the motor's three-phase stator current; employing a multi-criteria fusion method to diagnose whether a phase-loss fault has occurred in the motor; if so, switching the normal operation mode to a fault-tolerant control mode and simultaneously completing the inverter topology switch; in the phase-loss fault-tolerant operation mode, performing recursive estimation of the motor operating parameters for the current cycle using an extended Kalman filter algorithm to output real-time identification parameters; and substituting the real-time identification parameters into the phase-loss fault-tolerant operation model to obtain the optimal voltage vector for each operating cycle.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, and more specifically, to a phase-loss-tolerant fault control method for permanent magnet synchronous motors that takes into account parameter variations. Background Technology

[0002] Permanent magnet synchronous motors are widely used in fields with high reliability requirements, such as electric vehicles, aerospace, and industrial servo systems, due to their high power density, high efficiency, and excellent control performance. In these applications, the continuous and reliable operation of the motor is crucial. Once a fault occurs (especially a phase loss fault), if fault-tolerant control cannot be implemented in time, it may lead to motor shutdown, excessive torque pulsation, or even failure of the entire drive system, resulting in serious economic losses or safety accidents.

[0003] Traditional fault-tolerant control strategies, such as fault-tolerant schemes under direct torque control (DTC) or field-oriented control (FOC), typically maintain motor operation after a phase loss is detected by reconstructing the current reference value or modifying the switching table. However, these methods are mostly based on a key assumption: that the mathematical model parameters of the motor (such as stator inductance and permanent magnet flux linkage) remain unchanged before and after the fault. In reality, a phase loss fault leads to distortion of the motor winding current distribution, a sharp increase in copper losses, and consequently, a significant increase in winding temperature. Simultaneously, the symmetry of the motor's magnetic circuit is disrupted after a phase loss, and the operating point of the magnetic circuit shifts significantly, causing the direct-axis inductance L... d quadrature axis inductance L q Parameters exhibit nonlinear drift, permanent magnet flux linkage ψ f Irreversible demagnetization can also occur due to increased temperature, resulting in time-varying parameters.

[0004] Predictive torque control (PTC) iterates through all candidate voltage vectors in each duty cycle, calculates predicted torque and flux linkage values ​​based on an internal predictive model, and selects the optimal voltage vector through cost function optimization. It boasts advantages such as fast dynamic response and ease of handling multiple constraints (e.g., current limiting and voltage constraints), and has been increasingly applied in the field of fault-tolerant control of permanent magnet synchronous motors. However, the control performance of PTC is highly dependent on the accuracy of the internal predictive model. When a phase loss causes time-varying drift in motor parameters, if the parameters in the predictive model are not corrected in time, severe model mismatch will occur, leading to increased torque prediction errors and flux linkage prediction errors, failure of cost function optimization, and ultimately problems such as increased torque ripple, increased speed fluctuations, and decreased system efficiency. It may even cause fault-tolerant control failure, preventing the motor from operating stably and continuously.

[0005] Therefore, developing a system capable of real-time tracking of time-varying motor parameters after phase loss, high-precision online parameter identification, dynamic correction and prediction model, and forming a closed-loop self-correction system has significant engineering application value and theoretical significance. Summary of the Invention

[0006] The problem solved by this invention is: how to promptly correct the parameters in the prediction model when a phase loss causes time-varying drift in motor parameters, thereby improving the stability of fault-tolerant control.

[0007] To address the aforementioned issues, this invention provides a phase-loss fault-tolerant control method for permanent magnet synchronous motors that considers parameter variations. The control method includes: real-time monitoring of the motor's three-phase stator current; employing a multi-criteria fusion method to diagnose whether a phase-loss fault has occurred in the motor; if so, switching from normal operation mode to fault-tolerant control mode and simultaneously completing inverter topology switching; in phase-loss fault-tolerant operation mode, performing recursive estimation of the motor's operating parameters for the current cycle using an extended Kalman filter algorithm to output real-time identification parameters; and substituting the real-time identification parameters into the phase-loss fault-tolerant operation model to obtain the optimal voltage vector for each operating cycle.

[0008] Compared with existing technologies, the technical effects achieved by this solution are as follows: the multi-criteria fusion method for diagnosing motor faults makes the determination of motor phase loss more accurate, avoids the false activation of fault-tolerant control mode, and the setting of fault-tolerant control mode activation and inverter topology switching makes the current operating mode of the motor more in line with the current working environment. The extended Kalman filter algorithm can effectively handle the state estimation and parameter identification problems of nonlinear systems and is suitable for the nonlinear model of permanent magnet synchronous motor after phase loss. Compared with traditional Kalman filtering and least squares method, it has the advantages of high identification, strong anti-interference ability, and the ability to achieve joint estimation of state and parameters. The acquisition of the optimal voltage vector allows the motor to select the optimal data for verification during the correction process, reducing the time to restore to the normal operating mode.

[0009] In one embodiment of the present invention, the three-phase stator current of the motor is monitored in real time, and a multi-criteria fusion method is used to diagnose whether the motor has a phase loss fault. Specifically, this includes: continuously sampling the two-phase stator current and calculating the third-phase current; performing a mean judgment of the motor based on the sliding average of the absolute values ​​of each phase current; obtaining the fluctuation amplitude of the instantaneous value of the third-phase current within a target period, and performing a waveform continuous judgment of the motor in combination with the waveforms of the two-phase stator currents; using a motor prediction model under normal operating conditions, calculating the residual between the predicted current value and the measured values ​​of each current, and performing a residual detection judgment of the motor based on the mean and variance of the residual; and determining that the motor has a phase loss fault when the mean judgment, waveform continuous judgment, and residual detection judgment all meet the fault conditions.

[0010] Compared with existing technologies, the technical effects achieved by adopting this technical solution are as follows: by combining three judgment methods—mean value judgment, waveform continuity judgment, and residual detection judgment—to determine phase loss faults, it avoids misjudgments caused by judging phase loss faults by a single indicator, ensures the accurate activation of the phase loss fault-tolerant operation mode, guarantees that the motor can operate under reasonable working conditions, and extends the service life of the motor.

[0011] In one embodiment of the present invention, under the phase-loss fault-tolerant operation mode, the motor operating parameters of the current cycle are combined with the extended Kalman filter algorithm to perform recursive estimation and output real-time identification parameters. Specifically, this includes: establishing a nonlinear state-space model for EKF filtering and defining an augmented state vector. The parameters to be identified are used as virtual state variables; based on the motor equations modified under the phase-loss fault-tolerant operation mode, discrete nonlinear state equations are constructed to obtain the noise covariance matrix; the augmented state vector and the noise covariance matrix are recursively estimated using the Kalman filter algorithm to output the real-time identification parameters.

[0012] Compared with existing technologies, the technical effects achieved by this solution are as follows: by using the parameters to be identified as virtual state variables, the joint estimation of motor state and parameters is realized, avoiding the accumulation of errors caused by identifying parameters alone. Furthermore, by using a recursive estimation method, real-time identification parameters are output in stages, providing strong data support for the correction of the phase-loss fault-tolerant operation mode and improving the correction efficiency of the motor.

[0013] In one embodiment of the present invention, a nonlinear state-space model for EKF filtering is established, and an augmented state vector is defined. The parameters to be identified are used as virtual state variables, specifically including the augmented state vector used in the extended Kalman filter algorithm. The system state equations are as follows: ;in, i d , i q These are the direct-axis and quadrature-axis components of the stator current. , For the rotor's electric angular velocity and position, L d , L q , ψ f The state of the parameter to be identified.

[0014] Compared with existing technologies, the technical effects achieved by adopting this technical solution are as follows: EKF utilizes the measurable voltage and current signals of the reconstructed system, combined with the motor state equation and observation equation corrected after phase loss, and accurately estimates all elements in the augmented state vector through recursive calculations of time-updated measurement updates, thereby improving the accuracy of real-time output identification results.

[0015] In one embodiment of the present invention, the augmented state vector and noise covariance matrix are recursively estimated using a Kalman filter algorithm to output real-time identification parameters. Specifically, this includes: setting a state estimate value based on the first working cycle of entering the phase-loss fault-tolerant operation mode; predicting the state estimate value and error covariance matrix of the current cycle based on the state estimate value of the previous working cycle; correcting the state estimate value and error covariance matrix of the current cycle based on the real-time observation value of the current cycle to obtain a corrected estimate value; outputting the updated corrected estimate value after each working cycle, and using a portion of the corrected estimate value as a real-time identification parameter, which is then fed back to the predictive controller in real time.

[0016] Compared with existing technologies, the technical effects achieved by adopting this technical solution are as follows: by setting the state estimate value in the first working cycle, the subsequent parameter adjustments are more consistent with the working state of the motor after phase loss. The adjustment range is determined by the change of the state estimate value in subsequent working cycles, and quantified by the error covariance matrix. The real-time observation value is used for comparison and correction, which further improves the rationality of the real-time identification parameters and facilitates the predictive controller to execute subsequent predictions.

[0017] In one embodiment of the present invention, the real-time identification parameters are substituted into the phase-loss fault-tolerant operation model to obtain the optimal voltage vector in each working cycle. Specifically, this includes: determining the candidate voltage set for the current cycle based on the real-time identification parameters of the current cycle; calculating each candidate voltage vector in the candidate voltage set individually to obtain the predicted values ​​of stator flux linkage, electromagnetic torque, and phase current peak value under the action of the candidate voltage vector in the next working cycle; calculating the cost function value corresponding to each candidate voltage vector; and selecting the candidate voltage vector that minimizes the cost function as the optimal voltage vector.

[0018] Compared with existing technologies, the technical effects achieved by this solution are as follows: the prediction model performs calculations based on the most reliable motor parameters in each cycle, thereby theoretically eliminating model mismatch errors caused by time-varying parameters. In each working cycle, the optimal vector that makes the future state of the system closest to the desired target is selected from a limited number of candidate voltage vectors, which greatly improves the system's correction efficiency.

[0019] In one embodiment of the present invention, the control method further includes: in the phase-loss fault-tolerant operation mode, during the initial transition period of the first working cycle, introducing a variable forgetting factor for the extended Kalman filter algorithm, and adjusting the variable forgetting factor according to the changes and fluctuations of the state estimate and the correction estimate.

[0020] Compared with existing technologies, the technical effects achieved by adopting this technical solution are as follows: the setting of the variable forgetting factor improves the parameter identification speed during the initial transition period after a phase loss fault occurs in the motor, and further improves the identification accuracy after stable operation.

[0021] In one embodiment of the present invention, the real-time identification parameters are substituted into the phase-loss fault-tolerant operation model to obtain the optimal voltage vector in each working cycle. The method further includes: constructing a cost function J, which comprehensively balances torque tracking accuracy, flux tracking accuracy, and system safety. The cost function J is constructed as follows: ;in, For the torque setpoint, Given a value for the magnetic flux, This is the predicted value of electromagnetic torque. This is the predicted value of the stator flux linkage. and These are the weighting coefficients. This is the predicted peak value of the phase current.

[0022] Compared with existing technologies, the technical effects achieved by adopting this technical solution are as follows: by constructing a cost function and combining it with the priority theory derivation of the weight coefficient to control the objective, the rationality of multi-objective optimization is ensured.

[0023] In one embodiment of the present invention, a phase loss fault-tolerant control method for a permanent magnet synchronous motor further includes: applying an optimal voltage vector to the motor and continuously monitoring the three-phase stator current of the motor; when the motor still has a phase loss fault, the fault-tolerant control mode performs closed-loop self-correction operation; when the motor does not have a phase loss fault, the fault-tolerant control mode is switched to normal operation mode.

[0024] Compared with existing technologies, the technical effects achieved by adopting this technical solution are as follows: a closed-loop self-correction loop of "online parameter identification - prediction model update - control optimization execution" is constructed from a theoretical level to ensure that the system can continuously adapt to the time-varying characteristics of parameters under phase loss faults and maintain high-performance fault-tolerant control effects in the long term. Attached Figure Description

[0025] Figure 1 A flowchart of the phase loss-tolerant fault control method for permanent magnet synchronous motors considering parameter variations, as described in this invention; Figure 2 This is an overall block diagram of the phase loss fault-tolerant operation mode of the present invention; Figure 3This is a schematic diagram of the online parameter identification module based on extended Kalman filtering of the present invention. Figure 4 This is a flowchart illustrating the adaptive update of the prediction model and the optimization of the cost function in this invention. Detailed Implementation To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0026] [First Embodiment] See Figure 1 This application provides a phase loss-tolerant control method for a permanent magnet synchronous motor that considers parameter variations. The control method includes: S100: Real-time monitoring of the three-phase stator current of the motor, and use of multi-criteria fusion method to diagnose whether the motor has a phase loss fault; S200, if so, switch the normal operation mode to the fault-tolerant control mode and simultaneously complete the inverter topology switch; S300: In the phase loss fault-tolerant operation mode, the motor operating parameters of the current cycle are combined with the extended Kalman filter algorithm to perform recursive estimation and output real-time identification parameters. S400: Substitute the real-time identification parameters into the phase loss fault-tolerant operation model to obtain the optimal voltage vector in each working cycle.

[0027] See Figure 2 In steps S100 and S200, the sampling frequency of the three-phase stator current of the motor is not less than 20 times the fundamental frequency of the motor to ensure the integrity of the current signal. By calculating the average value of each phase current within one fundamental cycle and combining it with instantaneous waveform and residual analysis, the occurrence of a phase loss fault is diagnosed. Once a fault is confirmed, a fault flag signal Fault_Flag=1 is immediately generated, triggering the system to enter fault-tolerant operation mode. If no fault is detected, the normal operation mode is maintained, and the phase current is continuously monitored.

[0028] System reconstruction includes hardware topology reconstruction and software model reconstruction. The hardware topology is switched to a four-switch fault-tolerant topology, and the software model is switched to the mathematical model M of the fault-tolerant control mode. fault .

[0029] Hardware topology reconfiguration method: Block the drive signal of the original faulty phase (phase A) bridge arm and trigger the redundant switch to switch the inverter to the preset four-switch fault-tolerant topology (composed of four switches of phase B and phase C bridge arms).

[0030] Software Model and Coordinate Reconstruction Method: Coordinate Transformation Overload: The matrix used for Clark transformation in the control algorithm is in standard three-phase form. Switching to a specific form based on the remaining B and C phases : Control model switching: The motor prediction model inside the predictive torque controller is switched from the normal three-phase full-sequence type normal operation mode. M normal Switch to fault-tolerant control mode for asymmetrical two-phase systems after phase A loss. M fault ...

[0031] After entering the fault-tolerant control mode, the inverter is switched to the preset fault-tolerant topology (a four-switch topology using the remaining two phase windings), and the mathematical model inside the controller is switched from the normal three-phase full-bridge model to the mathematical model corresponding to the fault-tolerant topology and the phase-loss operation of the remaining windings.

[0032] In step S300, after the fault-tolerant control mode is activated, the online parameter identification module based on Extended Kalman Filter (EKF) is started. This module runs in parallel with the fault-tolerant predictive controller to ensure the real-time performance of parameter identification.

[0033] In step S400, the obtained real-time identification parameters are substituted into the voltage equation and electromagnetic torque equation of the phase loss fault-tolerant operation model to dynamically update the internal prediction model of the model predictive controller in order to correct the model parameter mismatch caused by phase loss and thermal effects. The updated prediction model is used to calculate the torque and flux prediction values ​​of each candidate voltage vector. The optimal voltage vector is selected through a multi-objective cost function to drive the fault-tolerant inverter to execute.

[0034] The multi-criteria fusion method for diagnosing motor faults makes the determination of motor phase loss faults more accurate and avoids the false activation of fault-tolerant control mode. The activation of fault-tolerant control mode and the setting of inverter topology switching make the current operating mode of the motor more consistent with the current working environment. The extended Kalman filter algorithm can effectively handle the state estimation and parameter identification problems of nonlinear systems. It is suitable for the nonlinear model of permanent magnet synchronous motor after phase loss. Compared with traditional Kalman filtering and least squares method, it has the advantages of high identification, strong anti-interference ability, and the ability to achieve joint estimation of state and parameters. The acquisition of the optimal voltage vector allows the motor to select the optimal data for verification during the correction process, reducing the time to restore to the normal operating mode.

[0035] Furthermore, the three-phase stator current of the motor is monitored in real time, and a multi-criteria fusion method is used to diagnose whether the motor has experienced a phase loss fault, specifically including: S110. Continuously sample the two-phase stator current and calculate the third-phase current; S120. Perform motor mean value judgment based on the sliding average of the absolute values ​​of each phase current; S130. Obtain the fluctuation range of the instantaneous value of the third phase current within the target period, and perform continuous waveform judgment of the motor in combination with the waveform of the two-phase stator current. S140. Using the motor prediction model under normal operating conditions, calculate the residual between the predicted current value and the measured current values, and perform residual detection and judgment of the motor based on the mean and variance of the residual. S150. When the mean value judgment, waveform continuity judgment and residual detection judgment all meet the fault conditions, it is determined that the motor has a phase loss fault.

[0036] In step S110, the system continuously samples the two-phase stator current. i b and i c And calculate the third phase current. .

[0037] In step S120, the mean value is determined as follows: the moving average of the absolute values ​​of the phase currents within one fundamental cycle is calculated. If satisfied (Right now <0.5A) and ,in The threshold is close to zero (0.5A). If it is a proportional coefficient, then the trigger flag is activated. .

[0038] In step S130, the waveform duration is determined as follows: when the working cycle length is 100 μs, the target cycle length is typically 800 μs, i.e., 8 working cycles, for detection. i a Does the instantaneous value remain continuously within the interval [-a, +a] (where a is the measurement noise band, i.e., [-0.3A, +0.3A])? i b and i c The waveforms are opposite and their amplitudes are approximately equal. If this condition is met, the flag is triggered. Flag_Wave =1, this judgment is used to distinguish between phase loss faults and instantaneous current fluctuations and measurement noise.

[0039] In step S140, the residual detection and judgment are performed as follows: using the motor prediction model under normal operating conditions, the residual between the predicted current value and the measured current value is calculated. , , Phase loss faults will cause severe model mismatch, resulting in a significant increase in residuals and a periodic pattern. The mean of the residuals is calculated by monitoring their statistical characteristics. With variance ,when and When this happens, the flag Flag_Residual is set to 1.

[0040] In step S150, when all three criteria above are met, i.e. Flag_Energy , Flag_Wave When both Flag_Residual and Fault_Flag are equal to 1, set Fault_Flag=1 and the system enters the fault-tolerant control system.

[0041] By combining three judgment methods—mean value judgment, waveform continuity judgment, and residual detection judgment—to determine phase loss faults, misjudgments caused by judging phase loss faults based on a single indicator are avoided. This ensures the accurate activation of the phase loss fault-tolerant operation mode, guarantees that the motor can operate under reasonable working conditions, and extends the motor's service life.

[0042] See Figure 3 Furthermore, in the phase-loss fault-tolerant operation mode, the motor operating parameters of the current cycle are combined with the extended Kalman filter algorithm to perform recursive estimation, and output real-time identification parameters, specifically including: S310. Establish a nonlinear state-space model for EKF filtering and define the augmented state vector. The parameters to be identified are treated as virtual state variables; S320. Based on the motor equations modified under the phase-loss fault-tolerant operation mode, a discrete nonlinear state equation is constructed to obtain the noise covariance matrix. S330. The augmented state vector and noise covariance matrix are recursively estimated using the Kalman filter algorithm to output real-time identification parameters.

[0043] In step S310, an augmented state vector containing the motor state vector and the parameters to be identified is constructed. The parameters to be identified are used as virtual state variables to achieve joint estimation of the motor state and parameters, avoiding error accumulation caused by identifying parameters alone. The motor state vector includes the direct-axis current. i d quadrature axis current i q Rotor electric angular velocity oh r Rotor position i r The parameters to be identified include direct-axis inductance. L d quadrature axis inductance L q Permanent magnet flux ψ fThe above three parameters are the key parameters affecting the accuracy of the prediction model, and they are also the parameters most affected by temperature and magnetic circuit distortion after phase loss.

[0044] In step S320, based on the motor equations corrected after phase loss, discrete nonlinear state equations are constructed: in, f (·) represents the nonlinear state transition function, the specific form of which is described by the core part of the following discretized equation (process noise). w ( k +1) is a 7-dimensional vector, with each component following a Gaussian distribution N(0, Q), where Q is the process, and the diagonal elements are preset as follows: , , , , , (Off-diagonal elements are 0): To control the voltage, Ts Where TL is the sampling period and TL is the load torque. w ( k )∼N(0,Q) is the process noise, which characterizes the model uncertainty.

[0045] The specific formula for the observation equation is as follows: Observation is dq shaft current, i.e. .

[0046] Observation matrix , To measure noise, R To measure the covariance matrix of the noise, .

[0047] In step S330, since the state equation is a nonlinear equation, the extended Kalman filter algorithm is used to obtain the Jacobian matrix F(k-1) by differentiating the state transition function f(·), thereby linearizing the nonlinear system and realizing recursive estimation.

[0048] By treating the parameters to be identified as virtual state variables, the joint estimation of motor state and parameters is achieved, avoiding the accumulation of errors caused by identifying parameters alone. Furthermore, through recursive estimation, real-time identified parameters are output in stages, providing strong data support for the correction of the phase-loss fault-tolerant operation mode and improving the correction efficiency of the motor.

[0049] Furthermore, a nonlinear state-space model for EKF filtering is established, and the augmented state vector is defined. The parameters to be identified are used as virtual state variables, specifically including: S311, the augmented state vector used in the extended Kalman filter algorithm The system state equations are as follows: ; in, i d , i q These are the direct-axis and quadrature-axis components of the stator current. , For the rotor's electric angular velocity and position, L d , L q , ψ f The state of the parameter to be identified.

[0050] In step S311, based on the assumption of slow time-varying parameters (i.e., the rate of parameter change is much lower than the working cycle, and the parameter is approximately constant within a single working cycle), the dynamic characteristics of the parameter to be identified satisfy: .

[0051] EKF utilizes the measurable voltage of the reconfigured system (inverter output) dq shaft voltage u d , u q ) and current signal (actual measurement) dq shaft current i d_meas , i q_meas By combining the motor state equations corrected after phase loss with the observation equations, and through recursive calculations updated by time-updated measurements, all elements in the augmented state vector are accurately estimated, and the identification results are output in real time. , , .

[0052] Furthermore, the augmented state vector and noise covariance matrix are recursively estimated using a Kalman filter algorithm to output real-time identification parameters, specifically including: S331. The first working cycle of entering the phase loss fault-tolerant operation mode sets the state estimate value, and the predictive controller predicts the state estimate value and error covariance matrix of the current cycle based on the state estimate value of the previous working cycle. S332. Based on the real-time observations of the current period, the state estimate and error covariance matrix of the current period are corrected to obtain the corrected estimate. S333. After each working cycle, output the updated corrected estimate and use part of the corrected estimate as a real-time identification parameter, and feed it back to the predictive controller in real time.

[0053] In step S331, during the first working cycle triggered by the fault ( k =0), set the initial state estimate. That is, the initial current, initial speed and position are 0, and the initial parameter values ​​are taken from the motor nameplate values; error covariance matrix. It is a 7x7 diagonal matrix, with the diagonal elements set by default: , , , , , , Off-diagonal elements are 0.

[0054] In step S332, based on the state estimate of the previous period, the state estimate and error covariance matrix of the current period are predicted: ; ; Where F(k-1) is the state transition function f(·) in The Jacobian matrix at point A, considering the slow-event characteristics of the parameter, has the following specific form: .

[0055] In step S333, based on the measured observation value z(k) of the current period, the state estimate and error covariance matrix are corrected, and the Kalman gain K(k) is calculated. The Kalman gain determines how to use the current measurement residual to correct all state estimates, including parameters. ; ; ; Where I is a 7x7 identity matrix; The residual is used to measure the deviation between the observed and predicted values, and is used to correct the state and parameter estimates.

[0056] Each work cycle outputs an updated state estimate. The following three elements are the parameter values ​​obtained from online identification: , , It is provided to the prediction controller in real time for updating the prediction model.

[0057] By setting the state estimate value in the first working cycle, subsequent parameter adjustments are made more consistent with the working state of the motor after a phase loss. The adjustment range is determined by the change in the state estimate value in subsequent working cycles, and quantified by the error covariance matrix. The real-time observation value is then compared and corrected, which further improves the rationality of the real-time identification parameters and facilitates the predictive controller to execute subsequent predictions.

[0058] See Figure 4 Furthermore, the real-time identified parameters are substituted into the phase loss fault-tolerant operation model to obtain the optimal voltage vector within each operating cycle, specifically including: S410. Determine the set of candidate voltages for the current cycle based on the real-time identification parameters of the current cycle. S420. Calculate each candidate voltage vector in the candidate voltage set separately to obtain the predicted values ​​of stator flux linkage, electromagnetic torque and phase current peak value under the action of the candidate voltage vector in the next working cycle. S430. Calculate the cost function value corresponding to each candidate voltage vector, and select the candidate voltage vector that minimizes the cost function as the optimal voltage vector.

[0059] In steps S410 to S430, in each work cycle k Before the prediction calculation begins, the controller reads the latest parameter estimates calculated by the EKF module from the shared memory area: ; For the first k The estimated vector (column vector) of motor parameters for each working cycle.

[0060] The above parameter estimates are injected into the phase loss fault-tolerant prediction model maintained within the predictive controller. M faul t Specifically, all discrete equations used for forward prediction in the controller (such as current prediction, flux prediction, and torque prediction equations) involve... , , The parameters were all replaced with , , .

[0061] For example, used to calculate candidate voltage vectors The next moment of action d The prediction equation for shaft current has been updated to: ; in, It is a vector corresponding dAxis voltage. Similarly, q The prediction equations for shaft current, flux linkage, and electromagnetic torque were all replaced with parameters simultaneously.

[0062] Based on the current four-switch fault-tolerant topology, the set of candidate voltage vectors for the finite control set is theoretically determined. The four alternative voltage vectors correspond to four switching combinations in the four-switch topology, which theoretically avoids invalid voltage vectors, reduces the amount of calculation, and ensures the real-time performance of the control.

[0063] Within each working cycle, based on the prediction model updated in step S333, the four candidate voltage vectors are substituted one by one into the current prediction equation, flux prediction equation, and torque prediction equation to theoretically calculate the next working cycle under the action of each voltage vector. k Stator flux prediction at time +1 Electromagnetic torque prediction value and the predicted peak value of phase current .

[0064] Iterate through all voltage vectors, calculate the cost function value for each voltage vector, and select the voltage vector that minimizes the cost function as the optimal voltage vector. V opt Once the optimal voltage vector is determined, a corresponding switching signal is generated. This switching signal corresponds one-to-one with the switching logic of the four-switch fault-tolerant topology, driving the switches of the fault-tolerant inverter to turn on and off according to the preset logic, outputting the optimal voltage vector, which acts on the B-phase and C-phase windings of the motor to achieve precise control of torque and flux linkage. The prediction model performs calculations based on the most reliable motor parameters at the time of each cycle, thereby theoretically eliminating model mismatch errors caused by time-varying parameters. Within each working cycle, the optimal vector that makes the future state of the system closest to the desired target is selected from a limited number of candidate voltage vectors, greatly improving the system's correction efficiency.

[0065] Furthermore, the control method also includes: in the phase-loss fault-tolerant operation mode, during the initial transition period of the first working cycle, introducing a variable forgetting factor for the extended Kalman filter algorithm, and adjusting the variable forgetting factor according to the changes and fluctuations of the state estimate and the correction estimate.

[0066] During the initial transition period after a fault is triggered, a variable forgetting factor is introduced into the extended Kalman filter algorithm. m The forgetting factor m Adaptive adjustment is performed based on the covariance trace of the estimated parameters: From the first working cycle to the Nth working cycle after the fault occurs (N ranges from 5 to 10, adaptively adjusted according to the working cycle length), a smaller value is set... m value( m=0.1~0.3), increasing the weight of measurement information, accelerating parameter identification, and quickly tracking parameter mutations; the parameter to be estimated enters the stable range (i.e., the fluctuation of the parameter estimate is less than the preset threshold). , After setting the value to 1%~3% of the parameter's rated value, it will automatically increase. m value( m =0.8~0.95), reducing the impact of measurement noise, smoothing the identification results, and improving identification accuracy.

[0067] The method for calculating the covariance trace is as follows: ,in , , The parameters to be identified are respectively L d , L q , The corresponding error covariance matrix P The diagonal element, when Adjustment m Decrease; at that time, adjust m Increase This is the preset covariance trace threshold.

[0068] The setting of the variable forgetting factor improves the parameter identification speed during the initial transition period after a phase loss fault in the motor, and further improves the identification accuracy after stable operation.

[0069] Furthermore, by substituting the real-time identified parameters into the phase-loss fault-tolerant operation model to obtain the optimal voltage vector within each operating cycle, the process also includes: constructing a cost function J, which comprehensively balances torque tracking accuracy, flux tracking accuracy, and system safety. The construction form of the cost function J is as follows: ; in, For the torque setpoint, Given a value for the magnetic flux, This is the predicted value of electromagnetic torque. This is the predicted value of the stator flux linkage. and These are the weighting coefficients. This is the predicted peak value of the phase current.

[0070] During execution, an adaptive weight adjustment strategy is adopted, dynamically adjusting the weights based on the magnitude of torque error and flux linkage error: when the torque error is large, the weight is increased. The weighting is increased to improve torque tracking speed; when the flux error is large, the weighting is increased. The weight is adjusted to ensure flux linkage stability, and when the peak current approaches the rated value, the weight is increased. The weighting limits the current value; The predicted peak phase current is used to constrain the current amplitude to protect the inverter switching devices and motor windings, preventing device damage or winding overheating due to excessive current.

[0071] The adaptive adjustment logic of the weighting coefficients is designed based on the theoretical changing trends of torque error and flux linkage error: when the torque error ( When the value is large, increase it appropriately. The weighting of the torque tracking speed is improved, theoretically suppressing torque ripple; when the flux linkage error ( When the value is large, increase The weighting is adjusted to ensure flux linkage stability; when the predicted peak current value is close to... When, increase The weighting is adjusted to strengthen current constraints and ensure safe system operation.

[0072] By constructing a cost function and combining it with the theoretical derivation of the priority of the objective by controlling the weight coefficients, the rationality of multi-objective optimization is ensured.

[0073] Furthermore, the fault-tolerant control method for phase loss of permanent magnet synchronous motor includes: applying the optimal voltage vector to the motor and continuously monitoring the three-phase stator current of the motor; when the motor still has a phase loss fault, the fault-tolerant control mode performs closed-loop self-correction operation; when the motor does not have a phase loss fault, the fault-tolerant control mode is switched to normal operation mode.

[0074] The optimal voltage vector obtained in step S430 is repeatedly input, and a new set of candidate voltages is obtained using the method in step S410. Steps S420 and S430 are executed again to obtain the updated optimal voltage vector. The update frequency is one working cycle. The update is repeated until the phase loss fault is cleared. When the phase loss fault is cleared, the fault-tolerant control mode is switched to the normal operation mode.

[0075] Theoretically, a closed-loop self-correction loop of "online parameter identification - prediction model update - control optimization execution" is constructed to ensure that the system can continuously adapt to the time-varying characteristics of parameters under phase loss faults and maintain high-performance fault-tolerant control effects in the long term.

[0076] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A phase-loss-tolerant fault control method for a permanent magnet synchronous motor considering parameter variations, characterized in that, The control method includes: The three-phase stator current of the motor is monitored in real time, and a multi-criteria fusion method is used to diagnose whether the motor has a phase loss fault. If so, switch the normal operation mode to the fault-tolerant control mode and simultaneously complete the inverter topology switch. In the phase loss fault-tolerant operation mode, the motor operating parameters of the current cycle are combined with the extended Kalman filter algorithm to perform recursive estimation and output real-time identification parameters; Substitute the real-time identification parameters into the phase loss fault-tolerant operation model to obtain the optimal voltage vector in each working cycle.

2. The method for phase loss-tolerant fault control of a permanent magnet synchronous motor according to claim 1, characterized in that, The real-time monitoring of the three-phase stator current of the motor employs a multi-criteria fusion method to diagnose whether the motor has experienced a phase loss fault, specifically including: Continuously sample the stator current of two phases and calculate the current of the third phase; The mean value of the motor is determined based on the sliding average of the absolute values ​​of the currents in each phase; Within the target period, the fluctuation range of the instantaneous value of the third phase current is obtained, and the waveform of the motor is continuously judged in combination with the waveform of the two phase stator currents. Using a motor prediction model under normal operating conditions, the residuals between the predicted current value and the measured current values ​​are calculated, and the residual detection judgment of the motor is performed based on the mean and variance of the residuals. When the mean judgment, waveform continuity judgment and residual detection judgment all meet the fault conditions, it is determined that the motor has experienced the phase loss fault.

3. The method for phase loss-tolerant fault control of a permanent magnet synchronous motor according to claim 2, characterized in that, In the phase-loss fault-tolerant operation mode, the motor operating parameters of the current cycle are recursively estimated using an extended Kalman filter algorithm to output real-time identification parameters, specifically including: Establish a nonlinear state-space model for EKF filtering and define the augmented state vector. The parameters to be identified are treated as virtual state variables; Based on the motor equations corrected under the phase-loss fault-tolerant operation mode, a discrete nonlinear state equation is constructed to obtain the noise covariance matrix. The augmented state vector and the noise covariance matrix are subjected to the recursive estimation using the Kalman filter algorithm, and the real-time identification parameters are output.

4. The method for phase loss-tolerant fault control of a permanent magnet synchronous motor according to claim 3, characterized in that, The nonlinear state-space model for EKF filtering is established, and the augmented state vector is defined. The parameters to be identified are used as virtual state variables, specifically including: The extended Kalman filter algorithm uses an augmented state vector. The system state equations are as follows: ； in, i d , i q These are the direct-axis and quadrature-axis components of the stator current. , For the rotor's electric angular velocity and position, L d , L q , ψ f The state of the parameter to be identified.

5. The method for phase loss-tolerant fault control of a permanent magnet synchronous motor according to claim 3, characterized in that, The step of performing the recursive estimation of the augmented state vector and the noise covariance matrix using the Kalman filter algorithm to output the real-time identification parameters specifically includes: The state estimate is set according to the first working cycle of entering the phase-loss fault-tolerant operation mode, and the predictive controller predicts the state estimate and error covariance matrix of the current cycle according to the state estimate of the previous working cycle. Based on the real-time observations of the current period, the state estimate and the error covariance matrix of the current period are corrected to obtain the corrected estimate. After each working cycle, the updated corrected estimate is output, and a portion of the corrected estimate is used as the real-time identification parameter and fed back to the prediction controller in real time.

6. The method for phase loss-tolerant fault control of a permanent magnet synchronous motor according to claim 5, characterized in that, The step of substituting the real-time identification parameters into the phase loss fault-tolerant operation model to obtain the optimal voltage vector in each working cycle specifically includes: The candidate voltage set for the current period is determined based on the real-time identification parameters of the current period; Each candidate voltage vector in the candidate voltage set is calculated individually to obtain the predicted values ​​of stator flux linkage, electromagnetic torque, and phase current peak value under the action of the candidate voltage vector in the next working cycle. Calculate the cost function value corresponding to each candidate voltage vector, and select the candidate voltage vector that minimizes the cost function as the optimal voltage vector.

7. The method for phase loss-tolerant fault control of a permanent magnet synchronous motor according to claim 5, characterized in that, The control method further includes: In the phase-deficient fault-tolerant operation mode, during the initial transition period of the first working cycle, a variable forgetting factor is introduced for the extended Kalman filter algorithm, and the variable forgetting factor is adjusted according to the changes and fluctuations of the state estimate and the corrected estimate.

8. The method for phase loss-tolerant fault control of a permanent magnet synchronous motor according to claim 6, characterized in that, The step of substituting the real-time identification parameters into the phase-loss fault-tolerant operation model to obtain the optimal voltage vector in each working cycle also includes: A cost function J is constructed, which comprehensively balances torque tracking accuracy, flux tracking accuracy, and system safety. The construction form of the cost function J is as follows: ； in, For the torque setpoint, Given a value for the magnetic flux, The electromagnetic torque prediction value is... The stator flux linkage prediction value is... and These are the weighting coefficients. This is the predicted peak value of the phase current.

9. The method for phase-loss fault-tolerant control of a permanent magnet synchronous motor according to any one of claims 1 to 8, characterized in that, The control method further includes: The optimal voltage vector is applied to the motor, and the three-phase stator current of the motor is continuously monitored; When the motor still has the phase loss fault, the fault-tolerant control mode performs closed-loop self-correction operation; When the motor does not have the phase loss fault, the fault-tolerant control mode is switched to the normal operation mode.

Images