Motor driving direction fault detection and compensation method and device

Abstract

The invention relates to the technical field of motor driving, and discloses a motor driving direction fault detection and compensation method and device. The method comprises the following steps: collecting state data of a target vehicle and constructing a sensor fusion vector; calculating a compensation torque value based on the state data and the sensor fusion vector; based on the compensation torque value, a boost module and a super capacitor combination in a two-stage voltage stabilizing circuit are driven to carry out voltage pre-adjustment, and a stable bus voltage is obtained; and generating a motor control signal based on the compensation torque value and the stable bus voltage, and sending the motor control signal to a motor controller to execute torque compensation and voltage stable adjustment operation. Self-adaptive compensation torque calculation under different working conditions is achieved, predictive adjustment on instantaneous power requirements is achieved through coordinated work of the boost module and the super capacitor combination, voltage fluctuation in the torque compensation process is effectively restrained, and then effective execution of torque compensation and voltage stability adjustment is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of motor driving, in particular to a motor driving direction fault detection and compensation method and device. BACKGROUND

[0002] The motor driving direction fault includes multiple fault modes such as Hall sensor failure, phase sequence error, power device breakdown, etc. These faults often occur suddenly, and if they cannot be detected and compensated in time, they will lead to serious consequences such as vehicle loss of control and overturning. The traditional method usually uses single-parameter monitoring based on current or speed, which cannot accurately identify complex driving direction faults and is prone to misjudgment under dynamic driving conditions. At the same time, the existing fault compensation strategy mostly uses fixed parameter control, lacks self-adaptive ability to different load, slope and other working conditions, and has limited compensation effect, which is difficult to guarantee the stable operation of the vehicle in the face of sudden faults. SUMMARY

[0003] The main purpose of the present application is to provide a motor driving direction fault detection and compensation method and device, which realizes adaptive compensation torque calculation under different working conditions, realizes predictive regulation of instantaneous power demand through the coordinated work of the boost module and the super capacitor combination, effectively suppresses the voltage fluctuation in the torque compensation process, and further ensures the effective execution of torque compensation and voltage stable regulation.

[0004] To achieve the above purpose, the present application provides a motor driving direction fault detection and compensation method, comprising the following steps: Collecting the state data of the target vehicle and constructing a sensor fusion vector; Calculating a compensation torque value based on the state data and the sensor fusion vector; Driving the boost module and the super capacitor combination in the two-stage voltage stabilization circuit based on the compensation torque value to perform voltage pre-regulation, and obtaining a stable bus voltage; Generating a motor control signal based on the compensation torque value and the stable bus voltage, and sending the motor control signal to the motor controller to perform torque compensation and voltage stable regulation operation.

[0005] Optionally, in the first implementation manner of the first aspect of the present application, the collecting the state data of the target vehicle and constructing a sensor fusion vector comprises: Pretreating the three-axis acceleration data and three-axis angular velocity data collected by the inertial measurement unit, the three-phase sequence data collected by the Hall sensor, the real-time current data collected by the bus current sensor, and the pulse data collected by the wheel speed encoder to obtain an original data set; Inputting the original data set into a Kalman filter for state updating to obtain state data; performing fast Fourier transform on the original data set to obtain a feature data matrix; setting a first weight coefficient of an inertial measurement unit, a second weight coefficient of a Hall sensor, a third weight coefficient of a current sensor, and a fourth weight coefficient of an encoder, and performing weighted combination on the feature data matrix to obtain a sensor fusion vector.

[0006] Optionally, in a second implementation manner of the first aspect of the present application, the compensation torque value is calculated based on the state data and the sensor fusion vector, comprising: extracting a posture angle and an angular velocity component from the state data to construct a quaternion state vector, and inputting the quaternion state vector into a quaternion differential equation to obtain a real-time pose parameter; performing quaternion dot product operation on the real-time pose parameter and a standard pose parameter of an expected driving trajectory to obtain a pose deviation angle; generating a driving direction abnormal signal when the pose deviation angle exceeds a preset angle threshold; triggering a fault mode recognition process according to the driving direction abnormal signal, determining a corresponding fault type by comparing the sensor fusion vector with the matching similarity of Hall failure, phase sequence error, power device failure, electromagnetic interference, mechanical jamming, overcurrent protection, encoder failure and power supply abnormality; calculating a compensation torque value according to the fault type and the sensor fusion vector.

[0007] Optionally, in a third implementation manner of the first aspect of the present application, the compensation torque value is calculated according to the fault type and the sensor fusion vector, comprising: performing fuzzy operation on the fault type and the sensor fusion vector as input variables to obtain corresponding fuzzy input values; matching fuzzy inference rules in a rule layer based on the fuzzy input values and performing linear combination to obtain a basic compensation torque under a current working condition; performing real-time correction on the basic compensation torque and an error signal of an actual torque output to obtain an adaptive parameter combination, and performing working condition compensation on the adaptive parameter combination to obtain a compensation torque value.

[0008] Optionally, in a fourth implementation manner of the first aspect of the present application, the compensation torque value is used to drive the voltage pre-regulation of the combination of the boost module and the super capacitor in the two-stage voltage stabilization circuit to obtain a stable bus voltage, comprising: performing power calculation on the compensation torque value and load current data to obtain instantaneous power demand data; performing voltage change prediction based on the instantaneous power demand data to obtain a voltage fluctuation prediction value; According to the voltage fluctuation prediction value, a duty cycle adjustment amount of a boost module in the two-stage voltage stabilization circuit is calculated, and a super capacitor combination in the two-stage voltage stabilization circuit is controlled to perform charging and discharging operations, so as to obtain a pre-regulation output voltage; The pre-regulation output voltage is input into a digital control synchronous step-down converter to perform current mode control, so as to obtain a stable bus voltage.

[0009] Optionally, in the fifth implementation manner of the first aspect of the present application, the calculation of the duty cycle adjustment amount of the boost module in the two-stage voltage stabilization circuit according to the voltage fluctuation prediction value, and the control of the super capacitor combination in the two-stage voltage stabilization circuit to perform the charging and discharging operations, so as to obtain the pre-regulation output voltage, comprises: The voltage fluctuation prediction value is subtracted from a current bus voltage to obtain a voltage deviation amount, and the voltage deviation amount is converted into the duty cycle adjustment amount of the boost module in the two-stage voltage stabilization circuit; When the duty cycle adjustment amount is a duty cycle increase, it is determined that the power demand is rising, and a super capacitor discharging instruction is generated; when the duty cycle adjustment amount is a duty cycle decrease, it is determined that the power demand is falling, and a super capacitor charging instruction is generated; The duty cycle adjustment amount is input into the boost module as a PWM control signal, a power switch tube is driven to perform on-off control according to the duty cycle adjustment amount, so as to obtain a boost output voltage; According to the super capacitor discharging instruction or the super capacitor charging instruction, the super capacitor combination in the two-stage voltage stabilization circuit is synchronously driven to perform energy release or storage operations, so as to obtain an instantaneous output power; The instantaneous output power and the boost output voltage are electrically connected in parallel to obtain the pre-regulation output voltage.

[0010] Optionally, in the sixth implementation manner of the first aspect of the present application, the synchronous driving of the super capacitor combination in the two-stage voltage stabilization circuit to perform the energy release or storage operations according to the super capacitor discharging instruction or the super capacitor charging instruction, so as to obtain the instantaneous output power, comprises: According to the super capacitor discharging instruction, a discharging time sequence control of the super capacitor combination in the two-stage voltage stabilization circuit is started, so as to obtain a super capacitor discharging current output; According to the super capacitor charging instruction, a charging time sequence control of the super capacitor combination in the two-stage voltage stabilization circuit is started, so as to obtain a super capacitor charging current input; Real-time power data of each super capacitor in the super capacitor combination is calculated according to the super capacitor discharging current output or the super capacitor charging current input; The real-time power data is subjected to vector superposition operation, so as to obtain the instantaneous output power.

[0011] Optionally, in a seventh implementation form of the first aspect of the present application, the generating the motor control signal based on the compensation torque value and the stable bus voltage, and sending the motor control signal to a motor controller to perform torque compensation and voltage stabilization adjustment operation, comprises: performing torque-current conversion on the compensation torque value to obtain a target current instruction, and calculating a PWM modulation parameter based on the stable bus voltage; coordinating the target current instruction with the PWM modulation parameter to obtain a three-phase PWM control instruction; packaging the three-phase PWM control instruction as a motor control signal, and sending the motor control signal to a motor controller through a communication interface to perform corresponding torque compensation and voltage stabilization adjustment operation.

[0012] Optionally, in an eighth implementation form of the first aspect of the present application, the coordinating the target current instruction with the PWM modulation parameter to obtain a three-phase PWM control instruction, comprises: calculating a maximum output current amplitude under the stable bus voltage, and performing amplitude limiting processing on the target current instruction when the target current instruction exceeds the maximum output current amplitude to obtain an amplitude-limited current instruction; calculating a three-phase target current component based on the amplitude-limited current instruction and a switching frequency in the PWM modulation parameter; inputting the three-phase target current component into a current loop controller to perform PI adjustment operation to obtain three-phase duty cycle control data, and generating corresponding three-phase PWM control instructions based on the three-phase duty cycle control data.

[0013] The present application also provides a motor driving direction fault detection and compensation device, comprising: a collection module configured to collect state data of a target vehicle and construct a sensor fusion vector; a calculation module configured to calculate a compensation torque value based on the state data and the sensor fusion vector; a voltage pre-adjustment module configured to perform voltage pre-adjustment on a boost module and a super capacitor combination in a two-stage voltage stabilization circuit based on the compensation torque value to obtain a stable bus voltage; an execution module configured to generate a motor control signal based on the compensation torque value and the stable bus voltage, and send the motor control signal to a motor controller to perform torque compensation and voltage stabilization adjustment operation.

[0014] In summary, the technical scheme provided by the application fuses the data of the inertial measurement unit, the Hall sensor, the current sensor and the encoder through the Kalman filter, constructs a high-precision sensor fusion vector, and has higher fault recognition accuracy and anti-interference ability compared with a single sensor detection method. The quaternion differential equation is used to solve the pose change rate, the singularity problem of the traditional Euler angle method is avoided, the vehicle body pose deviation can be stably calculated in the full attitude range, the fuzzy operation based on the fault type and the sensor fusion vector is combined with the online parameter update of the recursive least square algorithm, the adaptive compensation torque calculation under different working conditions is realized, the predictive adjustment of the instantaneous power demand is realized through the coordinated work of the boost module and the super capacitor combination, the voltage fluctuation in the torque compensation process is effectively suppressed, and the power supply stability of the control system is ensured. The coordination processing flow of the target current instruction and the PWM modulation parameter is established, the optimization generation of the motor control signal is realized through the amplitude limiting processing and the three-phase current component calculation, and the effective execution of the torque compensation and the voltage stable adjustment is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is a motor driving direction fault detection and compensation method step schematic diagram in an embodiment of the application. Figure 2 is a motor driving direction fault detection and compensation device structure block diagram in an embodiment of the application.

[0016] The realization of the object, functional characteristics and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0017] In order to make the object, technical scheme and advantages of the application more clear and explicit, the application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and not to limit the application.

[0018] With reference to Figure 1 The embodiment provides a motor driving direction fault detection and compensation method, which comprises the following steps: S1, collecting state data of a target vehicle and constructing a sensor fusion vector; Specifically, in the vehicle control system, the three-axis acceleration data and three-axis angular velocity data of the inertial measurement unit output, the three-phase signal phase sequence provided by the Hall sensor, the instantaneous current value fed back by the bus current sensor, and the pulse count data output by the wheel speed encoder are continuously collected at a high sampling frequency (such as 2 kHz), and the multi-source raw measurement information is uniformly structured and preprocessed, the preprocessing process includes time sequence alignment based on multi-channel synchronization mechanism, abnormal point elimination and linear interpolation correction based on sliding window, forming a raw data set containing complete time sequence and consistent sampling step. The raw data set is input into the Kalman filter based on the state space model, and the state prediction and correction update are performed through recursive form, wherein the state vector includes physical quantities such as position, velocity, acceleration, attitude angle and angular velocity, and the covariance matrices of process noise and observation noise are dynamically adjusted according to the measurement error statistical characteristics of each sensor. After the Kalman filter iteration operation, the state data output is obtained. At the same time of state update, the fast Fourier transform processing is performed on each signal data of the raw data set, the spectral feature parameters such as frequency domain main frequency, harmonic component and amplitude frequency power spectrum are extracted using a sliding window with a length of 256 points and setting an overlap rate of 50%, and a feature data matrix containing multiple time-frequency feature dimensions is constructed. According to the preset weight coefficient configuration of each sensor, for example, setting the first weight coefficient of the inertial measurement unit as 0.4, the second weight coefficient of the Hall sensor as 0.3, the third weight coefficient of the current sensor as 0.2, and the fourth weight coefficient of the wheel speed encoder as 0.1, the feature vectors corresponding to each type of sensor channel are weighted and processed, and the weighted sum fusion operation is performed, the weight compensation and collaborative combination of multi-source information in the feature dimension are completed, and a high-dimensional sensor fusion vector reflecting the current comprehensive state characteristics of the target vehicle is obtained.

[0019] S2, calculating a compensation torque value based on the state data and the sensor fusion vector; Specifically, the attitude angle information and angular velocity components are extracted from the state data, a quaternion state vector at the current time is constructed in the form of quaternion combination, and the quaternion state vector is input into the quaternion differential equation model based on the rigid body rotation differential structure for integral solution. The iterative derivation of the attitude evolution is completed at the discrete time step through the Runge-Kutta fourth-order numerical algorithm, and the real-time pose parameter vector describing the spatial orientation of the vehicle is obtained. The real-time pose parameter is dot product operated with the quaternion expression form of the standard driving attitude defined from the navigation planning module or the trajectory reference library, and the pose deviation angle between the current attitude and the target attitude is calculated according to the geometric meaning of the quaternion. The closer the dot product value is to 1, the more consistent the direction is. The deviation angle is obtained through the inverse cosine function conversion. When the deviation angle exceeds the preset angle threshold (for example, 8°), a driving direction abnormal signal is generated, which is used as the trigger condition of the fault recognition module to activate the fault recognition process. The sensor fusion vector constructed at the current time is compared with the fault feature fingerprint vector of each type of fault trained in the offline stage in the Euclidean distance or cosine similarity in the high-dimensional space. The typical fault modes such as Hall failure, phase sequence error, power device fault, electromagnetic interference, mechanical jamming, overcurrent protection, encoder fault and power supply abnormality are matched one by one. When the matching similarity exceeds the corresponding confidence threshold, the fault type is determined to be correct. According to the fault type, the current fusion vector and the input-response mapping rule in the corresponding fault model are jointly applied in the neural fuzzy inference structure. According to the pose deviation angle, angular velocity error and fault-related characteristic quantity as input, through the steps of fuzzy membership function evaluation, fuzzy rule activation, weight normalization and linear post-processing, the required compensation torque value is calculated in real time to correct the current driving abnormal state.

[0020] S3, based on the compensation torque value, driving the boost module and super capacitor combination in the dual-stage voltage regulation circuit to pre-regulate the voltage to obtain a stable bus voltage; Specifically, the compensation torque value is multiplied by the load current data in real time to obtain instantaneous power demand data in the current control period. Voltage change prediction is performed based on the instantaneous power demand data. A model predictive control architecture is constructed based on the coupling relationship between power, voltage and time, and the voltage change trend in the future several sampling periods is calculated using a forward difference method to obtain voltage fluctuation prediction values with time foresight, which comprehensively considers the bus voltage drop trend caused by the current power surge and the time window when the super capacitor is about to trigger discharge. The control system calculates the PWM duty cycle adjustment amount of the boost module switch tube according to the predicted voltage fluctuation change, combines the topology parameters and modulation model of the boost module, and applies the PWM duty cycle offset as a feedforward adjustment signal to the high-frequency synchronous boost controller. The super capacitor combination is jointly controlled to charge or discharge in advance according to the demand, so that the entire boost link outputs a pre-adjusted voltage close to the target bus voltage level before the actual power disturbance occurs, thereby improving the response speed and effectively suppressing transient voltage drop. The pre-adjusted output voltage is input into the digital control synchronous step-down converter located in the rear stage of the two-stage voltage stabilizing structure. In the step-down controller, a current mode control strategy is used to construct a voltage-inductor current inner loop regulation link, which is combined with the transfer function parameters designed in the compensation network to perform closed-loop regulation, and outputs a bus voltage with stable amplitude, low ripple and high response speed.

[0021] S4, generating a motor control signal based on the compensation torque value and the stable bus voltage, and sending the motor control signal to a motor controller to perform torque compensation and voltage stabilization adjustment operations.

[0022] Specifically, the real-time compensation torque value is taken as an input, combined with the torque constant and the number of pole pairs of the current motor, and a standard electromagnetic torque-current mapping model is used to convert the compensation torque value into torque-current, to obtain target d-q axis current components as target current commands, wherein the d-axis component is used to adjust the flux linkage stability and the q-axis component directly determines the output torque size. At the same time, the bus voltage value, which has been output by the double-stage voltage stabilizing circuit and stabilized in a preset range (such as 24V±0.3V), is taken as a DC side reference input, combined with the SVPWM space vector modulation principle and the three-phase bridge arm driving model to calculate the PWM modulation parameters required in the current control period, including the PWM carrier frequency, the dead time setting and the initial value of the three-phase pulse duty cycle, to ensure that the motor driver has sufficient modulation depth and energy control margin when performing high-frequency control. The target current command and the PWM modulation parameter are coordinated and processed, the target current command is input into the current loop PI controller, the modulation vector reference value is formed through current error closed-loop adjustment, and the modulation vector reference value and the SVPWM calculation module in the modulation parameter are jointly used to generate the projection time of the three-phase voltage vector in the sector, thereby deriving the three-phase PWM control command to ensure that the modulation waveform is highly consistent with the target current demand. The three-phase PWM control command is packaged into a data structure conforming to the motor control communication protocol, such as a CAN-FD, SPI or UART format motor control signal package, and is sent to the power stage driving module inside the motor controller through the communication interface. After receiving the three-phase PWM control command, the driving module executes the corresponding gate control logic to drive the three-phase inverter bridge arm, thereby completing the real-time application of the compensation torque and the duty cycle dynamic control in the voltage regulation process, so that the motor can recover to the desired attitude in a very short time after the direction fault occurs, and the driving system can continuously work under stable voltage conditions.

[0023] In one example, state data of a target vehicle is collected and a sensor fusion vector is constructed, including: The three-axis acceleration data and three-axis angular velocity data collected by the inertial measurement unit, the three-phase sequence data collected by the Hall sensor, the real-time current data collected by the bus current sensor, and the pulse data collected by the wheel speed encoder are preprocessed to obtain an original data set; The original data set is input into a Kalman filter for state updating to obtain state data; The original data set is subjected to fast Fourier transform to obtain a feature data matrix; The first weight coefficient of the inertial measurement unit, the second weight coefficient of the Hall sensor, the third weight coefficient of the current sensor, and the fourth weight coefficient of the encoder are set, and the feature data matrix is combined by weighting to obtain a sensor fusion vector.

[0024] In this example, the raw signal data of key components such as the inertial measurement unit, the Hall sensor, the bus current sensor, and the wheel speed encoder are synchronously acquired under the control of the system sampling frequency, among which the inertial measurement unit outputs six-dimensional vector data containing three-axis acceleration and three-axis angular velocity at a high frequency, representing the current linear motion state and angular momentum characteristics of the vehicle, the Hall sensor outputs a three-phase signal sequence with a fixed phase difference, encoded in the form of 120° electrical angle, for determining the motor rotor phase sequence and rotation direction, the bus current sensor provides real-time power supply current feedback data with high electrical dynamic sensitivity, and the wheel speed encoder records the wheel angular velocity and relative displacement information through pulse counting. The above-mentioned various raw data are uniformly preprocessed after collection, including format standardization, noise suppression, and time alignment, etc. The preprocessing procedures include outlier rejection, moving average filtering, linear interpolation completion, and synchronous resampling based on inter-channel timestamps, so that the sensor signals from different physical structures and different sampling frequencies are integrated into a consistent structure, time-synchronized, and non-hollow raw data set. The raw data set is input into the Kalman filter, which predicts and updates the state vector of the vehicle according to the state space model, and the state variables include position, velocity, acceleration, attitude angle, and angular velocity, etc. The prediction step of Kalman filtering is based on the state estimation value and input control quantity at the last time, the update step fuses the current raw observation data and the prediction result, and combines the process noise covariance matrix and the observation noise covariance matrix in the sensor model to dynamically adjust the Kalman gain, so as to realize the optimal estimation of the state variable. The state data output by the filter has strong stability and noise suppression capability; at the same time, the frequency domain analysis operation is performed on the raw data set, and the fast Fourier transform algorithm is used to transform and process each channel signal to extract the amplitude spectrum, phase spectrum, main frequency position, harmonic intensity, and energy distribution, etc. of the signal in the 0~500Hz frequency band, so as to obtain a feature data matrix with unified structure and dimension alignment, reflecting the frequency spectrum change behavior of each sensor channel in different time windows, for identifying the fault mode or dynamic characteristic fluctuation in the motor drive system. According to the reliability, response speed, and information density of each sensor in actual application, the weight coefficients of each channel are set, among which the inertial measurement unit is set as the first weight coefficient 0.4 due to its high frequency, high precision, and six-dimensional output characteristics, the Hall sensor is set as the second weight coefficient 0.3 due to its high sensitivity to phase error and direction reversal, the current sensor is set as the weight 0.2 due to its rapid response and representativeness to overload and short circuit faults, and the encoder is set as the fourth weight coefficient 0.1 as an auxiliary positioning channel. The data corresponding to each channel in the feature data matrix are multiplied by the corresponding weight coefficients and subjected to vector weighted summation to complete the fusion processing of different channel signal features and form a high-dimensional sensor fusion vector.

[0025] The original data set is input into the Kalman filter for state updating to obtain state data, including: based on the real-time noise characteristics of each sensor, the process noise covariance matrix of the Kalman filter is adaptively adjusted, the noise covariance parameters reflecting the real-time reliability of the sensor are obtained by calculating the inertial measurement unit temperature drift coefficient, the Hall sensor signal quality factor, the current sensor dynamic range and the encoder pulse stability index; the noise covariance parameters are input into the adaptive Kalman filtering algorithm for state prediction calculation, the vehicle body position, speed and attitude angle are recursively estimated according to the system dynamic model and the control input, the time-varying parameters of the state transition matrix are set to adapt to the nonlinear motion characteristics of the electric scooter, and the state prediction result containing uncertainty information is obtained; based on the residual analysis of the state prediction result and the actual sensor measurement value, the filter gain is adaptively adjusted, the real-time correction amount of the Kalman gain matrix is calculated, the trust weight of the corresponding sensor is reduced when the sensor measurement noise increases, and the optimal Kalman gain parameter after dynamic adjustment is obtained; the optimal Kalman gain parameter is applied to the state update equation for fusion calculation, the predicted state and the measurement value are weighted and combined, the credibility change of the system state estimation is reflected through the recursive update of the covariance matrix, and the filtering state vector with adaptive noise suppression capability is obtained; based on the filtering state vector, data quality evaluation and abnormality detection are performed, the eigenvalue threshold of the state covariance matrix is set to identify sensor failure or data anomaly, and when data anomaly is detected, the sensor redundancy switching mechanism is started, and the state data is obtained.

[0026] In one example, the compensation torque value is calculated based on the state data and the sensor fusion vector, including: A quaternion state vector is constructed by extracting the attitude angle and angular velocity components from the state data, and the quaternion state vector is input into a quaternion differential equation for solving to obtain real-time pose parameters; The real-time pose parameters are subjected to quaternion dot product operation with standard pose parameters of the expected driving trajectory to obtain a pose deviation angle; When the pose deviation angle exceeds a preset angle threshold, a driving direction abnormality signal is generated; A fault mode recognition process is triggered according to the driving direction abnormality signal, and the corresponding fault type is determined by comparing the sensor fusion vector with the matching similarity of Hall failure, phase sequence error, power device failure, electromagnetic interference, mechanical jamming, overcurrent protection, encoder failure and power supply abnormality; The compensation torque value is calculated according to the fault type and the sensor fusion vector.

[0027] In this example, the Euler angle posture of the current moment is extracted from the state data, that is, the roll angle, pitch angle and yaw angle, and the three-axis angular velocity components, where the angular velocity is derived from the gyroscope part of the inertial measurement unit, and the posture angle is obtained by fusing the accelerometer and magnetometer through complementary filtering or extended Kalman filtering. The posture angle is converted to quaternion expression form through three-dimensional rotation relationship, the quaternion state vector of the current moment is constructed, and the angular velocity vector is introduced as an input variable to introduce the quaternion differential equation for pose update calculation. The differential process is integrated by numerical methods such as Runge-Kutta fourth-order algorithm in a small time step to obtain continuous real-time pose parameters, which describe the direction state of the vehicle in three-dimensional space, and are more suitable for large-scale rotation and can effectively avoid gimbal lock problem compared with Euler angle expression form. The real-time pose quaternion is multiplied by the standard pose parameter provided by the trajectory planning module to obtain the rotation consistency measurement value between the two, and the dot product result is converted to the pose deviation angle by the inverse cosine function to evaluate the deviation degree between the current actual driving posture and the ideal trajectory posture. When the pose deviation angle exceeds the preset angle threshold, it is considered that the vehicle currently has a driving direction abnormality, and a driving direction abnormality signal is generated accordingly, which is used as the starting condition of the fault mode recognition process. After the fault recognition process is triggered, the sensor fusion vector generated at the current moment is compared with the multi-class typical fault feature fingerprint vector one by one, and the matching degree between each fault model is evaluated in turn by using similarity calculation methods such as Euclidean distance, cosine similarity or Mahalanobis distance. The matching objects include eight typical driving direction related faults, such as Hall failure, phase sequence error, power device failure, electromagnetic interference, mechanical jamming, overcurrent protection, encoder failure and power supply abnormality. Each type of fault is trained by a large amount of data in the offline stage to form a unique feature vector as a recognition reference. When the matching similarity of any one type of fault exceeds the confidence threshold, it is determined as this type of fault, and the current fault type is automatically marked by the system. According to the fault type, fault location and time-frequency domain feature parameters contained in the current sensor fusion vector, a compensation decision basis is constructed, which is input into an adaptive neural fuzzy inference system running online. The pose deviation angle, angular velocity error, Hall phase displacement, current disturbance intensity, etc. are used as input variables to activate the fuzzy rule set and update the parameters online combined with historical data. The compensation torque value required at the current moment is output through the fuzzy inference process.

[0028] In one example, the compensation torque value is calculated according to the fault type and the sensor fusion vector, comprising: Fuzzy operation is performed on the fault type and the sensor fusion vector as input variables to obtain corresponding fuzzy input values; Based on the fuzzy input values, the fuzzy inference rules are matched in the rule layer, and linear combination is performed to obtain the basic compensation torque under the current working condition; The error signal of the basic compensation torque and the actual torque output is corrected in real time to obtain an adaptive parameter combination, and the adaptive parameter combination is compensated for working conditions to obtain a compensation torque value.

[0029] In this example, the specific fault type is encoded into a form of input label recognizable by the system, and is input into the fuzzy inference system generated by the high-dimensional sensor fusion vector of the current cycle. The fuzzy inference system adopts an adaptive neural fuzzy inference system architecture, wherein the first layer is a fuzzification layer, which maps the input variables to corresponding fuzzy input values through a preset Gaussian membership function or a triangular membership function. The fault type is converted into a set of fuzzy labels with a single peak as a discrete logical variable, while the key parameters in the sensor fusion vector, such as attitude deviation, angular velocity disturbance, current abnormal amplitude or phase shift, are mapped into continuous fuzzy intervals through function transformation to construct a fuzzy input space. In the second layer rule layer, the fuzzy input values are matched and calculated with the fuzzy rule set set in the rule library. The rules are expressed in the form of “if-then”, such as “if the angular velocity is high and the current fluctuation is large and the fault type is phase sequence error, then output a medium-high value of compensation torque”. Each rule corresponds to a set of premise conditions and a fuzzy output label. The rule activation degree is scored and given a corresponding weight according to the input fuzzy value, and parallel activation of multiple rules is realized. In the third layer normalization layer, the weights of all activated fuzzy rules are normalized to ensure numerical consistency. In the fourth layer defuzzification layer, linear combination calculation is performed on the output values of each rule and its corresponding weight to obtain a basic compensation torque value representing the current working condition. The basic compensation torque is compared with the actual torque output signal of the motor in real time, and the torque error signal obtained by calculation is input into the adaptive adjustment mechanism as the adjustment amount. The adaptive adjustment mechanism adjusts the parameter weights in the fuzzy system based on the recursive least squares method or the incremental gradient descent algorithm to form a set of dynamically updated adaptive parameter combinations, which include real-time correction of the membership function shape, rule activation weight and output linear coefficient, and enable the fuzzy inference system to maintain high adaptability and high robustness when facing various external disturbances such as load, slope, voltage state or transmission resistance. The corrected adaptive parameter combination is applied to the current fuzzy inference structure to re-operate the original input variables, and a working condition compensation model related to environmental factors is added on this basis. The working condition compensation model adjusts the output torque based on the vehicle load, road slope, tire pressure state or temperature factor to obtain a compensation torque value.

[0030] The error signal of the actual torque output is corrected in real time with the basic compensation torque to obtain an adaptive parameter combination, including: collecting the current motor driving direction deviation angle, vehicle body attitude deviation angle and load change as nonlinear control variables, constructing a multi-dimensional state vector containing motor torque fluctuation, attitude angular velocity and load disturbance to obtain nonlinear state parameters reflecting the dynamic characteristics of the electric scooter; inputting the nonlinear state parameters and the error signal of the basic compensation torque into a nonlinear adaptive control algorithm, dynamically adjusting the control parameters by setting a nonlinear gain function and an adaptive law, including the under-actuated characteristics and time-varying disturbances of the motor driving system in the controller design to obtain a control law parameter containing a nonlinear compensation term; based on the grey wolf optimization algorithm, the weight parameters of the back propagation neural network are online tuned, the hidden layer weights and output layer biases of the neural network are iteratively optimized by setting the search mechanism and fitness function of the grey wolf group, and the neural network parameter combination optimized by the grey wolf algorithm is obtained; the neural network parameter combination optimized by the grey wolf algorithm is input into the adaptive controller for gain parameter tuning, the controller gain is updated in real time according to the current system state and external disturbance intensity, the problem of insufficient adaptability of fixed gain parameters to time-varying system signals is eliminated, and the adaptive parameter combination with online learning ability is obtained; based on the adaptive parameter combination with online learning ability, the basic compensation torque is nonlinearly corrected, the fast convergence control of the motor driving direction deviation is combined with the vehicle body attitude stability requirement, and the adaptive parameter combination is obtained.

[0031] In one example, the boost module and super capacitor combination in the two-stage voltage stabilization circuit are driven based on the compensation torque value to perform voltage pre-regulation, and a stable bus voltage is obtained, including: The compensation torque value and the load current data are calculated to obtain instantaneous power demand data; Based on the instantaneous power demand data, voltage change prediction is performed to obtain a voltage fluctuation prediction value; According to the voltage fluctuation prediction value, the duty cycle adjustment amount of the boost module in the two-stage voltage stabilization circuit is calculated, and the super capacitor combination in the two-stage voltage stabilization circuit is controlled to perform charging and discharging operation to obtain a pre-regulation output voltage; The pre-regulation output voltage is input into the digital control synchronous step-down converter for current mode control to obtain a stable bus voltage.

[0032] In this example, the compensation torque value is power-operated with real-time load current data, including mapping torque demand to corresponding current load through electromagnetic power relationship, and combining the actual instantaneous current amplitude flowing through the bus to participate in dynamic power evaluation, to calculate the instantaneous power demand data in the current control period. Based on the historical power data sequence in a short time window, a sliding time domain power prediction model is constructed, the voltage trend in future sampling periods is judged through first-order difference, second-order change rate and weighted moving average analysis of power data, and the voltage fluctuation prediction value is output, reflecting the bus voltage drop amplitude or voltage overshoot trend caused by compensation torque change in a very short time in the future. The calculation result will directly affect the adjustment strategy of the front-stage boost module. The voltage fluctuation prediction value is input into the control logic of the boost stage, according to the control characteristics and topology parameters of the switch tube in the boost circuit, the optimal duty cycle adjustment amount is calculated, the optimal duty cycle determines the dynamic change of the energy conversion efficiency in the boost loop, and the set high-frequency PWM carrier is used for pulse modulation control. At the same time, according to the current voltage gap and power demand, the state of the super capacitor combination is judged. If the current bus voltage has a significant downward trend and the super capacitor is in the fully charged state, the discharge path is triggered and discharged at a high current to increase the energy of the boost input end. If the bus voltage is stable but the subsequent prediction shows that the power will soon rise, the charging channel of the super capacitor is activated in advance, and part of the energy is stored in the super capacitor in a fast cycle manner, thereby forming a closed-loop energy regulation mechanism, and the pre-regulation output voltage after boost regulation and super capacitor charge-discharge combination control is output. The pre-regulation output voltage is input into the digital control synchronous buck converter, which adopts current mode control strategy, in which the inner loop uses inductance current detection signal to realize fast response, and the outer loop uses output voltage feedback as the main adjustment target to form a double-loop control structure. The double-loop control structure realizes stable closed-loop regulation through a compensation network, which contains proportional gain, zero frequency and pole frequency parameter adjustment to ensure that the system has good phase margin and gain margin, so as to ensure the high precision and fast recovery ability of the voltage output when facing different load dynamic changes, and outputs a stable bus voltage.

[0033] The voltage fluctuation prediction value is obtained by performing voltage change prediction based on instantaneous power demand data, including: constructing an electrical system dynamic model containing motor load characteristics, bus capacitance energy storage and system impedance parameters, converting the instantaneous power demand into a dynamic response equation of the bus voltage by establishing a transfer function relationship between the power demand and the voltage response, and obtaining a voltage prediction model reflecting the electrical characteristics of the system; setting the prediction time domain and control time domain parameters based on the model predictive control theory, taking the power demand change trend in the next ten sampling periods as the prediction input, and performing multi-step prediction calculation on the voltage fluctuation trajectory by a rolling optimization algorithm to obtain a voltage prediction sequence covering short-term and medium-term changes; establishing an external disturbance model containing load mutation, temperature change and device aging, estimating the unmodeled dynamics of the system in real time by setting a disturbance observer, and incorporating a disturbance compensation term into the voltage prediction calculation to correct the deviation between the ideal model and the actual system, thereby obtaining a corrected prediction model considering the influence of external disturbances; using the recursive least squares algorithm to identify the parameters of the corrected prediction model online, updating the model parameters according to the matching error of the historical voltage measurement data and the power demand data, and setting a forgetting factor to adapt to the time-varying nature of the system characteristics, thereby obtaining a real-time prediction model with parameter adaptive capability; inputting the current instantaneous power demand data into the real-time prediction model to calculate the voltage fluctuation, ensuring that the prediction result meets the voltage safety range requirement by using a constraint optimization algorithm, limiting the prediction value that exceeds the safety threshold, and obtaining the voltage fluctuation prediction value.

[0034] In one example, the duty cycle adjustment amount of the boost module in the two-stage voltage stabilizing circuit is calculated according to the voltage fluctuation prediction value, and the super capacitor combination in the two-stage voltage stabilizing circuit is controlled to perform charging and discharging operations to obtain a pre-regulated output voltage, including: The voltage fluctuation prediction value is subtracted from the current bus voltage to obtain a voltage deviation amount, and the voltage deviation amount is converted into a duty cycle adjustment amount of the boost module in the two-stage voltage stabilizing circuit; When the duty cycle adjustment amount is a duty cycle increase, it is determined that the power demand is rising and a super capacitor discharging instruction is generated, and when the duty cycle adjustment amount is a duty cycle decrease, it is determined that the power demand is falling and a super capacitor charging instruction is generated; The duty cycle adjustment amount is input as a PWM control signal to the boost module to drive the power switch tube to perform on-off control according to the duty cycle adjustment amount, thereby obtaining a boost output voltage; The super capacitor combination in the two-stage voltage stabilizing circuit is synchronously driven according to the super capacitor discharging instruction or the super capacitor charging instruction to perform energy release or storage operations, thereby obtaining an instantaneous output power; The instantaneous output power and the boost output voltage are electrically connected in parallel to obtain a pre-regulated output voltage.

[0035] In this example, the voltage fluctuation prediction value of the prediction module output is obtained from the previous control cycle, reflecting the voltage fluctuation trend of the bus voltage in the future very short time window due to load mutation, compensation torque change or system energy feedback; At the same time, the actual value of the current bus voltage is collected in real time, and the difference between the current bus voltage value and the predicted voltage value is calculated to obtain the voltage deviation, where the positive deviation represents the future voltage drop and the negative deviation represents the future voltage rise. The voltage deviation is input into the controller of the boost module, and based on the preset voltage-duty cycle mapping relationship, the deviation is converted into the duty cycle adjustment amount of the boost module in real time, which is used to adjust the duty cycle of the high-frequency PWM signal in the boost circuit, thereby controlling the conduction time of the main power switch tube. The controller determines the dynamic power state according to the numerical direction of the duty cycle adjustment amount. When the duty cycle adjustment amount is positive, that is, the duty cycle is increased, it is determined that the current power demand is rising, and a super capacitor discharge instruction is generated to prepare to release a large current through the super capacitor to support the boost circuit to maintain the output voltage stable; Conversely, when the duty cycle adjustment amount is negative, that is, the duty cycle needs to be reduced, it is determined that the load power demand is decreasing, and a super capacitor charging instruction is generated to fully utilize the voltage redundancy to recover energy and maintain sufficient power of the energy storage module. The duty cycle value is input into the PWM generation unit of the boost module, and the corresponding high-frequency control waveform is modulated in real time, and the main power switch tube in the boost topology is driven to control the on-off, the conduction time and the off time of the power switch tube are controlled, the energy conversion efficiency of the boost inductor is adjusted, and the stable controlled boost output voltage is output. In this process, the super capacitor group acts according to the charge and discharge instructions. If it is a discharge instruction, the discharge path is turned on to conduct current to the boost inductor port, forming a low-internal-resistance large-power instantaneous power supply support, so that the boost voltage quickly responds to the target value, improving the adaptability of the system to power disturbance; If it is a charging instruction, the control conduction path is switched to the super capacitor charging circuit, and the constant current or segmented constant voltage charging process is performed through the current limiting controller to prolong the service life of the super capacitor and protect the bus voltage from being affected by the charging disturbance. The energy release or storage action corresponds to the output instantaneous power value, which is calculated by the product of current and voltage, and is combined in parallel with the voltage signal output by the boost module in the electrical structure to form the pre-regulated voltage at the output end.

[0036] In one example, the super capacitor group in the two-stage voltage stabilizing circuit is driven according to the super capacitor discharge instruction or the super capacitor charging instruction to perform energy release or storage operation to obtain instantaneous output power, comprising: starting the discharge timing control of the super capacitor group in the two-stage voltage stabilizing circuit according to the super capacitor discharge instruction to obtain the super capacitor discharge current output; starting the charging timing control of the super capacitor group in the two-stage voltage stabilizing circuit according to the super capacitor charging instruction to obtain the super capacitor charging current input; According to the super capacitor discharge current output or the super capacitor charging current input, the real-time power data of each super capacitor in the super capacitor combination is calculated; The real-time power data is subjected to vector superposition operation to obtain the instantaneous output power.

[0037] In this example, after detecting the super capacitor discharge instruction issued by the boost controller, the discharge timing control logic of the super capacitor module is entered, which is executed by the discharge management unit inside the digital controller. By scheduling the discharge channels of each single super capacitor in the capacitor group, according to the preset power demand level and voltage recovery target, the discharge sequence, conduction time and current limiting strategy of the discharge channel are determined. The controller selects to start the discharge circuit of part or all super capacitor units in series or parallel according to the level of the current required output power, and dynamically adjusts the discharge duty cycle combined with the bus voltage drop trend to ensure that the capacitor discharge process can stably support the energy demand of the boost circuit. At the same time when the controller starts the discharge logic, the low-impedance loop is connected to the path from the super capacitor to the front end of the boost inductor, so that the super capacitor releases the stored energy in a very short time with a large current. The discharge current output signal of each super capacitor unit is obtained in real time through the current sampling unit, and the super capacitor discharge current output vector is constructed. When the system judges that the load power demand decreases or the voltage appears over-regulation phenomenon, the controller generates a super capacitor charging instruction according to the feedback of the boost circuit, and starts the charging timing control module of the super capacitor combination. The charging timing control module determines whether the charging path is turned on and whether the charging rate is limited according to the current state of charge, voltage margin and system energy surplus of the super capacitor. The controller controls the MOSFET or relay on each super capacitor charging path to realize group-by-group charging scheduling, and sets a reasonable charging current curve using the constant current or constant voltage mode to ensure efficient energy recovery during the entire charging process without overheating or life degradation caused by large current charging. During the charging control process, the charging current of each super capacitor unit is sampled in real time through the high-precision sampling module to obtain the super capacitor charging current input vector, which is used to describe the current dynamic behavior of the capacitor module absorbing energy. According to the discharge current output vector or the charging current input vector, and combined with the instantaneous terminal voltage of each super capacitor, the real-time power data of each super capacitor unit is calculated. The real-time power data is indexed by the capacitor number to form a multi-dimensional power data array, which reflects the specific energy level of each group of super capacitors at the current time. The power data of each super capacitor is subjected to vector superposition operation, that is, the numerical values of all single power values are added to obtain the instantaneous output power of the entire super capacitor combination.

[0038] The super capacitor discharge current output is obtained by starting the discharge timing control of the super capacitor combination in the two-stage voltage stabilizing circuit according to the super capacitor discharge instruction, including: establishing a hierarchical energy management architecture of three parallel super capacitors, decomposing the overall power demand into power distribution instructions of each super capacitor through the hierarchical control structure of the main control layer, the coordination control layer and the execution control layer, performing load balancing distribution according to the residual capacity and health state of each super capacitor, and obtaining a hierarchical power distribution strategy; based on the power distribution strategy, the discharge sequence of the three super capacitors is optimized, the equivalent series resistance, capacity attenuation rate and temperature characteristics of each super capacitor are calculated, the super capacitor with the best performance is preferentially selected to undertake the main power output task, and the optimal discharge timing scheme considering the difference in device characteristics is obtained; the pulse width modulation technology is used to accurately control the discharge current of each super capacitor, the reference pulse width of the first super capacitor, the phase delay pulse width of the second super capacitor and the amplitude modulation pulse width of the third super capacitor are set, the gradient output of the discharge current is realized through the dynamic adjustment of the duty cycle, and the current waveform control signal with smooth transition is obtained; a multi-objective optimization control algorithm for super capacitor temperature rise and voltage balance is established, the temperature constraint function and the voltage balance constraint function are set, the temperature rise and voltage imbalance degree are minimized under the premise of meeting the power output requirement, the Lagrange multiplier method is used to solve the multi-constraint optimization problem, and the optimal control parameters considering performance and safety are obtained; based on the optimal control parameters, the super capacitor discharge control circuit is driven to control the switch tube conduction, the stored energy of each super capacitor is released to the bus according to the set timing through high-frequency switching action, the amplitude and waveform quality of the discharge current are monitored in real time, and the super capacitor discharge current output is obtained.

[0039] In one example, the motor control signal is generated based on the compensation torque value and the stable bus voltage, and the motor control signal is sent to the motor controller to perform torque compensation and voltage stabilization adjustment operations, including: The compensation torque value is converted into a target current instruction, and the PWM modulation parameters are calculated based on the stable bus voltage; The target current instruction and the PWM modulation parameters are coordinated to obtain a three-phase PWM control instruction; The three-phase PWM control instruction is packaged into a motor control signal, which is sent to the motor controller through a communication interface to perform corresponding torque compensation and voltage stabilization adjustment operations.

[0040] In this example, the compensation torque value is input into the torque-current mapping module, which maps the compensation torque value to the corresponding target current value based on the structural parameters of the controlled motor, including the number of pole pairs, stator inductance, motor type (such as permanent magnet synchronous motor or brushless direct current motor), and torque constant, etc. key indicators, through the established electromagnetic model, the compensation torque value is converted into the corresponding target current value expressed in the form of d-axis and q-axis coordinate system components, where the d-axis current is used to adjust the magnetic flux, and the q-axis current mainly determines the output torque size. In the drive system using vector control strategy, the accuracy of q-axis current is directly related to the accuracy of compensation execution. Synchronously read the real-time value of the current stable bus voltage from the front-stage voltage stabilization module, the system controller inputs the stable bus voltage as one of the input variables into the PWM modulation parameter generation module, which calculates the optimal modulation ratio, carrier frequency, dead time compensation amount, and modulation region position under the current modulation period based on the space vector PWM (SVPWM) or sine PWM strategy, combined with the proportional relationship between the target output voltage amplitude and the bus voltage, to ensure that the PWM modulation does not appear under-modulation or over-modulation phenomenon, and maintains the maximum efficiency within the safe working interval of the power device. The target current instruction and PWM modulation parameter are input into the current regulation loop for coordinated processing, and the dynamic correction of the current error is realized through the closed-loop PI or PID controller. The target current instruction is compared with the actual motor feedback current, the current error is calculated, and the modulation voltage instruction vector is calculated through the proportional and integral elements. The modulation voltage instruction vector and the PWM modulation parameter jointly act on the SVPWM modulation module to complete the matching of the modulation reference vector and the sector position, the carrier comparison and the conduction time allocation, generate three-phase PWM control instructions that meet the current operating state of the motor, and control the power switching devices of the motor three-phase bridge arm to realize the regulation of the phase voltage. The three-phase PWM control instructions are packaged into a standard motor control data structure, which is protocol layer encapsulated according to the type of communication bus (such as CAN, SPI, UART or RS-485, etc.), including frame header, address field, data load, CRC check bit, etc. Form a motor control signal frame, and then send it to the motor controller receiving end through the communication interface. After the motor controller receives the motor control signal, the data is parsed and the three-phase PWM duty cycle instruction is extracted, the power device performs the corresponding conduction and cutoff operation, realizes the dynamic control of the motor three-phase current, completes the torque compensation control based on the target current instruction, and ensures the smooth and stable voltage regulation process through the control of the modulation ratio.

[0041] In one example, the target current instruction and the PWM modulation parameter are coordinated to obtain a three-phase PWM control instruction, including: The maximum output current amplitude under the stable bus voltage is calculated, and when the target current instruction exceeds the maximum output current amplitude, the target current instruction is limited to obtain a limited current instruction; The three-phase target current component is calculated based on the limited current instruction and the switching frequency in the PWM modulation parameter; The three-phase target current component is input into the current loop controller for PI adjustment operation to obtain three-phase duty cycle control data, and the corresponding three-phase PWM control instruction is generated based on the three-phase duty cycle control data.

[0042] In this example, the stable bus voltage is matched with the inverter DC bus-AC side maximum current conversion model, which includes motor back electromotive force constant, power device conduction voltage drop, inductance current limiting capability and thermal protection parameters, etc. According to these parameters, the maximum allowable current amplitude that the system can safely output under the stable bus voltage level is estimated. At the same time, the target current instruction calculated from the torque compensation logic module is compared, and if the target current value does not exceed the above-mentioned maximum allowable amplitude, it is directly used as the input reference signal for current regulation; if the target current instruction exceeds the maximum output current amplitude, the limiting processing logic is enabled, and the target current is truncated to within the maximum safe amplitude by performing upper and lower limit constraint operations on the target current instruction, to obtain the limited current instruction. According to the limited current instruction and the switching frequency in the PWM modulation parameter, the three-phase target current component in the three-phase abc coordinate system is calculated. The three-phase target current component is input into the current loop controller as a set value for closed-loop regulation. The controller uses a PI structure, which continuously iteratively outputs a modulation voltage reference value by inputting the error between the target current value and the current feedback value into the proportional and integral sections. The proportional regulation quickly responds to error changes, and the integral regulation eliminates steady-state error. The combination of the two enables the controller to compress the current error to a minimum in a very short time, avoiding torque instability or noise oscillation caused by current deviation. The PI regulation result is mapped in real time to three-phase voltage duty cycle reference values, i.e. three-phase duty cycle control data, corresponding to U, V, and W three-phase drive channels. The three-phase duty cycle control data is input into the SVPWM or SPWM modulation module to generate the corresponding PWM pulse waveform according to the modulation region and carrier waveform. This process includes conduction time calculation, carrier comparison, dead time compensation, and modulation boundary correction, etc. three steps to form a three-phase PWM control instruction.

[0043] Reference Figure 2The embodiment provides a motor driving direction fault detection and compensation device, comprising: The acquisition module 1 is used for acquiring state data of a target vehicle and constructing a sensor fusion vector; The calculation module 2 is used for calculating a compensation torque value based on the state data and the sensor fusion vector; The voltage pre-regulation module 3 is used for driving a boost module and a super capacitor combination in a two-stage voltage stabilization circuit to perform voltage pre-regulation based on the compensation torque value, so as to obtain a stable bus voltage; The execution module 4 is used for generating a motor control signal based on the compensation torque value and the stable bus voltage, and sending the motor control signal to a motor controller to perform torque compensation and voltage stabilization adjustment operations.

[0044] In the embodiment, the specific implementation of each unit in the device embodiment is described above in the method embodiment, and will not be repeated here.

[0045] The technical scheme provided by the application fuses the data of the inertial measurement unit, the Hall sensor, the current sensor and the encoder through the Kalman filter, constructs a high-precision sensor fusion vector, and has higher fault recognition accuracy and anti-interference ability compared with the single sensor detection method. The quaternion differential equation is used to solve the pose change rate, the singularity problem of the traditional Euler angle method is avoided, the vehicle body pose deviation can be stably calculated in the full attitude range, the fuzzy operation based on the fault type and the sensor fusion vector is combined with the online parameter update of the recursive least square algorithm, the adaptive compensation torque calculation under different working conditions is realized, the predictive adjustment of the instantaneous power demand is realized through the coordinated work of the boost module and the super capacitor combination, the voltage fluctuation in the torque compensation process is effectively inhibited, and the power supply stability of the control system is ensured. The coordination processing flow of the target current instruction and the PWM modulation parameter is established, the amplitude limiting processing and the three-phase current component calculation are performed, the motor control signal is generated, the effective execution of the torque compensation and the voltage stabilization adjustment is ensured.

[0046] It should be noted that in this paper, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, device, article or method including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, device, article or method. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, device, article or method including the element.

[0047] The above merely describes preferred embodiments of the present application, and is not intended to limit the patent scope of the present application, and any equivalent structure or equivalent process conversion, or direct or indirect application in other related technical fields, which are made by using the content of the present application specification and drawings, are also included in the patent protection scope of the present application.

Claims

1. A method of motor drive direction fault detection and compensation, characterized by, The method comprises the following steps: collecting state data of a target vehicle and constructing a sensor fusion vector; calculating a compensation torque value based on the state data and the sensor fusion vector; driving a boost module and a super capacitor combination in a two-stage voltage stabilization circuit based on the compensation torque value to perform voltage pre-regulation, and obtaining a stable bus voltage; generating a motor control signal based on the compensation torque value and the stable bus voltage, and sending the motor control signal to a motor controller to perform torque compensation and voltage stabilization adjustment operations.

2. The motor drive direction fault detection and compensation method of claim 1, wherein, The method of collecting state data of a target vehicle and constructing a sensor fusion vector comprises the following steps: preprocessing three-axis acceleration data and three-axis angular velocity data collected by an inertial measurement unit, three-phase sequence data collected by a Hall sensor, real-time current data collected by a bus current sensor, and pulse data collected by a wheel speed encoder to obtain an original data set; inputting the original data set into a Kalman filter for state updating to obtain state data; performing fast Fourier transform on the original data set to obtain a feature data matrix; setting a first weight coefficient of the inertial measurement unit, a second weight coefficient of the Hall sensor, a third weight coefficient of the current sensor, and a fourth weight coefficient of the encoder, and performing weighted combination on the feature data matrix to obtain a sensor fusion vector.

3. The motor drive direction fault detection and compensation method of claim 1, wherein, The method of calculating a compensation torque value based on the state data and the sensor fusion vector comprises the following steps: extracting an attitude angle and an angular velocity component from the state data to construct a quaternion state vector, and inputting the quaternion state vector into a quaternion differential equation for solving to obtain real-time pose parameters; performing quaternion dot product operation on the real-time pose parameters and standard pose parameters of an expected driving trajectory to obtain a pose deviation angle; generating a driving direction abnormal signal when the pose deviation angle exceeds a preset angle threshold; triggering a fault mode recognition process according to the driving direction abnormal signal, determining a corresponding fault type by comparing the sensor fusion vector with the matching similarity of Hall failure, phase sequence error, power device failure, electromagnetic interference, mechanical jamming, overcurrent protection, encoder failure, and power supply abnormality; calculating a compensation torque value according to the fault type and the sensor fusion vector.

4. The motor drive direction fault detection and compensation method of claim 3, wherein, The method of calculating a compensation torque value according to the fault type and the sensor fusion vector comprises the following steps: performing fuzzy operation on the fault type and the sensor fusion vector as input variables to obtain corresponding fuzzy input values; matching fuzzy reasoning rules in a rule layer based on the fuzzy input values, and performing linear combination to obtain a basic compensation torque under a current working condition; performing real-time correction on the basic compensation torque and an error signal of actual torque output to obtain an adaptive parameter combination, and performing working condition compensation on the adaptive parameter combination to obtain a compensation torque value.

5. The motor drive direction fault detection and compensation method of claim 1, wherein, The method of driving a boost module and a super capacitor combination in a two-stage voltage stabilization circuit based on the compensation torque value to perform voltage pre-regulation and obtain a stable bus voltage comprises the following steps: performing power calculation on the compensation torque value and load current data to obtain instantaneous power demand data; performing voltage change prediction based on the instantaneous power demand data to obtain a voltage fluctuation prediction value; According to the voltage fluctuation prediction value, a duty cycle adjustment amount of a boost module in the two-stage voltage stabilization circuit is calculated, and a super capacitor combination in the two-stage voltage stabilization circuit is controlled to perform charging and discharging operations, so as to obtain a pre-regulation output voltage; The pre-regulation output voltage is input into a digital control synchronous step-down converter to perform current mode control, so as to obtain a stable bus voltage.

6. The motor drive direction fault detection and compensation method of claim 5, wherein, The method according to the voltage fluctuation prediction value, a duty cycle adjustment amount of a boost module in the two-stage voltage stabilization circuit is calculated, and a super capacitor combination in the two-stage voltage stabilization circuit is controlled to perform charging and discharging operations, so as to obtain a pre-regulation output voltage, includes: The voltage fluctuation prediction value is subtracted from the current bus voltage to obtain a voltage deviation amount, and the voltage deviation amount is converted into a duty cycle adjustment amount of a boost module in the two-stage voltage stabilization circuit; When the duty cycle adjustment amount is a duty cycle increase, it is determined that the power demand is rising, and a super capacitor discharging instruction is generated; when the duty cycle adjustment amount is a duty cycle decrease, it is determined that the power demand is falling, and a super capacitor charging instruction is generated; The duty cycle adjustment amount is input into the boost module as a PWM control signal to drive a power switch tube to perform on-off control according to the duty cycle adjustment amount, so as to obtain a boost output voltage; According to the super capacitor discharging instruction or the super capacitor charging instruction, the super capacitor combination in the two-stage voltage stabilization circuit is synchronously driven to perform energy release or storage operations, so as to obtain an instantaneous output power; The instantaneous output power and the boost output voltage are electrically connected in parallel to obtain the pre-regulation output voltage.

7. The motor drive direction fault detection and compensation method of claim 6, wherein, The method according to the super capacitor discharging instruction or the super capacitor charging instruction, the super capacitor combination in the two-stage voltage stabilization circuit is synchronously driven to perform energy release or storage operations, so as to obtain an instantaneous output power, includes: According to the super capacitor discharging instruction, a discharging time sequence control of the super capacitor combination in the two-stage voltage stabilization circuit is started, so as to obtain a super capacitor discharging current output; According to the super capacitor charging instruction, a charging time sequence control of the super capacitor combination in the two-stage voltage stabilization circuit is started, so as to obtain a super capacitor charging current input; Real-time power data of each super capacitor in the super capacitor combination is calculated according to the super capacitor discharging current output or the super capacitor charging current input; The real-time power data is subjected to vector superposition operation to obtain the instantaneous output power.

8. The motor drive direction fault detection and compensation method of claim 1, wherein, The method of generating a motor control signal based on the compensation torque value and the stable bus voltage, and sending the motor control signal to a motor controller to perform torque compensation and voltage stabilization adjustment operations, includes: The compensation torque value is subjected to torque current conversion to obtain a target current instruction, and a PWM modulation parameter is calculated based on the stable bus voltage; The target current instruction and the PWM modulation parameter are coordinated to obtain a three-phase PWM control instruction; The three-phase PWM control instruction is packaged as a motor control signal, which is sent to a motor controller through a communication interface to perform corresponding torque compensation and voltage stabilization adjustment operations.

9. The motor drive direction fault detection and compensation method of claim 8, wherein, The method of coordinating the target current instruction and the PWM modulation parameter to obtain a three-phase PWM control instruction, includes: The maximum output current amplitude under the stable bus voltage is calculated, and when the target current instruction exceeds the maximum output current amplitude, the target current instruction is limited to obtain a limited current instruction; Based on the limited current instruction and the switching frequency in the PWM modulation parameter, three-phase target current components are calculated; The three-phase target current components are input into a current loop controller for PI adjustment operation to obtain three-phase duty cycle control data, and corresponding three-phase PWM control instructions are generated based on the three-phase duty cycle control data.

10. A motor drive direction fault detection and compensation apparatus, characterized by, The motor drive direction fault detection and compensation method comprises the following steps: A collection module is configured to collect state data of a target vehicle and construct a sensor fusion vector; A calculation module is configured to calculate a compensation torque value based on the state data and the sensor fusion vector; A voltage pre-regulation module is configured to drive a boost module and a super capacitor combination in a two-stage voltage stabilization circuit to perform voltage pre-regulation based on the compensation torque value, thereby obtaining a stable bus voltage; An execution module is configured to generate a motor control signal based on the compensation torque value and the stable bus voltage, and send the motor control signal to a motor controller to perform torque compensation and voltage stabilization adjustment operations.

Images