Real-time fault detection method and system for new energy vehicle

Abstract

The invention relates to the technical field of electrical control system monitoring, and discloses a real-time fault detection method and system for a new energy automobile. The method comprises the steps of collecting voltage and current control signals of a battery management system and a motor control system to calculate a working condition feature vector, establishing a fast and slow dual-time scale state observer, calculating an adaptive gain matrix through a mahalanobis distance and a Gaussian membership function to correct the slow observer to obtain a health parameter estimated value, and judging a sudden fault by using a fast observer residual error and analyzing a health parameter trend to judge a gradual change fault. The method solves the problems that the estimation precision is reduced and the false alarm rate is increased due to the fact that the observer gain cannot adapt to multi-working-condition changes in the prior art.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of electrical control system monitoring, in particular to a real-time fault detection method and system for new energy vehicles. BACKGROUND

[0002] The real-time fault detection method for new energy vehicles is a key technology to ensure the safe operation of vehicles. In the prior art, a fault detection method based on fixed thresholds is generally used, which sets upper and lower threshold values of parameters such as voltage, current and temperature in the battery management system and the motor control system, and triggers a fault alarm when the monitored parameters exceed the preset range. Another type of prior art uses a fault detection method based on a state observer, which establishes a state observer by building a mathematical model of the electrical system, uses observation residuals to determine whether the system has a fault, or uses machine learning algorithms such as support vector machines and neural networks to establish a fault classification model by training historical fault data to identify fault types. These methods can achieve certain fault detection functions in laboratory environments and specific operating conditions, providing a technical basis for the safe monitoring of new energy vehicles.

[0003] However, the prior art has significant shortcomings. The detection method based on fixed thresholds is difficult to adapt to parameter fluctuations of new energy vehicles in different operating conditions such as acceleration, braking and climbing, and the increased voltage drop caused by the battery internal resistance in the large current discharge condition is easily misjudged as a voltage abnormality fault, and the reverse flow of current in the energy recovery condition easily triggers an abnormal alarm of the current direction, resulting in a high false alarm rate. The method based on a fixed gain state observer causes the observation error to increase due to model mismatch when the operating condition is switched, and when the vehicle is switched from a constant speed operating condition to an emergency acceleration operating condition, the observer gain parameters cannot be adjusted in time to adapt to changes in the dynamic characteristics of the system, resulting in false peaks in the observation residuals triggering false alarms. The existing machine learning methods are highly dependent on the quality and integrity of the training data, and have insufficient recognition ability for rare fault modes and new types of faults, and the complex deep learning algorithms have large computational loads and long response times, making it difficult to achieve real-time processing in the limited computing power of the vehicle embedded system, and the fault judgment cannot be completed within the control cycle, affecting the timely response of the emergency protection strategy.

[0004] Even if the state observer method is used to track the running state of the electrical components, the prior art still faces the dual challenges of multi-working condition adaptability and gradual fault recognition. Due to the complex and variable operation conditions of new energy vehicles, the fixed observer gain matrix cannot maintain optimal performance in all working conditions. When the working condition characteristics change, the estimation accuracy of the observer for the health parameters such as the battery internal resistance and the motor winding resistance decreases significantly, which not only leads to a decrease in the detection sensitivity of sudden faults, but more seriously, the slow degradation trend of the health parameters cannot be accurately tracked. The performance degradation of electrical components is often a gradual process. The battery internal resistance gradually increases due to cycle aging, the motor winding insulation deteriorates due to temperature cycling, and the power device characteristics drift due to repeated switching stress. These gradual processes have small parameter changes in the early stage, which are submerged in normal working condition fluctuations and measurement noise. The traditional threshold judgment method can only detect faults when the parameters are severely out of bounds, at which time the electrical components have already reached the near failure state and missed the time window for preventive maintenance. Therefore, the prior art cannot solve the problem of adaptive optimization of the observer gain in multi-working condition environment, nor can it realize early trend recognition of small gradual signals. SUMMARY

[0005] The present application provides a new energy vehicle real-time fault detection method and system, which is used to solve the problem that the observer gain cannot adapt to the change of multi-working conditions, leading to a decrease in estimation accuracy and an increase in false alarm rate in the prior art, and the problem that the traditional method cannot recognize the performance gradual fault of electrical components and cannot realize early warning.

[0006] In a first aspect, the present application provides a new energy vehicle real-time fault detection method, which comprises: Step S1: Collecting the voltage control signal, current control signal of the battery management system and three-phase current control signal of the motor control system, and calculating the working condition characteristic vector according to the voltage control signal, current control signal and control instruction; Step S2: Establishing a fast time scale state observer according to the voltage control signal and current control signal, and establishing a slow time scale observer taking the battery internal resistance parameter, motor winding resistance parameter and power device on-state voltage drop parameter as the extended state according to the voltage control signal and current control signal; Step S3: Calculating the Mahalanobis distance of the working condition characteristic vector and each preset working condition clustering center, substituting the Mahalanobis distance into the Gaussian membership function to calculate the membership degree of each working condition category, taking the membership degree as a weight coefficient to weight and sum the observer gain matrix corresponding to each working condition category to obtain an adaptive gain matrix, and using the adaptive gain matrix to correct the state of the slow time scale observer to obtain the battery internal resistance parameter estimation value, motor winding resistance parameter estimation value and power device on-state voltage drop parameter estimation value; Step S4: determining an electrical element sudden failure by using the observation residual of the fast time scale state observer, performing trend analysis on the battery internal resistance parameter estimate, motor winding resistance parameter estimate and power device on-state voltage drop parameter estimate to determine an electrical element performance degradation failure, and generating a fault detection result.

[0007] In a second aspect, the application provides a real-time fault detection system for a new energy vehicle, which comprises: A collection module configured to collect a voltage control signal and a current control signal of a battery management system and a three-phase current control signal of a motor control system, and calculate a working condition feature vector according to the voltage control signal, the current control signal and a control instruction; A generation module configured to establish a fast time scale state observer according to the voltage control signal and the current control signal, and establish a slow time scale observer taking a battery internal resistance parameter, a motor winding resistance parameter and a power device on-state voltage drop parameter as extended states according to the voltage control signal and the current control signal; A calculation module configured to calculate Mahalanobis distances between the working condition feature vector and preset working condition clustering centers, calculate a membership degree of each working condition category by substituting the Mahalanobis distances into a Gaussian membership function, weight and sum observer gain matrices corresponding to each working condition category by taking the membership degrees as weight coefficients to obtain an adaptive gain matrix, and perform state correction on the slow time scale observer by using the adaptive gain matrix to obtain a battery internal resistance parameter estimate, a motor winding resistance parameter estimate and a power device on-state voltage drop parameter estimate; A determination module configured to determine an electrical element sudden failure by using the observation residual of the fast time scale state observer, perform trend analysis on the battery internal resistance parameter estimate, the motor winding resistance parameter estimate and the power device on-state voltage drop parameter estimate to determine an electrical element performance degradation failure, and generate a fault detection result.

[0008] In a third aspect, a real-time fault detection device for a new energy vehicle is provided, which comprises a memory and at least one processor, the memory stores instructions, and the at least one processor invokes the instructions in the memory to enable the real-time fault detection device for a new energy vehicle to perform the real-time fault detection method for a new energy vehicle described above.

[0009] In a fourth aspect, a computer readable storage medium is provided, which stores instructions, and when the instructions are run on a computer, the computer is enabled to perform the real-time fault detection method for a new energy vehicle described above.

[0010] The technical scheme provided in the application realizes multi-dimensional feature representation of the running state of the electric control system of the new energy vehicle by collecting the voltage control signal of the battery management system, the current control signal and the three-phase current control signal of the motor control system, and calculating the working condition feature vector according to the control signals. The working condition feature vector integrates parameters reflecting different physical meanings such as battery power, vehicle speed, acceleration, motor torque and state of charge into a unified feature space representation, thereby providing a complete state information basis for subsequent working condition recognition and adaptive adjustment. A double-observer architecture of a fast-time-scale state observer and a slow-time-scale extended state observer is established, thereby realizing time scale decoupling of control performance tracking and health parameter estimation. The fast observer adopts a reduced-order linear model to focus on high-frequency tracking of control-related states, and has low computational complexity and does not affect the real-time performance of the control system. The slow observer adopts an extended state model to indirectly estimate the battery internal resistance parameter, the motor winding resistance parameter and the power device on-resistance parameter as extended states, which are difficult to directly measure. The health parameter change is calculated by measuring the residual error in the reverse direction. The method breaks through the limitation of traditional methods that can only monitor measurable electrical quantities, and realizes non-destructive online evaluation of the internal health state of electrical elements.

[0011] The Mahalanobis distance normalizes and decorrelates the feature space by introducing the covariance matrix, eliminates the non-uniform influence of different dimension parameters on distance calculation, accurately measures the real similarity between the current working condition and each typical working condition, and converts the distance measurement into a probability sense membership by using the Gaussian membership function, thereby realizing the transformation from hard classification to soft classification of the working condition class. The membership continuously and smoothly changes when the working condition is switched, thereby avoiding the mutation of the classification boundary. The weighted fusion algorithm performs convex combination on the observer gain matrix corresponding to each working condition according to the membership, and obtains an adaptive gain matrix. The observation weight of each health parameter is dynamically adjusted according to the working condition feature. In the large current working condition, the observation gain of the battery internal resistance is automatically increased to quickly track the change of the internal resistance. In the small current working condition, the observation gain is automatically reduced to suppress the measurement noise interference. The method solves the problems of estimation accuracy reduction and false alarm rate increase of the fixed gain observer in the multi-working condition environment. The observation residual error of the fast observer is used to realize millisecond-level rapid detection of sudden faults such as sensor failure and power device short circuit. The health parameter estimation value output by the slow observer is subjected to weighted least squares linear regression fitting to extract the degradation slope. In combination with the double condition criterion of the normalized distance, early warning of the gradual degradation fault of the electrical element is realized. The early warning caused by only relying on the degradation rate is avoided, and the late detection caused by only relying on the parameter threshold is also avoided. When the electrical element is in the accelerated degradation stage and is about to break through the health boundary, early warning is triggered in time, thereby providing a sufficient time window for predictive maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on these drawings.

[0013] Figure 1 An embodiment of the real-time fault detection method for new energy vehicles in the present application; Figure 2 An embodiment of the real-time fault detection method for new energy vehicles in the present application; Figure 3 An embodiment of the real-time fault detection method for new energy vehicles in the present application; DETAILED DESCRIPTION

[0014] The present application provides a real-time fault detection method and system for new energy vehicles. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0015] For the sake of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 An embodiment of the real-time fault detection method for new energy vehicles in the present application includes: Step S1: Collecting the voltage control signal, current control signal of the battery management system and three-phase current control signal of the motor control system, and calculating the working condition characteristic vector according to the voltage control signal, current control signal and control instruction; The collected voltage control signal and current control signal are from the control loop feedback of the electrical control system, the voltage control signal is collected by the voltage sensor at the positive and negative poles of the battery pack, the current control signal is collected by the Hall current sensor for the main loop current of the battery pack, and the three-phase current control signal is collected by the three-phase current sensor of the motor stator winding. The Clarke-Park coordinate transformation converts the current signal in the three-phase static coordinate system into the direct-axis component and the quadrature-axis component in the two-phase rotating coordinate system, and the quadrature-axis current component is proportional to the motor torque. The working condition characteristic vector combines the multi-dimensional parameters reflecting the power demand of the vehicle and the load state of the electrical system, forms a characteristic space representing the current operating condition, and the combination mode of each parameter under different working conditions is significantly different.

[0016] Step S2: establishing a fast time scale state observer according to the voltage control signal and the current control signal, and establishing a slow time scale observer taking the battery internal resistance parameter, the motor winding resistance parameter, and the power device on-state voltage drop parameter as extended states according to the voltage control signal and the current control signal; The fast time scale state observer adopts a reduced-order linear model, and the state transition matrix is derived from the linearization of the first-order RC equivalent circuit model of the battery and the d-q axis voltage equation of the permanent magnet synchronous motor, describing the dynamic evolution law of the battery voltage and current and the motor current in the control period. The slow time scale observer takes the health parameters as extended state variables, the battery internal resistance parameter reflects the trend of the increase of the ohmic internal resistance of the battery with aging, the motor winding resistance parameter reflects the resistance change of the stator winding due to temperature rise and insulation aging, and the power device on-state voltage drop parameter reflects the deterioration of the on-state characteristics of the IGBT power module due to the increase of junction temperature and chip degradation. The extended state observer inversely calculates the change of these parameters through the residual error between the measured value and the predicted value.

[0017] Step S3: calculating the Mahalanobis distance of the working condition characteristic vector and each preset working condition clustering center, substituting the Mahalanobis distance into the Gaussian membership function to calculate the membership degree of each working condition category, taking the membership degree as a weight coefficient to weight and sum the observer gain matrix corresponding to each working condition category to obtain an adaptive gain matrix, and using the adaptive gain matrix to correct the state of the slow time scale observer to obtain the estimated value of the battery internal resistance parameter, the estimated value of the motor winding resistance parameter, and the estimated value of the power device on-state voltage drop parameter; In the Mahalanobis distance calculation, the inverse matrix of the covariance matrix normalizes and decorrelates the feature space, eliminating the dimensional influence of the components of the working condition feature vector in the numerical scale. The fuzzy parameter in the Gaussian membership function controls the membership decay rate, realizing the smooth transition of the membership during the working condition switching. The weighted summation process continuously interpolates the observer gain matrix corresponding to multiple discrete working conditions, and the adaptive gain matrix dynamically adjusts the observation weight of each extended state variable of the slow observer according to the current working condition feature. The state correction process adds the product of the adaptive gain matrix and the measurement residual to the predicted state, and the measurement residual is the actual measured terminal voltage minus the predicted terminal voltage calculated according to the current state estimation and health parameter estimation, and the health parameter estimation value is updated through residual feedback.

[0018] Step S4: Determine the sudden failure of the electrical element using the observation residual of the fast time scale state observer, perform trend analysis on the battery internal resistance parameter estimation value, motor winding resistance parameter estimation value and power device on-state voltage drop parameter estimation value to determine the performance degradation failure of the electrical element, and generate a fault detection result.

[0019] Specifically, the observation residual generated by the fast time scale state observer includes voltage residual and current residual, and when the weighted two norm of the residual exceeds the dynamic threshold and lasts for multiple control periods, it is determined as a sudden failure such as sensor failure or power device short circuit. Trend analysis performs weighted least squares linear regression fitting on the sequence of health parameter estimation values, and the slope term in the regression coefficient represents the change rate of the health parameter with time, i.e. the degradation slope. The normalized distance is calculated by dividing the difference between the health parameter estimation value and the preset health boundary by the health range width. When the absolute value of the degradation slope exceeds the degradation rate threshold and the normalized distance is less than the safety margin, it is determined that the electrical element is in the performance accelerated degradation stage, and the gradual failure warning flag is triggered. The fault detection result includes the fault type code and the current value of the corresponding health parameter.

[0020] In a specific embodiment, step S1 includes: The total current of the battery pack is collected by the current sensor, and the terminal voltage of the battery pack is collected by the voltage sensor to obtain the current control signal and the voltage control signal. The three-phase current of the motor stator is collected by the three-phase current sensor to obtain the three-phase current control signal. The acceleration pedal opening, brake pedal opening and real-time vehicle speed are read, the battery power is calculated as the product of the terminal voltage of the battery pack and the total current of the battery pack, the vehicle acceleration is calculated as the difference between the current vehicle speed and the vehicle speed at the previous time divided by the control period, the Clarke-Park coordinate transformation is performed on the three-phase current control signal to obtain the quadrature axis current component, and the motor torque is calculated according to the quadrature axis current component, the number of motor pole pairs and the permanent magnet flux linkage. The battery power, real-time vehicle speed, vehicle acceleration, motor torque and state of charge are combined to form a working condition feature vector.

[0021] Specifically, the current control signal and the voltage control signal come from the control loop feedback of the battery management system, the current sensor measures the total current of the battery pack main loop in real time using the Hall effect principle, and the voltage sensor directly measures the terminal voltage between the positive and negative electrodes of the battery pack. The three-phase current control signal is collected by the current sensor installed on the stator winding A, B and C of the motor, which reflects the real-time control state of the motor controller. Clarke-Park coordinate transformation is a standard coordinate system conversion method in the field of motor control. Clarke transformation converts three-phase static coordinate system (A-B-C) to two-phase static coordinate system (α-β), and Park transformation further converts two-phase static coordinate system to two-phase rotating coordinate system (d-q), where d-axis is the same direction as rotor magnetic field, and q-axis is orthogonal to rotor magnetic field. The current component of the cross-axis directly produces electromagnetic torque, and the motor torque is calculated by multiplying the motor pole number by the permanent magnet flux linkage, which reflects the current output capacity and load state of the motor.

[0022] The working condition feature vector combines the five parameters of battery power, real-time vehicle speed, vehicle acceleration, motor torque and state of charge to form a point in a five-dimensional feature space, which uniquely represents the current operating condition of the new energy vehicle electrical control system. The battery power reflects the direction and intensity of energy flow in the electrical system, positive value indicates discharging condition and negative value indicates charging condition. Real-time vehicle speed and vehicle acceleration reflect the motion state of the vehicle, motor torque reflects the load demand of the drive system, and state of charge reflects the remaining capacity of the battery. Different combinations of these five parameters correspond to different typical working conditions, such as high power, high torque and low speed for climbing condition, low power, medium speed and zero acceleration for constant speed cruising condition, and negative power, negative torque and positive speed for braking energy recovery condition.

[0023] The current working condition feature vector is compared with the typical working condition cluster centers obtained by real vehicle data clustering in advance. Through Mahalanobis distance calculation and Gaussian membership function conversion, the membership degree of the current working condition to each typical working condition category is obtained. The membership degree reflects the similarity between the current electrical control system operating state and each typical working condition. Based on the membership degree, the observer gain matrix corresponding to each working condition is weighted and fused, so that the gain parameters of the slow time scale observer can be adjusted in real time according to the working condition. For example, in the large current discharging condition, the gain matrix will increase the observation weight of the battery internal resistance parameter, because the internal resistance change has more significant influence on the terminal voltage under large current, and in the small current condition, the gain matrix will reduce the internal resistance observation weight to avoid measurement noise interference, solving the problem of estimation accuracy decline and false alarm rate rise caused by model mismatch in fixed gain observer under multi-working condition environment.

[0024] In a specific embodiment, the Mahalanobis distance between the operating condition feature vector and each preset operating condition cluster center in step S3 comprises: reading the preset operating condition cluster center and the covariance matrix of each cluster; calculating the difference vector between the operating condition feature vector and the i-th operating condition cluster center; performing matrix multiplication operation on the difference vector and the inverse matrix of the covariance matrix of the i-th cluster, and then performing matrix multiplication operation on the transpose of the difference vector, and performing square root operation on the operation result to obtain the Mahalanobis distance between the operating condition feature vector and the i-th operating condition cluster center, which is used to identify the current operating condition of the electrical control system.

[0025] Specifically, the preset operating condition cluster center and the covariance matrix are obtained by K-means clustering analysis of real vehicle operating data in the offline stage. The cluster center is the mean of all sample operating condition feature vectors in this type of operating condition, and the covariance matrix describes the fluctuation range and correlation of each feature parameter in this type of operating condition. The difference vector is obtained by subtracting the corresponding elements of the i-th cluster center from each element of the current operating condition feature vector. This vector reflects the deviation of the current operating condition from the typical operating condition in the feature space. The inverse matrix of the covariance matrix performs weighted transformation on the difference vector, eliminating the influence of different dimensions and fluctuation amplitudes of each feature parameter. For example, when a new energy vehicle is running, the numerical range of the battery power is in the order of tens of thousands of watts, and the range of the vehicle speed is in the order of tens of kilometers per hour. If the Euclidean distance is directly calculated, the large numerical parameter will dominate the distance calculation result. The inverse matrix of the covariance matrix normalizes the contribution weight of each dimension feature to the distance.

[0026] The complete process of matrix multiplication operation is to first multiply the difference vector by the inverse matrix of the covariance matrix to obtain an intermediate vector. This intermediate vector is the representation of the difference vector in the normalized feature space after inverse transformation by the covariance matrix. Then, the intermediate vector is multiplied by the transpose of the difference vector to obtain a scalar, which represents the distance square in the normalized feature space. The square root of the scalar is obtained to obtain the Mahalanobis distance. This calculation process essentially transforms the Euclidean distance in the original feature space into a weighted distance considering the feature distribution characteristics through the covariance matrix. When a certain feature parameter fluctuates greatly in this type of operating condition, its contribution to the distance will be automatically de-weighted. When two feature parameters are strongly correlated, the Mahalanobis distance will automatically eliminate redundant information to avoid repeated calculation.

[0027] The core improvement of Mahalanobis distance compared with traditional Euclidean distance is to introduce the covariance matrix to adaptively correct the feature space, which is of great significance for the working condition recognition of new energy vehicle electrical control system, because the change characteristics of each electrical parameter are significantly different under different working conditions, for example, the battery power and motor torque change dramatically while the vehicle speed changes relatively flat under the condition of sudden acceleration, and the vehicle speed remains stable while the power and torque fluctuate slightly under the condition of constant speed cruising. If Euclidean distance is used, it cannot distinguish the difference in parameter distribution within the working condition, resulting in misjudgment of working condition recognition. Mahalanobis distance captures the inherent statistical characteristics of each working condition category through the covariance matrix, and can accurately identify the running working condition of the current electrical control system, solving the problem of low working condition recognition accuracy in multi-working condition environment by using fixed threshold or Euclidean distance method.

[0028] Figure 2 A schematic diagram of dynamic changes of Mahalanobis distance between the working condition feature vector and the preset working condition cluster center in the embodiment of the present application. Figure 2 The figure shows the change curves of Mahalanobis distance between the current working condition feature vector and the sudden acceleration working condition cluster center and the constant speed cruising working condition cluster center with time during the working condition switching process of the new energy vehicle. In the time 0-10s stage, the vehicle is in the constant speed cruising working condition, at this time the constant speed cruising working condition Mahalanobis distance remains at a low level of about 2-2.5, while the sudden acceleration working condition Mahalanobis distance is higher, about 8-10, indicating that the current working condition is highly matched with the constant speed cruising working condition. In the time 10-25s stage, the vehicle gradually transits from the constant speed cruising working condition to the sudden acceleration working condition, the sudden acceleration working condition Mahalanobis distance rapidly decreases from about 8 to about 3, while the constant speed cruising working condition Mahalanobis distance rises from about 2 to about 6, the two curves intersect at time about 13s, indicating the critical point of working condition switching. After time 25s, the vehicle enters the stable sudden acceleration working condition, the sudden acceleration working condition Mahalanobis distance stabilizes at 5-6, and the constant speed cruising working condition Mahalanobis distance stabilizes at 2.5-3.5. The figure directly shows that the Mahalanobis distance calculation method can accurately identify the running working condition of the current electrical control system, and provide continuous and smooth distance measurement during working condition switching.

[0029] In a specific embodiment, the Mahalanobis distance is substituted into the Gaussian membership function in step S3 to calculate the membership degree of each working condition category, including: The Mahalanobis distance is squared to obtain the Mahalanobis distance square value, the Mahalanobis distance square value is divided by twice the fuzzification parameter, the result is taken as a negative value and then exponentiated to obtain the exponential term of the i-th working condition category; The exponential terms of all working condition categories are summed to obtain the total sum of the exponential terms; The exponential term of the i-th working condition category is divided by the total sum of the exponential terms to obtain the membership degree of the i-th working condition category, which is used to quantify the matching degree of the electrical control system with each typical running working condition.

[0030] Specifically, the Mahalanobis distance square value is obtained by squaring the Mahalanobis distance calculated in the previous step. The squaring operation amplifies the distance difference, making the distinction between similar and dissimilar working conditions more obvious. The fuzzification parameter controls the decay rate of the Gaussian membership function. The parameter determines how quickly the membership decreases as the Mahalanobis distance increases. The larger the parameter value, the smoother the membership function, and the smoother the transition between working conditions. The smaller the parameter value, the steeper the membership function, and the more responsive it is to changes in working conditions. Dividing the Mahalanobis distance square value by twice the fuzzification parameter and taking the negative value forms the exponential part of the Gaussian function. The negative value ensures that the exponential operation result is between zero and one. When the Mahalanobis distance is zero, the exponential term takes the maximum value of one, indicating a perfect match to the working condition. As the Mahalanobis distance increases, the exponential term gradually decays to zero, indicating a decrease in similarity to the working condition. The exponential operation uses the natural exponential function, converting the linear distance metric to a nonlinear similarity metric. This nonlinear mapping conforms to human cognitive laws of similarity, which decreases exponentially rather than linearly with increasing distance.

[0031] The sum of the exponential terms for all working condition categories is obtained, which is used as a normalization factor to ensure that the sum of the memberships of all working condition categories is equal to one, satisfying the normalization condition in probability theory. Dividing the exponential term of the i-th working condition category by the sum of the exponential terms gives the membership of that working condition. The membership value ranges from zero to one. The closer the value is to one, the more similar the current electrical control system operating state is to the typical working condition. The closer the value is to zero, the lower the similarity. This membership calculation method based on the Gaussian function has smoothing properties. When the vehicle transitions from one working condition to another, the membership of each working condition category changes continuously rather than abruptly, avoiding the boundary jump problem in hard classification methods.

[0032] The membership quantifies the matching degree of the electrical control system to each typical operating working condition. This quantification result is directly used in the subsequent observer gain matrix weighted fusion. The higher the membership of a working condition category, the greater the weight of its corresponding gain matrix in the fusion process, making the final adaptive gain matrix closer to the optimal gain configuration of the current actual working condition. For example, when the vehicle is in a transition state between rapid acceleration and climbing working conditions, the membership of the rapid acceleration working condition is 0.6, the membership of the climbing working condition is 0.3, and the sum of the memberships of other working conditions is 0.1. The adaptive gain matrix will mainly be weighted and synthesized by the gain matrices of the rapid acceleration and climbing working conditions in a ratio of 6 to 3. The soft switching mechanism can smoothly adapt to changes in working conditions, avoiding oscillations in the state estimation of the slow time scale observer caused by sudden changes in the gain matrix, and solving the problem of increased estimation error and rising false alarm rate caused by gain mismatch in traditional fixed gain observers during working condition switching.

[0033] Figure 3Fig. 2 is a schematic diagram of dynamic changes of membership degrees of each working condition category calculated based on the Gaussian membership function in the embodiments of the present application. Figure 3 Fig. 2 shows the curves of the membership degrees of the rapid acceleration working condition and the constant speed cruising working condition calculated by the Gaussian membership function during the working condition switching process of the new energy vehicle. In the time period of 0-10 seconds, the vehicle is in the constant speed cruising working condition, the membership degree of the constant speed cruising working condition is kept between 0.75 and 0.8, and the membership degree of the rapid acceleration working condition is kept between 0.15 and 0.2, and the sum of the two is about 1, which meets the membership degree normalization condition. In the time period of 10-25 seconds, the vehicle gradually transits from the constant speed cruising working condition to the rapid acceleration working condition, the membership degree of the rapid acceleration working condition smoothly rises from about 0.15 to about 0.85, and the membership degree of the constant speed cruising working condition smoothly drops from about 0.75 to about 0.12, and the two curves intersect at about 0.5 at the time of about 17 seconds, which indicates the midpoint of the working condition switching. After the time of 25 seconds, the vehicle stably enters the rapid acceleration working condition, the membership degree of the rapid acceleration working condition is stably kept between 0.85 and 0.9, and the membership degree of the constant speed cruising working condition is stably kept at about 0.12. The figure directly shows that, after the Gaussian membership function converts the Mahalanobis distance into the membership degree in the sense of probability, the membership degree presents a continuous and smooth change instead of a sudden change during the working condition switching, avoids the boundary jumping problem of the hard classification method, provides smooth weight coefficients for the subsequent weighted fusion of the observer gain matrix, and realizes the dynamic optimization of the adaptive gain matrix according to the working condition characteristics.

[0034] In a specific embodiment, the adaptive gain matrix is obtained by weighting and summing the observer gain matrices corresponding to each working condition category with the membership degrees as the weight coefficients in step S3, which includes: performing a matrix multiplication operation on the membership degree of the i-th working condition category and the observer gain matrix corresponding to the i-th working condition category to obtain the i-th weighted gain matrix; performing a matrix addition operation on the weighted gain matrices of all working condition categories to obtain the adaptive gain matrix, which is used to adjust the observation weight of the slow time scale observer on the electrical element health parameter.

[0035] Specifically, the observer gain matrix is calibrated separately for each operating condition category in the offline stage. The calibration process injects the measured data of the operating condition as excitation in the simulation environment to minimize the weighted root mean square value of the observation error as the optimization objective, and uses the particle swarm optimization algorithm to search for the optimal gain matrix parameters under the operating condition. The gain matrix parameters corresponding to different operating condition categories differ significantly. For example, the battery internal resistance has a greater impact on the terminal voltage under the rapid acceleration operating condition, so the gain element corresponding to the battery internal resistance in the gain matrix of this operating condition has a larger value. Under the constant speed cruising operating condition, the current is small and the internal resistance has a relatively weak impact, so the gain element has a smaller value. Matrix multiplication operation multiplies the membership scalar with each element of the gain matrix to obtain a weighted gain matrix that retains the structure of the original gain matrix but all elements are scaled by the membership. The higher the membership of the operating condition, the larger the value of the elements of the weighted gain matrix, and the more significant the contribution to the final fusion result.

[0036] The matrix addition operation adds the corresponding elements of the weighted gain matrices of all operating condition categories one by one. Since the sum of all memberships is equal to one, the adaptive gain matrix obtained after weighted summation is a convex combination of the optimal gain matrices of each operating condition, which mathematically guarantees that it is located in the convex space spanned by the optimal gain matrices of each operating condition. The adaptive gain matrix dynamically adjusts according to the current operating condition characteristics. When the vehicle switches from the constant speed operating condition to the rapid acceleration operating condition, the membership of the rapid acceleration operating condition gradually increases and the membership of the constant speed operating condition gradually decreases. The adaptive gain matrix smoothly transitions to the optimal gain configuration of the rapid acceleration operating condition. This continuous change avoids sudden changes in the gain matrix, which can cause jumps in the state estimation of the observer.

[0037] The adaptive gain matrix is used to adjust the observation weight of the slow time scale observer on the health parameters of the electrical elements. The gain element in the gain matrix for the battery internal resistance parameter determines the strength of the observer in correcting the internal resistance estimate value based on the terminal voltage residual error. The larger the gain element value, the faster the tracking response of the observer to the parameter but the weaker the noise immunity. The smaller the gain element value, the slower the tracking response but the smoother the estimation result. The adaptive gain matrix automatically balances this trade-off relationship according to the operating condition. In the large current operating condition, the internal resistance observation gain is increased to quickly capture the internal resistance change. In the small current operating condition, the internal resistance observation gain is reduced to suppress the measurement noise interference, thereby achieving dynamic optimization of the health parameter estimation accuracy and robustness under different operating conditions and solving the contradiction that the traditional fixed gain observer cannot simultaneously consider the performance of multiple operating conditions.

[0038] In a specific embodiment, the fast time scale state observer is established in step S2, including: Defining the fast state vector to include the battery pack terminal voltage, the battery pack total current, the motor current direct axis component, and the motor current quadrature axis component; The state transition matrix and the control input matrix of the fast state vector are established according to the battery equivalent circuit model and the motor d-q axis mathematical model, and the discrete state equation of the fast time scale state observer is constructed. The measurement matrix is established according to the voltage control signal and the current control signal, and the measurement equation of the fast time scale state observer is constructed. The observer gain matrix is designed by the pole placement method, and the update equation of the fast state vector estimate is established to track the real-time running state of the electrical control system.

[0039] Specifically, the four state variables contained in the fast state vector are all key electrical quantities that directly affect the control performance in the electrical control system. The battery pack terminal voltage reflects the output capability of the battery, the total current of the battery pack reflects the energy flow intensity, the d-axis component of the motor current controls the magnetic field strength, and the q-axis component of the motor current generates the electromagnetic torque. These four variables constitute the core state space of the electrical control of the new energy vehicle power system. The state transition matrix describes the dynamic evolution law of these state variables in a control period. The battery equivalent circuit model uses a first-order RC model to express the battery terminal voltage as the open-circuit voltage minus the ohmic internal resistance voltage drop and then minus the polarization internal resistance voltage drop. The motor d-q axis mathematical model expresses the stator voltage equation in the rotating coordinate system as the algebraic sum of the resistance voltage drop, the inductance voltage drop and the back electromotive force. By linearizing these continuous time differential equations near the operating point and discretizing them by the Euler method or the zero-order hold method, the state transition matrix of the discrete state equation is obtained. The matrix elements contain the functions of the battery internal resistance, the battery polarization resistance, the inductance and the sampling period. The control input matrix describes the driving effect of the d-axis voltage command and the q-axis voltage command output by the motor controller on the state variables. This matrix relates the control input to the rate of change of the state variables, and constitutes a complete discrete state equation.

[0040] The measurement matrix establishes the mapping relationship between the state variables and the sensor measurement values. Since the battery pack terminal voltage and the total current of the battery pack can be directly measured, the identity matrix form is used for the corresponding state variables in the measurement matrix. The d-q axis components of the motor current cannot be directly measured and need to be calculated through three-phase current Clarke-Park transformation, and the measurement equation maps the state vector into the measurement value vector through the measurement matrix. The measurement noise term is introduced in the measurement equation to represent the random error of the sensor. The noise term is assumed to be zero-mean Gaussian white noise, and the covariance matrix is obtained by offline statistical sensor data fluctuation characteristics. The measurement equation of the fast time scale state observer is a linear equation, so that the observer can be realized in the Luenberger observer framework with high computational efficiency.

[0041] The pole placement method designs the gain matrix by selecting the position of the observer eigenvalues, which determine the convergence speed and dynamic response characteristics of the observation error. The smaller the modulus of the eigenvalue, the faster the convergence of the observation error, but the more sensitive it is to measurement noise. The larger the modulus of the eigenvalue, the slower the convergence of the observation error, but the stronger the robustness. When configuring the poles, the observer eigenvalues are configured inside the unit circle and the modulus is smaller than the modulus of the state transition matrix eigenvalues, ensuring that the observer converges faster than the system dynamic change speed, while avoiding noise amplification caused by too small eigenvalues. The update equation of the fast state vector estimate uses a prediction-correction structure. First, the predicted state is calculated based on the state transition matrix and control input. Then, the final estimate is obtained by correcting the predicted state using the gain matrix based on the residual between the measured value and the predicted measured value. This update equation is executed once every control period, tracking the real-time changes of the state variables of the electrical control system and providing accurate state feedback to the controller. At the same time, it generates observation residuals for fault detection. When the observation residual exceeds the threshold value, it indicates that the actual measurement value deviates from the model predicted value, indicating that the electrical element has an abnormality.

[0042] In a specific embodiment, the trend analysis of the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-state voltage drop parameter estimate in step S4 determines the performance degradation fault of the electrical element, including: A sliding time window is established to store the historical sequence of the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-state voltage drop parameter estimate; A weighted least squares linear regression fitting is performed on the historical sequence. By constructing a design matrix and a measurement vector, the regression coefficient vector is solved, and the degradation slope representing the performance degradation rate of the electrical element is extracted; The normalized distance between the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-state voltage drop parameter estimate and the pre-set health boundary is calculated; It is determined whether the absolute value of the degradation slope exceeds the degradation rate threshold and whether the normalized distance is less than the safety margin. When both conditions are met, the gradual fault warning flag of the performance degradation fault of the electrical element is triggered.

[0043] Specifically, the sliding time window stores the health parameter estimates output by the slow-time-scale observer in a first-in-first-out queue structure. The window length is set to a fixed number of observation periods. Each time the slow observer completes a state update, the latest health parameter estimate enters the window, and the oldest estimate is removed from the window. The window always maintains a fixed number of historical data points. The historical sequence records the evolution trajectory of the battery internal resistance parameter, the motor winding resistance parameter, and the power device on-state voltage drop parameter over a period of time. These parameters show a slow upward or downward trend due to the aging and performance degradation of the electrical element. By analyzing the trend of the historical sequence, it can be determined whether the electrical element is in an accelerated performance degradation phase.

[0044] The weighted least squares linear regression fitting approximates the historical sequence as a linear function, and the slope of the linear function reflects the rate of change of the parameter over time, i.e. the degradation rate. The first column of the design matrix is all 1s, corresponding to the intercept term of the linear function, and the second column is the time index, corresponding to the slope term of the linear function. Each row of the design matrix is also multiplied by the weight coefficient at the corresponding time, and the weight coefficient increases over time so that the influence of recent data on the regression result is greater than that of distant data, thereby improving the sensitivity of trend analysis to parameter changes. The measurement vector is composed of the sequence of health parameter estimate values within the window, and each estimate value is also multiplied by the corresponding weight coefficient. The regression coefficient vector is solved by the normal equation, which multiplies the transpose of the design matrix with the design matrix to obtain the coefficient matrix, multiplies the transpose of the design matrix with the measurement vector to obtain the constant vector, and multiplies the inverse matrix of the coefficient matrix with the constant vector to obtain the regression coefficient vector. The second element of the regression coefficient vector is the degradation slope. A positive degradation slope indicates that the health parameter increases over time, such as an increase in battery internal resistance, and a negative degradation slope indicates that the health parameter decreases over time, such as a decrease in power device on-state voltage drop. The larger the absolute value of the slope, the faster the performance degradation.

[0045] The preset health boundary is determined according to the design life and safety margin of the electrical element. The upper limit of the health boundary of the battery internal resistance is a certain multiple of the factory nominal value, the upper limit of the health boundary of the motor winding resistance is set based on the insulation class and temperature rise limit, and the health boundary of the power device on-state voltage drop is set according to the maximum rated value in the chip manual. The normalized distance calculates the difference between the current health parameter estimate value and the health boundary, and then divides it by the normal variation range of the health parameter, so that the distances of health parameters with different dimensions are comparable. The normalized distance value is between 0 and 1, and the smaller the value, the closer the health parameter is to the boundary and enters the risk area.

[0046] The degradation rate threshold is set as the typical change rate of the health parameter in the normal aging process. When the absolute value of the degradation slope exceeds the threshold, it indicates that the electrical element is experiencing abnormal accelerated degradation rather than normal slow aging. The safety margin is set as the critical value of the normalized distance. When the normalized distance is less than the margin, it indicates that the health parameter has entered the warning interval near the health boundary. The double-condition criterion requires that the degradation slope exceeds the threshold and the normalized distance is less than the margin at the same time. This design avoids the limitations of single-condition criteria, and does not trigger a warning when only the degradation rate is fast but the parameter value is not close to the boundary, avoiding premature alarms. It also does not trigger a warning when only the parameter value is close to the boundary but the degradation rate is slow, avoiding false positives for normal aging. Only when the electrical element is in an accelerated degradation state and is about to break through the health boundary, the gradual failure warning flag is triggered, achieving precise early identification of gradual failures such as battery aging, motor winding degradation, and power device performance attenuation, and providing a predictive maintenance time window for vehicle operators.

[0047] The new energy vehicle real-time fault detection method in the embodiment of the application is described above, and the new energy vehicle real-time fault detection system in the embodiment of the application is described below, and one embodiment of the new energy vehicle real-time fault detection system in the embodiment of the application comprises: The acquisition module is configured to acquire a voltage control signal and a current control signal of a battery management system and a three-phase current control signal of a motor control system, and to calculate a working condition feature vector according to the voltage control signal, the current control signal, and a control instruction; The generation module is configured to establish a fast-time-scale state observer according to the voltage control signal and the current control signal, and to establish a slow-time-scale observer taking a battery internal resistance parameter, a motor winding resistance parameter, and a power device on-resistance parameter as extended states according to the voltage control signal and the current control signal; The calculation module is configured to calculate Mahalanobis distances between the working condition feature vector and preset working condition clustering centers, to calculate a membership degree of each working condition category by substituting the Mahalanobis distances into a Gaussian membership function, to perform weighted summation on observer gain matrices corresponding to each working condition category by taking the membership degree as a weight coefficient to obtain an adaptive gain matrix, and to perform state correction on the slow-time-scale observer by using the adaptive gain matrix to obtain a battery internal resistance parameter estimate value, a motor winding resistance parameter estimate value, and a power device on-resistance parameter estimate value; The determination module is configured to determine an electrical element sudden fault by using an observation residual of the fast-time-scale state observer, to determine an electrical element performance degradation fault by performing trend analysis on the battery internal resistance parameter estimate value, the motor winding resistance parameter estimate value, and the power device on-resistance parameter estimate value, and to generate a fault detection result.

[0048] The application further provides a computer readable storage medium, which can be a nonvolatile computer readable storage medium or a volatile computer readable storage medium, and the computer readable storage medium stores instructions, and the instructions enable a computer to perform the steps of the new energy vehicle real-time fault detection method when the instructions are run on the computer.

[0049] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system, and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein.

[0050] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for enabling a new energy vehicle real-time fault detection device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (read-only memory, ROM), a random access memory (random access memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0051] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A new energy vehicle real-time fault detection method, characterized in that, The method comprises: Step S1: collecting a voltage control signal, a current control signal of a battery management system and a three-phase current control signal of a motor control system, and calculating a working condition characteristic vector according to the voltage control signal, the current control signal and a control instruction; Step S2: establishing a fast-time-scale state observer according to the voltage control signal and the current control signal, and establishing a slow-time-scale observer taking a battery internal resistance parameter, a motor winding resistance parameter and a power device on-resistance parameter as extended states according to the voltage control signal and the current control signal; Step S3: calculating Mahalanobis distances between the working condition characteristic vector and preset working condition clustering centers, substituting the Mahalanobis distances into a Gaussian membership function to calculate a membership degree of each working condition category, taking the membership degree as a weight coefficient to perform weighted summation on observer gain matrices corresponding to each working condition category to obtain an adaptive gain matrix, and performing state correction on the slow-time-scale observer by using the adaptive gain matrix to obtain a battery internal resistance parameter estimation value, a motor winding resistance parameter estimation value and a power device on-resistance parameter estimation value; Step S4: determining an electrical element sudden failure by using an observation residual of the fast-time-scale state observer, determining an electrical element performance degradation failure by performing trend analysis on the battery internal resistance parameter estimation value, the motor winding resistance parameter estimation value and the power device on-resistance parameter estimation value, and generating a fault detection result.

2. The real-time fault detection method for new energy vehicles according to claim 1, characterized in that, The step S1 comprises: collecting a total current of a battery pack by using a current sensor and collecting an end voltage of the battery pack by using a voltage sensor to obtain the current control signal and the voltage control signal; collecting three-phase stator currents of a motor by using three-phase current sensors to obtain the three-phase current control signal; reading an accelerator pedal opening degree, a brake pedal opening degree and a real-time vehicle speed, calculating a battery power as a product of the end voltage of the battery pack and the total current of the battery pack, calculating a vehicle acceleration as a difference between a current vehicle speed and a previous vehicle speed divided by a control period, performing Clarke-Park coordinate transformation on the three-phase current control signal to obtain a quadrature-axis current component, and calculating a motor torque according to the quadrature-axis current component, a motor pole pair number and a permanent magnet flux linkage; constructing the battery power, the real-time vehicle speed, the vehicle acceleration, the motor torque and a state of charge as the working condition characteristic vector.

3. The real-time fault detection method for new energy vehicles according to claim 1, characterized in that, The step S3 of calculating the Mahalanobis distances between the working condition characteristic vector and the preset working condition clustering centers comprises: reading preset working condition clustering centers and covariance matrices of respective clusters; calculating a difference vector between the working condition characteristic vector and an i-th working condition clustering center; performing matrix multiplication operation on the difference vector and an inverse matrix of the covariance matrix of the i-th cluster, and then performing matrix multiplication operation on the difference vector and a transpose of the difference vector, performing square root operation on an operation result, and obtaining the Mahalanobis distance between the working condition characteristic vector and the i-th working condition clustering center, which is used to identify a running working condition currently taken by the electrical control system.

4. The real-time fault detection method for new energy vehicles according to claim 3, characterized in that, The step S3 of substituting the Mahalanobis distances into the Gaussian membership function to calculate the membership degree of each working condition category comprises: The Mahalanobis distance is squared to obtain a Mahalanobis distance square value, the Mahalanobis distance square value is divided by twice the fuzzification parameter, a negative value is taken of the result, and an exponential operation is performed to obtain an exponential term of the i-th operating condition category; The exponential terms of all operating condition categories are summed to obtain an exponential term sum; The exponential term of the i-th operating condition category is divided by the exponential term sum to obtain a membership degree of the i-th operating condition category, which is used to quantify the matching degree of the electrical control system and each typical operating condition.

5. The real-time fault detection method for new energy vehicles according to claim 4, characterized in that, The membership degree in the step S3 is used as a weight coefficient to perform weighted summation on the observer gain matrix corresponding to each operating condition category to obtain an adaptive gain matrix, including: The membership degree of the i-th operating condition category is multiplied by the observer gain matrix corresponding to the i-th operating condition category to obtain an i-th weighted gain matrix; The weighted gain matrices of all operating condition categories are subjected to matrix addition operation to obtain the adaptive gain matrix, which is used to adjust the observation weight of the slow time scale observer on the electrical element health parameter.

6. The real-time fault detection method for new energy vehicles according to claim 1, characterized in that, The fast time scale state observer in the step S2 includes: The fast state vector is defined to include the battery pack terminal voltage, the battery pack total current, the motor current direct axis component, and the motor current quadrature axis component; The state transition matrix and the control input matrix of the fast state vector are established according to the battery equivalent circuit model and the motor d-q axis mathematical model to construct the discrete state equation of the fast time scale state observer; The measurement matrix is established according to the voltage control signal and the current control signal to construct the measurement equation of the fast time scale state observer; The observer gain matrix is designed by the pole placement method to establish the update equation of the fast state vector estimate, which is used to track the real-time operating state of the electrical control system.

7. The real-time fault detection method for new energy vehicles according to claim 1, characterized in that, The trend analysis of the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-state voltage drop parameter estimate is performed to determine the electrical element performance degradation fault in the step S4, including: A sliding time window is established to store the historical sequence of the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-state voltage drop parameter estimate; The historical sequence is subjected to weighted least squares linear regression fitting, a design matrix and a measurement vector are constructed, a regression coefficient vector is solved, and a degradation slope representing the electrical element performance degradation rate is extracted; The normalized distance of the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-state voltage drop parameter estimate from a preset health boundary is calculated; It is determined whether the absolute value of the degradation slope exceeds a degradation rate threshold and whether the normalized distance is less than a safety margin, and when both conditions are met, a gradual fault warning flag of the electrical element performance degradation fault is triggered.

8. A new energy vehicle real-time fault detection system, characterized in that, The new energy vehicle real-time fault detection system for implementing the new energy vehicle real-time fault detection method in any one of claims 1-7 includes: An acquisition module is configured to acquire a voltage control signal, a current control signal of a battery management system, and a three-phase current control signal of a motor control system, and calculate an operating condition feature vector according to the voltage control signal, the current control signal, and a control instruction; The generating module is configured to establish a fast-time-scale state observer according to the voltage control signal and the current control signal, and establish a slow-time-scale observer taking the battery internal resistance parameter, the motor winding resistance parameter, and the power device on-resistance parameter as extended states according to the voltage control signal and the current control signal; The calculating module is configured to calculate Mahalanobis distances between the working condition characteristic vector and each preset working condition clustering center, calculate the membership degrees of each working condition category by substituting the Mahalanobis distances into a Gaussian membership function, weight and sum the observer gain matrices corresponding to each working condition category by taking the membership degrees as weight coefficients to obtain an adaptive gain matrix, and perform state correction on the slow-time-scale observer by using the adaptive gain matrix to obtain the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-resistance parameter estimate; The determining module is configured to determine an electrical element sudden failure by using an observation residual of the fast-time-scale state observer, determine an electrical element performance degradation failure by performing trend analysis on the battery internal resistance parameter estimate, the motor winding resistance parameter estimate, and the power device on-resistance parameter estimate, and generate a fault detection result.

9. A new energy vehicle real-time fault detection device, characterized in that, The computer program is run on the processor to implement the real-time fault detection method for new energy vehicles according to any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is run on the processor to implement the real-time fault detection method for new energy vehicles according to any one of claims 1 to 7.

Images