Multi-module Collaborative Fault Detection and Warning Method for Hydrogen Electric Vehicles

Abstract

The present invention discloses a multi-module collaborative fault detection and warning method for hydrogen-electric electric vehicles, which relates to the technical field of electric vehicle fault detection. The method includes: determining multiple functional modules of the hydrogen-electric electric vehicle; collecting the multiple functional modules through an in-vehicle sensing network to obtain multiple groups of sensing data sets; performing pattern recognition on the multiple groups of sensing data sets based on a pattern classification algorithm to output a real-time matching working mode; constructing a module cooperation relationship database; calling the module cooperation relationship database to identify a module cooperation relationship diagram corresponding to the real-time matching working mode; and performing collaborative fault detection on the multiple groups of sensing data sets according to the module cooperation relationship diagram to output a fault warning signal corresponding to the real-time matching working mode. It solves the technical problem in the prior art that there is a lack of a comprehensive fault detection means for the collaborative relationship of multiple functional modules, resulting in untimely and incomplete identification of electric vehicle faults, and achieves the technical effect of improving the comprehensiveness and accuracy of fault identification.

Description

Technical Field

[0001] This application relates to the technical field of electric vehicle fault detection, and particularly to a multi-module collaborative fault detection and warning method for hydrogen-electric vehicles. Background Art

[0002] With the rapid development of hydrogen fuel cell and electric vehicle technologies, hydrogen-electric vehicles have been increasingly widely used in transportation due to their high efficiency and environmental friendliness. However, hydrogen-electric vehicles contain multiple complex functional modules, such as power systems, energy storage systems, fuel cell systems, intelligent control systems, etc. These functional modules are tightly integrated and work together through an in-vehicle sensing network. To ensure the safety and reliability of vehicle operation, how to detect vehicle faults in real-time and collaboratively has become a key issue. Traditional methods mainly independently monitor a single functional module, usually collecting module operation data based on the in-vehicle sensing network. However, due to the large variety of data, the accuracy of pattern classification and fault identification is relatively low, and it is impossible to comprehensively evaluate the collaborative relationship between modules, thus affecting the vehicle's working mode. For example, the collaborative efficiency between the fuel cell system and the energy storage system determines vehicle energy management, and it is difficult to handle the complex working conditions with high integration between modules, making it impossible to conduct comprehensive and accurate fault detection.

[0003] In the current related technologies, there is a lack of comprehensive fault detection means for the collaborative relationship of multiple functional modules, resulting in the technical problems of untimely and incomplete fault identification of electric vehicles. Summary of the Invention

[0004] This application provides a multi-module collaborative fault detection and warning method for hydrogen-electric vehicles, which solves the technical problems in the prior art of lacking comprehensive fault detection means for the collaborative relationship of multiple functional modules, resulting in untimely and incomplete fault identification of electric vehicles, and achieves the technical effect of improving the comprehensiveness and accuracy of fault identification.

[0005] This application provides a multi-module collaborative fault detection and warning method for hydrogen-electric vehicles, including: determining multiple functional modules of the hydrogen-electric vehicle, where the multiple functional modules are connected to the in-vehicle sensing network; collecting the multiple functional modules through the in-vehicle sensing network to obtain multiple groups of sensing data sets; performing pattern recognition on the multiple groups of sensing data sets based on a pattern classification algorithm to output a real-time matching working mode; constructing a module cooperation relationship database, where the module cooperation relationship database stores the cooperation relationships of the multiple functional modules under each working mode; calling the module cooperation relationship database to identify the module cooperation relationship diagram corresponding to the real-time matching working mode; performing collaborative fault detection on the multiple groups of sensing data sets according to the module cooperation relationship diagram, and outputting a fault warning signal corresponding to the real-time matching working mode.

[0006] In a possible implementation, the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle further performs the following processes: extracting features from the multiple groups of sensing data sets to output a sensing feature vector; inputting the sensing feature vector into a trained random forest classification model, performing pattern recognition according to the random forest classification model, and obtaining a real-time matching working mode output by the random forest classification model; training the random forest classification model includes training multiple decision trees to convergence according to training data, where the training data includes obtaining a historical sensing sample data set and label data identifying the mode type of the hydrogen-electric vehicle.

[0007] In a possible implementation, the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle further performs the following processes: analyzing the multiple groups of sensing data sets using a sliding window to obtain a pattern duration; obtaining a stability index according to the difference between the pattern duration and a preset duration threshold; if the stability index is greater than or equal to a preset stability index, performing pattern recognition on the multiple groups of sensing data sets based on a pattern classification algorithm.

[0008] In a possible implementation, the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle further performs the following processes: determining a first switching working mode according to the multiple mode switching probabilities, where the first switching working mode is the working mode with the highest probability among the multiple mode switching probabilities; calling the module cooperation relationship database to identify the module cooperation relationship diagram corresponding to the first switching working mode; performing collaborative fault detection on the multiple groups of continuously monitored sensing data sets according to the module cooperation relationship diagram.

[0009] In a possible implementation, the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle further performs the following processes: identifying the healthy sensing data sample corresponding to the module cooperation relationship diagram; comparing the multiple groups of sensing data sets according to the healthy sensing data sample to determine a first faulty functional module; obtaining a first fault index and multiple collaborative fault indices corresponding to the first faulty functional module; performing weighted summation according to the first fault index and the multiple collaborative fault indices to output a fault level, and converting the fault level to output a fault warning signal.

[0010] In a possible implementation, the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle further performs the following processes: combining the module cooperation relationship diagram, and outputting a fault propagation probability matrix according to the connection relationship between the first faulty functional module and other functional modules; calculating multiple collaborative fault indices according to the fault propagation probability matrix.

[0011] In a possible implementation, the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle further performs the following processing: obtaining the mode switching combinations of the real-time matching working mode according to the module cooperation relationship database; identifying multiple switching risks corresponding to the mode switching combinations according to the fault warning signal; screening out the marked mode switching combinations with switching risks greater than the preset switching risk in the multiple switching risks; and generating a switching warning signal when the received mode switching request belongs to the marked mode switching combinations.

[0012] In a possible implementation, the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle further performs the following processing: detecting the repair index of the real-time matching working mode according to the switching warning signal; if the repair index does not meet the preset repair index, performing mode switching optimization in the module cooperation relationship database with the preset switching risk, and outputting an optimal switching mode solution.

[0013] It is intended to determine multiple functional modules of the hydrogen-electric vehicle through the multi-module collaborative fault detection and warning method proposed in this application; collect multiple functional modules through an in-vehicle sensor network to obtain multiple groups of sensor data sets; perform pattern recognition on the multiple groups of sensor data sets based on a pattern classification algorithm, and output a real-time matching working mode; construct a module cooperation relationship database; call the module cooperation relationship database to identify the module cooperation relationship diagram corresponding to the real-time matching working mode; perform collaborative fault detection on the multiple groups of sensor data sets according to the module cooperation relationship diagram, and output a fault warning signal corresponding to the real-time matching working mode. This solves the technical problem in the prior art of lacking a comprehensive fault detection means for the collaborative relationship of multiple functional modules, resulting in untimely and incomplete fault identification of electric vehicles, and achieves the technical effect of improving the comprehensiveness and accuracy of fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments of the present invention will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the operations in the front or below do not necessarily need to be executed precisely in sequence. On the contrary, according to needs, they can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0015] Figure 1 It is a schematic flowchart of the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle provided by the embodiment of the present application;

[0016] Figure 2 It is a schematic flowchart of outputting a real-time matching working mode in the multi-module collaborative fault detection and warning method for the hydrogen-electric vehicle provided by the embodiment of the present application. Detailed Implementation Modes

[0017] The above description is only an overview of the technical solution of this application. In order to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. In order to make the above and other purposes, features, and advantages of this application more obvious and understandable, the specific implementation modes of this application are specifically given below.

[0018] In order to make the purpose, technical solution, and advantages of this application clearer, the following will further describe this application in detail with reference to the accompanying drawings. The described embodiments should not be regarded as limitations of this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

[0019] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The terms "first\second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, product, or server including a series of steps does not necessarily have to be limited to those steps clearly listed, but may include other steps not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application.

[0020] The embodiment of this application provides a multi-module collaborative fault detection and early warning method for hydrogen-electric vehicles, as Figure 1 shown, the method includes:

[0021] Step S100, determine multiple functional modules of the hydrogen-electric vehicle, wherein the multiple functional modules are connected to the vehicle-mounted sensing network.

[0022] Preferably, multiple functional modules of the hydrogen-electric vehicle are determined, that is, multiple system modules with specific functions in the hydrogen-electric vehicle that perform data exchange and communication through the in-vehicle sensing network are obtained. The multiple functional modules are all connected to the in-vehicle sensing network through various sensors, data buses, and communication interfaces to form an integrated and interconnected network system, enabling each module to collect sensing data in real time and transmit it to the in-vehicle control system or other modules. For example, through the in-vehicle sensing network, the fuel cell system module can transmit its status information to the control system, and the control system can in turn adjust the operating modes of the motor and the energy storage system, thus achieving collaborative work. The functional modules may include, but are not limited to, the power system module, including the motor, control unit, etc., which is related to the driving ability and motion performance of the vehicle; the fuel cell system module, which is responsible for the storage of hydrogen, conversion into electrical energy, and supply to the motor and other modules for use; the energy storage system module, such as a battery pack, which stores the electrical energy obtained from the fuel cell system or the energy recovered through regenerative braking, etc.; the thermal management system module, which regulates the temperatures of modules such as the battery pack and the fuel cell to ensure that the system operates at the optimal working temperature; the in-vehicle control system module, which is responsible for the overall control, monitoring, and data processing of the vehicle and may include a power control unit (ECU), a vehicle status monitoring system, etc.; the safety monitoring system module, including sensors and a monitoring system, which is responsible for monitoring the safety of the vehicle, such as hydrogen leakage detection, over-temperature alarm, etc.

[0023] Step S200, collect the multiple functional modules through the in-vehicle sensing network to obtain multiple sets of sensing data sets.

[0024] Preferably, various types of sensor data are collected and obtained from each functional module of the hydrogen-electric vehicle through the in-vehicle sensing network, reflecting the working status, performance indicators, and possible fault characteristics of each module. Specifically, different types of sensors are installed on each functional module to monitor the operating status of each module, which may include temperature sensors (for monitoring the temperature of the battery, fuel cell, and power system), current and voltage sensors (for monitoring the battery voltage, current, and power output), hydrogen leakage sensors (for detecting whether there is leakage in the hydrogen system), pressure sensors (monitoring the pressure of the gas system and fuel cell system), vibration sensors (for detecting abnormal vibrations in the mechanical system), and rotational speed sensors (monitoring the motor and wheel speeds), etc. The in-vehicle sensing network collects data from each functional module in real time through the sensors, generates multiple sets of sensing data sets, and reflects the operating status, performance changes, and potential fault signs of each module in real time. The data may include data at different time periods and different working conditions, such as data sets of the power system (such as motor output power and torque changes), data sets of the energy storage system (such as the charge and discharge status of the battery and the battery health status), data sets of the fuel cell system (such as hydrogen pressure and fuel cell current output), data sets of the safety system (such as temperature, hydrogen concentration, and leakage alarm), etc. Through pattern recognition, the operating status and mutual relationship of different modules are analyzed to comprehensively understand the real-time operating status and collaborative efficiency of multiple functional modules of the hydrogen-electric vehicle and perform more accurate fault warnings on the vehicle.

[0025] Step S300, perform pattern recognition on the multiple sets of sensing data sets based on the pattern classification algorithm, and output the real-time matching working mode.

[0026] Preferably, a pattern classification algorithm is used to analyze and process the sensing data sets collected from multiple functional modules, so as to identify the working mode of the current hydrogen-electric vehicle and output the matching real-time working mode. Specifically, the multiple sensing data sets come from different functional modules (such as the power system, fuel cell system, energy storage system, etc.), and may vary in terms of time series, data type, and feature dimension. The pattern classification algorithm analyzes and processes multiple groups of sensing data sets to extract effective feature information. For example, considering that the working environment of an electric vehicle is dynamically changing, the pattern classification algorithm processes data with time series features and extracts the changes in the working states of the modules at different time points; by analyzing the data of each sensor, the algorithm extracts the key features (such as temperature, voltage, pressure, etc.) that affect the working mode and fuses them to generate a feature vector suitable for classification; then, according to the input sensing data set, the state of the current system is matched to a predefined working mode. The working mode is the possible operating state of the vehicle under different working conditions, and may include but is not limited to the normal working mode, where all functional modules are in a normal state and the data indicators are within the preset range; the high-load working mode, such as when the motor power output reaches the maximum and the fuel cell system operates under high pressure; the fault warning mode, where an abnormality in a certain functional module causes a decline in the overall vehicle performance, such as too high battery temperature or abnormal hydrogen pressure in the fuel cell; the energy recovery mode, when the electric vehicle is in the braking state, the energy storage system enters the energy recovery mode; through the real-time processing of the sensing data, the pattern classification algorithm can accurately judge the current working mode of the vehicle, match it with the pre-set working mode, and output the working mode that matches the current sensing data set, that is, the current working state of the vehicle. For example, if it is recognized that the current motor and fuel cell are operating under high load, the algorithm will output the high-load working mode.

[0027] Preferably, the pattern classification algorithm is a technique in machine learning or data mining, aiming to identify the patterns or categories represented by the data according to the characteristics of the input data set. It may include supervised learning algorithms, such as support vector machine (SVM), decision tree, random forest, neural network, etc. To identify different working modes by training a model, it is usually necessary to label the data, that is, use the sensing data corresponding to known faults and normal working modes as the training set to train a classifier; unsupervised learning algorithms, such as clustering algorithms (such as K-means, DBSCAN, etc.), automatically discover the potential patterns in the data through the clustering analysis of the sensing data, which is suitable for situations where data does not need to be labeled in advance; deep learning methods, such as convolutional neural network (CNN) and recurrent neural network (RNN), etc. These methods can process high-dimensional data and automatically extract features from it, especially suitable for processing the time series characteristics and non-linear relationships in sensor data.

[0028] Further, asFigure 2 As shown, step S300 further includes step S310 of extracting features from the multiple groups of sensing data sets to output a sensing feature vector; step S320 of inputting the sensing feature vector into a trained random forest classification model, performing pattern recognition according to the random forest classification model, and obtaining a real-time matching working mode output by the random forest classification model; step S330 of training the random forest classification model includes obtaining by training multiple decision trees to convergence according to training data, where the training data includes a historical sensing sample data set and label data identifying the mode types of hydrogen-electric vehicles.

[0029] Preferably, extracting features from multiple groups of sensing data sets to determine multiple representative features that can describe the current vehicle state, that is, the state features of each functional module of the electric vehicle, and then generating a sensing feature vector, which is a condensed representation of multiple sensor data and is a data format used to input into a machine learning model. Obtaining training data includes a historical sensing sample data set and label data identifying the mode types of hydrogen-electric vehicles. Among them, the historical sensing sample data set is a historical sensing data set collected through an on-vehicle sensing network, which contains multiple groups of sensing data collected at different time periods and under different working conditions. The label data records the working mode labels corresponding to each group of sensing data of the hydrogen-electric vehicle mode type. Then, based on the training data, a random forest classification model is trained. Specifically, during the training process, the random forest model will construct multiple decision trees. Each decision tree will learn how to classify according to the input features through recursive splitting based on the historical sensing data and label data. The splitting basis of the decision tree is usually to select the feature that can best distinguish different classes (for example, temperature change, power output, vibration frequency, etc.). As the training progresses, each tree will gradually improve its classification ability until a certain convergence condition is met (such as tree depth limit, maximum number of splits, etc.), and finally a complete decision tree is generated. When the construction of the tree is completed, the random forest will determine the final classification result through the voting of all trees and can accurately predict different working modes. Finally, the sensing feature vector is input into the trained random forest classification model, and classification is performed through the decision trees of the random forest to identify the real-time matching working mode. Specifically, the input sensing feature vector represents the state of the vehicle at the current moment, and the random forest classification model will judge the working mode in which the vehicle is currently located (such as normal working mode, high-load mode, fault warning mode, etc.) according to these features, and the output result is the real-time matching working mode predicted by the model.

[0030] Further, before step S300, it includes step A: analyzing the multiple groups of sensing data sets using a sliding window to obtain the pattern duration; step B: obtaining a stability index according to the difference between the pattern duration and a preset duration threshold; step C: if the stability index is greater than or equal to a preset stability index, performing pattern recognition on the multiple groups of sensing data sets based on a pattern classification algorithm.

[0031] Preferably, a sliding window is a data processing method used for segmental analysis of time series data. The window moves on the data sequence, covering a certain number of consecutive data points each time. The size of the sliding window (i.e., the time span) is usually determined by preset parameters. Specifically, multiple groups of sensing data (such as temperature, voltage, pressure, etc.) are segmented into several small segments according to the time series using the sliding window. The characteristics of the data are analyzed within each window to determine whether the working state of the vehicle within the current window belongs to a certain pattern (such as a normal pattern, a high-load pattern, etc.), thereby obtaining the pattern duration, that is, the continuous duration length of a certain working pattern in time. Then, according to the difference between the pattern duration and the preset duration threshold, a stability index is calculated. Among them, the preset duration threshold defines the shortest time that a certain pattern needs to last. The stability index measures the relationship between the actual duration of a certain pattern and the preset threshold. For example, if the stability index ≥ 0, it indicates that the pattern duration has reached or exceeded the preset threshold, and the current pattern can be considered stable. If the stability index ≤ 0, it indicates that the pattern duration is insufficient, which may be an instantaneous change or an abnormality, and pattern recognition is not performed temporarily. Finally, based on the stability index, it is decided whether to perform pattern recognition. Only when the stability index ≥ a preset stability index (such as 0 or a positive value), it is considered that the current pattern is worthy of further pattern recognition, and finally pattern recognition is performed to determine the specific working pattern in which the vehicle is currently located, and it serves as the basis for fault detection and early warning.

[0032] Further, step C further includes step C1: if the stability index is less than the preset stability index, performing pattern switching prediction according to the multiple groups of sensing data sets and outputting multiple pattern switching probabilities; step C2: determining a first switching working mode according to the multiple pattern switching probabilities, where the first switching working mode is the working mode with the highest probability among the multiple pattern switching probabilities; step C3: calling the module cooperation relationship database to identify the module cooperation relationship diagram corresponding to the first switching working mode; step C4: performing collaborative fault detection on the multiple groups of continuously monitored sensing data sets according to the module cooperation relationship diagram.

[0033] Preferably, when the stability index is less than the preset threshold, it indicates that the duration of the current mode is not sufficient to consider the system stable, meaning that the working state of the vehicle is in dynamic change. For example, it switches from the "normal mode" to the "high load mode", or from the "high load mode" to the "fault warning mode". In the unstable state, by analyzing the dynamic change trend of the sensing data, predict the next working mode that the current mode may switch to. Specifically, use multiple groups of sensing data sets as input data, including time series of parameters such as temperature, voltage, pressure, vibration, etc. Based on a model trained with historical data (such as Markov chain, neural network, state machine, etc.), calculate the probability of switching from the current state to other states, and output the probabilities of the current mode switching to each possible target mode. For example, the probability of switching to the "high load mode" is 60%, the probability of switching to the "fault warning mode" is 30%, and the probability of switching to the "energy saving mode" is 10%.

[0034] Preferably, select the working mode with the highest switching probability from multiple switching working modes as the first switching working mode, that is, the next working mode most likely to be entered. Then call the module cooperation relationship database to extract the cooperation relationship diagram of each functional module in this mode. Among them, the cooperation relationship diagram includes the cooperation rules of each module in the current mode (such as energy flow, signal interaction, etc.) and the data dependency relationship between modules (such as the energy distribution between the battery and the fuel cell, the temperature regulation relationship between the temperature control system and the energy storage system, etc.). By understanding the relationships of each module, it is possible to judge which modules may be affected by anomalies, thereby locating the fault source. Finally, perform cooperative fault detection on multiple groups of continuously monitored sensing data sets according to the module cooperation relationship diagram. Specifically, in the first switching working mode, use the module cooperation relationship diagram to comprehensively analyze the sensing data of multiple functional modules to determine whether there are faults or anomalies. For example, according to the module cooperation rules defined in the cooperation relationship diagram, analyze whether the data of each module conforms to the normal working state. If the anomaly of a certain module causes the data of other modules to exceed the normal range, it is judged as a cooperative fault, which enhances the adaptability of the electric vehicle to the dynamic environment, improves the accuracy of fault detection, and thus improves the safety and reliability of vehicle operation.

[0035] Step S400, construct a module cooperation relationship database, and the module cooperation relationship database stores the cooperation relationships of the multiple functional modules in each working mode.

[0036] Preferably, a module cooperation relationship database is constructed to record the cooperative relationship of multiple functional modules of a hydrogen-electric vehicle in different working modes, that is, the cooperation relationship of multiple functional modules in each working mode is stored, including the cooperation methods and mutual influences of each module under different working conditions. Specifically, the database will store the cooperation relationship between each module (such as the power system, fuel cell system, energy storage system, etc.) in each working mode. For example, in the normal working mode, modules such as the power system, fuel cell system, and energy storage system may need to maintain balanced operation and cooperate with each other to maintain the best energy efficiency. For example, the fuel cell outputs electricity to supply the motor, and the battery provides additional power support when needed; in the high-load mode, the power system may need to exert greater power, and the fuel cell system and energy storage system need to adjust their outputs and charging states according to the load conditions; in the fault warning mode, if a certain module fails, the database will record how other modules respond or adjust in this fault mode and how to cooperate with the faulty module to ensure the safety of the whole vehicle; the module cooperation relationship may also include complex interaction relationships, such as power sharing (how the power system and energy storage system cooperate to provide power or how to recover energy), temperature control (the fuel cell system and energy storage system may need to share the temperature control system), and data sharing (each module shares data in real time through a sensor network, such as battery voltage, output current of the fuel cell, and rotational speed of the power system). Through the module cooperation relationship database, the vehicle can quickly query and match the cooperation relationship of each module according to the current working mode during real-time operation, further enhancing the synergy effect of each functional module of the electric vehicle. If a certain module fails, the database can provide the cooperation relationship between this module and other modules, thereby helping to determine the scope of influence of the fault, so as to quickly diagnose and send a warning signal when the fault occurs.

[0037] Step S500, call the module cooperation relationship database to identify the module cooperation relationship diagram corresponding to the real-time matching working mode.

[0038] Preferably, during the fault detection of the electric vehicle, the constructed module cooperation relationship database is utilized, and according to the real-time identified working mode, the corresponding module cooperation relationship diagram is searched for and generated to display the collaborative working relationship among the functional modules in the current working mode. Specifically, the module cooperation relationship diagram is a graphical tool used to display the relationships and interaction methods among the functional modules of the hydrogen-electric vehicle in a specific working mode, usually presented in the form of a chart or a network. Each module is connected together by edges (relationships), and the weights and connection methods of the edges represent the collaborative working rules among the modules. The nodes in the module cooperation relationship diagram represent each functional module (such as the power system, energy storage system, fuel cell system, etc.), and the edges represent the cooperation relationships among these modules. In different working modes, the cooperation relationships among the modules may change, that is, the connection methods and states among the nodes in the diagram will also change with the change of the working mode. Through the sensor network, the vehicle will detect the current working mode (such as normal working mode, high-load mode, fault warning mode, etc.) in real time. According to the real-time matching working mode, the specific cooperation relationships corresponding to each functional module in this mode are matched from the module cooperation relationship database. By calling the real-time matched cooperation relationship diagram, it is possible to more clearly understand the working states and cooperation situations among the modules in a specific mode, so as to quickly conduct fault diagnosis through the cooperation relationship diagram; it can also help analyze the synergistic effects among different modules and ensure that the vehicle can operate efficiently and safely in various working modes.

[0039] Step S600, perform collaborative fault detection on the multiple groups of sensing data sets according to the module cooperation relationship diagram, and output a fault warning signal corresponding to the real-time matching working mode.

[0040] Preferably, based on the collaborative working mode between modules shown in the module cooperation relationship diagram, comprehensive collaborative fault detection and analysis are performed on multiple sensing data sets to discover potential module faults or abnormal behaviors, and a fault warning signal for the current working mode is generated according to the analysis results. Specifically, the sensing data sets from multiple modules (such as temperature, voltage, current, rotational speed, etc.) are subjected to correlation analysis according to the definition of the module cooperation relationship diagram to determine whether the data relationship between modules conforms to the normal collaborative mode. For example, in the high-load mode, the fuel cell and the battery should output relatively high power at the same time. If the sensing data shows that the fuel cell output is abnormally low while the battery is overloaded, it indicates that there is an abnormality in the collaborative cooperation. Then, the sensing data of each module is analyzed separately to detect abnormalities (such as too high temperature, abnormal current fluctuations, etc.), and it is analyzed whether there is a causal relationship between the abnormalities between modules. For example, if the power output of the fuel cell is insufficient, whether it causes the power output of the power system to decrease. Based on the detection results, the faulty module or fault mode is determined, that is, the module where the problem occurs is quickly located, the influence range on other modules is determined, and the detected abnormal behavior is matched with the predefined fault mode, and then a fault warning signal corresponding to the real-time matching working mode is output, that is, a warning message generated according to the detection results, which is used to remind the driver or maintenance personnel to pay attention to the abnormal state of the current vehicle. The warning signal should indicate which module (such as the fuel cell system, energy storage system, etc.) has a problem and describe the specific fault phenomenon, such as "insufficient power output", "too high temperature", "abnormal pressure", etc., and at the same time indicate the working mode when the fault occurs, such as "abnormality of the fuel cell system under the high-load mode". It can be output in various forms, such as warning lights on the dashboard or sound alarms, and a detailed fault report is displayed on the central control screen. By collaboratively analyzing the data of multiple modules and using the module cooperation relationship diagram, false alarms and missed alarms are reduced, and the accuracy and comprehensiveness of fault detection are improved.

[0041] Further, step S600 further includes step S610 of identifying a healthy sensing data sample corresponding to the module cooperation relationship diagram; step S620 of comparing the multiple groups of sensing data sets according to the healthy sensing data sample to determine a first faulty functional module; step S630 of obtaining a first fault index and multiple collaborative fault indexes corresponding to the first faulty functional module; and step S640 of performing weighted summation according to the first fault index and the multiple collaborative fault indexes to output a fault level, and converting and outputting a fault warning signal according to the fault level.

[0042] Preferably, the faulty module is determined, the degree of the fault is evaluated, and a fault warning signal is generated by comparing the health data with the sensing data. Specifically, the health sensing data samples corresponding to the cooperation relationship diagram of the recognition module are identified. That is, for each working mode (such as the normal mode, the fault warning mode, etc.), through the analysis of the cooperation relationship diagram of the modules, the collaborative working relationship between the functional modules in these modes, and the expected sensing data during normal operation are obtained, including the sensor data of each functional module under the normal operating conditions of the vehicle, which represents the standard behavior of the system in the healthy state, that is, the normal working data of each module without faults or abnormalities. The current multiple sets of sensing data sets (real-time data from different functional modules) are compared with the previously identified health sensing data samples. For example, it is detected whether the sensor data exceeds a predetermined safety threshold, whether the data trend deviates from the normal mode, or the mean, variance, fluctuation, etc. of the data are calculated to determine whether the current data is abnormal and whether the current data matches the normal range in the health data samples, so as to determine the first faulty functional module.

[0043] Preferably, the first fault index corresponding to the first faulty functional module and multiple collaborative fault indexes are obtained. Among them, the first fault index refers to the independent fault characteristics related to the first faulty functional module, which may include changes in parameters such as the temperature, pressure, voltage, power, etc. of the functional module. The collaborative fault index refers to the fault signals related to other modules, which may be caused by problems in the collaborative effect between the modules. For example, the collaborative fault index of the energy storage system and the fuel cell. The energy storage system may be overcharged or over-discharged due to the unstable output of the fuel cell, resulting in a decline in performance; the collaborative fault index of the temperature control system and the power system. The failure of the temperature control system may cause the power system to overheat, thus affecting the power performance of the vehicle. The weighted sum is performed according to the first fault index and the multiple collaborative fault indexes, that is, according to the severity of the impact of each fault index on the system fault, a certain weight is assigned, and the weighted sum is output as the fault level, indicating the severity of the fault. The higher the fault level, the more serious the fault, and immediate measures may be required. Finally, a fault warning signal is output according to the fault level. For example, if the fault level is low and the impact is small, the vehicle can continue to run; if the fault level is medium, it is recommended to monitor or take preventive measures; if the fault level is high, an emergency warning is issued, which may cause the system to shut down or require immediate maintenance.

[0044] Further, step S630 further includes step S631, in combination with the cooperation relationship diagram of the modules, an output fault propagation probability matrix is obtained according to the connection relationship between the first faulty functional module and other functional modules; step S632, multiple collaborative fault indexes are calculated according to the fault propagation probability matrix.

[0045] Preferably, by analyzing the connection relationships and fault propagation paths among modules, a fault propagation probability matrix is constructed, and based on this matrix, the collaborative fault indicators among various modules are calculated. Specifically, according to the connection relationships (the mutual influence and dependence among functional modules) between the first faulty functional module (the module determined to be faulty after comparing sensing data with healthy samples) and other functional modules, and based on historical data, the connection strength among modules, and the fault propagation paths, the probability of a fault propagating from one module to another is calculated, that is, a fault propagation probability matrix is established, which describes the probability that a fault may propagate from the first faulty functional module to other modules. Each element in the matrix represents the probability that another module is affected after a fault occurs in one functional module. The fault propagation probability matrix can quantitatively describe the fault propagation risks among different modules and help understand the potential impact scope of faults. Finally, based on the fault propagation probability matrix, multiple collaborative fault indicators are calculated, that is, the probability of collaborative faults occurring among multiple modules is evaluated through the fault propagation probability matrix (the collaborative fault indicators are calculated by weighted summation). For example, for the collaborative fault indicator of an energy storage system, if the probability of the energy storage system being affected is high when the battery fails, the collaborative fault indicator is calculated according to the probability of the battery fault propagating to the energy storage system. This reflects the fault linkage among modules, can help the system identify potential fault chains and risks in advance, and thus provide more accurate fault warnings, diagnoses, and optimization measures to improve the reliability and safety of the system.

[0046] Further, step S600 further includes step S650 of obtaining the mode switching combinations corresponding to the real-time matching working mode according to the module cooperation relationship database. Step S660 of identifying multiple switching risks corresponding to the mode switching combinations according to the fault warning signal. Step S670 of screening out the identified mode switching combinations with switching risks greater than the preset switching risks among the multiple switching risks. Step S680 of generating a switching warning signal when the received mode switching request belongs to the identified mode switching combinations.

[0047] Preferably, obtain the mode switching combinations that match the real-time working mode from the module cooperation relationship database, that is, the combinations of all target modes that the vehicle may switch from the current working mode to, and consider the feasibility and conditions of each switching. Then, according to the fault warning signal, identify multiple switching risks corresponding to the mode switching combinations. The switching risk refers to the operating risks that the system may face when switching from the current working mode to the target mode, including aggravated module failures, enlarged coordination problems, or reduced energy efficiency, etc. For example, if the current working mode is the "high-load mode", switching to the "energy-saving mode" may cause the load of the power system to drop rapidly, thus aggravating the power balance problem between the battery and the fuel cell, and the switching risk may be relatively high; screen the identified mode switching combinations with a switching risk greater than the preset switching risk. Among them, the preset switching risk is a threshold for distinguishing low-risk switching and high-risk switching. Among all the mode switching combinations, screen out the mode switching with a risk value greater than the preset switching risk according to the risk assessment result; the mode switching request can come from the driver's operation instruction (such as switching to a certain mode) or the vehicle automatic adjustment system (such as automatically switching to the energy-saving mode). When the received switching request belongs to the identified high-risk mode switching combination, generate a switching warning signal, display a high-risk switching warning on the dashboard, and suggest avoiding the switching or taking buffer measures (such as delaying the switching, gradually switching, etc.), thereby avoiding high-risk operations and improving safety and reliability.

[0048] Further, step S680 further includes step S681 of detecting the repair index of the real-time matching working mode according to the switching warning signal; step S682 of, if the repair index does not meet the preset repair index, performing mode switching optimization in the module cooperation relationship database with the preset switching risk, and outputting an optimal solution for the switching mode.

[0049] Preferably, according to the switching warning signal (indicating that the current mode switching may have a relatively high risk, for example, the states of some modules may not be suitable for performing the switching operation), the repair metrics for the real-time matching working mode are detected. The repair metrics are a set of key parameters that measure the health status of the current working mode, representing the minimum performance or state standards that the system needs to achieve before the switching operation. For example, the temperature of the fuel cell must fall back to a certain range, the state of charge (SOC) of the energy storage system needs to reach a certain level, and the load fluctuation of the power system needs to be stabilized within a safe range. If the repair metrics do not meet the preset repair metrics, that is, it is detected that the repair metrics of the current working mode do not reach the preset requirements, it indicates that the state of the system is not sufficient to safely complete the switching operation to the target mode. On the premise of ensuring safety and based on the principle of minimizing risk, using the current working mode (real-time matching mode), the target working mode (the mode to be switched to), and the mode switching risks and conditions recorded in the module cooperation relationship database as input parameters, an optimization algorithm (such as dynamic programming, shortest path algorithm, genetic algorithm, etc.) is used to calculate the optimal switching path, evaluate all possible intermediate mode combinations (such as from "normal mode" to "energy-saving mode", and then to "high-load mode"), and perform a weighted sum of the switching risks of each path, and output the switching path with the lowest risk, which is marked as the optimal solution of the switching mode and output. That is, in the case where the current repair metrics are not met, the mode switching scheme with the lowest risk determined after optimization calculation is obtained. This not only improves the reliability of vehicle operation, but also reduces the failure risk caused by unsafe mode switching and enhances the vehicle intelligent control ability.

[0050] The above specific embodiments do not constitute a limitation to the protection scope of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of this application shall be included within the protection scope of this application. In some cases, the actions or steps recorded in this application can be executed in a different order from that in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A multi-module coordinated fault detection and early warning method for hydrogen-electric vehicles, characterized in that: The method comprises: Determining a plurality of functional modules of a hydrogen-electric vehicle, wherein the plurality of functional modules are connected to an onboard sensor network; The plurality of functional modules are collected through the vehicle-mounted sensor network to obtain a plurality of sensor data sets; Analyzing the plurality of sensor data sets using a sliding window to obtain a pattern duration; Obtaining a stability index according to a difference between the mode duration and a preset duration threshold; If the stability index is greater than or equal to a preset stability index, performing pattern recognition on the multiple sets of sensor data sets based on a pattern classification algorithm; Performing feature extraction on the multiple sets of sensor data sets and outputting sensor feature vectors; Inputting the sensor feature vector into a trained random forest classification model, performing pattern recognition according to the random forest classification model, and obtaining a real-time matching working mode output by the random forest classification model; Training the random forest classification model includes training a plurality of decision trees to convergence based on training data, wherein the training data includes obtaining a historical sensor sample data set and label data identifying a hydrogen electric vehicle mode type; Constructing a module coordination relationship database, wherein the module coordination relationship database stores coordination relationships of the plurality of functional modules in each working mode; Calling the module coordination relationship database to identify the module coordination relationship diagram corresponding to the real-time matching working mode; Performing collaborative fault detection on the multiple sensor data sets according to the module coordination relationship diagram, and outputting a fault warning signal corresponding to the real-time matching working mode; If the stability index is less than the preset stability index, performing mode switching prediction according to the multiple sets of sensor data sets, and outputting multiple mode switching probabilities; Determining a first switching working mode according to the multiple mode switching probabilities, wherein the first switching working mode is the working mode with the highest probability among the multiple mode switching probabilities; The module coordination relationship database is called to identify the module coordination relationship diagram corresponding to the first switching working mode.

2. The multi-module coordinated fault detection and early warning method for a hydrogen-electric vehicle as claimed in claim 1, characterized in that: According to the module coordination relationship diagram, collaborative fault detection is performed on the multiple sensor data sets, and a fault warning signal corresponding to the real-time matching working mode is output, the method comprising: Identify the health sensor data samples corresponding to the module coordination relationship diagram; Comparing the plurality of sensor data sets according to the healthy sensor data samples to determine a first faulty functional module; Obtaining a first fault indicator and a plurality of coordinated fault indicators corresponding to the first fault function module; A weighted sum is taken according to the first fault indicator and the multiple coordinated fault indicators to output a fault level, and a fault warning signal is output according to the conversion of the fault level.

3. The multi-module coordinated fault detection and early warning method for a hydrogen-electric vehicle as claimed in claim 2, characterized in that: Methods for obtaining multiple coordinated fault indicators include: Combined with the module coordination relationship diagram, output a fault propagation probability matrix according to the connection relationship between the first fault function module and other function modules; A plurality of coordinated fault indicators are calculated according to the fault propagation probability matrix.

4. The multi-module coordinated fault detection and early warning method for a hydrogen-electric vehicle as claimed in claim 1, characterized in that: After outputting the fault warning signal corresponding to the real-time matching working mode, the method further includes: According to the module coordination relationship database, a mode switching combination of the real-time matching working mode is obtained; identifying, according to the fault warning signal, a plurality of switching risks corresponding to the mode switching combination; Selecting a marking mode switching combination having a greater than a preset switching risk from the plurality of switching risks; When the received mode switching request belongs to the identified mode switching combination, a switching warning signal is generated.

5. The multi-module coordinated fault detection and early warning method for a hydrogen-electric vehicle as claimed in claim 4, characterized in that: When the received mode switching request belongs to the identified mode switching combination, after generating the switching warning signal, the method further includes: According to the switching warning signal, detecting a repair index of the real-time matching working mode; If the repair index does not meet the preset repair index, the preset switching risk is used to perform mode switching optimization in the module coordination relationship database, and an optimal switching mode solution is output.

Images