A fuel cell hydrogen supply system fault detection method, device and medium

Abstract

The application discloses a fuel cell hydrogen supply system fault detection method, equipment and medium, and relates to the technical field of hydrogen fuel cell fault detection. The method collects far-field sensor data and near-field sensor data of a fuel cell hydrogen supply system, identifies a vehicle operating condition, adopts a neural network model to output far-field sensor threshold values and near-field sensor threshold values under the current condition, and sequentially judges whether the fuel cell hydrogen supply system has hydrogen supply abnormalities, water management abnormalities and thermal management abnormalities according to the far-field sensor data, the near-field sensor data and the corresponding threshold values, so as to finally determine a fault detection result of the fuel cell hydrogen supply system according to a hydrogen supply detection result, a water management detection result and a thermal management detection result. The fuel cell hydrogen supply system fault detection method, equipment and medium provided by the application can realize comprehensive and accurate detection of faults of the fuel cell hydrogen supply system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hydrogen fuel cell fault detection, in particular to a fuel cell hydrogen supply system fault detection method, device and medium. BACKGROUND

[0002] As a clean energy conversion device, hydrogen fuel cell has the advantages of high efficiency and zero emission, and has received widespread attention in recent years. The hydrogen supply system is one of the core components of the hydrogen fuel cell, and its performance directly affects the overall efficiency and safety of the fuel cell. However, the hydrogen supply system may occur a variety of faults in actual operation, such as hydrogen leakage, insufficient flow, abnormal pressure, etc. If these faults cannot be detected and handled in time, it may lead to performance degradation, safety hazards and even shutdown of the fuel cell.

[0003] The existing fuel cell hydrogen supply system fault detection method mainly relies on a single sensor or a single monitoring parameter, such as only relying on a hydrogen flow sensor or a pressure sensor for monitoring. This method has the following disadvantages: 1. Low accuracy, single sensor is easily affected by environmental interference, self-failure and other factors, resulting in inaccurate detection results. 2. Limited range, single monitoring parameter cannot fully reflect the state of the hydrogen supply system, and some faults may be missed. SUMMARY

[0004] The purpose of the present application is to provide a fuel cell hydrogen supply system fault detection method, device and medium, which can realize comprehensive and accurate detection of fuel cell hydrogen supply system faults.

[0005] To achieve the above purpose, the present application provides the following solutions.

[0006] In a first aspect, the present application provides a fuel cell hydrogen supply system fault detection method, comprising:

[0007] Collecting far-field sensor data and near-field sensor data of the fuel cell hydrogen supply system; the far-field sensor data includes far-field hydrogen concentration data and far-field infrared thermal imaging data; the near-field sensor data includes near-field pressure data, near-field flow data, near-field humidity data, near-field liquid level data and near-field temperature data;

[0008] Identifying the vehicle operating condition, and outputting the far-field sensor threshold and the near-field sensor threshold under the current operating condition by using a neural network model; the far-field sensor threshold includes a far-field hydrogen concentration threshold; the near-field sensor threshold includes a near-field pressure drop threshold, a near-field flow drop threshold, a near-field humidity threshold, a near-field liquid level threshold and a near-field temperature threshold;

[0009] According to the far-field hydrogen concentration data, the near-field pressure data and the near-field flow data, and corresponding threshold values, it is judged whether the hydrogen supply of the fuel cell hydrogen supply system is abnormal, and a hydrogen supply detection result is obtained;

[0010] If the hydrogen supply is normal, according to the near-field humidity data and the near-field liquid level data, and corresponding threshold values, it is judged whether the water management of the fuel cell hydrogen supply system is abnormal, and a water management detection result is obtained;

[0011] If the water management is normal, according to the far-field infrared thermal imaging data and the near-field temperature data, and corresponding threshold values, it is judged whether the thermal management of the fuel cell hydrogen supply system is abnormal, and a thermal management detection result is obtained;

[0012] According to the hydrogen supply detection result, the water management detection result and the thermal management detection result, a fault detection result of the fuel cell hydrogen supply system is determined.

[0013] Optionally, the method for determining the neural network model comprises:

[0014] Collect historical data of the vehicle under different working condition categories to construct a historical data set; the working condition categories include: starting working condition, accelerating working condition, uniform speed working condition, idling working condition and braking working condition; the historical data includes: working condition data, sensor data and sensor threshold values; the working condition data includes: brake pedal opening degree, accelerator pedal opening degree and vehicle speed;

[0015] Taking the working condition category, the working condition data and the sensor data as input, and taking the sensor threshold value as output, a multi-layer perception neural network is constructed, and the multi-layer perception neural network is trained by using an optimization algorithm according to the historical data set, so as to obtain a trained neural network model.

[0016] Optionally, according to the far-field hydrogen concentration data, the near-field pressure data and the near-field flow data, and corresponding threshold values, it is judged whether the hydrogen supply of the fuel cell hydrogen supply system is abnormal, and a hydrogen supply detection result is obtained, which comprises:

[0017] According to the far-field hydrogen concentration data and the far-field hydrogen concentration threshold value, it is judged whether the far-field hydrogen concentration is too high, and a first judgment result is obtained;

[0018] If the first judgment result is yes, according to the near-field pressure data and the near-field pressure drop threshold value, it is judged whether the pressure drop exceeds the threshold value, and a second judgment result is obtained;

[0019] If the second judgment result is yes, a detection result that there is a potential hydrogen leakage in the fuel cell hydrogen supply system is obtained, and a first fusion function is used to calculate a hydrogen leakage probability;

[0020] If the second judgment result is no, according to the near-field flow data and the near-field flow drop threshold value, it is judged whether the flow is too low, and a third judgment result is obtained;

[0021] If the third determination result is yes, a detection result that there is a potential hydrogen leakage or pressure sensor fault in the fuel cell hydrogen supply system is obtained, and a second fusion function is used to calculate a hydrogen leakage or pressure sensor fault probability;

[0022] If the third determination result is no, a detection result that there is a potential hydrogen concentration sensor fault in the fuel cell hydrogen supply system is obtained, and a third fusion function is used to calculate a hydrogen concentration sensor fault probability;

[0023] If the first determination result is no, it is determined according to the near-field flow data and the near-field flow drop threshold value whether the flow measurement value is too low, and a fourth determination result is obtained;

[0024] If the fourth determination result is yes, it is determined according to the near-field pressure data whether the pressure measurement value is too high, and a fifth determination result is obtained;

[0025] If the fifth determination result is yes, a detection result that there is a potential pipeline blockage in the fuel cell hydrogen supply system is obtained, and a fourth fusion function is used to calculate a pipeline blockage probability;

[0026] If the fifth determination result is no, a detection result that there is a potential hydrogen storage bottle opening failure or hydrogen concentration sensor fault in the fuel cell hydrogen supply system is obtained, and a fifth fusion function is used to calculate a hydrogen storage bottle opening failure or hydrogen concentration sensor fault probability;

[0027] If the fourth determination result is no, a detection result that the hydrogen supply of the fuel cell hydrogen supply system is normal is obtained.

[0028] Optionally, it is determined according to the near-field humidity data and the near-field liquid level data and corresponding thresholds whether there is an abnormal water management in the fuel cell hydrogen supply system, and a water management detection result is obtained, including:

[0029] It is determined according to the near-field humidity data and the near-field humidity threshold value whether the humidity measurement value is abnormal, and a sixth determination result is obtained; wherein the humidity measurement value abnormality includes humidity being too high and humidity being too low;

[0030] If the sixth determination result is yes and the humidity is too high, it is determined according to the near-field liquid level data and the near-field liquid level threshold value whether the liquid level is too high, and a seventh determination result is obtained;

[0031] If the seventh determination result is yes, a detection result that there is a potential drain valve fault in the fuel cell hydrogen supply system is obtained, and a sixth fusion function is used to calculate a drain valve fault probability;

[0032] If the seventh determination result is no, a detection result that there is a potential humidifier fault in the fuel cell hydrogen supply system is obtained, and a seventh fusion function is used to calculate a humidifier fault probability;

[0033] If the sixth determination result is yes and the humidity is too low, a detection result that there is a potential humidifier fault or purge valve fault in the hydrogen supply system of the fuel cell is obtained, and a probability of the humidifier fault or purge valve fault is calculated using an eighth fusion function;

[0034] If the sixth determination result is no, a detection result that water management is normal in the hydrogen supply system of the fuel cell is obtained.

[0035] Optionally, whether there is thermal management abnormality in the hydrogen supply system of the fuel cell is determined according to the far-field infrared thermal imaging data and the near-field temperature data and corresponding threshold values, and a thermal management detection result is obtained, including:

[0036] Whether the temperature measurement value is too high is determined according to the near-field temperature data and a near-field temperature threshold value, and an eighth determination result is obtained;

[0037] If the eighth determination result is yes, whether there is local high temperature is determined according to the far-field infrared thermal imaging data, and a ninth determination result is obtained;

[0038] If the ninth determination result is yes, a detection result that there is a potential coolant blockage in the hydrogen supply system of the fuel cell is obtained, and a probability of the coolant blockage is calculated using a ninth fusion function;

[0039] If the ninth determination result is no, a detection result that there is a potential cooling circulation pump fault in the hydrogen supply system of the fuel cell is obtained, and a probability of the cooling circulation pump fault is calculated using a tenth fusion function;

[0040] If the eighth determination result is no, a detection result that thermal management is normal in the hydrogen supply system of the fuel cell is obtained.

[0041] Optionally, the hydrogen supply detection result includes: hydrogen leakage, pressure sensor fault, hydrogen concentration sensor fault, pipeline blockage, hydrogen storage bottle opening failure, and normal hydrogen supply;

[0042] An expression for calculating a probability of hydrogen leakage using a first fusion function is:

[0043]

[0044] wherein P is the probability of hydrogen leakage, C is an actual far-field hydrogen concentration, C is a far-field hydrogen concentration threshold value, ΔP is an actual near-field pressure drop value, ΔP is a near-field pressure drop threshold value, t is time, and α, β, γ, and δ are a first set of weight coefficients; leak th th

[0045] An expression for calculating a probability of hydrogen leakage or pressure sensor fault using a second fusion function is:

[0046] ​​​

[0047] wherein P potential is a hydrogen leakage or pressure sensor failure probability, F is an actual near field flow value, F th is a near field flow drop threshold, w1, w2 and w3 are second group weight coefficients;

[0048] The expression for calculating the hydrogen concentration sensor failure probability using a third fusion function is:

[0049]

[0050] wherein P sensor is a hydrogen concentration sensor failure probability, a and b are third group weight coefficients;

[0051] The expression for calculating the pipe blockage probability using a fourth fusion function is:

[0052]

[0053] wherein P block is a pipe blockage probability, P is an actual near field pressure, P high is a near field pressure threshold, c and d are fourth group weight coefficients;

[0054] The expression for calculating the hydrogen storage bottle opening failure or hydrogen concentration sensor failure probability using a fifth fusion function is:

[0055]

[0056] wherein P open is a hydrogen storage bottle opening failure or hydrogen concentration sensor failure probability, P normal is a near field pressure normal value, e and f are fifth group weight coefficients.

[0057] Optionally, the water management detection result includes: a drain valve failure, a humidifier failure, a purge valve failure, and water management normal;

[0058] The expression for calculating the drain valve failure probability using a sixth fusion function is:

[0059]

[0060] wherein P drain is a drain valve failure probability, H is an actual near field humidity, H th is a near field humidity threshold, L is an actual near field liquid level, L th is a near field liquid level threshold, k1 and k2 are sixth group weight coefficients;

[0061] The expression for calculating the humidifier failure probability using a seventh fusion function is:

[0062]

[0063] wherein, P humidifier is the humidifier failure probability, L low is a specific value of the liquid level being lower than the threshold value, and m1 and m2 are the seventh group of weight coefficients;

[0064] The expression for calculating the humidifier failure or purge valve failure probability using the eighth fusion function is:

[0065]

[0066] wherein, P humidifier-purge is the humidifier failure or purge valve failure probability, r1, s1 and t1 are the eighth group of weight coefficients, and u is an adjustment parameter.

[0067] Optionally, the thermal management detection result includes: cooling liquid blockage, cooling circulating pump failure, and normal thermal management;

[0068] The expression for calculating the cooling liquid blockage probability using the ninth fusion function is:

[0069]

[0070] wherein, P coolant-block is the cooling liquid blockage probability, T is the actual near-field temperature, T th is the near-field temperature threshold value, I local is the local high temperature feature parameter of the far-field infrared thermal imaging, and q1 and q2 are the ninth group of weight coefficients.

[0071] The expression for calculating the cooling circulating pump failure probability using the tenth fusion function is:

[0072]

[0073] wherein, P pump is the cooling circulating pump failure probability, I global is the overall high temperature feature parameter of the far-field infrared thermal imaging, and q3 and q4 are the tenth group of weight coefficients.

[0074] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the fuel cell hydrogen supply system failure detection method described above.

[0075] In a third aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the fuel cell hydrogen supply system failure detection method described above.

[0076] According to the specific embodiments provided in the application, the application discloses the following technical effects:

[0077] The application provides a fuel cell hydrogen supply system fault detection method, device and medium, by collecting far-field sensor data and near-field sensor data of the fuel cell hydrogen supply system and identifying the vehicle operating condition, and outputting the far-field sensor threshold and the near-field sensor threshold in the current condition by using the neural network model, according to the far-field sensor data, the near-field sensor data and the corresponding threshold, whether the fuel cell hydrogen supply system exists hydrogen supply abnormality, water management abnormality and thermal management abnormality is judged in turn, and finally the fault detection result of the fuel cell hydrogen supply system is determined according to the hydrogen supply detection result, the water management detection result and the thermal management detection result. The application can automatically adjust the sensor threshold according to the different conditions of the fuel cell, adapt to various operating conditions, improve the accuracy and reliability of fault detection, wherein the far-field sensor monitors the overall state, and the near-field sensor monitors the local details, which can realize comprehensive monitoring, the far-field sensor data and the near-field sensor data are fused, and the fault probability can be more accurately evaluated. BRIEF DESCRIPTION OF DRAWINGS

[0078] In order to more clearly illustrate the technical solutions in the embodiments of the application or the related art, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0079] Figure 1 The fuel cell hydrogen supply system fault detection method flowchart provided by the application.

[0080] Figure 2 The fuel cell hydrogen supply system fault detection flowchart provided by the application. DETAILED DESCRIPTION

[0081] The technical solutions in the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0082] In the monitoring of fuel cell systems, near-field sensors and far-field sensors each have their characteristics and advantages. Near-field sensors are deployed in a small range near the monitored equipment or system, focusing on obtaining fine parameters of key parts of the equipment, providing direct basis for local state evaluation of the system. Far-field sensors are arranged at a relatively distant position from the monitored object, used to obtain system overall state or surrounding environment related information, which helps to grasp the running state of the system in a larger range and the mutual relationship with the surrounding environment.

[0083] Near-field and far-field sensor linkage is a technical strategy in monitoring and control systems that comprehensively uses near-field sensors and far-field sensors to make them interact with each other and exchange information, thereby achieving more comprehensive and accurate monitoring and control. Near-field and far-field sensors realize data interaction and collaborative work through a fusion function, which takes the data of near-field and far-field sensors as input for fusion, and processes these inputs through a specific algorithm to calculate the fault probability.

[0084] The purpose of the present application is to provide a fuel cell hydrogen supply system fault detection method, device and medium, which can realize comprehensive and accurate detection of fuel cell hydrogen supply system faults through near-field and far-field sensor data fusion.

[0085] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0086] In an exemplary embodiment, the present application provides a fuel cell hydrogen supply system fault detection method. In the embodiment of the present application, as shown in Figure 1 the method comprises the following steps 1 to 6.

[0087] Step 1: Collecting far-field sensor data and near-field sensor data of the fuel cell hydrogen supply system. The far-field sensor data includes far-field hydrogen concentration data and far-field infrared thermal imaging data. The near-field sensor data includes near-field pressure data, near-field flow data, near-field humidity data, near-field liquid level data and near-field temperature data.

[0088] Among them, the far-field hydrogen concentration sensor is used to obtain the hydrogen concentration data in the environment of the fuel cell hydrogen supply system, the far-field infrared thermal imaging sensor is used to obtain the overall thermal distribution data of the fuel cell hydrogen supply system, the near-field pressure sensor is used to obtain the hydrogen pressure value inside the hydrogen supply system, the near-field flow sensor is used to obtain the hydrogen flow value inside the hydrogen supply system, the near-field humidity sensor is used to obtain the humidity value of the hydrogen entering the anode of the fuel cell, the near-field liquid level sensor is used to obtain the liquid level height value in the fuel cell water separator, and the near-field temperature sensor is used to obtain the temperature value of the hydrogen supply system.

[0089] Step 2: Identify the vehicle operating condition, and output the far-field sensor threshold value and the near-field sensor threshold value in the current condition by using the neural network model. The far-field sensor threshold value includes: a far-field hydrogen concentration threshold value. The near-field sensor threshold value includes: a near-field pressure drop threshold value, a near-field flow rate drop threshold value, a near-field humidity threshold value, a near-field liquid level threshold value, and a near-field temperature threshold value.

[0090] wherein the category of the vehicle operating condition includes: a start-up condition, an acceleration condition, a constant speed condition, an idle condition, and a braking condition. The vehicle operating condition is determined by the brake pedal opening degree, the accelerator pedal opening degree, and the vehicle speed, specifically: monitoring the speed (V) of the vehicle in real time, determining whether the vehicle is in a stationary state, i.e., whether the speed (V) is 0 km / h. If the vehicle is in a stationary state, further detect the accelerator pedal opening degree (APO), if APO is greater than 0%, determine that the current condition is "start-up". If APO is 0%, determine that the current condition is "idle". If the vehicle speed (V) is not 0 km / h, further detect the brake pedal opening degree (BPO), if BPO is greater than 0%, determine that the current condition is "braking". If BPO is 0% and APO remains stable, further monitor the change of the vehicle speed (V), if the vehicle speed remains relatively stable, determine that the current condition is "constant speed". If the vehicle speed (V) is increasing, determine that the current condition is "acceleration".

[0091] Preferably, the determination method of the neural network model comprises: collecting historical data of the vehicle in different condition categories, constructing a historical data set; taking the condition category, the condition data, and the sensor data as input, and taking the sensor threshold value as output, constructing a multi-layer perception neural network, and training the multi-layer perception neural network by using an optimization algorithm according to the historical data set to obtain a trained neural network model. The historical data includes: condition data, sensor data, and sensor threshold value. The condition data includes: brake pedal opening degree, accelerator pedal opening degree, and vehicle speed.

[0092] Specifically, vehicle operation data including brake pedal opening degree, accelerator pedal opening degree and vehicle speed are collected; vehicle operation conditions are determined according to the vehicle operation data, and one-hot encoding is performed for five conditions of starting, accelerating, constant speed, idling and braking. For example, the starting condition is encoded as [1, 0, 0, 0, 0], the accelerating condition is [0, 1, 0, 0, 0], the constant speed condition is [0, 0, 1, 0, 0], the idling condition is [0, 0, 0, 1, 0], and the braking condition is [0, 0, 0, 0, 1]. For each condition, relevant sensor data are collected, and the collected data are preprocessed, including removing outliers and data normalization. A multilayer perceptron (MLP) neural network is constructed, which consists of an input layer, a hidden layer and an output layer. The input layer inputs data training sample set, data verification sample set and data test sample set. The hidden layer learns and trains according to the data training sample set to obtain a trained neural network model. According to the real-time monitored brake pedal opening degree, accelerator pedal opening degree and vehicle speed, the current condition is identified, the real-time monitored condition data and the condition category code (one-hot encoding form) are input into the trained neural network model, and the model outputs the threshold of each sensor under the current condition.

[0093] For each condition, relevant sensor data are collected, including far-field hydrogen concentration data, far-field infrared thermal imaging data, near-field pressure data, near-field flow data, near-field humidity data, near-field liquid level data, etc. The collected data are preprocessed, including removing outliers (for example, using 3 times standard deviation method, if the data point is outside the range of mean ± 3 times standard deviation, it is considered as an outlier for correction or deletion) and data normalization, mapping the data to the interval [0, 1], and the normalization formula is where x is the original data, x min and x max are the minimum and maximum values of the feature, respectively.

[0094] The multi-layer perceptron (MLP) neural network is composed of an input layer, a hidden layer, and an output layer. The number of input layer nodes is determined according to the input data, including the working condition category (using one-hot encoding, with the starting condition being [1, 0, 0, 0, 0], the accelerating condition being [0, 1, 0, 0, 0], the constant speed condition being [0, 0, 1, 0, 0], the idling condition being [0, 0, 0, 1, 0], and the braking condition being [0, 0, 0, 0, 1], with a total of 5 nodes) and key working condition data (brake pedal opening, accelerator pedal opening, and vehicle speed, with a total of 3 nodes) and key sensor data (far-field hydrogen concentration data, far-field infrared thermal imaging data, near-field pressure data, near-field flow data, near-field humidity data, near-field liquid level data, and near-field temperature data, with a total of 7 nodes), so the number of input layer nodes is 15. The hidden layer is set to 2 layers, with the first hidden layer having 16 nodes and the second hidden layer having 8 nodes. The number of output layer nodes is the same as the number of sensors whose thresholds need to be adjusted. The hidden layer activation function is selected as the ReLU (Rectified Linear Unit) function ReLU(x1) = max(0, x1), and the output layer activation function is selected as the Softplus function Softplus(x2) = ln(1 + e x2 ) according to the nature of the threshold.

[0095] A large amount of historical data under different working conditions is collected, including the pedal opening, vehicle speed, and corresponding sensor data and ideal threshold values in the above-mentioned working condition judgment basis. The data is divided into a training set, a validation set, and a test set in a ratio of 70%, 20%, and 10%, respectively.

[0096] The stochastic gradient descent (SGD) algorithm and its variants (such as the Adam optimization algorithm) are used to optimize the weights and biases of the neural network. The loss function is selected as the Mean Square Error (MSE) function where n is the number of training samples, y i is the true sensor threshold value of the i-th sample, and is the sensor threshold value predicted by the neural network model for the i-th sample. A suitable learning rate (such as 0.001) and training iteration number (such as 2000 times) are set. During training, the Early Stopping method is used to prevent overfitting, i.e., when the loss function value on the validation set no longer decreases or the decrease is less than a certain threshold (such as 0.0001), training is stopped. The L2 regularization (weight decay) method is used to add a regularization term to the loss function (where λ1 is the regularization coefficient and w i is the weight), limiting the model complexity and improving the generalization ability of the model.

[0097] Step 3: Based on the far-field hydrogen concentration data, near-field pressure data, near-field flow data, and corresponding thresholds, determine whether there is an abnormality in the hydrogen supply system of the fuel cell, and obtain the hydrogen supply detection results.

[0098] Step 3 specifically includes: determining whether the far-field hydrogen concentration is too high based on far-field hydrogen concentration data and a far-field hydrogen concentration threshold, obtaining a first judgment result; if the first judgment result is yes, determining whether the pressure drop exceeds the threshold based on near-field pressure data and a near-field pressure drop threshold, obtaining a second judgment result; if the second judgment result is yes, obtaining a detection result indicating a potential hydrogen leak in the fuel cell hydrogen supply system, and calculating the probability of hydrogen leak using a first fusion function; if the second judgment result is no, determining whether the flow rate is too low based on near-field flow rate data and a near-field flow rate drop threshold, obtaining a third judgment result; if the third judgment result is yes, obtaining a detection result indicating a potential hydrogen leak or pressure sensor malfunction in the fuel cell hydrogen supply system, and calculating the probability of hydrogen leak or pressure sensor malfunction using a second fusion function; if the third judgment result is no, obtaining a detection result indicating a potential hydrogen leak or pressure sensor malfunction in the fuel cell hydrogen supply system. The system detects a hydrogen concentration sensor malfunction and calculates its probability using a third fusion function. If the first judgment is negative, it determines whether the measured flow rate is too low based on near-field flow data and the near-field flow rate drop threshold, resulting in a fourth judgment. If the fourth judgment is positive, it determines whether the measured pressure is too high based on near-field pressure data, resulting in a fifth judgment. If the fifth judgment is positive, it detects potential pipeline blockage in the fuel cell hydrogen supply system and calculates the probability of pipeline blockage using a fourth fusion function. If the fifth judgment is negative, it detects potential issues with the opening of the hydrogen storage tank or a hydrogen concentration sensor malfunction in the fuel cell hydrogen supply system and calculates the probability of these issues using a fifth fusion function. If the fourth judgment is negative, it detects normal hydrogen supply in the fuel cell hydrogen supply system.

[0099] First, a far-field hydrogen concentration threshold C is set based on a neural network model. th Near-field pressure drop threshold ΔP th and near-field flow drop threshold F th The far-field hydrogen concentration data is compared with the near-field pressure and flow rate data. If the far-field hydrogen concentration data is higher than C... th Then, further analysis of the near-field pressure data is needed. If the near-field pressure data decreases by more than ΔP... th At this point, the far-field hydrogen concentration data and the near-field pressure data corroborate each other, indicating a high probability of hydrogen leakage in the fuel cell hydrogen supply system. The data fusion model calculates the hydrogen leakage probability P based on the degree of deviation and trend of the two data. leak =f(CC) th ,ΔP-ΔPth ), where C is the actual far-field hydrogen concentration, ΔP is the actual near-field pressure drop, and f is a first fusion function determined based on an empirical or machine learning model.

[0100] Preferably, the following is an example of a fusion function to calculate the hydrogen leak probability:

[0101]

[0102] where P leak is the hydrogen leak probability, C is the actual far-field hydrogen concentration, C th is the far-field hydrogen concentration threshold, ΔP is the actual near-field pressure drop, and ΔP th is the near-field pressure drop threshold. and represent the degree of deviation of the far-field hydrogen concentration and the near-field pressure drop from the threshold, respectively. and represent the change trend (time t derivative) of the far-field hydrogen concentration and the near-field pressure drop, respectively. α, β, γ, δ are the first group of weight coefficients, which can be adjusted according to actual situation and experience to reflect the importance of the degree of deviation and the change trend in the calculation of the failure probability.

[0103] Again, if the far-field hydrogen concentration data is higher than C th but the near-field pressure data has not dropped by more than ΔP th , the near-field flow data is further analyzed. If the near-field flow data has dropped by more than F th , the data fusion model calculates the potential failure probability, i.e. the hydrogen leak or pressure sensor failure probability P potential , by integrating the far-field hydrogen concentration data, the near-field pressure data, and the near-field flow data. Considering that the pressure has not dropped significantly, the flow data has a relatively high weight in the calculation at this time, P potential = f (C - C th , ΔP, F - F th ), where F is the actual near-field flow, and f is a second fusion function.

[0104] Preferably, the following is an example of a fusion function to calculate the hydrogen leak or pressure sensor failure probability:

[0105]

[0106] where P potential is the hydrogen leak or pressure sensor failure probability, C is the far-field hydrogen concentration value, C th is the far-field hydrogen concentration threshold, ΔP is the actual near-field pressure drop, ΔP th is the near-field pressure drop threshold, F is the actual near-field flow value, and F this the near-field flow rate drop threshold value. w1, w2, w3 are the second set of weight coefficients, respectively representing the weight of the degree of deviation of the far-field hydrogen concentration, the degree of deviation of the near-field pressure drop value, and the degree of deviation of the near-field flow rate drop value. Since the flow rate data weight is relatively high when the pressure does not drop significantly, w3> w1 and w3> w2 can be set. The specific weight coefficient values can be adjusted according to actual conditions and experience.

[0107] If the near-field flow rate data does not drop by more than F th , the data fusion model combines the far-field hydrogen concentration data and the near-field flow rate data to determine whether there is a hydrogen concentration sensor failure, and the hydrogen concentration sensor failure probability P sensor = h(C-C th , F), where h is a third fusion function.

[0108] Preferably, the following is an example of a fusion function to calculate the hydrogen concentration sensor failure probability:

[0109]

[0110] where P sensor is the hydrogen concentration sensor failure probability, C is the actual far-field hydrogen concentration, C th is the far-field hydrogen concentration threshold value, F is the actual near-field flow rate, F th is the near-field flow rate drop threshold value. a and b are the third set of weight coefficients, respectively representing the weight of the degree of deviation of the far-field hydrogen concentration and the degree of deviation of the near-field flow rate drop value.

[0111] If the far-field hydrogen concentration data does not show an anomaly, the judgment is mainly based on the near-field flow rate and pressure data.

[0112] When the near-field flow rate data is lower than F th , if the near-field pressure data is higher than the set pressure threshold value P high , the data fusion model calculates the pipeline blockage probability P block = k(F-F th , P-P high ) according to the flow rate and pressure data, where P is the actual near-field pressure, and k is a fourth fusion function.

[0113] Preferably, the following is an example of a fusion function to calculate the pipeline blockage probability:

[0114]

[0115] where P block is the pipeline blockage probability, F is the actual near-field flow rate, F th is the near-field flow rate drop threshold value, P is the actual near-field pressure, and P highare fourth set of weight coefficients, respectively representing the weight of the deviation of the near-field flow drop value and the weight of the degree that the near-field pressure is higher than the threshold value.

[0116] If the near-field pressure data is not higher than P high , the data fusion model comprehensively judges whether there is a hydrogen storage bottle opening defect or a hydrogen concentration sensor failure based on the near-field flow and pressure data, and calculates the failure probability P open = m(F-F th , P), where m is a fifth fusion function.

[0117] Preferably, the following is an example of a fusion function to calculate the probability of hydrogen storage bottle opening defect or hydrogen concentration sensor failure:

[0118]

[0119] where P open is the probability of hydrogen storage bottle opening defect or hydrogen concentration sensor failure, F is the actual near-field flow, F th is the near-field flow drop threshold value, P is the actual near-field pressure, and P normal is the near-field pressure value under normal circumstances, i.e. the near-field pressure normal value. e and f are the fifth set of weight coefficients, respectively representing the weight of the deviation of the near-field flow drop value and the weight of the deviation of the near-field pressure from the normal.

[0120] Step 4: If the hydrogen supply is normal, the fuel cell hydrogen supply system is judged based on the near-field humidity data and the near-field liquid level data and the corresponding threshold values to determine whether there is an abnormal water management, and a water management detection result is obtained.

[0121] Step 4 specifically includes: judging whether the humidity measurement value is abnormal based on the near-field humidity data and the near-field humidity threshold value to obtain a sixth judgment result; wherein the humidity measurement value abnormality includes excessive humidity and insufficient humidity; if the sixth judgment result is yes and the humidity is excessive, then judging whether the liquid level is too high based on the near-field liquid level data and the near-field liquid level threshold value to obtain a seventh judgment result; if the seventh judgment result is yes, a detection result that there is a potential drain valve failure in the fuel cell hydrogen supply system is obtained, and a sixth fusion function is used to calculate the drain valve failure probability; if the seventh judgment result is no, a detection result that there is a potential humidifier failure in the fuel cell hydrogen supply system is obtained, and a seventh fusion function is used to calculate the humidifier failure probability; if the sixth judgment result is yes and the humidity is insufficient, a detection result that there is a potential humidifier failure or purge valve failure in the fuel cell hydrogen supply system is obtained, and an eighth fusion function is used to calculate the humidifier failure or purge valve failure probability; if the sixth judgment result is no, a detection result that the water management of the fuel cell hydrogen supply system is normal is obtained.

[0122] First, the near-field humidity threshold value Hth and near-field liquid level threshold L th .

[0123] If the near-field humidity measurement value exceeds H th If the near-field liquid level sensor reading is higher than L th The data fusion model combines humidity and liquid level data to calculate the failure probability P of the drain valve. drain =n(HH) th ,LL th ), where H is the actual near-field humidity, L is the actual near-field liquid level, and n is the sixth fusion function, indicating that the drain valve is likely to malfunction.

[0124] Preferably, the following is an example of a fusion function to calculate a drain valve fault:

[0125]

[0126] Where H is the actual near-field humidity, H th Here, L represents the near-field humidity threshold, and L represents the actual near-field liquid level. th This represents the near-field liquid level threshold. k1 and k2 are the sixth set of weighting coefficients, representing the weights of the humidity deviation and liquid level deviation, respectively.

[0127] If the near-field liquid level sensor reading is lower than L t h, then calculate the humidifier failure probability P humidifier =p(HH) th ,L), where p is the seventh fusion function, indicating that the humidifier is likely to malfunction.

[0128] Preferably, the following is an example of a fusion function to calculate the humidifier failure probability:

[0129]

[0130] Among them, P humidifier H represents the humidifier failure probability, where H is the actual near-field humidity. th L is the humidity threshold. low It is a specific value where the liquid level is below the threshold. m1 and m2 are the seventh set of weighting coefficients, which respectively represent the weight of the degree of humidity deviation and the degree of liquid level deviation (in the direction below the threshold).

[0131] If the near-field humidity measurement value is lower than H th At that time, the data fusion model only judged the probability P of humidifier failure or purge valve failure based on humidity data. humidifier-purge =q(H), where q is the eighth fusion function, because the contribution of the liquid level data to the judgment of these two faults is relatively small at this time.

[0132] Preferably, the following is an example of a fusion function to calculate humidifier failure or purge valve failure probability:

[0133]

[0134] where P is the humidifier failure or purge valve failure probability, H is the actual near-field humidity, H is the humidity threshold. r1, s1 and t1 are the eighth set of weight coefficients, and u is an adjustment parameter. humidifier-purge th

[0135] Step 5: If water management is normal, determine whether the fuel cell hydrogen supply system has thermal management abnormalities according to the far-field infrared thermal imaging data and the near-field temperature data and the corresponding threshold values, and obtain a thermal management detection result.

[0136] Step 5 specifically includes: determining whether the temperature measurement value is too high according to the near-field temperature data and the near-field temperature threshold value, and obtaining an eighth determination result; if the eighth determination result is yes, determining whether there is a local high temperature according to the far-field infrared thermal imaging data, and obtaining a ninth determination result; if the ninth determination result is yes, obtaining a detection result that there is a potential coolant blockage in the fuel cell hydrogen supply system, and calculating a coolant blockage probability using a ninth fusion function; if the ninth determination result is no, obtaining a detection result that there is a potential cooling circulation pump failure in the fuel cell hydrogen supply system, and calculating a cooling circulation pump failure probability using a tenth fusion function; if the eighth determination result is no, obtaining a detection result that the thermal management of the fuel cell hydrogen supply system is normal.

[0137] First, the near-field temperature threshold T is set based on the above neural network model th .

[0138] If the near-field temperature sensor detects that the temperature exceeds T th , further auxiliary judgment is made according to the far-field infrared thermal imaging data.

[0139] If the infrared thermal imaging data shows that the fuel cell hydrogen supply system has a local high temperature, the data fusion model calculates the coolant blockage failure probability P coolant-block = r (T - T th , I local ), where T is the actual near-field temperature, I local is a local high temperature characteristic parameter of far-field infrared thermal imaging, and r is a ninth fusion function.

[0140] Preferably, the following is an example of a fusion function to calculate coolant blockage failure probability:

[0141]

[0142] where P coolant-block ​​T represents the probability of coolant blockage, and T represents the actual near-field temperature. th I is the near-field temperature threshold. local These are the local high-temperature characteristic parameters of far-field infrared thermal imaging. q1 and q2 are the ninth set of weighting coefficients, representing the weight of the temperature deviation and the local high-temperature characteristic parameters, respectively.

[0143] If infrared thermal imaging data shows that the fuel cell hydrogen supply system has an overall high temperature, calculate the probability P of the cooling circulation pump failure. pump =s(TT) th ,I global ), where I global represents the overall high-temperature characteristic parameters of far-field infrared thermal imaging, and s is the tenth fusion function.

[0144] Preferably, the following is an example of a fusion function to calculate the failure probability of the cooling circulation pump:

[0145]

[0146] Among them, P pump The probability of failure of the cooling circulation pump is given by T, where T is the actual near-field temperature. th I is the near-field temperature threshold. global These are the overall high-temperature characteristic parameters for far-field infrared thermal imaging. q3 and q4 are the tenth set of weighting coefficients, representing the degree of temperature deviation and the weight of the overall high-temperature characteristic parameters, respectively.

[0147] Step 6: Determine the fault detection results of the fuel cell hydrogen supply system based on the hydrogen supply test results, water management test results, and thermal management test results.

[0148] To facilitate understanding of the scheme in this application, a specific example is provided below. For example... Figure 2 As shown, the fault detection process for the fuel cell hydrogen supply system provided in this application includes the following steps:

[0149] S10: Collects data from the far-field hydrogen concentration sensor, far-field infrared thermal imaging sensor, near-field pressure sensor, flow sensor, humidity sensor, liquid level sensor, and temperature sensor, respectively.

[0150] S11: Determine whether the hydrogen concentration in the fuel cell hydrogen supply system environment is too high based on the far-field hydrogen concentration data. If yes, proceed to S12; otherwise, proceed to S17.

[0151] S12: Determine whether the hydrogen pressure in the fuel cell hydrogen supply system is too high based on the near-field pressure sensor data. If yes, proceed to S13; otherwise, proceed to S14.

[0152] S13: Calculate the probability of hydrogen leakage in the fuel cell hydrogen supply system.

[0153] S14: Determine if the hydrogen flow in the fuel cell hydrogen supply system is too low based on near field flow sensor data. If yes, go to S15. If no, go to S16.

[0154] S15: Calculate the probability that the fuel cell hydrogen supply system has a hydrogen leak or a near field pressure sensor malfunction.

[0155] S16: Calculate the probability that the fuel cell hydrogen supply system has a hydrogen concentration sensor malfunction.

[0156] S17: Determine if the hydrogen flow in the fuel cell hydrogen supply system is too low based on near field flow sensor data. If yes, go to S18. If no, go to S21.

[0157] S18: Determine if the hydrogen pressure in the fuel cell hydrogen supply system is too high based on near field pressure sensor data. If yes, go to S19. If no, go to S20.

[0158] S19: Calculate the probability that the fuel cell hydrogen supply system has a hydrogen supply line blockage.

[0159] S20: Calculate the probability that the fuel cell hydrogen supply system has a hydrogen storage cylinder that is not opening properly or a hydrogen concentration sensor malfunction.

[0160] S21: Determine if the hydrogen humidity entering the fuel cell anode is too high or too low based on near field humidity sensor data. If too high, go to S22. If too low, go to S23.

[0161] S22: Get the result of the determination that the hydrogen humidity entering the fuel cell anode is too high.

[0162] S23: Get the result of the determination that the hydrogen humidity entering the fuel cell anode is too low.

[0163] S24: Determine if the liquid level in the fuel cell hydrogen supply system water and gas separator is too high based on near field liquid level sensor data. If yes, go to S26. If no, go to S27.

[0164] S25: Calculate the probability that the fuel cell hydrogen supply system has a humidifier malfunction or a purge valve malfunction.

[0165] S26: Calculate the probability that the fuel cell hydrogen supply system has a drain valve malfunction.

[0166] S27: Calculate the probability that the fuel cell hydrogen supply system has a humidifier malfunction.

[0167] S28: Determine if the hydrogen temperature in the fuel cell hydrogen supply system is too high based on near field temperature sensor data. If yes, go to S29. If no, end the fault determination process.

[0168] S29: judging whether the temperature distribution image obtained by the infrared thermal imaging sensor indicates an overall high temperature or a local high temperature of the fuel cell hydrogen supply pipeline, if the temperature distribution image indicates a local high temperature, entering S30, if the temperature distribution image indicates an overall high temperature, entering S31.

[0169] S30: obtaining a detection result that the fuel cell hydrogen supply system has a local high temperature.

[0170] S31: obtaining a detection result that the fuel cell hydrogen supply system has an overall high temperature.

[0171] S32: calculating a probability of a coolant blockage of the fuel cell.

[0172] S33: calculating a probability of a cooling system pump failure of the fuel cell.

[0173] In summary, the present application provides a fuel cell hydrogen supply system fault detection method, based on adaptive threshold near-field and far-field sensor linkage, the system overall state is monitored by the far-field sensor, and the local details are monitored by the near-field sensor, comprehensive monitoring is realized, and the far-field sensor data and the near-field sensor data are combined, the fault area can be more accurately located, and the reliability of the detection result is improved.

[0174] In an exemplary embodiment, the present application further provides a computer device including a memory and a processor, the memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0175] In an exemplary embodiment, the present application further provides a computer readable storage medium storing a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0176] In an exemplary embodiment, the present application further provides a computer program product including a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0177] In the present application, all actions of obtaining signals, information or data are performed in compliance with the corresponding data protection regulations and policies of the country where the device is located, and with the authorization of the corresponding device owner. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data need to comply with relevant legal regulations.

[0178] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0179] The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0180] The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, they should be considered as the scope of the present application.

[0181] The principles and implementation manners of the present application are described herein by using specific examples, and the above examples are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manners and application ranges will have changes. In conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A fault detection method for a fuel cell hydrogen supply system, characterized in that, include: Collect far-field and near-field sensor data from the fuel cell hydrogen supply system; The far-field sensor data includes: far-field hydrogen concentration data and far-field infrared thermal imaging data; the near-field sensor data includes: near-field pressure data, near-field flow rate data, near-field humidity data, near-field liquid level data, and near-field temperature data. The system identifies the vehicle's operating conditions and uses a neural network model to output the far-field sensor thresholds and near-field sensor thresholds under the current operating conditions. The far-field sensor thresholds include: far-field hydrogen concentration threshold; the near-field sensor thresholds include: near-field pressure drop threshold, near-field flow drop threshold, near-field humidity threshold, near-field liquid level threshold, and near-field temperature threshold. Based on far-field hydrogen concentration data, near-field pressure data, near-field flow data, and corresponding thresholds, it is determined whether there is an abnormality in the hydrogen supply of the fuel cell hydrogen supply system, and the hydrogen supply detection results are obtained. If the hydrogen supply is normal, the near-field humidity data, near-field liquid level data and corresponding thresholds are used to determine whether there is an abnormality in the water management of the fuel cell hydrogen supply system, and the water management test results are obtained. If the water management is normal, the thermal management of the fuel cell hydrogen supply system is determined based on the far-field infrared thermal imaging data, near-field temperature data, and corresponding thresholds, and the thermal management detection results are obtained. The fault detection results of the fuel cell hydrogen supply system are determined based on the hydrogen supply test results, water management test results, and thermal management test results. The method for determining the neural network model includes: Historical data of the vehicle under different operating conditions are collected to construct a historical dataset; the operating conditions include: starting condition, acceleration condition, constant speed condition, idling condition and braking condition; the historical data includes: operating condition data, sensor data and sensor thresholds; the operating condition data includes: brake pedal opening, accelerator pedal opening and vehicle speed. Using operating condition category, operating condition data, and sensor data as inputs, and sensor thresholds as outputs, a multilayer perceptron neural network is constructed. Based on historical datasets, an optimization algorithm is used to train the multilayer perceptron neural network to obtain a trained neural network model.

2. The method for fault detection of a fuel cell hydrogen supply system according to claim 1, characterized in that, Based on far-field hydrogen concentration data, near-field pressure data, and near-field flow rate data, along with corresponding thresholds, the system is assessed to determine if there are any abnormalities in the hydrogen supply of the fuel cell hydrogen supply system. The resulting hydrogen supply detection results include: Based on the far-field hydrogen concentration data and the far-field hydrogen concentration threshold, it is determined whether the far-field hydrogen concentration is too high, and the first judgment result is obtained; If the first judgment result is yes, then the pressure drop is determined based on the near-field pressure data and the near-field pressure drop threshold to determine whether the pressure drop exceeds the threshold, and the second judgment result is obtained; If the second judgment result is yes, then the detection result of potential hydrogen leakage in the fuel cell hydrogen supply system is obtained, and the probability of hydrogen leakage is calculated using the first fusion function. If the second judgment result is negative, then the near-field flow data and the near-field flow drop threshold are used to determine whether the flow is too low, and a third judgment result is obtained. If the third judgment result is yes, then the detection result of potential hydrogen leakage or pressure sensor failure in the fuel cell hydrogen supply system is obtained, and the probability of hydrogen leakage or pressure sensor failure is calculated by the second fusion function. If the third judgment result is negative, the detection result of potential hydrogen concentration sensor failure in the fuel cell hydrogen supply system is obtained, and the probability of hydrogen concentration sensor failure is calculated using the third fusion function. If the first judgment result is negative, then the fourth judgment result is obtained by judging whether the flow measurement value is too low based on the near-field flow data and the near-field flow drop threshold. If the fourth judgment result is yes, then the pressure measurement value is judged to be too high based on the near-field pressure data, and the fifth judgment result is obtained; If the fifth judgment result is yes, then the detection result of potential pipeline blockage in the fuel cell hydrogen supply system is obtained, and the probability of pipeline blockage is calculated using the fourth fusion function. If the fifth judgment result is negative, the detection result of potential hydrogen storage tank opening failure or hydrogen concentration sensor failure in the fuel cell hydrogen supply system is obtained, and the probability of hydrogen storage tank opening failure or hydrogen concentration sensor failure is calculated by the fifth fusion function. If the fourth judgment result is negative, then the detection result of normal hydrogen supply in the fuel cell hydrogen supply system is obtained.

3. The method for fault detection of a fuel cell hydrogen supply system according to claim 2, characterized in that, Based on near-field humidity data, near-field liquid level data, and corresponding thresholds, the system determines whether there are any water management anomalies in the fuel cell hydrogen supply system, and obtains the water management detection results, including: Based on near-field humidity data and near-field humidity threshold, the sixth judgment result is obtained to determine whether the humidity measurement value is abnormal; among them, abnormal humidity measurement value includes excessively high humidity and excessively low humidity. If the sixth judgment result is yes, and the humidity is too high, then the liquid level is judged to determine whether the liquid level is too high based on the near-field liquid level data and the near-field liquid level threshold, and the seventh judgment result is obtained. If the result of the seventh judgment is yes, then the detection result of potential drain valve failure in the fuel cell hydrogen supply system is obtained, and the probability of drain valve failure is calculated using the sixth fusion function. If the result of the seventh judgment is negative, the detection result of potential humidifier failure in the fuel cell hydrogen supply system is obtained, and the probability of humidifier failure is calculated using the seventh fusion function. If the sixth judgment result is yes and the humidity is too low, then the detection result of potential humidifier failure or purge valve failure in the fuel cell hydrogen supply system is obtained, and the probability of humidifier failure or purge valve failure is calculated by the eighth fusion function. If the result of the sixth judgment is negative, then the test result of normal water management in the fuel cell hydrogen supply system is obtained.

4. The method for fault detection of a fuel cell hydrogen supply system according to claim 3, characterized in that, Based on far-field infrared thermal imaging data and near-field temperature data, along with corresponding thresholds, the system is assessed for thermal management anomalies in the fuel cell hydrogen supply system. The resulting thermal management detection results include: Based on the near-field temperature data and the near-field temperature threshold, it is determined whether the temperature measurement value is too high, and the eighth judgment result is obtained. If the eighth judgment result is yes, then the existence of local high temperature is determined based on the far-field infrared thermal imaging data, and the ninth judgment result is obtained; If the result of the ninth judgment is yes, then the detection result of potential coolant blockage in the fuel cell hydrogen supply system is obtained, and the probability of coolant blockage is calculated using the ninth fusion function. If the result of the ninth judgment is negative, the detection result of potential cooling circulation pump failure in the fuel cell hydrogen supply system is obtained, and the failure probability of the cooling circulation pump is calculated using the tenth fusion function. If the result of the eighth judgment is negative, then the detection result of normal thermal management in the fuel cell hydrogen supply system is obtained.

5. The method for fault detection of a fuel cell hydrogen supply system according to claim 2, characterized in that, The hydrogen supply detection results include: hydrogen leakage, pressure sensor failure, hydrogen concentration sensor failure, pipeline blockage, improper opening of hydrogen storage cylinder, and normal hydrogen supply. The expression for calculating the probability of hydrogen leakage using the first fusion function is as follows: Among them, P leak Let C be the probability of hydrogen leakage, and C be the actual far-field hydrogen concentration. th ΔP represents the far-field hydrogen concentration threshold, and ΔP represents the actual near-field pressure drop. th Here, t represents the near-field pressure drop threshold, α, β, γ, and δ are the first set of weighting coefficients; The expression for calculating the probability of hydrogen leakage or pressure sensor failure using the second fusion function is as follows: Among them, P potential F represents the probability of hydrogen leakage or pressure sensor failure, and F is the actual near-field flow rate. th The threshold for near-field flow decrease is defined, and w1, w2, and w3 are the second set of weighting coefficients. The expression for calculating the failure probability of the hydrogen concentration sensor using the third fusion function is as follows: Among them, P sensor denoted as the failure probability of the hydrogen concentration sensor, and a and b are the third set of weighting coefficients; The expression for calculating the probability of pipeline blockage using the fourth fusion function is as follows: Among them, P block Let P be the probability of pipe blockage, and P be the actual near-field pressure. high Here, c and d represent the near-field pressure threshold, and c and d are the fourth set of weighting coefficients. The expression for calculating the probability of a faulty hydrogen storage tank opening or a malfunctioning hydrogen concentration sensor using the fifth fusion function is as follows: Among them, P open P represents the probability of a faulty hydrogen storage tank opening or a malfunctioning hydrogen concentration sensor. normal The near-field pressure is the normal value, and e and f are the weighting coefficients of the fifth group.

6. The fault detection method for a fuel cell hydrogen supply system according to claim 3, characterized in that, The water management test results include: drain valve failure, humidifier failure, purge valve failure, and normal water management. The expression for calculating the failure probability of the drain valve using the sixth fusion function is as follows: Among them, P drain H represents the probability of drain valve failure, and H represents the actual near-field humidity. th Here, L represents the near-field humidity threshold, and L represents the actual near-field liquid level. th The near-field liquid level threshold is defined by k1 and k2, which are the sixth set of weighting coefficients. The expression for calculating the humidifier failure probability using the seventh fusion function is as follows: Among them, P humidifier L represents the probability of humidifier failure. low For a specific value where the liquid level is below the threshold, m1 and m2 are the seventh set of weighting coefficients; The expression for calculating the probability of humidifier failure or purge valve failure using the eighth fusion function is as follows: Among them, P humidifier-purge The probability of humidifier failure or purge valve failure is represented by r1, s1, and t1, which are the weighting coefficients of the eighth group, and u is the adjustment parameter.

7. The method for fault detection of a fuel cell hydrogen supply system according to claim 4, characterized in that, The thermal management test results include: coolant blockage, cooling circulation pump failure, and normal thermal management. The expression for calculating the probability of coolant blockage using the ninth fusion function is as follows: Among them, P coolant-block T represents the probability of coolant blockage, and T represents the actual near-field temperature. th I is the near-field temperature threshold. local These are the local high-temperature characteristic parameters of far-field infrared thermal imaging, and q1 and q2 are the ninth set of weighting coefficients. The expression for calculating the failure probability of the cooling circulation pump using the tenth fusion function is as follows: Among them, P pump For the failure probability of the cooling circulation pump, I global q3 and q4 are the overall high-temperature characteristic parameters of far-field infrared thermal imaging, and the tenth set of weighting coefficients.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the fault detection method for a fuel cell hydrogen supply system according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the fault detection method for a fuel cell hydrogen supply system as described in any one of claims 1-7.

Images