A method and device for intelligent operation monitoring of PEM fuel cells

Abstract

This invention discloses an intelligent operation monitoring method and device for PEM fuel cells, relating to the field of intelligent monitoring technology. The method includes preprocessing voltage and temperature data to eliminate dimensional differences and numerical range interference between the original data, and then employing a dual-path analysis strategy on the preprocessed data: firstly, obtaining a first failure rate through data fault analysis; secondly, obtaining a second failure rate by substituting the image encoding and flattening reconstruction operation into a lightweight long short-term network model, and finally obtaining the operation status monitoring result based on the first and second failure rates. This allows for multi-dimensional evaluation of the operation status and improves the accuracy of operation monitoring.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring technology, specifically relating to an intelligent operation monitoring method and device for PEM fuel cells. Background Technology

[0002] PEM fuel cells, with their high energy conversion efficiency, rapid response, and zero emissions, play a crucial role in new energy vehicles, distributed energy stations, and other fields. As their application scenarios continue to expand, accurate monitoring of the operating status of PEM fuel cells has become a core requirement for ensuring energy supply stability and equipment safety. During operation, faults such as abnormal voltage fluctuations and sudden temperature changes not only degrade battery performance but may also lead to safety accidents. Existing PEM fuel cell operation monitoring technologies have significant shortcomings in accuracy. Traditional threshold judgment methods rely solely on comparing single voltage or temperature data with fixed thresholds, failing to consider the coupling relationships between data and the impact of changes in operating conditions.

[0003] Patent CN116995280A discloses an intelligent monitoring method for the operating status of a hydrogen fuel cell, relating to the field of battery monitoring. The method includes: obtaining a predetermined set of monitoring variables for a first hydrogen fuel cell, the predetermined set including M predetermined monitoring variables; constructing M battery state variable sensing and monitoring modules; obtaining a set of monitoring data for the M variables based on the M battery state variable sensing and monitoring modules; performing deviation analysis on the set of monitoring data for the M variables based on a variable state deviation analysis function to obtain the deviation analysis results for the M variables; detecting the deviation analysis results for the M variables using a battery state detection module to obtain the battery state detection results; and obtaining a battery state warning signal when the battery state detection results do not meet preset battery state constraints. This solves the technical problems of low accuracy and low efficiency in the state monitoring and warning of hydrogen fuel cells in existing technologies. However, in practical applications, due to insufficient consideration of the coupling relationship between voltage and temperature data and failure to adapt to the impact of changes in operating conditions, it is difficult to further improve the accuracy of operation monitoring in complex operating scenarios. Summary of the Invention

[0004] The purpose of this invention is to solve the problem of low accuracy in monitoring operating status, and to propose an intelligent operation monitoring method and device for PEM fuel cells.

[0005] In a first aspect of this invention, a smart operation monitoring method for a PEM fuel cell is first proposed, the method comprising:

[0006] Simultaneously collect voltage and temperature data from the PEM fuel cell;

[0007] The voltage data and temperature data are preprocessed respectively to obtain effective voltage data and effective temperature data;

[0008] A first failure rate is obtained by performing data fault analysis on the effective voltage data and the effective temperature data.

[0009] A two-dimensional image is obtained by performing two-dimensional data conversion on the effective voltage data and the effective temperature data through image encoding operations.

[0010] The two-dimensional image is flattened and reconstructed to obtain a temporal feature sequence. The temporal feature sequence is then substituted into a lightweight long short-term network model to obtain a second failure rate.

[0011] The operational status monitoring results are obtained based on the first failure rate and the second failure rate.

[0012] Optionally, the voltage data and the temperature data can be normalized to obtain normalized voltage data and normalized temperature data.

[0013] The normalized voltage data and the normalized temperature data are divided into sub-voltage datasets and sub-temperature datasets by using a preset sliding window.

[0014] The effective voltage data and effective temperature data are obtained by replacing the average value of the sub-voltage data and sub-temperature data within each window with the data points within the window.

[0015] Optionally, the effective voltage data and the effective temperature data can be divided into sub-effective voltage datasets and sub-effective temperature datasets by using a preset sliding window;

[0016] Calculate the voltage slope and temperature slope of the sub-effective voltage data and sub-effective temperature data within each window. If the absolute value of the voltage slope is greater than the preset voltage slope threshold or the absolute value of the temperature slope is greater than the preset temperature slope threshold, then mark the data points within the window and obtain the number of fault points.

[0017] The ratio of the total number of fault points to the total number of data points is taken as the first failure rate.

[0018] Optionally, the effective voltage data and the effective temperature data can be arranged in timestamp order to obtain a one-dimensional sequence set;

[0019] Normalize the one-dimensional data in the one-dimensional sequence set to obtain a normalized dataset;

[0020] A two-dimensional feature matrix is ​​obtained by performing a preset inverse cosine normalization operation on the normalized data in the normalized dataset.

[0021] A two-dimensional image is obtained by mapping the values ​​in the two-dimensional feature matrix to preset pixel grayscale color values.

[0022] Optionally, a low-rank approximation matrix can be obtained by performing a compression matrix operation on the weight matrix of the Long Short-Term Memory network.

[0023] The low-rank approximation matrix replaces the original weight matrix of the long short-term network in the calculation to obtain a lightweight long short-term network.

[0024] The temporal feature sequence is processed step by step using a lightweight long short-term network to obtain a forward state vector and a reverse state vector, respectively, starting from the beginning of the temporal feature sequence in a forward direction and reversing from the end of the temporal feature sequence.

[0025] The positive state vector and the negative state vector are concatenated to obtain the fused state vector;

[0026] The second failure rate is obtained by performing a nonlinear transformation on the fused feature vector using a preset monitoring formula.

[0027] In a second aspect of the invention, an intelligent operation monitoring device for a PEM fuel cell is provided, comprising:

[0028] The data acquisition module is used to acquire voltage and temperature data of the PEM fuel cell; the voltage and temperature data are data synchronously acquired at a preset period before the current moment;

[0029] The preprocessing module is used to preprocess the voltage data and the temperature data respectively to obtain effective voltage data and effective temperature data;

[0030] The first failure rate generation module is used to perform data failure analysis on the effective voltage data and the effective temperature data to obtain a first failure rate.

[0031] A two-dimensional image generation module is used to perform two-dimensional data conversion on the effective voltage data and the effective temperature data through image encoding operations to obtain a two-dimensional image.

[0032] The second failure rate generation module is used to perform a flattening dimension reconstruction operation on the two-dimensional image to obtain a temporal feature sequence, and to substitute the temporal feature sequence into a lightweight long short-term network model to obtain a second failure rate.

[0033] The detection result generation module is used to obtain the operation status monitoring results based on the first failure rate and the second failure rate.

[0034] Optionally, the preprocessing module includes:

[0035] The first normalization module is used to normalize the voltage data and the temperature data to obtain normalized voltage data and normalized temperature data.

[0036] The first data segmentation module is used to divide the normalized voltage data and the normalized temperature data into sub-voltage datasets and sub-temperature datasets through a first preset sliding window;

[0037] The data replacement module is used to replace the data points in each window with the average value of the sub-voltage data and sub-temperature data to obtain the effective voltage data and effective temperature data.

[0038] Optionally, the first failure rate generation module includes:

[0039] The second data segmentation module is used to divide the effective voltage data and the effective temperature data into sub-effective voltage datasets and sub-effective temperature datasets through a second preset sliding window;

[0040] The marking module is used to calculate the voltage slope and temperature slope of sub-effective voltage data and sub-effective temperature data in each window. If the absolute value of the voltage slope is greater than the preset voltage slope threshold or the absolute value of the temperature slope is greater than the preset temperature slope threshold, the data points in the window are marked and the number of fault points is obtained.

[0041] The first failure rate module is used to determine the ratio of the number of all failure points to the number of data points as the first failure rate.

[0042] Optionally, the two-dimensional image generation module includes:

[0043] The sorting module is used to arrange the effective voltage data and the effective temperature data in timestamp order to obtain a one-dimensional sequence set;

[0044] The second normalization module is used to normalize the one-dimensional data in the one-dimensional sequence set to obtain a normalized dataset.

[0045] The third normalization module is used to perform a preset inverse cosine normalization operation on the normalized data in the normalized dataset to obtain a two-dimensional feature matrix.

[0046] The mapping module is used to map the values ​​in the two-dimensional feature matrix to preset pixel grayscale color values ​​to obtain a two-dimensional image.

[0047] Optionally, the second failure rate generation module includes:

[0048] The compression module is used to perform matrix compression operations on the weight matrix of the Long Short-Term Memory network to obtain a low-rank approximate matrix.

[0049] The replacement module is used to replace the original weight matrix of the long short-term network with the low-rank approximation matrix in the calculation to obtain a lightweight long short-term network.

[0050] The bidirectional processing module is used to process the temporal feature sequence step by step using a lightweight long short-term network to obtain a forward state vector and a reverse state vector, respectively, from the beginning of the temporal feature sequence in a forward direction and from the end of the temporal feature sequence in a reverse direction.

[0051] The splicing module is used to splice the positive state vector and the negative state vector to obtain a fused state vector;

[0052] The second failure rate module is used to perform a nonlinear transformation on the fused feature vector using a preset monitoring formula to obtain a second failure rate.

[0053] The beneficial effects of this invention are as follows: Voltage and temperature data of the PEM fuel cell are collected simultaneously; the voltage and temperature data are preprocessed to obtain effective voltage and effective temperature data; data fault analysis is performed on the effective voltage and effective temperature data to obtain a first failure rate; two-dimensional data conversion is performed on the effective voltage and effective temperature data through image encoding to obtain a two-dimensional image; a flattening and dimensional reconstruction operation is performed on the two-dimensional image to obtain a time-series feature sequence, which is then substituted into a lightweight long short-term network model to obtain a second failure rate; and the operating status monitoring results are obtained based on the first and second failure rates. By preprocessing the voltage and temperature data to eliminate dimensional differences and numerical range interference between the original data, a dual-path analysis strategy is adopted on the preprocessed data: firstly, the first failure rate is obtained through data fault analysis; secondly, the second failure rate is obtained by substituting the image encoding and flattening / reconstruction operation into a lightweight long short-term network model. Finally, the operating status monitoring results are obtained based on the first and second failure rates, thereby evaluating the operating status from multiple dimensions and improving the accuracy of operating monitoring. Attached Figure Description

[0054] The invention will now be further described with reference to the accompanying drawings.

[0055] Figure 1 A flowchart illustrating an intelligent operation monitoring method for a PEM fuel cell provided in an embodiment of the present invention;

[0056] Figure 2 A flowchart of another intelligent operation monitoring method for PEM fuel cells provided in an embodiment of the present invention;

[0057] Figure 3 A framework diagram of an intelligent operation monitoring device for a PEM fuel cell provided in an embodiment of the present invention; Detailed Implementation

[0058] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0059] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] This invention provides an intelligent operation monitoring method for PEM fuel cells. See also... Figure 1 , Figure 1 A flowchart illustrating an intelligent operation monitoring method for a PEM fuel cell provided in an embodiment of the present invention. The method includes the following steps:

[0061] S101 collects voltage and temperature data from the PEM fuel cell;

[0062] S102, preprocess the voltage data and temperature data respectively to obtain effective voltage data and effective temperature data;

[0063] S103, perform data fault analysis on effective voltage data and effective temperature data to obtain the first failure rate;

[0064] S104, a two-dimensional image is obtained by performing two-dimensional data conversion on effective voltage data and effective temperature data through image encoding operation;

[0065] S105, flatten the two-dimensional image to reconstruct the temporal feature sequence, and substitute the temporal feature sequence into the lightweight long short-term network model to obtain the second failure rate.

[0066] S106, the operating status monitoring results are obtained based on the first failure rate and the second failure rate.

[0067] Based on the intelligent operation monitoring method for PEM fuel cells provided in this invention, voltage and temperature data are preprocessed to eliminate dimensional differences and numerical range interference between the original data. Then, a dual-path analysis strategy is adopted for the preprocessed data: first, a first failure rate is obtained through data fault analysis; second, a second failure rate is obtained by substituting the image encoding and flattening reconstruction operation into a lightweight long short-term network model. Finally, the operation status monitoring results are obtained based on the first and second failure rates, thereby evaluating the operation status from multiple dimensions and improving the accuracy of operation monitoring.

[0068] In one implementation, the flattening dimension reconstruction operation involves performing a fixed-size convolution operation on the two-dimensional image to obtain a multi-channel feature map, then performing max pooling on the multi-channel feature map to obtain a fixed-dimensional feature map, and finally obtaining a temporal feature sequence by traversing each element of the fixed-dimensional feature map and arranging them in order.

[0069] In one implementation, the comprehensive failure rate is calculated using a first failure rate and a second failure rate. The comprehensive failure rate is calculated as follows: Comprehensive failure rate = α * first failure rate + (1 − α) * second failure rate. If the comprehensive failure rate is greater than or equal to a threshold A, the operating status is abnormal. If the comprehensive failure rate is less than a threshold A, the operating status is normal. The threshold A is determined by technical personnel.

[0070] In one implementation, by synchronously collecting voltage and temperature data, the dynamic change trend of the battery over a period of time can be fully captured, avoiding the random errors of single-point data.

[0071] In one implementation, the raw data may contain outliers caused by sensor errors or system fluctuations, and preprocessing can improve data reliability.

[0072] In one implementation, a preset sliding window is used to calculate the slope of voltage and temperature, and the fault points are marked by comparing them with a threshold. This quickly locks in the abnormal trend of the data and quantifies the proportion of fault points as the first fault rate, providing an intuitive indicator for rapid assessment of the operating status. The voltage and temperature data are converted into two-dimensional images to realize the conversion of time domain to spatial features. The advantages of image visualization are used to highlight the data distribution pattern. Combined with inverse cosine normalization and pixel mapping, the feature discrimination is enhanced, and the fault identification capability is improved.

[0073] In one implementation, one-dimensional time-series data is converted into a two-dimensional image, which preserves the temporal correlation and spatial features of the data, providing richer input features for deep learning models. Image encoding can compress high-dimensional time-series data into a two-dimensional matrix, reducing data redundancy while retaining key features, improving the computational efficiency of subsequent models, and integrating the failure rates obtained from traditional computation and deep learning to achieve complementary rapid initial judgment and deep diagnosis. This comprehensively improves the accuracy of operational status assessment, reduces false alarms and missed alarms, and provides a reliable basis for monitoring results.

[0074] In one embodiment, voltage data and temperature data are normalized to obtain normalized voltage data and normalized temperature data.

[0075] The normalized voltage data and normalized temperature data are divided into sub-voltage datasets and sub-temperature datasets by using a first preset sliding window.

[0076] The effective voltage data and effective temperature data are obtained by replacing the average value of the sub-voltage data and sub-temperature data within each window with the data points within the window.

[0077] In one implementation, all data points for voltage and temperature data are iterated through, and the mean voltage μ1 and mean temperature μ2 are calculated respectively. Then, the standard deviations of voltage and temperature data σ1 and σ2 are calculated. For each voltage data point Vi, the normalized value Vi = For each temperature data point Tj, calculate the normalized value Tj= .

[0078] In one implementation, the size n of the first preset sliding window is determined by the technician. Starting from the first data point in the normalized voltage data, the average value of the first to nth data points is calculated and replaced with the original values ​​of the first to nth data points. The window slides one data point to the right, and the average value of the second to (n+1)th data points is calculated and replaced with the values ​​of the second to (n+1)th data points. This process is repeated until all data is traversed. The processing operation for the normalized temperature data is similar.

[0079] In one embodiment, the effective voltage data and effective temperature data are divided into sub-effective voltage datasets and sub-effective temperature datasets by using a second preset sliding window;

[0080] Calculate the voltage slope and temperature slope of the sub-effective voltage data and sub-effective temperature data within each window. If the absolute value of the voltage slope is greater than the preset voltage slope threshold or the absolute value of the temperature slope is greater than the preset temperature slope threshold, then mark the data points within the window and obtain the number of fault points.

[0081] The ratio of the total number of fault points to the total number of data points is taken as the first failure rate.

[0082] In one implementation, the second preset sliding window m, the preset voltage threshold, and the preset temperature threshold are determined by technicians. The step size is the position of one data point. For the effective voltage data V1, V2, ..., Vm within each window, the voltage slope can be calculated by fitting the data with the straight line V = a*t + b, where a is the voltage slope, and the time indices corresponding to the m data points within the window are t1, t2, ..., tm. m Then the slope a = The calculation of the temperature slope for the effective temperature data T1, T2, ..., Tm within each window is similar.

[0083] In one implementation, a sliding window is used to calculate segment by segment, which can capture sudden changes or abnormal accelerated changes in data in real time. The fixed window size can balance the sensitivity of short-term fluctuations with the stability of long-term trends. When the absolute value of the slope exceeds the threshold, the data points in the window are marked, which can directly locate the time interval of the fault and achieve accurate positioning.

[0084] In one embodiment, see Figure 2 ,exist Figure 1 Following step S104, the method further includes:

[0085] S1041, Arrange the effective voltage data and effective temperature data in the order of timestamps to obtain a one-dimensional sequence set;

[0086] S1042, Normalize the one-dimensional data in the one-dimensional sequence set to obtain a normalized dataset;

[0087] S1043, Perform a preset inverse cosine normalization operation on the normalized data in the normalized dataset to obtain a two-dimensional feature matrix;

[0088] S1044, performs pixel grayscale mapping on the values ​​in the two-dimensional feature matrix to obtain a two-dimensional image.

[0089] In one implementation, sorting by timestamp refers to arranging the collected effective voltage and effective temperature data in the order of voltage and temperature according to the time sequence of each data point to obtain a one-dimensional sequence set.

[0090] In one implementation, the normalization operation can be linear normalization, Z-Score normalization, etc., and the preset inverse cosine normalization operation is obtained based on the staff's historical experience.

[0091] In one implementation, the matrix numerical range is linearly mapped to the 0-255 interval, and the grayscale value can be calculated using the formula: grayscale value = .

[0092] In one implementation, the normalized data meets the input requirements of the machine learning model, which can reduce oscillations during gradient updates, improve model training efficiency and feature extraction effect. Inverse cosine normalization can transform the linear relationship of one-dimensional data into a non-linear distribution in two-dimensional space, expose the features that were hidden in the original data. After mapping the values ​​to grayscale values, the two-dimensional image can intuitively show the distribution pattern of the data, which not only preserves the temporal information of the original data, but also mines the hidden correlation features through spatial transformation and image processing.

[0093] In one embodiment, a low-rank approximation matrix is ​​obtained by performing a compression matrix operation on the weight matrix of the Long Short-Term Memory network.

[0094] A lightweight long short-term network is obtained by replacing the original weight matrix of the long short-term network with a low-rank approximation matrix in the calculation.

[0095] Lightweight long short-term network is used to process the temporal feature sequence step by step, starting from the beginning of the temporal feature sequence in the forward direction and in the reverse direction from the end of the temporal feature sequence, to obtain the forward state vector and the reverse state vector.

[0096] The fused state vector is obtained by concatenating the forward state vector and the reverse state vector.

[0097] The second failure rate is obtained by performing a nonlinear transformation on the fused feature vector using a preset monitoring formula.

[0098] The fused state vector is obtained by concatenating the forward state vector and the reverse state vector.

[0099] The second failure rate is obtained by performing a nonlinear transformation on the fused feature vector using a preset monitoring formula.

[0100] In one implementation, after inputting the temporal feature sequence, the process proceeds forward step by step from the beginning of the sequence: First, singular value decomposition is performed on the weight matrix, breaking it down into a left singular matrix, a singular value matrix, and a right singular matrix. Then, the top k largest singular values ​​are selected (dynamically adjusted by technicians based on actual conditions). These selected singular values ​​are used to reconstruct a low-rank approximation matrix, which is then used in the calculation to generate a forward state vector. The same steps are performed backward step by step from the end of the temporal feature sequence, but the singular value retention ratio of the weights at the last time step is dynamically increased to strengthen the capture of key information in the reverse temporal sequence, generating a reverse state vector. Compared to lightweight long short-term memory networks, in terms of matrix processing mechanisms, traditional long short-term memory networks use full-rank matrices directly in the calculation, with the weight matrix dimension being W∈ The previous method had high parameter redundancy, computational complexity of O(mn), and no decomposition optimization measures. It relied on post-processing methods such as model compression or quantization, which could easily damage the network structure. The new method performs singular value decomposition on the forward and reverse weight matrices and reconstructs the low-rank matrix by retaining the first k largest singular values, reducing the computational complexity to O(mk+kn).

[0101] In one implementation, the second failure rate = σ ), where σ can be an activation function such as Sigmoid or ReLU, and 2 indicates that there are two subspace characteristic voltages and temperatures. The importance coefficients represent the features of the i-th subspace. These are the feature weighting coefficients of the i-th subspace. It is the offset of the i-th subspace. , , Determined by the historical experience of technical personnel, These are local feature fragments extracted from the original data. For example, fuel cell time series data can be divided into multiple subspaces based on voltage trends, temperature trends, and voltage-temperature correlations.

[0102] In one implementation, one-dimensional time-series data is transformed into high-dimensional features containing causal relationships. This retains the advantages of LSTM in processing long sequences and breaks through the bottleneck of unidirectional information flow through a bidirectional mechanism. It achieves high-precision identification of complex operating status monitoring patterns with low computational cost, while providing interpretable operating status analysis basis. It is especially suitable for PEM fuel cell systems with extremely high requirements for safety and reliability.

[0103] Based on the same inventive concept, this invention also provides an intelligent operation monitoring device for a PEM fuel cell. See also Figure 3 , Figure 3 A framework diagram of an intelligent operation monitoring device for a PEM fuel cell provided in an embodiment of the present invention includes:

[0104] The data acquisition module is used to acquire voltage and temperature data of the PEM fuel cell;

[0105] The preprocessing module is used to preprocess the voltage data and temperature data respectively to obtain effective voltage data and effective temperature data;

[0106] The first failure rate generation module is used to perform data failure analysis on effective voltage data and effective temperature data to obtain the first failure rate.

[0107] The two-dimensional image generation module is used to convert effective voltage data and effective temperature data into two-dimensional images through image encoding operations.

[0108] The second failure rate generation module is used to perform a flattening and dimensional reconstruction operation on the two-dimensional image to obtain a temporal feature sequence, and then substitute the temporal feature sequence into a lightweight long short-term network model to obtain the second failure rate.

[0109] The detection result generation module is used to obtain the operational status monitoring results based on the first failure rate and the second failure rate.

[0110] Among them, the voltage data and temperature data are data that were synchronously collected at a preset period before the current moment.

[0111] Based on the intelligent operation monitoring device for PEM fuel cells provided in this embodiment of the invention, the device preprocesses voltage and temperature data to eliminate dimensional differences and numerical range interference between the original data. Then, it employs a dual-path analysis strategy on the preprocessed data: first, a first failure rate is obtained through data fault analysis; second, a second failure rate is obtained by substituting the image encoding and flattening reconstruction operation into a lightweight long short-term network model. Finally, the operation status monitoring results are obtained based on the first and second failure rates, thereby evaluating the operation status from multiple dimensions and improving the accuracy of operation monitoring.

[0112] In one embodiment, the preprocessing module includes:

[0113] The first normalization module is used to normalize voltage and temperature data to obtain normalized voltage and temperature data.

[0114] The first data segmentation module is used to divide the normalized voltage data and normalized temperature data into sub-voltage datasets and sub-temperature datasets through a first preset sliding window.

[0115] The data replacement module is used to replace the data points in each window with the average value of the sub-voltage data and sub-temperature data to obtain the effective voltage data and effective temperature data.

[0116] In one embodiment, the first failure rate generation module includes:

[0117] The second data segmentation module is used to divide the effective voltage data and effective temperature data into sub-effective voltage datasets and sub-effective temperature datasets through a second preset sliding window;

[0118] The marking module is used to calculate the voltage slope and temperature slope of sub-effective voltage data and sub-effective temperature data in each window. If the absolute value of the voltage slope is greater than the preset voltage slope threshold or the absolute value of the temperature slope is greater than the preset temperature slope threshold, the data points in the window are marked and the number of fault points is obtained.

[0119] The first failure rate module is used to determine the ratio of the number of all failure points to the number of data points as the first failure rate.

[0120] In one embodiment, the two-dimensional image generation module includes:

[0121] The sorting module is used to arrange the effective voltage data and effective temperature data in timestamp order to obtain a one-dimensional sequence set;

[0122] The second normalization module is used to normalize the one-dimensional data in the one-dimensional sequence set to obtain a normalized dataset;

[0123] The third normalization module is used to perform a preset inverse cosine normalization operation on the normalized data in the normalized dataset to obtain a two-dimensional feature matrix.

[0124] The mapping module is used to map the values ​​in the two-dimensional feature matrix to preset pixel grayscale color values ​​to obtain a two-dimensional image.

[0125] In one embodiment, the second failure rate generation module includes:

[0126] The compression module is used to perform matrix compression operations on the weight matrix of the Long Short-Term Memory network to obtain a low-rank approximate matrix.

[0127] The replacement module is used to replace the original weight matrix of the long short-term network with a low-rank approximation matrix to obtain a lightweight long short-term network.

[0128] The bidirectional processing module is used to process the temporal feature sequence step by step using a lightweight long short-term network to obtain the forward state vector and the reverse state vector, respectively, from the beginning of the temporal feature sequence in the forward direction and from the end of the temporal feature sequence in the reverse direction.

[0129] The splicing module is used to splice the forward state vector and the reverse state vector to obtain the fused state vector;

[0130] The second failure rate module is used to obtain the second failure rate by performing a nonlinear transformation on the fused feature vector through a preset monitoring formula.

[0131] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for intelligent operation monitoring of a PEM fuel cell, characterized in that, The method includes: Simultaneously collect voltage and temperature data from the PEM fuel cell; The voltage data and temperature data are preprocessed respectively to obtain effective voltage data and effective temperature data; A first failure rate is obtained by performing data fault analysis on the effective voltage data and the effective temperature data. A two-dimensional image is obtained by performing two-dimensional data conversion on the effective voltage data and the effective temperature data through image encoding operations. The two-dimensional image is flattened and reconstructed to obtain a temporal feature sequence. The temporal feature sequence is then substituted into a lightweight long short-term network model to obtain a second failure rate. The operational status monitoring results are obtained based on the first failure rate and the second failure rate; The temporal feature sequence is obtained by flattening and reconstructing the dimensions of the two-dimensional image. Substituting the temporal feature sequence into a lightweight long short-term network model yields the second failure rate, which includes: A low-rank approximate matrix is ​​obtained by performing a compression matrix operation on the weight matrix of the Long Short-Term Memory network. The low-rank approximation matrix replaces the original weight matrix of the long short-term network in the calculation to obtain a lightweight long short-term network. The temporal feature sequence is processed step by step using a lightweight long short-term network from the beginning of the temporal feature sequence in the forward direction and from the end of the temporal feature sequence in the reverse direction to obtain a forward state vector and a reverse state vector. The positive state vector and the negative state vector are concatenated to obtain the fused state vector; The second failure rate is obtained by performing a nonlinear transformation on the fused feature vector using a preset monitoring formula.

2. The intelligent operation monitoring method for a PEM fuel cell according to claim 1, characterized in that, Preprocessing the voltage data and temperature data to obtain effective voltage data and effective temperature data includes: The voltage data and the temperature data are normalized to obtain normalized voltage data and normalized temperature data; The normalized voltage data and the normalized temperature data are divided into sub-voltage datasets and sub-temperature datasets by using a first preset sliding window. The effective voltage data and effective temperature data are obtained by replacing the average value of the sub-voltage data and sub-temperature data within each window with the data points within the window.

3. The intelligent operation monitoring method for a PEM fuel cell according to claim 1, characterized in that, Data fault analysis of the effective voltage data and the effective temperature data yields a first fault rate, including: The effective voltage data and the effective temperature data are divided into sub-effective voltage datasets and sub-effective temperature datasets by using a second preset sliding window; Calculate the voltage slope and temperature slope of sub-effective voltage data and sub-effective temperature data within each window. If the absolute value of the voltage slope is greater than the preset voltage slope threshold or the absolute value of the temperature slope is greater than the preset temperature slope threshold, then mark the data points within the window and obtain the number of fault points. The ratio of the total number of fault points to the total number of data points is taken as the first failure rate.

4. The intelligent operation monitoring method for a PEM fuel cell according to claim 1, characterized in that, The two-dimensional image is obtained by performing two-dimensional data conversion on the effective voltage data and the effective temperature data through image encoding operations, including: The effective voltage data and the effective temperature data are arranged in timestamp order to obtain a one-dimensional sequence set; Normalize the one-dimensional data in the one-dimensional sequence set to obtain a normalized dataset; A two-dimensional feature matrix is ​​obtained by performing a preset inverse cosine normalization operation on the normalized data in the normalized dataset. A two-dimensional image is obtained by mapping the values ​​in the two-dimensional feature matrix to preset pixel grayscale color values.

5. An intelligent operation monitoring device for a PEM fuel cell, characterized in that, The device includes: The data acquisition module is used to simultaneously acquire voltage and temperature data of the PEM fuel cell; The preprocessing module is used to preprocess the voltage data and the temperature data respectively to obtain effective voltage data and effective temperature data; The first failure rate generation module is used to perform data failure analysis on the effective voltage data and the effective temperature data to obtain a first failure rate. A two-dimensional image generation module is used to perform two-dimensional data conversion on the effective voltage data and the effective temperature data through image encoding operations to obtain a two-dimensional image. The second failure rate generation module is used to perform a flattening dimension reconstruction operation on the two-dimensional image to obtain a temporal feature sequence, and to substitute the temporal feature sequence into a lightweight long short-term network model to obtain a second failure rate. The detection result generation module is used to obtain the operation status monitoring result based on the first failure rate and the second failure rate; The second failure rate generation module includes: The compression module is used to perform matrix compression operations on the weight matrix of the Long Short-Term Memory network to obtain a low-rank approximate matrix. The replacement module is used to replace the original weight matrix of the long short-term network with the low-rank approximation matrix in the calculation to obtain a lightweight long short-term network. The bidirectional processing module is used to process the temporal feature sequence step by step using a lightweight long short-term network to obtain a forward state vector and a reverse state vector, respectively, from the beginning of the temporal feature sequence in a forward direction and from the end of the temporal feature sequence in a reverse direction. The splicing module is used to splice the positive state vector and the negative state vector to obtain a fused state vector; The second failure rate module is used to perform a nonlinear transformation on the fused feature vector using a preset monitoring formula to obtain a second failure rate.

6. The intelligent operation monitoring device for a PEM fuel cell according to claim 5, characterized in that, The preprocessing module includes: The first normalization module is used to normalize the voltage data and the temperature data to obtain normalized voltage data and normalized temperature data. The first data segmentation module is used to divide the normalized voltage data and the normalized temperature data into sub-voltage datasets and sub-temperature datasets through a first preset sliding window; The data replacement module is used to replace the data points in each window with the average value of the sub-voltage data and sub-temperature data to obtain the effective voltage data and effective temperature data.

7. The intelligent operation monitoring device for a PEM fuel cell according to claim 5, characterized in that, The first failure rate generation module includes: The second data segmentation module is used to divide the effective voltage data and the effective temperature data into sub-effective voltage datasets and sub-effective temperature datasets through a second preset sliding window; The marking module is used to calculate the voltage slope and temperature slope of sub-effective voltage data and sub-effective temperature data in each window. If the absolute value of the voltage slope is greater than the preset voltage slope threshold or the absolute value of the temperature slope is greater than the preset temperature slope threshold, the data points in the window are marked and the number of fault points is obtained. The first failure rate module is used to determine the ratio of the number of all failure points to the number of data points as the first failure rate.

8. The intelligent operation monitoring device for a PEM fuel cell according to claim 5, characterized in that, The 2D image generation module includes: The sorting module is used to arrange the effective voltage data and the effective temperature data in timestamp order to obtain a one-dimensional sequence set; The second normalization module is used to normalize the one-dimensional data in the one-dimensional sequence set to obtain a normalized dataset. The third normalization module is used to perform a preset inverse cosine normalization operation on the normalized data in the normalized dataset to obtain a two-dimensional feature matrix. The mapping module is used to map the values ​​in the two-dimensional feature matrix to preset pixel grayscale color values ​​to obtain a two-dimensional image.

Images