CN113359037A - Fuel cell fault diagnosis method based on BP neural network - Google Patents

Fuel cell fault diagnosis method based on BP neural network Download PDF

Info

Publication number
CN113359037A
CN113359037A CN202110652603.XA CN202110652603A CN113359037A CN 113359037 A CN113359037 A CN 113359037A CN 202110652603 A CN202110652603 A CN 202110652603A CN 113359037 A CN113359037 A CN 113359037A
Authority
CN
China
Prior art keywords
data
fault
neural network
fuel cell
matrix
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110652603.XA
Other languages
Chinese (zh)
Inventor
李宁
郭泽林
袁铁江
杨金成
张伟
王永超
杨永建
白银平
王海磊
谢珍
于静
王丽娟
费守河
李航
黄琰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Marketing Service Center Of State Grid Xinjiang Electric Power Co Ltd Capital Intensive Center Metering Center
Dalian University of Technology
Electric Power Research Institute of State Grid Xinjiang Electric Power Co Ltd
Original Assignee
Marketing Service Center Of State Grid Xinjiang Electric Power Co Ltd Capital Intensive Center Metering Center
Dalian University of Technology
Electric Power Research Institute of State Grid Xinjiang Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Marketing Service Center Of State Grid Xinjiang Electric Power Co Ltd Capital Intensive Center Metering Center, Dalian University of Technology, Electric Power Research Institute of State Grid Xinjiang Electric Power Co Ltd filed Critical Marketing Service Center Of State Grid Xinjiang Electric Power Co Ltd Capital Intensive Center Metering Center
Priority to CN202110652603.XA priority Critical patent/CN113359037A/en
Publication of CN113359037A publication Critical patent/CN113359037A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/378—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC] specially adapted for the type of battery or accumulator
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Fuel Cell (AREA)

Abstract

A fuel cell fault diagnosis method based on BP neural network, normalize the data that the sensor gathers at first, make it distribute in [0,1] interval; considering that fault data does not necessarily obey normal distribution, introducing Box-Cox transformation to perform normalization processing on the data, and performing feature extraction on the normalized data by adopting linear discriminant analysis to screen fault features so as to realize dimension reduction on the fault data; and (3) taking the extracted feature vector as an input layer variable of the BP neural network, taking the fault type of the fuel cell as an output layer variable, and introducing a heuristic method to determine the optimal number of nodes of the hidden layer so as to obtain a diagnosis result taking the fault type as the output variable.

Description

Fuel cell fault diagnosis method based on BP neural network
Technical Field
The present invention relates to a fuel cell fault diagnosis method.
Background
Fuel cells have been widely used as a new clean energy source using hydrogen as a raw material in various fields including traffic, energy storage, and the like. However, the method has the problems of short service life, poor reliability and the like, and the large-scale industrial development of the method is limited. In the operation process of the fuel cell, flooding and membrane drying are faults which often occur, and the flooding means that liquid water in the galvanic pile is continuously accumulated so as to block a gas diffusion layer, a catalyst layer and even a gas flow passage, so that the phenomenon of smooth proceeding of electrochemical reaction in the galvanic pile is influenced; the membrane dryness refers to the phenomenon that the hydration of a membrane electrode is blocked, the conductivity is reduced and the impedance of the membrane is increased due to insufficient liquid water in the galvanic pile, and when the membrane dryness has serious faults, the local high temperature of the galvanic pile can be caused, so that the service life of a fuel cell is shortened. While if these faults occur, they can cause a reduction in the performance of the automation system or of the vehicle equipped with the fuel cell, and can lead to irreparable consequences.
The current mainstream methods for the fault diagnosis of the fuel cell include two major categories, model-based and non-model-based. The fault diagnosis method based on the model needs to establish a corresponding model according to the physical process of the fuel cell system, and realizes fault diagnosis and separation through residual analysis between a model simulation result and the output of an actual system. The non-model-based fault diagnosis method is used for acquiring fault information of the fuel cell system on the basis of an expert system, a signal processing method or a mixed method of the expert system and the signal processing method and the application of a visualization technology. Compared with the fault diagnosis method based on the model, the method does not need to establish a fuel cell system model, and the process of fault diagnosis and separation is replaced by human reasoning activities simulated by an expert system, but the fault type is predefined. The non-model fault diagnosis method is in the future along with the rapid development of artificial intelligence, and is widely applied to the field of fault diagnosis in engineering at present. The advantages of real-time and linearity of the method promote its evolution into one of the main methods of fuel cell system fault diagnosis.
Disclosure of Invention
The invention provides a fuel cell fault diagnosis method based on a BP neural network, aiming at the defects of the prior art in the fuel cell fault diagnosis research. Aiming at the characteristics of high dimensionality and complex structure of fuel cell fault data, the fuel cell fault data acquired from a sensor is subjected to normalization processing, so that a result is mapped between [0,1 ]; b, introducing Box-Cox transformation to carry out normalization processing on the data to obtain normal distribution data which is easy to linearly discriminate, analyze and process; performing feature extraction on the normalized data, projecting the data at high latitude to low dimensionality, and ensuring the minimum intra-class variance and the maximum inter-class variance after projection; the extracted feature vector is used as an input layer variable of a BP neural network, the fault type of the fuel cell is used as an output layer variable, a heuristic method is introduced to determine the optimal number of nodes of a hidden layer, and a diagnosis result with the fault type as the output variable is obtained, and the method comprises the following steps:
1. analyzing the fault data information and the fault type of the fuel cell, and carrying out normalization processing on the fault data of the fuel cell acquired from the sensor;
2. introducing a Box-Cox method to perform normal conversion on the normalized data, so that the converted data obey normal distribution;
3. performing dimensionality reduction on the normalized data by adopting a linear discriminant analysis method to obtain a new feature vector;
4. the characteristic vector is used as an input variable of a neural network input layer, and the fault type of the fuel cell is used as an output variable of an output layer;
5. introducing a heuristic method, determining the optimal number of nodes of the hidden layer, and constructing a BP neural network;
6. and training the BP neural network to obtain a diagnosis result with the fault type as an output variable.
In step 1, the fuel cell fault data information includes, but is not limited to: voltage, current, power, inlet flow rates of the cathode and the anode, inlet pressures of the cathode and the anode, outlet pressures of the cathode and the anode, inlet temperatures of the cathode and the anode, outlet temperatures of the galvanic pile, temperature and power of the heater and the like.
The fault types mainly comprise water flooding, membrane dry water and other water management faults.
Given a set of original samples of dimension n x1,x2,x3,...,xnAnd dispersion normalization of data, namely linear transformation of original data, so that the result is mapped between 0 and 1, wherein the transformation principle is as follows:
Figure BDA0003111559530000021
wherein x ismaxRepresenting the maximum value, x, of the data in the original sample setminRepresenting the minimum, x, of the data in the original sample setiThe sample value of the ith dimension is represented, i represents the serial number of a certain sample in the original sample set with the dimension of n, and the value range is [1, n]And x represents data after dispersion normalization processing.
In the step 2, considering that the fault data of the fuel cell does not necessarily conform to normal distribution, in order to convert the data into normal distribution data suitable for linear discriminant analysis processing, a BOX-COX method is introduced to perform normal distribution conversion on the normalized data.
The general form of the Box-Cox transform is:
Figure BDA0003111559530000031
in the formula, y (lambda) is a new variable after Box-Cox transformation, y is an original continuous dependent variable, and lambda is a transformation parameter. The above transformation requires that the original variable y is positive, if the original variable y is negative, a constant a is added to all original data to make (y + a) positive, and then the above transformation is performed.
The parameter λ in the Box-Cox transform can be given by a maximum likelihood estimate, typically, with respect to the parameters β, σ constructed2And carrying out maximum likelihood estimation on the likelihood function to obtain the maximum value of the likelihood function as:
Figure BDA0003111559530000032
in the formula, Lmax(λ) and L represent parametersMaximum likelihood function of lambda, beta, sigma2Respectively representing the mean and variance of normal distribution, MSE (lambda) representing the mean-squared error of the model, J being the transformed Jacobian determinant, n being the dimension of the vector y, beta, sigma2Respectively, mean and variance of a normal distribution.
In the step 3, Linear Discriminant Analysis (LDA) is a supervised data dimension reduction technology, which can project data at high latitude onto low dimension and ensure that the intra-class variance is minimum and the inter-class variance is maximum after projection.
Given a data sample matrix X ═ X obeying a normal distribution1,x2,x3,...,xm]∈Rn×mWhere n and m represent the dimension and number of the data sample matrix, R, respectivelyn×mRepresenting a matrix of dimension n x m, Rn×dRepresenting a matrix of dimension n x d, RdRepresenting a data sample with dimension d, the objective of Linear Discriminant Analysis (LDA) is to train a linear transformation matrix W epsilon R with dimension n x dn×dAnd making the high-dimensional data x be equal to RnMapping to low-dimensional data y ∈ Rd：
y＝WTx
Formula y ═ WTIn X, X is classified as X ═ pi1,π2,...,πc](ii) a Wherein c represents the number of categories; piiIs a data set of category i with dimension n × ni；niIndicating the number of data samples in category i.
Obtaining an optimal linear transformation matrix by solving the following formula:
minTr[(WTStW)-1(WTSwW)]
wherein W represents a linear transformation matrix, WTRepresenting the transpose of a linearly varying matrix, Tr representing the locus of the matrix, minTr representing the minimum of the locus of the matrix, SwRepresents an intra-class scattering matrix, StThe whole class scattering matrix is represented, and the calculation mode is as follows:
Figure BDA0003111559530000033
Figure BDA0003111559530000034
wherein the content of the first and second substances,
Figure BDA0003111559530000041
represents the mean of the samples in the category i,
Figure BDA0003111559530000042
represents the mean of all samples, T represents the matrix transpose, n represents the number of samples, c represents the number of classes, x represents the samples, πiIs a data set of category i.
The solution of the objective function can be simplified to solve the following eigenvalue decomposition problem:
St -1SwW＝WΛ
wherein Λ is St -1SwThe feature matrix of (2).
Calculating S using the Lagrange multiplier methodt -1SwAnd the corresponding eigenvectors (w)1,w2,...,wd) And obtaining the projection matrix W.
In step 4, the input/output vector determining method of the neural network is as follows:
and selecting the low-dimensional vector obtained after feature selection as an input vector of the neural network, and selecting several typical fault types of the fuel cell as output variables of the neural network.
In the step 5, the established fault diagnosis model has 3 layers, namely an input layer, a hidden layer and an output layer, wherein the input layer is used for reading input data, the middle layer is used for training and processing data, and the output layer is used for outputting results. The optimal node number of the hidden layer is determined by adopting a heuristic method, and the method comprises the following steps:
firstly, a starting point x is set1And starting step h0Starting point x1And starting step h0Can be calculated by the following formula:
x1＝log2n
h0＝(n-n2)/3
wherein x is1Is the number of initial nodes of the hidden layer, n is the number of input units, n2Is the number of output units.
The contrast point x is calculated by the following formula2：
x2＝x1+h0
Calculating the neural network prediction error values f corresponding to the two points in sequence1(x) And f2(x) Determining the position of the next point according to the relationship between the two values,
Figure BDA0003111559530000043
x3for the position of the next point until the optimum point x is foundiI.e. the optimal number of hidden layer nodes.
The built fault diagnosis model has 3 layers which are an input layer, a hidden layer and an output layer respectively. The number of nodes of the input layer is the dimensionality of the feature vector, and the number of nodes of the output layer is the number of fault categories; the input layer is used for reading in input data, the middle layer is used for training and processing data, and the output layer is used for outputting results.
In the step 6, the BP neural network is a multilayer feedforward network trained according to an error inverse propagation algorithm, an operation mechanism of the BP neural network comprises two processes of forward propagation and feedback modification, and a fault diagnosis method of the BP neural network comprises the following steps:
and training and testing the BP neural network by using the processed fault data to obtain a diagnosis result with the fault type as an output variable.
The invention introduces a 'heuristic' method to improve the determination method of the number of nodes in the middle layer, namely the step 5.
Drawings
FIG. 1 is a flow chart of the fuel cell fault diagnosis method based on BP neural network of the invention;
fig. 2 is a structural diagram of a neural network.
Detailed Description
The invention is further described with reference to the following drawings and detailed description.
As shown in fig. 1, the fuel cell fault diagnosis method based on the BP neural network of the present invention has the following processes:
1. analyzing the fault data information and the fault type of the fuel cell, and carrying out normalization processing on the fault data of the fuel cell acquired from the sensor;
2. introducing a Box-Cox method to perform normal conversion on the normalized data, so that the converted data obey normal distribution;
3. performing dimensionality reduction on the normalized data by adopting a linear discriminant analysis method to obtain a new feature vector;
4. the characteristic vector is used as an input variable of a neural network input layer, and the fault type of the fuel cell is used as an output variable of an output layer;
5. introducing a heuristic method, determining the optimal number of nodes of the hidden layer, and constructing a BP neural network;
6. as shown in fig. 2, the BP neural network is trained, wherein the input layer is used for reading in a fault feature vector, the intermediate layer is used for training and processing data, and the output layer is used for outputting a fault diagnosis result to finally obtain a diagnosis result with a fault type as an output variable.
In conclusion, the invention applies the improved linear discriminant analysis and neural network method to the fault diagnosis of the fuel cell. The method comprises the steps of firstly carrying out normalization and normalization processing on fuel cell fault data, then carrying out fault data feature extraction through linear discriminant analysis, using extracted vectors for an input layer of a neural network, determining the optimal node number of a hidden layer by adopting a heuristic method, and effectively applying the neural network to fault diagnosis research of the fuel cell after constructing and training the neural network.

Claims (7)

1. A fuel cell fault diagnosis method based on BP neural network is characterized in that after normalization processing is carried out on fault data obtained from a fuel cell system sensor, an improved linear discriminant analysis method is adopted for feature extraction, and further the neural network is utilized for fault diagnosis, and the method comprises the following steps:
step 1: analyzing the fault data information and the fault type of the fuel cell, and carrying out normalization processing on the fault data of the fuel cell acquired from the sensor;
step 2: introducing a Box-Cox method to perform normal conversion on the normalized data, so that the converted data obey normal distribution;
and step 3: performing dimensionality reduction on the normalized data by adopting a linear discriminant analysis method to obtain a new feature vector;
and 4, step 4: the characteristic vector is used as an input variable of a neural network input layer, and the fault type of the fuel cell is used as an output variable of an output layer;
and 5: introducing a heuristic method, determining the optimal number of nodes of the hidden layer, and constructing a BP neural network;
step 6: and training the BP neural network to obtain a diagnosis result with the fault type as an output variable.
2. The BP neural network-based fuel cell fault diagnosis method according to claim 1, wherein in the step 1, the fuel cell fault data acquired from the sensor is normalized as follows:
given a set of original samples of dimension n x1,x2,x3,...,xnAnd dispersion normalization of data, namely linear transformation of original data, so that the result is mapped between 0 and 1, wherein the transformation principle is as follows:
Figure FDA0003111559520000011
wherein x ismaxRepresenting the maximum value, x, of the data in the original sample setminRepresenting the minimum, x, of the data in the original sample setiThe sample value representing the ith dimension, and x represents the data after dispersion normalization processing.
3. The BP neural network-based fuel cell fault diagnosis method according to claim 1, wherein in the step 2, the normalized data normal conversion method is as follows:
introducing a BOX-COX method to perform normal distribution conversion on the normalized data;
the general form of the Box-Cox transform is:
Figure FDA0003111559520000012
in the formula, y (lambda) is a new variable after Box-Cox transformation, y is an original continuous dependent variable, and lambda is a transformation parameter;
the transformation requires that the value of an original variable y is positive, if the value of the original variable y is negative, a constant a is added to all original data to enable the (y + a) to be a positive value, and then the transformation is carried out;
the parameter λ in the Box-Cox transform can be given by a maximum likelihood estimate, typically, with respect to the parameters β, σ constructed2And carrying out maximum likelihood estimation on the likelihood function to obtain the maximum value of the likelihood function as:
Figure FDA0003111559520000021
in the formula, Lmax(λ) and L represent the maximum likelihood function of the parameter λ, β, σ2Respectively representing the mean and variance of normal distribution, MSE (lambda) representing the mean square error of the model, J being the transformed Jacobian determinant, and n being the dimension of the vector y.
4. The BP neural network-based fuel cell fault diagnosis method according to claim 1, wherein in the step 3, the fault data feature extraction method is as follows:
given a data sample matrix X ═ X obeying a normal distribution1,x2,x3,...,xm]∈Rn×mWherein n and m are respectively represented byShowing the dimensionality and the number of a data sample matrix, and training an n x d-dimensional linear transformation matrix W epsilon Rn×dAnd making the high-dimensional data x be equal to RnMapping to low-dimensional data y ∈ Rd：
y＝WTx
In the above formula, X is classified as X ═ pi1,π2,...,πc](ii) a Wherein c represents the number of categories; piiIs a data set of category i with dimension n × ni；niRepresenting the number of data samples in category i;
obtaining an optimal linear transformation matrix by solving the following formula:
minTr[(WTStW)-1(WTSwW)]
wherein W represents a linear transformation matrix, WTRepresenting the transpose of a linearly varying matrix, Tr representing the locus of the matrix, SwRepresents an intra-class scattering matrix, StThe scattering matrix of the whole class is expressed, minTr represents the minimum value of the matrix track, and the calculation mode is as follows:
Figure FDA0003111559520000022
Figure FDA0003111559520000023
wherein the content of the first and second substances,
Figure FDA0003111559520000024
represents the mean of the samples in the category i,
Figure FDA0003111559520000025
represents the mean of all samples, T represents the matrix transpose, n represents the number of samples, c represents the number of classes, x represents the samples, πiIs a data set of category i.
The solution of the objective function can be simplified to solve the following eigenvalue decomposition problem:
St -1SwW＝WΛ
wherein Λ is St -1SwA feature matrix of (a);
calculating S using the Lagrange multiplier methodt -1SwMaximum d eigenvalues and corresponding eigenvectors (w)1,w2,...,wd) And obtaining the projection matrix W.
5. The method for diagnosing the fault of the fuel cell based on the BP neural network as claimed in claim 1, wherein in the step 4, the low-dimensional vector obtained after feature selection is selected as an input vector of the neural network, and several typical fault types of the fuel cell are selected as output variables of the neural network.
6. The BP neural network-based fuel cell fault diagnosis method according to claim 1, wherein in the step 5, the optimal node number of the hidden layer is determined as follows:
firstly, a starting point x is set1And starting step h0Starting point x1And starting step h0Calculated by the following formula:
x1＝log2 n
h0＝(n-n2)/3
wherein x is1Is the number of initial nodes of the hidden layer, n is the number of input units, n2Is the number of output units;
the contrast point x is calculated by the following formula2：
x2＝x1+h0
Calculating the neural network prediction error values f corresponding to the two points in sequence1(x) And f2(x) Determining the position of the next point according to the relationship between the two values,
Figure FDA0003111559520000031
x3for the position of the next point until the optimum point x is foundiThe number of nodes of the hidden layer is the optimal number;
the built fault diagnosis model has 3 layers which are an input layer, a hidden layer and an output layer respectively. The input layer is used for reading in input data, the middle layer is used for training and processing data, and the output layer is used for outputting results.
7. The BP neural network-based fuel cell fault diagnosis method according to claim 1, wherein in the step 6, the number of nodes of the input layer is a dimension of a feature vector, and the number of nodes of the output layer is a fault category number; and training and testing the BP neural network by using the processed fault data to obtain a diagnosis result with the fault type as an output variable.
CN202110652603.XA 2021-06-11 2021-06-11 Fuel cell fault diagnosis method based on BP neural network Pending CN113359037A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110652603.XA CN113359037A (en) 2021-06-11 2021-06-11 Fuel cell fault diagnosis method based on BP neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110652603.XA CN113359037A (en) 2021-06-11 2021-06-11 Fuel cell fault diagnosis method based on BP neural network

Publications (1)

Publication Number Publication Date
CN113359037A true CN113359037A (en) 2021-09-07

Family

ID=77533912

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110652603.XA Pending CN113359037A (en) 2021-06-11 2021-06-11 Fuel cell fault diagnosis method based on BP neural network

Country Status (1)

Country Link
CN (1) CN113359037A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113884903A (en) * 2021-10-19 2022-01-04 中国计量大学 Battery fault diagnosis method based on multilayer perceptron neural network
CN115084593A (en) * 2022-05-31 2022-09-20 同济大学 Fuel cell fault diagnosis method based on nonlinear impedance spectrum
CN115084593B (en) * 2022-05-31 2024-05-31 同济大学 Fuel cell fault diagnosis method based on nonlinear impedance spectrum

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110794360A (en) * 2019-10-22 2020-02-14 中国电力科学研究院有限公司 Method and system for predicting fault of intelligent electric energy meter based on machine learning

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110794360A (en) * 2019-10-22 2020-02-14 中国电力科学研究院有限公司 Method and system for predicting fault of intelligent electric energy meter based on machine learning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘嘉蔚, 中国电机工程学报, vol. 39, no. 12, pages 3614 - 3620 *
孙弋清, 中国优秀硕士学位论文全文数据库 (工程科技Ⅱ辑), pages 11 - 12 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113884903A (en) * 2021-10-19 2022-01-04 中国计量大学 Battery fault diagnosis method based on multilayer perceptron neural network
CN113884903B (en) * 2021-10-19 2023-08-18 中国计量大学 Battery fault diagnosis method based on multi-layer perceptron neural network
CN115084593A (en) * 2022-05-31 2022-09-20 同济大学 Fuel cell fault diagnosis method based on nonlinear impedance spectrum
CN115084593B (en) * 2022-05-31 2024-05-31 同济大学 Fuel cell fault diagnosis method based on nonlinear impedance spectrum

Similar Documents

Publication Publication Date Title
Liu et al. Prognostics methods and degradation indexes of proton exchange membrane fuel cells: A review
Liu et al. Remaining useful life prediction of PEMFC based on long short-term memory recurrent neural networks
Jiang et al. A comparative study of different features extracted from electrochemical impedance spectroscopy in state of health estimation for lithium-ion batteries
Huo et al. Performance prediction of proton-exchange membrane fuel cell based on convolutional neural network and random forest feature selection
Zuo et al. Degradation prediction model for proton exchange membrane fuel cells based on long short-term memory neural network and Savitzky-Golay filter
Ma et al. A hybrid transfer learning scheme for remaining useful life prediction and cycle life test optimization of different formulation Li-ion power batteries
Mao et al. Investigation of polymer electrolyte membrane fuel cell internal behaviour during long term operation and its use in prognostics
Zhang et al. A short-and long-term prognostic associating with remaining useful life estimation for proton exchange membrane fuel cell
Liu et al. Sequence fault diagnosis for PEMFC water management subsystem using deep learning with t-SNE
Sun et al. Short-term performance degradation prediction of a commercial vehicle fuel cell system based on CNN and LSTM hybrid neural network
Chen et al. Prognosis of fuel cell degradation under different applications using wavelet analysis and nonlinear autoregressive exogenous neural network
Wu et al. A hybrid prognostic model applied to SOFC prognostics
Nagulapati et al. Machine learning based fault detection and state of health estimation of proton exchange membrane fuel cells
CN115598536A (en) PEMFC fault diagnosis method based on fuzzy C-means clustering and probabilistic neural network
CN113359037A (en) Fuel cell fault diagnosis method based on BP neural network
Xiao et al. Water management fault diagnosis for proton-exchange membrane fuel cells based on deep learning methods
Wang et al. An efficient state-of-health estimation method for lithium-ion batteries based on feature-importance ranking strategy and PSO-GRNN algorithm
Liu et al. The life prediction of PEMFC based on group method of data handling with Savitzky–Golay smoothing
Yuan et al. Fault Diagnosis of Fuel Cells by a Hybrid Deep Learning Network Fusing Characteristic Impedance
CN117312939A (en) SOFC system working condition identification method based on deep learning
CN116031453A (en) On-line estimation method for characteristic frequency impedance of proton exchange membrane fuel cell
Lv et al. Deep learning-based fault diagnosis and Electrochemical Impedance Spectroscopy frequency selection method for Proton Exchange Membrane Fuel Cell
Sun et al. Fault diagnosis method for proton exchange membrane fuel cell system based on digital twin and unsupervised domain adaptive learning
Dong et al. Online adaptive humidity monitoring method for proton exchange membrane fuel cell based on fuzzy C-means clustering and online sequence extreme learning machine
Rubio et al. Failure mode diagnosis in proton exchange membrane fuel cells using local electrochemical noise

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination