CN115000471A - Fuel cell catalyst layer prediction-analysis-optimization method based on machine learning - Google Patents
Fuel cell catalyst layer prediction-analysis-optimization method based on machine learning Download PDFInfo
- Publication number
- CN115000471A CN115000471A CN202210538054.8A CN202210538054A CN115000471A CN 115000471 A CN115000471 A CN 115000471A CN 202210538054 A CN202210538054 A CN 202210538054A CN 115000471 A CN115000471 A CN 115000471A
- Authority
- CN
- China
- Prior art keywords
- catalyst layer
- fuel cell
- optimization
- parameter
- model
- 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
Links
- 239000003054 catalyst Substances 0.000 title claims abstract description 67
- 238000005457 optimization Methods 0.000 title claims abstract description 64
- 239000000446 fuel Substances 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000010801 machine learning Methods 0.000 title claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 20
- 238000013178 mathematical model Methods 0.000 claims abstract description 20
- 239000012528 membrane Substances 0.000 claims abstract description 13
- 230000002068 genetic effect Effects 0.000 claims abstract description 12
- 230000003197 catalytic effect Effects 0.000 claims description 55
- BASFCYQUMIYNBI-UHFFFAOYSA-N platinum Substances [Pt] BASFCYQUMIYNBI-UHFFFAOYSA-N 0.000 claims description 47
- 238000012549 training Methods 0.000 claims description 21
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 claims description 12
- 229910052799 carbon Inorganic materials 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 12
- 229910052697 platinum Inorganic materials 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 238000010206 sensitivity analysis Methods 0.000 claims description 10
- 238000012795 verification Methods 0.000 claims description 10
- 230000008878 coupling Effects 0.000 claims description 9
- 238000010168 coupling process Methods 0.000 claims description 9
- 238000005859 coupling reaction Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 9
- 239000002245 particle Substances 0.000 claims description 9
- 230000035945 sensitivity Effects 0.000 claims description 9
- 238000009826 distribution Methods 0.000 claims description 7
- 230000009471 action Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 108090000623 proteins and genes Proteins 0.000 claims description 6
- 230000010287 polarization Effects 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 229920000554 ionomer Polymers 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000013461 design Methods 0.000 abstract description 8
- 238000004458 analytical method Methods 0.000 abstract description 7
- 238000011161 development Methods 0.000 description 4
- 238000009792 diffusion process Methods 0.000 description 4
- 238000003487 electrochemical reaction Methods 0.000 description 4
- 239000007789 gas Substances 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000007788 liquid Substances 0.000 description 2
- 239000001301 oxygen Substances 0.000 description 2
- 229910052760 oxygen Inorganic materials 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 230000005514 two-phase flow Effects 0.000 description 2
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 1
- DSVGQVZAZSZEEX-UHFFFAOYSA-N [C].[Pt] Chemical compound [C].[Pt] DSVGQVZAZSZEEX-UHFFFAOYSA-N 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000001257 hydrogen Substances 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01M—PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY
- H01M8/00—Fuel cells; Manufacture thereof
- H01M8/04—Auxiliary arrangements, e.g. for control of pressure or for circulation of fluids
- H01M8/04298—Processes for controlling fuel cells or fuel cell systems
- H01M8/04305—Modeling, demonstration models of fuel cells, e.g. for training purposes
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01M—PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY
- H01M8/00—Fuel cells; Manufacture thereof
- H01M8/04—Auxiliary arrangements, e.g. for control of pressure or for circulation of fluids
- H01M8/04298—Processes for controlling fuel cells or fuel cell systems
- H01M8/04992—Processes for controlling fuel cells or fuel cell systems characterised by the implementation of mathematical or computational algorithms, e.g. feedback control loops, fuzzy logic, neural networks or artificial intelligence
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/30—Hydrogen technology
- Y02E60/50—Fuel cells
Landscapes
- Engineering & Computer Science (AREA)
- General Chemical & Material Sciences (AREA)
- Sustainable Development (AREA)
- Sustainable Energy (AREA)
- Manufacturing & Machinery (AREA)
- Chemical & Material Sciences (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Electrochemistry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Fuel Cell (AREA)
- Automation & Control Theory (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
Abstract
The invention provides a fuel cell catalyst layer prediction-analysis-optimization method based on machine learning, which aims at the problems of strong parameter nonlinear relation and large analysis and optimization difficulty in a proton exchange membrane fuel cell catalyst layer, takes a fuel cell mathematical model as a data driving source, applies a machine learning algorithm to develop a data driving model to realize rapid prediction and quantitative characteristic analysis of cell performance under different catalyst layer structures, and further couples a genetic algorithm to realize multi-target multi-parameter optimization of catalyst layer structure parameters; the method can reveal the influence rule of the catalyst layer parameters on the battery performance from the global angle, and provides guidance for the actual catalyst layer design; the multi-objective multi-parameter optimization can accurately and comprehensively carry out comprehensive optimization on all parameters, and obtain multi-objective comprehensive optimization values which are difficult to obtain by the traditional optimization method.
Description
Technical Field
The invention relates to the technical field of fuel cells, in particular to a prediction-analysis-optimization method of a fuel cell catalyst layer based on machine learning
Background
The proton exchange membrane fuel cell is important energy conversion equipment due to the advantages of high energy density, compact structure, zero emission and the like. During the operation of the hydrogen-oxygen proton exchange membrane fuel cell, hydrogen and oxygen flow into the electrode through the anode flow field and the cathode flow field respectively, and reach the surface of the catalyst in the catalytic layer to generate electrochemical reaction, so that chemical energy is directly converted into electric energy. The catalytic layer is used as a place where electrochemical reaction of the fuel cell occurs, the internal structure design of the catalytic layer directly influences the discharge performance, stability and durability of the cell, and the analysis and design of the catalytic layer with strong electro-catalytic performance and good thermal stability are the key points for promoting the industrialization of the proton exchange membrane fuel cell.
In the working process of the fuel cell, multi-scale and multi-physical processes such as gas-liquid two-phase flow, heat mass transmission, electrochemical reaction and the like are involved in the catalyst layer, the nonlinear relation of internal parameters is strong due to the mutual coupling among the multi-physical fields, and the difficulty is high when the structural parameters in the catalyst layer are analyzed and optimized. In the traditional catalytic layer parameter analysis optimization, a single variable method is the most common method, namely, other parameters are fixed, and only one parameter is changed at a time to analyze the influence of the parameter on the performance of the battery. When the method is applied, only the result of the research parameter at a certain determined point can be obtained, and the parameter global analysis in the whole solution space cannot be realized. Meanwhile, the single variable method is mainly suitable for solving the problem of strong linear correlation, and the result accuracy is low when the problem of strong nonlinear relation is analyzed, so that the method is not suitable for analyzing the structure parameters of the catalytic layer.
In the catalytic layer structure optimization research, the qualitative optimization of the structure parameters based on the analysis result of a single variable method is also the most commonly used method, the method often obtains a parameter optimization direction or an ideal interval, an accurate global optimal solution is difficult to obtain, and the multi-objective optimization of the battery performance cannot be realized. In part of research on design optimization of catalytic layer structure parameters by using an optimization algorithm, the peak power density of a battery is often taken as a unique optimization target, and multi-target optimization of the battery from multiple aspects cannot be obtained so as to obtain a catalytic layer structure with higher comprehensive performance. Meanwhile, because the feature screening of the optimized parameters is not carried out, a plurality of parameters with low sensitivity cause model oscillation in the optimization process, so that the optimization algorithm is slow in convergence, and the optimization efficiency is greatly limited.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a fuel cell catalyst layer prediction-analysis-optimization method based on machine learning.
In order to achieve the purpose, the invention adopts the technical scheme that:
s1, building a mathematical model to obtain a data driving source;
s2, selecting the catalyst layer structure parameters and the battery output target to generate a data set;
s3, preprocessing a data set and establishing a data driving model;
s4, battery performance rapid prediction and parameter quantitative sensitivity analysis based on the data driving model;
s5, coupling the data driving model and the catalytic layer multi-target multi-parameter optimization of the genetic algorithm;
and S6, verifying the structural parameters of the catalytic layer obtained by optimization.
The step S1 specifically includes the following steps: aiming at multi-scale and multi-physical phenomena in the proton exchange membrane fuel cell, a two-dimensional two-phase non-isothermal fuel cell mathematical model is built by using a coupling control equation, model accuracy verification is completed by comparing the mathematical model with a polarization curve of an experimental test, and the verified mathematical model is used as a data driving source to assist data driving of the model.
The step S2 specifically includes the following steps: selecting 9 catalytic layer structure parameters as input characteristics of a data-driven model:
X＝(m Pt ,x Pt/C ,x Ion/C ,r Agg ,r C ,r Pt ,dp CL ,H CL ,θ CL )
where X is the input parameter set, m Pt For catalytic layer platinum loading, x Pt/C Is the ratio of platinum to carbon in the catalyst layer, x Ion/C Is a catalyst layerMass ratio of internal ionomer to carbon support, r Agg Radius of catalyst agglomerates, r C Is the carbon particle radius, r Pt Is the platinum particle radius, dp CL Pore diameter in the catalyst layer H CL Thickness of catalytic layer, theta CL A catalytic layer contact angle;
in order to describe the performance of the fuel cell in three aspects of maximum load capacity, maximum discharge capacity and thermal uniformity, the peak power density, the limiting current density and the standard deviation of the temperature distribution of the catalytic layer at the peak point of the cell are selected as optimization targets:
Y＝(P max ,I Lim ,σ T,CL )
wherein Y is the output parameter set, P max Is the peak power density, I Lim Is the limiting current density, σ T,CL Is the standard deviation of the temperature distribution in the catalytic layer;
determining the variation range of the input parameters: m is Pt ∈(0.025,0.5)mg·m -2 ，x Pt/C ∈(0.033,0.5)，x Ion/C ∈(0.2,1.3)， r Agg ∈(60,1000)nm，r C ∈(10,1000)nm，r Pt ∈(1,5)nm，dp CCL ∈(10,200)nm，H CL ∈(1,40)μm，θ CL E (92,150) °; in order to ensure the rationality of the structure of the catalyst layer of the whole fuel cell and prevent the failure of the catalyst layer structure caused by error accumulation, the limiting condition that the ratio of the radius of aggregates to the radius of carbon particles in the catalyst layer of the fuel cell is more than 4 is set;
according to the variation range of each optimized parameter, determining that each input parameter randomly varies within the range and obeys the mean value mu i Is the median, variance of the interval
Normal distribution of (2); and sequentially bringing each group of randomly generated input sets into a mathematical model, calculating corresponding optimized target values, and combining the optimized target values and the optimized target values to obtain a data driving data set.
The step S3 specifically includes the following steps: performing normalization pretreatment on the obtained data set, mapping each parameter data to an interval of 0-1 to eliminate the influence of different magnitude levels among different data, dividing the data set into a training set and a test set according to the proportion of 0.75:0.25, and respectively using the training set and the test set for data-driven model training and model verification;
selecting an integrated learning algorithm extreme gradient lifting tree (XGboost) to carry out data driving model training, and carrying out model accuracy verification by using a test set; and debugging the hyper-parameters in the XGboost of different training targets by using a grid search method in the training process to complete the configuration of the XGboost algorithm.
The step S4 specifically includes the following steps: a direct mapping relation between the catalytic layer structure and multiple targets of the battery is constructed by applying a data driving model obtained by training, and the rapid prediction of the battery peak power density, the limiting current density and the peak point temperature uniformity under different catalytic layer structures is realized;
extracting the characteristic influence degree in the data driving model, combining the characteristic influence degree with the action direction in the Pearson correlation coefficient to obtain a quantitative action rule of the input characteristic on an output target, and realizing parameter sensitivity analysis of the structural parameters of the catalytic layer on the peak power density, the limiting current density and the temperature uniformity of the battery; and performing parameter sensitivity grading and core feature screening on the input features based on the parameter sensitivity analysis result.
The step S5 specifically includes the following steps: configuring a genetic algorithm according to an optimization task, and determining the number of populations, the number of elite, the maximum algebra of the populations, the stagnant algebra and the convergence residual error; normalizing the generated population genes according to a mapping mode in an original data set, sequentially substituting the population genes into data driving models under different optimization targets to predict the performance of the battery, and weighting and summing the predicted optimization targets under normalization to obtain the individual fitness:
W＝(w 1 ,w 2 ,w 3 )
wherein Fit is individual fitness, W is a parameter weight set, and Y is 0 Is a normalized output target data set;
the optimization target is to find a CCL structure which enables the electric rated power density to be larger, the limiting current density to be larger and the temperature nonuniformity to be smaller, the weights of the three output targets are equal, and the fitness function can be expressed as follows:
Fit＝-(P max,0 +I Lim,0 -σ T,CL,0 )
wherein, P max,0 、I Lim,0 And σ T,CL,0 Respectively obtaining normalized peak power density, normalized limit current density and normalized peak point temperature uniformity;
and (3) performing genetic operation according to the individual fitness, specifically comprising selection, crossing and variation, calculating to obtain a new population, performing convergence judgment according to the set cut-off condition, and obtaining an optimal parameter set, namely the catalytic layer structure obtained by comprehensive optimization.
The step S6 specifically includes the following steps: the obtained catalyst layer structure parameters are brought into a fuel cell mathematical model to complete model verification, and the catalyst layer structure of the proton exchange membrane fuel cell with high load capacity, strong discharge performance and good thermal uniformity is obtained. Compared with the prior art, the method aims at the catalytic layer of the fuel cell, and realizes the rapid prediction, quantitative analysis and multi-objective optimization of the structural parameters of the catalytic layer by coupling the genetic algorithm on the basis of the machine learning algorithm. The method has the advantages that the rapid prediction of the battery performance under different catalyst layer structures is realized, meanwhile, the influence rule of the catalyst layer structure parameters on the battery performance is disclosed, the multi-target multi-parameter comprehensive optimization is carried out on the catalyst layer structure parameters from multiple angles, and guidance is provided for the design and development of the actual fuel battery catalyst layer. The method can realize rapid prediction of the battery performance under different catalytic layer structures by using a data driving model developed by machine learning, and has obvious time cost advantage compared with the traditional experiment and simulation method; the data driving model can simultaneously realize the global quantitative sensitivity analysis of the catalyst layer structure parameters, and the parameter global sensitivity which is difficult to determine in the traditional single variable method is obtained; the data driving model is coupled with the genetic algorithm, so that multi-target multi-parameter comprehensive optimization can be performed on the structural parameters of the catalytic layer from multiple angles, and the catalytic layer comprehensive optimization structure which is difficult to determine in the traditional method is obtained.
Therefore, the method can quickly, accurately and comprehensively optimize the structural parameters of the catalyst layer in a multi-target manner, and provides guidance for the design and development of the actual fuel cell catalyst layer
Drawings
FIG. 1 is a schematic diagram of a process for applying the method of the present invention in an example;
FIG. 2 is a schematic diagram of a mathematical model domain of a two-dimensional two-phase non-isothermal PEMFC in an embodiment;
the notation in the figure is: 1. the fuel cell comprises an anode plate, 2, an anode fuel flow passage, 3, an anode gas diffusion layer, 4, an anode catalyst layer, 5, a proton exchange membrane, 6, a cathode catalyst layer, 7, a cathode gas diffusion layer, 8, a cathode fuel flow passage, 9, a cathode plate, 10, an anode microporous layer, 11 and a cathode microporous layer;
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments.
In this embodiment, a fuel cell catalyst layer prediction-analysis-optimization method based on machine learning as shown in fig. 1 is used to analyze and optimize the structural parameters of a cathode catalyst layer in a proton exchange membrane fuel cell, which is shown in fig. 2 and composed of 1, an anode plate, 2, an anode fuel flow channel, 3, an anode gas diffusion layer, 4, an anode catalyst layer, 5, a proton exchange membrane, 6, a cathode catalyst layer, 7, a cathode gas diffusion layer, 8, a cathode fuel flow channel, 9, a cathode plate, 10 anode microporous layers, 11, and a cathode microporous layer.
The optimization process mainly comprises the following steps:
and S1, building a mathematical model and obtaining a data driving source. Aiming at multi-scale and multi-physical phenomena in a proton exchange membrane fuel cell, a two-dimensional two-phase non-isothermal fuel cell mathematical model is built by using a coupling control equation, and a model control domain is shown in figure 1; the built data model uses a mass conservation equation, a momentum conservation equation, an energy conservation equation, an electronic conservation equation, a proton conservation equation and a Butler-Volmer equation, realizes the accurate calculation of multiple physical processes such as gas-liquid two-phase flow channels, heat and mass transfer, electrochemical reaction and the like in the battery, and couples a spherical aggregate model in the catalytic layer and a local oxygen transmission model to realize the accurate description of the multiple-scale physical phenomenon in the catalytic layer;
performing multi-physical field coupling based on the control equation, performing grid division, and performing discrete calculation by using a finite element analysis method to obtain a battery output polarization curve; and comparing the test result of the experimental fuel cell with the model output to verify the effectiveness of the mathematical model, wherein the verified mathematical model is used as a data driving source to assist the development of the data driving model.
And S2, selecting the structural parameters of the catalyst layer and the battery output target to generate a data set. To ensure that the selected parameters of the catalytic layer are operable, 9 structural parameters of the catalytic layer, which are often involved in the actual production design, are selected as input features of the data-driven model:
X＝(m Pt ,x Pt/C ,x Ion/C ,r Agg ,r C ,r Pt ,dp CL ,H CL ,θ CL )
where X is the input parameter set, m Pt For catalytic layer platinum loading, x Pt/C Is the ratio of platinum to carbon in the catalyst layer, x Ion/C Is the mass ratio of ionomer to carbon support in the catalytic layer, r Agg Radius of catalyst agglomerates, r C Is the carbon particle radius, r Pt Is the platinum particle radius, dp CL Pore diameter in the catalyst layer H CL Thickness of catalytic layer, theta CL A catalytic layer contact angle;
in order to describe the performance of the fuel cell in three aspects of maximum load capacity, maximum discharge capacity and thermal uniformity, the peak power density, the limiting current density and the maximum temperature difference of the catalytic layer at a peak point of the cell are selected as optimization targets:
Y＝(P max ,I Lim ,ΔT CL )
wherein Y is the output parameter set, P max Is the peak power density, I Lim Is the limiting current density, Δ T CL The maximum temperature difference in the catalytic layer;
to ensure the rationality of CCL structural parametersThe variation range of the input parameters needs to be set: m is Pt ∈(0.025,0.5)mg·m -2 ， x Pt/C ∈(0.033,0.5)，x Ion/C ∈(0.2,1.3)，r Agg ∈(60,1000)nm，r C ∈(10,1000)nm，r Pt ∈(1,5)nm， dp CCL ∈(10,200)nm，H CL ∈(1,40)μm，θ CL E (92,150) °; meanwhile, in order to ensure the rationality of the whole CCL structure and prevent failure of the catalyst layer structure caused by error accumulation, the porosity in the catalyst layer is controlled within the range of 0.05-0.95, and the ratio of the radius of the aggregate to the radius of the carbon particles is controlled to be more than 4;
according to the variation range of each optimized parameter, determining that each input parameter randomly varies within the range and obeys the mean value mu i Is the median, variance of the interval
Normal distribution of (2); and sequentially bringing each group of randomly generated input sets into a mathematical model, calculating corresponding optimized target values, and combining the optimized target values and the optimized target values to obtain a data driving data set.
And S3, preprocessing the data set and developing a data driving model. Performing normalization pretreatment on the obtained data set, mapping each parameter data to an interval of 0-1 to eliminate the influence of different magnitude levels among different data, dividing the data set into a training set and a test set according to the proportion of 0.75:0.25, and respectively using the training set and the test set for data-driven model training and model verification;
aiming at the problems of strong nonlinear relation and high training difficulty of parameters in the catalyst layer, an integrated learning algorithm extreme gradient lifting tree (XGboost) which considers model accuracy and interpretability is selected for training a data driving model, and a test set is used for verifying the model accuracy; and debugging the hyper-parameters in the XGboost of different training targets by using a grid search method in the training process, wherein the debugging specifically comprises the maximum tree depth, the learning rate, the minimum cotyledon weight, the regularization coefficient and the sampling rate, and obtaining the optimal hyper-parameter set to complete the XGboost algorithm configuration.
And S4, battery performance rapid prediction and parameter quantitative sensitivity analysis based on the data-driven model. Exercise stationThe obtained data-driven model constructs a direct mapping relation between the catalytic layer structure and multiple targets of the battery, and the developed data-driven model realizes rapid prediction of the peak power density, the limiting current density and the temperature uniformity of the peak point of the battery under different catalytic layer structures, and the prediction precision R 2 >0.95；
The trained data driving model can simultaneously realize sensitivity analysis on input features, the influence degree of each input feature on each output target is quantified, and the influence degree is combined with the action direction in the Pearson correlation coefficient, so that the quantitative action rule of the input features on the output targets can be obtained;
based on the parameter sensitivity analysis result, carrying out sensitivity grading on the input characteristics, wherein | S T |>0.15 is extremely sensitive, 0.15>|S T |>0.05 is quite sensitive, | S T |<0.05 is not sensitive; and determining the sensitivity of the characteristic according to the average sensitivity value of the input characteristic under different optimization targets, completing parameter sensitivity grading, selecting the characteristic with the characteristic sensitivity of quite sensitive and extremely sensitive as a core characteristic, and selecting the platinum loading capacity, the platinum-carbon ratio, the catalyst aggregate radius and the catalyst layer thickness in the cathode catalyst layer with the selected core characteristic.
And S5, coupling the data driving model and the catalytic layer multi-target multi-parameter optimization of the genetic algorithm. Further performing multi-target multi-parameter comprehensive optimization on the core characteristics of the selected catalytic layer by applying a genetic algorithm; according to the optimization task configuration genetic algorithm, in the embodiment, the population number is 200, the elite number is 2, the population maximum algebra is 300, the stagnation algebra is 50, and the convergence residual error is 1e -7 And applying parameter limiting conditions when generating the data set to ensure the reasonability of the population;
normalizing the generated population genes according to a mapping mode in an original data set, sequentially substituting the population genes into data driving models under different optimization targets to predict the performance of the battery, and weighting and summing the predicted optimization targets under normalization to obtain the fitness of an individual:
W＝(w 1 ,w 2 ,w 3 )
wherein Fit is individual fitness, W is a parameter weight set, and Y is 0 Outputting a normalized output target data set;
in the optimization process of the research, the optimization target is to find a CCL structure which enables the electrical rated power density to be larger, the limiting current density to be larger and the temperature nonuniformity to be smaller, the three output targets are equal in weight, and the fitness function can be expressed as:
Fit＝-(P max,0 +I Lim,0 -ΔT CL,0 )
wherein, P max,0 、I Lim,0 And Δ T CL,0 Respectively obtaining normalized peak power density, normalized limit current density and normalized peak point temperature uniformity;
in the embodiment, a random uniform distribution selection method (Stochastic uniform) is adopted as a selection function, a discrete recombination method (Scattered) is adopted as a cross function, the cross probability is 0.8, Gaussian variation (Gaussian) is adopted as a variation function, the variation probability is 0.2, a new population is obtained by calculation, convergence judgment is performed according to a set cut-off condition, and an optimal parameter set, namely a catalytic layer structure obtained by comprehensive optimization, is finally obtained.
And S6, verifying the structural parameters of the catalytic layer obtained by optimization. Finally, the obtained catalyst layer structure parameters are brought into a fuel cell mathematical model to complete model verification, the obtained cathode catalyst layer structure obviously improves the cell discharge performance, especially in a high current density section, the cell concentration polarization loss is obviously reduced, and the proton exchange membrane fuel cell catalyst layer structure with high load capacity, strong discharge performance and good thermal uniformity is obtained.
The method can realize the rapid prediction, quantitative analysis and multi-objective optimization of the structural parameters of the catalyst layer of the fuel cell, and provides guidance for the design and development of the actual catalyst layer of the fuel cell.
Claims (7)
1. A fuel cell catalyst layer prediction-analysis-optimization method based on machine learning is characterized by comprising the following steps:
s1, building a mathematical model to obtain a data driving source;
s2, selecting the catalyst layer structure parameters and the battery output target to generate a data set;
s3, preprocessing a data set and establishing a data driving model;
s4, battery performance rapid prediction and parameter quantitative sensitivity analysis based on the data driving model;
s5, coupling the data driving model and the catalytic layer multi-target multi-parameter optimization of the genetic algorithm;
and S6, verifying the structural parameters of the catalytic layer obtained by optimization.
2. The fuel cell catalyst layer prediction-analysis-optimization method based on machine learning according to claim 1, wherein the step S1 specifically includes the following steps: aiming at multi-scale and multi-physical phenomena in the proton exchange membrane fuel cell, a two-dimensional two-phase non-isothermal fuel cell mathematical model is built by using a coupling control equation, model accuracy verification is completed by comparing the mathematical model with a polarization curve of an experimental test, and the verified mathematical model is used as a data driving source to assist data driving of the model.
3. The fuel cell catalyst layer prediction-analysis-optimization method based on machine learning according to claim 1, wherein the step S2 specifically includes the following steps: selecting 9 catalytic layer structure parameters as input characteristics of a data-driven model:
X＝(m Pt ,x Pt/C ,x Ion/C ,r Agg ,r C ,r Pt ,dp CL ,H CL ,θ CL )
where X is the input parameter set, m Pt For catalytic layer platinum loading, x Pt/C Is the ratio of platinum to carbon in the catalyst layer, x Ion/C Is the mass ratio of ionomer to carbon support in the catalytic layer, r Agg Radius of catalyst agglomerates，r C Is the carbon particle radius, r Pt Is the platinum particle radius, dp CL Pore diameter in the catalyst layer H CL Thickness of catalytic layer, theta CL A catalytic layer contact angle;
in order to describe the performance of the fuel cell in three aspects of maximum load capacity, maximum discharge capacity and thermal uniformity, the peak power density, the limiting current density and the standard deviation of the temperature distribution of the catalytic layer at the peak point of the cell are selected as optimization targets:
Y＝(P max ,I Lim ,σ T,CL )
wherein Y is the output parameter set, P max Is the peak power density, I Lim Is the limiting current density, σ T,CL Is the standard deviation of the temperature distribution in the catalytic layer;
determining the variation range of the input parameters: m is Pt ∈(0.025,0.5)mg·m -2 ，x Pt/C ∈(0.033,0.5)，x Ion/C ∈(0.2,1.3)，r Agg ∈(60,1000)nm，r C ∈(10,1000)nm，r Pt ∈(1,5)nm，dp CCL ∈(10,200)nm，H CL ∈(1,40)μm，θ CL E (92,150) °; in order to ensure the rationality of the structure of the catalyst layer of the whole fuel cell and prevent the failure of the catalyst layer structure caused by error accumulation, the limiting condition that the ratio of the radius of aggregates to the radius of carbon particles in the catalyst layer of the fuel cell is more than 4 is set;
according to the variation range of each optimized parameter, determining that each input parameter randomly varies within the range and obeys the mean value mu i Is the median, variance of the interval
Normal distribution of (2); and sequentially bringing each group of randomly generated input sets into a mathematical model, calculating corresponding optimized target values, and combining the optimized target values and the optimized target values to obtain a data driving data set.
4. The fuel cell catalyst layer prediction-analysis-optimization method based on machine learning according to claim 1, wherein the step S3 specifically includes the following steps: performing normalization pretreatment on the obtained data set, mapping each parameter data to an interval of 0-1 to eliminate the influence of different magnitude levels among different data, dividing the data set into a training set and a test set according to the proportion of 0.75:0.25, and respectively using the training set and the test set for data-driven model training and model verification;
selecting an integrated learning algorithm extreme gradient lifting tree (XGboost) to carry out data-driven model training, and carrying out model accuracy verification by using a test set; and debugging the hyper-parameters in the XGboost of different training targets by using a grid search method in the training process to complete the configuration of the XGboost algorithm.
5. The fuel cell catalyst layer prediction-analysis-optimization method based on machine learning according to claim 1, wherein the step S4 specifically includes the following steps: a direct mapping relation between the catalytic layer structure and multiple targets of the battery is constructed by applying a data driving model obtained by training, and the rapid prediction of the battery peak power density, the limiting current density and the peak point temperature uniformity under different catalytic layer structures is realized;
extracting the characteristic influence degree in the data driving model, combining the characteristic influence degree with the action direction in the Pearson correlation coefficient to obtain a quantitative action rule of the input characteristic on an output target, and realizing parameter sensitivity analysis of the structural parameters of the catalytic layer on the peak power density, the limiting current density and the temperature uniformity of the battery; and performing parameter sensitivity grading and core feature screening on the input features based on the parameter sensitivity analysis result.
6. The fuel cell catalyst layer prediction-analysis-optimization method based on machine learning according to claim 1, wherein the step S5 specifically includes the following steps: configuring a genetic algorithm according to an optimization task, and determining the number of populations, the number of elite, the maximum algebra of the populations, the stagnation algebra and the convergence residual error; normalizing the generated population genes according to a mapping mode in an original data set, sequentially substituting the population genes into data driving models under different optimization targets to predict the performance of the battery, and weighting and summing the predicted optimization targets under normalization to obtain the individual fitness:
Fit＝WY 0 T
W＝(w 1 ,w 2 ,w 3 )
wherein Fit is individual fitness, W is a parameter weight set, and Y is 0 Outputting a normalized output target data set;
the optimization target is to find a CCL structure which enables the electric rated power density to be larger, the limiting current density to be larger and the temperature nonuniformity to be smaller, the weights of the three output targets are equal, and the fitness function can be expressed as follows:
Fit＝-(P max,0 +I Lim,0 -σ T,CL,0 )
wherein, P max,0 、I Lim,0 And σ T,CL,0 Respectively obtaining normalized peak power density, normalized limit current density and normalized peak point temperature uniformity;
and (3) performing genetic operation according to the individual fitness, specifically comprising selection, crossing and variation, calculating to obtain a new population, performing convergence judgment according to the set cut-off condition, and obtaining an optimal parameter set, namely the catalytic layer structure obtained by comprehensive optimization.
7. The fuel cell catalyst layer prediction-analysis-optimization method based on machine learning according to claim 1, wherein the step S6 specifically includes the following steps: the obtained catalyst layer structure parameters are brought into a fuel cell mathematical model to complete model verification, and the catalyst layer structure of the proton exchange membrane fuel cell with high load capacity, strong discharge performance and good thermal uniformity is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210538054.8A CN115000471A (en) | 2022-05-18 | 2022-05-18 | Fuel cell catalyst layer prediction-analysis-optimization method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210538054.8A CN115000471A (en) | 2022-05-18 | 2022-05-18 | Fuel cell catalyst layer prediction-analysis-optimization method based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115000471A true CN115000471A (en) | 2022-09-02 |
Family
ID=83027068
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210538054.8A Pending CN115000471A (en) | 2022-05-18 | 2022-05-18 | Fuel cell catalyst layer prediction-analysis-optimization method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115000471A (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150278704A1 (en) * | 2014-03-28 | 2015-10-01 | Korea Institute Of Energy Research | Method for optimization of fuel cells operating conditions using hybrid model |
| CN107463995A (en) * | 2017-08-01 | 2017-12-12 | 浙江大学 | Fuel cell Optimization Modeling method with Adaptive Genetic strategy RNA GA |
| CN107944072A (en) * | 2017-10-11 | 2018-04-20 | 天津大学 | The definite method of Proton Exchange Membrane Fuel Cells sensitivity to parameter |
| CN108681742A (en) * | 2018-04-12 | 2018-10-19 | 上海交通大学 | A kind of analysis method for analyzing driver driving behavioural characteristic sensibility |
| CN109992921A (en) * | 2019-04-12 | 2019-07-09 | 河北工业大学 | A kind of online soft sensor method and system of the coal-fired plant boiler thermal efficiency |
| CN111200141A (en) * | 2020-01-14 | 2020-05-26 | 东北电力大学 | Proton exchange membrane fuel cell performance prediction and optimization method based on deep belief network |
| AU2020101453A4 (en) * | 2020-07-23 | 2020-08-27 | China Communications Construction Co., Ltd. | An Intelligent Optimization Method of Durable Concrete Mix Proportion Based on Data mining |
| WO2021073036A1 (en) * | 2019-10-15 | 2021-04-22 | 江苏大学 | Real-time global optimization intelligent control system and method for fuel cell bus |
| CN113946995A (en) * | 2021-10-18 | 2022-01-18 | 上海神力科技有限公司 | Multi-objective fuel cell cooling flow channel optimization design method |
| CN114117953A (en) * | 2021-11-08 | 2022-03-01 | 武汉大学 | Hydrological model structure diagnosis method, runoff forecasting method and device |
| US20220067249A1 (en) * | 2020-02-05 | 2022-03-03 | Imagars Llc | Machine Learning to Accelerate Design of Energetic Materials |
| CN114447378A (en) * | 2022-02-10 | 2022-05-06 | 温州大学 | Parameter optimization method of proton exchange membrane fuel cell |
| CN114462305A (en) * | 2022-01-17 | 2022-05-10 | 青岛海洋地质研究所 | Argo buoy long-endurance energy consumption optimization and life estimation method |
-
2022
- 2022-05-18 CN CN202210538054.8A patent/CN115000471A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150278704A1 (en) * | 2014-03-28 | 2015-10-01 | Korea Institute Of Energy Research | Method for optimization of fuel cells operating conditions using hybrid model |
| CN107463995A (en) * | 2017-08-01 | 2017-12-12 | 浙江大学 | Fuel cell Optimization Modeling method with Adaptive Genetic strategy RNA GA |
| CN107944072A (en) * | 2017-10-11 | 2018-04-20 | 天津大学 | The definite method of Proton Exchange Membrane Fuel Cells sensitivity to parameter |
| CN108681742A (en) * | 2018-04-12 | 2018-10-19 | 上海交通大学 | A kind of analysis method for analyzing driver driving behavioural characteristic sensibility |
| CN109992921A (en) * | 2019-04-12 | 2019-07-09 | 河北工业大学 | A kind of online soft sensor method and system of the coal-fired plant boiler thermal efficiency |
| WO2021073036A1 (en) * | 2019-10-15 | 2021-04-22 | 江苏大学 | Real-time global optimization intelligent control system and method for fuel cell bus |
| CN111200141A (en) * | 2020-01-14 | 2020-05-26 | 东北电力大学 | Proton exchange membrane fuel cell performance prediction and optimization method based on deep belief network |
| US20220067249A1 (en) * | 2020-02-05 | 2022-03-03 | Imagars Llc | Machine Learning to Accelerate Design of Energetic Materials |
| AU2020101453A4 (en) * | 2020-07-23 | 2020-08-27 | China Communications Construction Co., Ltd. | An Intelligent Optimization Method of Durable Concrete Mix Proportion Based on Data mining |
| CN113946995A (en) * | 2021-10-18 | 2022-01-18 | 上海神力科技有限公司 | Multi-objective fuel cell cooling flow channel optimization design method |
| CN114117953A (en) * | 2021-11-08 | 2022-03-01 | 武汉大学 | Hydrological model structure diagnosis method, runoff forecasting method and device |
| CN114462305A (en) * | 2022-01-17 | 2022-05-10 | 青岛海洋地质研究所 | Argo buoy long-endurance energy consumption optimization and life estimation method |
| CN114447378A (en) * | 2022-02-10 | 2022-05-06 | 温州大学 | Parameter optimization method of proton exchange membrane fuel cell |
Non-Patent Citations (2)
| Title |
|---|
| 姚红光: "一带一路沿线国家物流网络的结构分析及深度融合研究", 30 November 2021, 北京：旅游教育出版社, pages: 276 - 277 * |
| 杜新;王路路;王金龙;吴利荣;: "一种考虑催化层结构参数的PEMFC气相模型", 电源技术, no. 03 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | Global parameters sensitivity analysis and development of a two-dimensional real-time model of proton-exchange-membrane fuel cells | |
| CN109324291B (en) | Prediction method for predicting service life of proton exchange membrane fuel cell | |
| CN111146478B (en) | Method for predicting residual service life of proton exchange membrane fuel cell stack | |
| Nanadegani et al. | On neural network modeling to maximize the power output of PEMFCs | |
| Yang et al. | Modeling and optimization for proton exchange membrane fuel cell stack using aging and challenging P systems based optimization algorithm | |
| CN103336867B (en) | Proton Exchange Membrane Fuel Cells model optimization disposal route | |
| Sun et al. | Proton exchange membrane fuel cell model parameter identification based on dynamic differential evolution with collective guidance factor algorithm | |
| CN112289385B (en) | Electrochemical impedance spectrum prediction method for high-power proton exchange membrane fuel cell stack | |
| CN114447378B (en) | Parameter optimization method for proton exchange membrane fuel cell | |
| CN113946995A (en) | Multi-objective fuel cell cooling flow channel optimization design method | |
| CN114943372A (en) | Method and device for predicting life of proton exchange membrane based on Bayesian recurrent neural network | |
| CN106709131A (en) | Parameter intelligent optimization method suitable for proton exchange membrane fuel cell model | |
| Liu et al. | Multi-objective optimization of proton exchange membrane fuel cell geometry and operating parameters based on three new performance evaluation indexes | |
| Deng et al. | Degradation prediction of PEMFCs using stacked echo state network based on genetic algorithm optimization | |
| Su et al. | Study of cell voltage uniformity of proton exchange membrane fuel cell stack with an optimized artificial neural network model | |
| Du et al. | A fault diagnosis model for proton exchange membrane fuel cell based on impedance identification with differential evolution algorithm | |
| Zhang et al. | Degradation prediction model of PEMFC based on multi-reservoir echo state network with mini reservoir | |
| Fu et al. | Proton membrane fuel cell stack performance prediction through deep learning method | |
| CN115000471A (en) | Fuel cell catalyst layer prediction-analysis-optimization method based on machine learning | |
| CN115642269B (en) | Solid oxide fuel cell structure and optimal design method thereof | |
| CN115547433A (en) | Modeling optimization method for proton exchange membrane electrolytic cell | |
| CN113540534B (en) | Starvation fault diagnosis method for proton exchange membrane fuel cell stack | |
| CN113078335B (en) | Performance analysis method and device of proton exchange membrane fuel cell and terminal equipment | |
| CN110059342A (en) | A kind of method for parameter estimation of lithium ion battery P2D model | |
| CN114415034A (en) | Lithium ion battery capacity prediction method based on internal mechanism analysis |
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 |