CN117725446B - SOFC (solid oxide fuel cell) stack performance decay prediction method and system based on GA-BP (GAs-liquid-phase reactor) - Google Patents
SOFC (solid oxide fuel cell) stack performance decay prediction method and system based on GA-BP (GAs-liquid-phase reactor) Download PDFInfo
- Publication number
- CN117725446B CN117725446B CN202410173816.8A CN202410173816A CN117725446B CN 117725446 B CN117725446 B CN 117725446B CN 202410173816 A CN202410173816 A CN 202410173816A CN 117725446 B CN117725446 B CN 117725446B
- Authority
- CN
- China
- Prior art keywords
- sofc
- voltage
- population
- probability
- individual
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000000446 fuel Substances 0.000 title claims abstract description 36
- 239000007787 solid Substances 0.000 title claims abstract description 27
- 239000007791 liquid phase Substances 0.000 title description 2
- 238000013528 artificial neural network Methods 0.000 claims abstract description 40
- 230000002068 genetic effect Effects 0.000 claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 18
- 230000008859 change Effects 0.000 claims abstract description 12
- 230000003044 adaptive effect Effects 0.000 claims description 34
- 230000035772 mutation Effects 0.000 claims description 26
- 238000002485 combustion reaction Methods 0.000 claims description 14
- 238000011156 evaluation Methods 0.000 claims description 12
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 9
- 230000015556 catabolic process Effects 0.000 claims description 6
- 238000006731 degradation reaction Methods 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 3
- 230000001105 regulatory effect Effects 0.000 claims 2
- 210000004027 cell Anatomy 0.000 description 32
- 230000006870 function Effects 0.000 description 14
- 238000012360 testing method Methods 0.000 description 10
- 210000000349 chromosome Anatomy 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 4
- 101150030229 nth gene Proteins 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 238000010248 power generation Methods 0.000 description 4
- 108090000623 proteins and genes Proteins 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 239000001257 hydrogen Substances 0.000 description 3
- 229910052739 hydrogen Inorganic materials 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000007789 gas Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000003345 natural gas Substances 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000013016 damping Methods 0.000 description 1
- 239000007772 electrode material Substances 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 150000002431 hydrogen Chemical class 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000002407 reforming Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- -1 temperature Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- 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
- Fuel Cell (AREA)
Abstract
The invention discloses a method and a system for predicting SOFC stack performance attenuation based on GA-BP, which relate to the technical field of stack performance attenuation prediction, and comprise the steps of firstly acquiring an SOFC prototype operation data set which comprises voltage and voltage influence parameters, and designing a BP neural network structure; training by using an SOFC prototype operation data set, and optimizing an initial weight and a threshold value of a BP neural network by using an improved genetic algorithm to obtain a GA-BP solid oxide fuel cell voltage prediction model; collecting the latest SOFC actual operation data, and inputting voltage influence parameters into a model to obtain a voltage predicted value; and obtaining a voltage change trend according to the voltage predicted value, and obtaining the voltage change trend as the stack performance attenuation trend under the same current condition. The invention combines the improved genetic algorithm with the BP neural network to find out the better parameter combination, considers the shutdown time and the shutdown times and improves the accuracy of the prediction result.
Description
Technical Field
The invention relates to the technical field of stack performance decay prediction, in particular to a method and a system for predicting SOFC stack performance decay based on GA-BP.
Background
Solid Oxide Fuel Cells (SOFCs) have become a well-established new generation of alternative energy power generation systems. SOFCs are considered to be applicable to automobiles, replacing conventional internal combustion engines, due to their many advantages of high efficiency, no emissions, low noise, etc. Fossil energy sources such as coal and petroleum have taken up the main body of energy sources in the beginning of the 21 st century, but this energy source structure has a number of problems. Natural gas is widely regarded as an efficient, clean and widely applicable energy source, and is an ideal alternative energy source. Natural gas has the advantages of hydrogen production by reforming, various storage modes, convenient transportation and the like. Solid Oxide Fuel Cells (SOFCs) are believed to be applicable to automobiles, replacing the ideal alternatives to conventional internal combustion engines. Unlike conventional forms of power generation, SOFCs produce electrical energy through the reaction of oxygen with hydrogen, and therefore their efficiency is not limited by the carnot cycle. Meanwhile, SOFC has been widely used with its characteristics of high efficiency, environmental protection, silence and reliability. In order to meet the industrialization requirement, the SOFC must be integrated with BOP components such as a tail gas combustion chamber, a heat exchanger, a reformer and the like to meet the power requirement of new energy automobiles and ships.
However, the service life and durability of SOFCs in the system environment cannot meet the commercial demands, which is one of the main problems facing the large-scale popularization of SOFCs. In addition, the operation time of the SOFC stack has close correlation with the service life and durability of the SOFC stack due to abnormal shutdown, and particularly the abnormal shutdown has great influence on the service life of the SOFC stack, and the power generation performance and structural characteristics of the cells in the stack are damaged. Particularly, abnormal shutdown factors, namely the temperature runaway condition occurs in the operation process of the SOFC system, and fuel which is not fully utilized is usually led into a combustion chamber for combustion utilization in the normal operation process of the fuel cell, but the temperature is rapidly increased in the combustion process, so that the temperature runaway is caused, the fuel cell automatically starts a self-shutdown protection measure, the current suddenly drops to 0, and the voltage is severely fluctuated. Therefore, predicting the damping characteristics of SOFC systems is of great importance for extending their lifetime and durability. The fuel cell voltage, as a common parameter of SOFC systems, can be measured directly using sensors. And the voltage may directly reflect the output performance of the fuel cell. Degradation of the cell performance necessarily results in voltage decay with constant current density. Therefore, the voltage is selected as a health index of the fuel cell, and the evaluation of the SOFC performance attenuation is realized through the prediction of the voltage.
In past research, there have been many machine learning methods applied to predict the voltage of Solid Oxide Fuel Cells (SOFCs), such as BP neural networks. However, the prediction accuracy of the BP neural network is not ideal, and there is a large error. Such errors may be due to the very complex operation of the SOFC cell involving interactions of various factors. For example, the electrode materials, temperature, gas flow, chemical reactions, and the like of an SOFC cell all have an impact on their voltage. Furthermore, SOFC cell performance may also be affected by noise and data imperfections, which may also lead to increased errors in machine learning method predictions. Therefore, a more accurate predictive model is needed to address this problem, taking more factors into account.
Disclosure of Invention
In view of the above, the invention provides a method and a system for predicting SOFC stack performance decay based on GA-BP.
In order to achieve the above object, the present invention provides the following technical solutions:
A SOFC stack performance decay prediction method based on GA-BP comprises the following steps:
Step 1, acquiring SOFC prototype operation data acquired from an SOFC system, and preprocessing the data to form an SOFC prototype operation data set; the SOFC prototype operation data comprise voltage and voltage influence parameters;
step 2, designing a BP neural network structure according to the voltage and the voltage influence parameters of the SOFC prototype operation data set;
Step 3, taking the voltage influence parameters in the SOFC prototype operation data set as input, taking the voltage as output, training, and optimizing the initial weight and the threshold value of the BP neural network by using an improved genetic algorithm to obtain a GA-BP solid oxide fuel cell voltage prediction model;
Step 4, acquiring SOFC actual operation data acquired from an SOFC system, and inputting voltage influence parameters into a GA-BP solid oxide fuel cell voltage prediction model to obtain a voltage prediction value;
and step 5, obtaining a voltage change trend according to the voltage predicted value, and obtaining the voltage change trend as the stack performance attenuation trend under the same current condition.
Optionally, in the step 1, the data preprocessing includes data normalization, and the formula is as follows:
；
Wherein y is the normalized SOFC prototype operation data, and y max=1,ymin =0; x is the running data of the original SOFC prototype to be normalized currently; x max and x min are the maximum value and the minimum value of the original SOFC prototype operation data to be processed in the SOFC prototype operation data set respectively.
Optionally, in step 1, the voltage influencing parameter is a system operating parameter that influences voltage, including a combustion chamber temperature, a pile temperature, a current, a load power, a system operating time, a system abnormal shutdown frequency and shutdown time data.
Optionally, in the step 2, the designed BP neural network structure is: the number of the input layer nodes is 7, the number of the hidden layer nodes is 3, and the number of the output layer nodes is 1.
Optionally, in the step 3, the improvement of the genetic algorithm includes:
the traditional roulette selection operation is changed to an index selection operator;
designing a traditional fixed crossover probability as an adaptive crossover probability;
The conventional fixed variation probability is designed as an adaptive variation probability.
Optionally, in the step 3, the method for optimizing the initial weight and the threshold of the BP neural network by using an improved genetic algorithm is as follows:
step 3.1, generating N initial strings by adopting a random method as an initial group, wherein each initial string is called an individual;
Step 3.2, calculating the fitness value F i of each individual of the ith generation of population according to the fitness function;
Step 3.3, selecting: selecting a child population from the parent population by adopting an index selection operator;
step 3.4, cross operation: adopting an adaptive crossover operator of adaptive crossover probability to carry out crossover operation on two individuals selected from the population;
step 3.5, mutation operation: adopting an adaptive mutation operator with adaptive mutation probability to select individuals from the population to carry out mutation operation;
step 3.6, judging whether the genetic algebra meets the termination condition, if so, stopping operation, and assigning the network initial weight and the threshold value corresponding to the optimal individual to the BP neural network; otherwise, returning to the step 3.2.
Optionally, after the GA-BP solid oxide fuel cell voltage prediction model is obtained in the step 3, a model evaluation is further performed on the GA-BP solid oxide fuel cell voltage prediction model, where the evaluation parameter indexes include: mean absolute error MAE, mean square error MSE, root mean square error RMSE, decision coefficient R 2.
A GA-BP based SOFC stack performance decay prediction system comprising:
The data set acquisition module is used for acquiring the SOFC prototype operation data acquired from the SOFC system and preprocessing the data to form an SOFC prototype operation data set; the SOFC prototype operation data comprise voltage and voltage influence parameters;
the neural network structure design module is used for designing the structure of the BP neural network according to the voltage and the voltage influence parameters of the SOFC prototype operation data set;
the model training and optimizing module is used for taking the voltage influence parameters in the SOFC prototype operation data set as input, taking the voltage as output, training, optimizing the initial weight and the threshold value of the BP neural network by using an improved genetic algorithm, and obtaining a GA-BP solid oxide fuel cell voltage prediction model;
The voltage prediction module is used for acquiring the SOFC actual operation data acquired from the SOFC system, inputting the voltage influence parameters into the GA-BP solid oxide fuel cell voltage prediction model, and obtaining a voltage prediction value;
And the electric pile performance attenuation prediction module is used for obtaining the voltage change trend according to the voltage predicted value, and the electric pile performance attenuation trend is obtained under the same current condition.
According to the technical scheme, the invention provides the SOFC stack performance attenuation prediction method and system based on GA-BP, and compared with the prior art, the SOFC stack performance attenuation prediction method and system based on GA-BP have the following beneficial effects:
According to the invention, a BP neural network structure is constructed according to the sample data of sensing and measuring, a genetic algorithm is adopted to optimize the neural network, an optimal individual generated after optimization is assigned to the BP neural network, a GA-BP solid oxide fuel cell voltage prediction model is obtained, future galvanic pile voltage can be predicted according to the model, and further the performance attenuation condition of a galvanic pile of the system is estimated. When the running time, the combustion chamber temperature, the pile temperature, the current, the load power and the corresponding voltage data of the SOFC system are used as data sets to carry out training models, the traditional BP neural network is easy to fall into a local optimal solution, and the genetic algorithm is used as a global optimization method to help overcome the problem. According to the improved GA-BP method, the improved genetic algorithm is combined with the BP neural network, so that the model can perform better global search, a better parameter combination is found, and the prediction accuracy is improved.
The invention also considers the stop accumulation effect, namely, the stop phenomenon caused by the temperature runaway possibly occurs in the SOFC system during the operation process, and the phenomenon has great influence on the prediction of the voltage. Compared with other models, the GA-BP solid oxide fuel cell voltage prediction model adds two important factors of shutdown time and shutdown times in the input layer, so that the model can more comprehensively consider the influence of the effects on the voltage, and the accuracy of a prediction result is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of SOFC system shutdown point versus voltage fluctuation;
FIG. 2 is a flow chart of the improved genetic algorithm of the present invention for optimizing BP neural network model;
Fig. 3 is a comparison of test results using test sets: (a) unmodified GA-BP; (b) the present invention improves GA-BP;
FIG. 4 is a graph of fitness of a GA-BP solid oxide fuel cell voltage prediction model as a function of algebraic change;
FIG. 5 shows the comparison of the predicted results and the actual results of the algorithm model: (a) MLR; (b) RBF; (c) BP; (d) LSTM; (e) PSO-BP; (f) the present invention improves GA-BP;
fig. 6 is a schematic diagram of steps of a method for predicting performance degradation of an SOFC stack according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
The embodiment discloses a SOFC stack performance attenuation prediction method based on GA-BP, which comprises the following steps of:
step 1, setting a sensor, acquiring SOFC prototype operation data from an SOFC system, and preprocessing the data to form an SOFC prototype operation data set; the SOFC prototype operation data comprise voltage and voltage influence parameters; the voltage influence parameters are system operation parameters which influence the voltage, and comprise combustion chamber temperature, stack temperature, current, load power, system operation time, abnormal shutdown times of the system and shutdown time data. Fig. 1 is a graph of the shutdown point and voltage fluctuation of the SOFC system, and it can be seen that the shutdown point has a close relationship with the voltage oscillation, so that the shutdown point information (shutdown time and shutdown frequency) is included in the model to help to realize accurate prediction of the voltage.
In a specific implementation, the SOFC prototype operation data sampling time interval may be set to 1 minute.
And 2, designing a BP neural network structure according to the voltage and the voltage influence parameters of the SOFC prototype operation data set, wherein the number of input layer nodes is 7, the number of hidden layer nodes is 3, and the number of output layer nodes is 1.
And 3, taking voltage influence parameters in the SOFC prototype operation data set as input, taking voltage as output, training, optimizing an initial weight and a threshold value of the BP neural network by using an improved genetic algorithm (Genetic Algorithm, GA), performing iterative optimization continuously according to the genetic algorithm, obtaining an optimal individual through continuous selection, crossing and mutation operation, and assigning the initial weight and the threshold value corresponding to the individual to the BP network to obtain the GA-BP solid oxide fuel cell voltage prediction model.
Among other things, the improvement of the genetic algorithm includes the following three aspects:
(1) Improving the selection operation of the genetic algorithm: the conventional roulette selection operation is changed to employ an index selection operator.
And a nonlinear selection operator, namely an exponential selection operator is adopted, and the calculation of the selection probability is introduced into a nonlinear function so as to enhance the diversity of the selection operation. The index selection operator is adopted to highlight individuals with higher fitness and increase the probability of being selected, so that the genetic algorithm is guided to converge to the global optimal solution more quickly. Compared with the traditional linear selection operator, the exponential selection operator has larger selection capacity, and can effectively avoid the problems of premature convergence and sinking into a local optimal solution. By introducing a nonlinear selection operator, the genetic algorithm can better explore the search space, and individuals with higher fitness are more likely to be selected while diversity is reserved in the population. According to the characteristics of a specific problem, the size of the selection capacity can be controlled by adjusting the parameter k, so that better searching performance can be obtained. In the exponential selection operator, the probability that an individual is selected is proportional to the exponential function of its fitness. Specifically, assuming that the fitness of the ith individual in the population is F i, the probability that individual is selected can be expressed as:
；
Wherein: k is an adjusting factor, N is the population size, For the probability of the ith individual being selected, F i the fitness value of the ith individual.
(2) Crossover operation of improved genetic algorithm: the conventional fixed crossover probability is designed as an adaptive crossover probability.
The adaptive crossover operator of the adaptive crossover probability is adopted to improve crossover operation, and the crossover probability is dynamically adjusted to improve the searching performance of the genetic algorithm. In conventional genetic algorithms, the crossover probability is usually fixed, but such fixed crossover probability may not be adaptable to the nature of the problem and the requirements of the optimization process.
In the self-adaptive crossover operator, crossover probability is adjusted according to the fitness value of the individual and the average fitness value of the population. The adaptivity is enhanced using a nonlinear function sigmoid function. The difference in fitness can be mapped into a more appropriate cross probability range. Specifically, if the fitness of the ith individual in the population is F i and the average fitness of the population is avg F,pc, the initial crossover probability of the ith individual may be expressed as:
；
Wherein, a is a nonlinear function adaptive adjustment factor, and p c is 0.5.
By introducing the self-adaptive crossover probability, the genetic algorithm can dynamically adjust the crossover probability according to the fitness of individuals, so that individuals with higher fitness are more likely to participate in crossover operation, and the diversity and searching capability of the population are increased.
The crossover scheme is designed as that the crossover formula of the mth chromosome Y m and the nth chromosome Y n at the p-th position is as follows:
；
wherein: is a random seed, and/> 。
(3) Modification of genetic algorithm: the conventional fixed variation probability is designed as an adaptive variation probability.
An adaptive mutation operator with adaptive mutation probability is adopted to improve mutation operation, and the search performance of a genetic algorithm is improved by dynamically adjusting the mutation probability; specifically, assuming that the fitness of the ith individual in the population is F i and the average fitness of the population is avg F,pm is the initial mutation probability, the mutation probability of the individual i can be expressed as:
；
wherein b is a nonlinear function adaptive adjustment factor, and p m is 0.45.
Variation design, nth gene of mth chromosomeThe variation formula of (2) is:
；
Wherein, Is a gene/>Upper bound of (2); /(I)Is a gene/>G is the iteration number, G is the maximum iteration number,/>Is a random seed, and/>。
In step 3, using the improved genetic algorithm, see fig. 2, a specific method for optimizing the initial weights and thresholds of the BP neural network (improved GA-BP) is:
step 3.1, generating N initial strings by adopting a random method as an initial group, wherein each initial string is called an individual;
and 3.2, calculating the fitness value F i of each individual of the ith generation of population according to the fitness function. The calculation method of the fitness function comprises the following steps:
；
Wherein F is the fitness value, N is the number of training set samples, x i is the actual value of the training set samples, and x i' is the predicted value of the training set samples.
Step 3.3, selecting: and selecting a child population from the parent population by adopting an index selection operator, wherein the probability of the selected individual is P i, and the calculation formula is as follows:
；
Wherein: k is an adjusting factor, N is the population size, For the probability of the ith individual being selected, F i the fitness value of the ith individual.
Step 3.4, cross operation: adopting an adaptive crossover operator of adaptive crossover probability to carry out crossover operation on two individuals selected from the population;
if the fitness of the ith individual in the population is F i and the average fitness of the population is avg F,pc, the initial crossover probability of the ith individual in the population is represented as:
；
Wherein, a is a nonlinear function adaptive adjustment factor, and p c is 0.5.
By introducing the self-adaptive crossover probability, the genetic algorithm can dynamically adjust the crossover probability according to the fitness of individuals, so that individuals with higher fitness are more likely to participate in crossover operation, and the diversity and searching capability of the population are increased.
The crossover scheme is designed as that the crossover formula of the mth chromosome Y m and the nth chromosome Y n at the p-th position is as follows:
；
wherein: is a random seed, and/> 。
Step 3.5, mutation operation: adopting an adaptive mutation operator with adaptive mutation probability to select individuals from the population to carry out mutation operation;
The fitness of the ith individual in the population is F i, the average fitness of the population is avg F,pm, and the initial mutation probability is expressed as:
；
wherein b is a nonlinear function adaptive adjustment factor, and p m is 0.45.
Variation design, nth gene of mth chromosomeThe variation formula of (2) is:
；
Wherein, Is a gene/>Upper bound of (2); /(I)Is a gene/>G is the iteration number, G is the maximum iteration number,/>Is a random seed, and/>。
Step 3.6, judging whether the genetic algebra meets the termination condition, if so, stopping operation, and assigning the network initial weight and the threshold value corresponding to the optimal individual to the BP neural network; otherwise, returning to the step 3.2.
Step 4, acquiring the latest SOFC actual operation data acquired from an SOFC system, and inputting voltage influence parameters into a GA-BP solid oxide fuel cell voltage prediction model to obtain a voltage prediction value;
and step 5, obtaining a voltage change trend according to the voltage predicted value, and obtaining the voltage change trend as the stack performance attenuation trend under the same current condition.
Optionally, in the step 1, the data preprocessing includes data normalization, and the formula is as follows:
；
Wherein y is the normalized SOFC prototype operation data, and y max=1,ymin =0; x is the running data of the original SOFC prototype to be normalized currently; x max and x min are the maximum value and the minimum value of the original SOFC prototype operation data to be processed in the SOFC prototype operation data set respectively.
In step 1 of this embodiment, the collected SOFC prototype operation data includes 9104 groups, and the variables included in the data include the combustion chamber temperature, the stack temperature, the current, the load power, the voltage, the system operation time, the number of abnormal shutdown times and the shutdown time. 68.35% of the 9104 set of data was divided into training data sets and 31.65% into test data sets.
The training data set is used for training the BP neural network in the step 3, the input value is the combustion chamber temperature, the pile temperature, the current, the load power, the system running time, the abnormal shutdown times and the shutdown time of the system of the data of the front 6222 group during training, and the output value is the voltage of the front 6222 group.
The test dataset was used to perform test evaluation on the generated GA-BP solid oxide fuel cell voltage prediction model. The input values of the test are the temperature of the combustion chamber, the temperature of the electric pile, the current, the load power, the running time of the system, the abnormal shutdown times and the shutdown time of the system, and the output values are the voltage of the rear 2882 group. The result pairs tested using the test set are shown in fig. 3, for example. According to the test result, the prediction effect of the GA-BP solid oxide fuel cell voltage prediction model is evaluated, and the parameter indexes used for evaluation comprise: mean absolute error MAE, mean square error MSE, root mean square error RMSE, decision coefficient R 2. Wherein MAE, MSE, RMSE are errors, and the smaller the numerical value is, the more accurate the representation model is. R 2 is used to measure the regression model,The closer R 2 is to 1, the better the model fitting effect is; MAE, MSE, RMSE and R 2 are defined as:
；
；
；
；
wherein: is the actual value of the fuel cell stack voltage; /(I) A predicted value of a fuel cell stack voltage; /(I)Is the average value of the fuel cell stack voltage, and n is the number of samples. The evaluation index of the results of the test using the test dataset is shown in table 1.
TABLE 1 evaluation and comparison of GA-BP before and after improvement Using an evaluation System
Example 2
Corresponding to the method, the invention also provides a SOFC stack performance attenuation prediction system based on GA-BP, which comprises the following steps:
The data set acquisition module is used for acquiring the SOFC prototype operation data acquired from the SOFC system and preprocessing the data to form an SOFC prototype operation data set; the SOFC prototype operation data comprise voltage and voltage influence parameters;
the neural network structure design module is used for designing the structure of the BP neural network according to the voltage and the voltage influence parameters of the SOFC prototype operation data set;
the model training and optimizing module is used for taking the voltage influence parameters in the SOFC prototype operation data set as input, taking the voltage as output, training, optimizing the initial weight and the threshold value of the BP neural network by using an improved genetic algorithm, and obtaining a GA-BP solid oxide fuel cell voltage prediction model;
The voltage prediction module is used for acquiring the SOFC actual operation data acquired from the SOFC system, inputting the voltage influence parameters into the GA-BP solid oxide fuel cell voltage prediction model, and obtaining a voltage prediction value;
And the electric pile performance attenuation prediction module is used for obtaining the voltage change trend according to the voltage predicted value, and the electric pile performance attenuation trend is obtained under the same current condition.
For the system module disclosed in the embodiment, since the system module corresponds to the method disclosed in the embodiment, the description is simpler, and the relevant points refer to the description of the method section.
Example 3
This example further demonstrates the effectiveness of the inventive protocol using the following detailed experiment.
Experiment platform: the processor is Intel i5-8250, and the memory is 8.0GB; the system is Windows10 (64 bits); the program version is MATLAB 2018b.
The experimental contents are as follows:
In a 1kw SOFC power generation system, experimental conditions were set to bypass air flow 144.35L/min, stack cathode air flow 253.41L/min, and hydrogen flow 8.15/L/min, at every minute intervals, 5 hours of operating data were collected, and the collected variables included combustor temperature, stack temperature, current, load power, voltage, system operating time, number of system shutdowns, and downtime, with this set of experimental data as verification data.
The parameter value of the BP neural network is optimized by adopting a genetic algorithm, the population scale is set to be 20, the maximum iteration times are 100, the input layer node of the BP network is 7, the hidden layer node is 3, the output layer node is 1, and the learning rate is 0.01. FIG. 4 is a graph of fitness of a GA-BP neural network as a function of algebraic variation.
The experiment selects a classical prediction model as a control: multiple Linear Regression (MLR), radial Basis Function (RBF), BP neural network, long-term short-term memory network (LSTM) and particle swarm optimized BP neural network (PSO-BP) were compared with the predictive performance of the method of the invention (improved GA-BP). FIG. 5 is a graph of predicted versus actual values of six models versus voltage. The comparison results of the evaluation indexes of the six models are shown in table 2.
Table 2 MLR,RBF,BP,LSTM,PSO-BP, improving various evaluation indexes of GA-BP
The analysis shows that the SOFC stack performance attenuation prediction method based on GA-BP provided by the invention can obtain lower prediction error than the existing method, and improves the voltage prediction precision. The proposal provided by the invention has the lowest error in the selected experimental data, and proves that the method has good applicability. Experiments show that the method has high prediction precision, compared with MLR, RBF, BP, LSTM, PSO-BP in the traditional prediction algorithm, the improved GA-BP has the smallest average absolute value error, mean square error and root mean square error of data prediction, the largest decision coefficient R 2, and the model reduces the calculated amount and shows better prediction performance.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (5)
1. The SOFC stack performance decay prediction method based on GA-BP is characterized by comprising the following steps of:
Step 1, acquiring SOFC prototype operation data acquired from an SOFC system, and preprocessing the data to form an SOFC prototype operation data set; the SOFC prototype operation data comprise voltage and voltage influence parameters; the voltage influence parameters are system operation parameters for influencing the voltage, and comprise combustion chamber temperature, stack temperature, current, load power, system operation time, abnormal shutdown times of the system and shutdown time data;
step 2, designing a BP neural network structure according to the voltage and the voltage influence parameters of the SOFC prototype operation data set;
Step 3, taking the voltage influence parameters in the SOFC prototype operation data set as input, taking the voltage as output, training, and optimizing the initial weight and the threshold value of the BP neural network by using an improved genetic algorithm to obtain a GA-BP solid oxide fuel cell voltage prediction model;
The genetic algorithm comprises the following steps:
the traditional roulette selection operation is changed to an index selection operator;
designing a traditional fixed crossover probability as an adaptive crossover probability;
Designing the traditional fixed variation probability as the adaptive variation probability;
Specifically, the method for optimizing the initial weight and the threshold value of the BP neural network by using the improved genetic algorithm comprises the following steps:
step 3.1, generating N initial strings by adopting a random method as an initial group, wherein each initial string is called an individual;
Step 3.2, calculating the fitness value F i of each individual of the ith generation of population according to the fitness function;
Step 3.3, selecting: the index selection operator is adopted in the parent population to select the child population, and the probability of the selected individual is P i as follows:
wherein: k is an adjusting factor, N is the population size, P i is the probability that the ith individual is selected, and F i is the fitness value of the ith individual;
Step 3.4, cross operation: adopting an adaptive crossover operator of adaptive crossover probability to carry out crossover operation on two individuals selected from a population, wherein the fitness of the ith individual in the population is F i, the average fitness of the population is avg F,pc, and the crossover probability of the individual i can be expressed as:
Wherein, a is a nonlinear function self-adaptive regulating factor, and p c is 0.5;
Step 3.5, mutation operation: adopting an adaptive mutation operator with adaptive mutation probability to select individuals from a population to perform mutation operation, wherein the fitness of the ith individual in the population is F i, the average fitness of the population is avg F,pm, and the mutation probability of the individual i can be expressed as:
Wherein b is a nonlinear function adaptive adjustment factor, and p m is 0.45;
Step 3.6, judging whether the genetic algebra meets the termination condition, if so, stopping operation, and assigning the network initial weight and the threshold value corresponding to the optimal individual to the BP neural network; otherwise, returning to the step 3.2;
Step 4, acquiring SOFC actual operation data acquired from an SOFC system, and inputting voltage influence parameters into a GA-BP solid oxide fuel cell voltage prediction model to obtain a voltage prediction value;
and step 5, obtaining a voltage change trend according to the voltage predicted value, and obtaining the voltage change trend as the stack performance attenuation trend under the same current condition.
2. The method for predicting the performance degradation of an SOFC stack based on GA-BP according to claim 1, wherein in step 1, the data preprocessing process includes data normalization, and the formula is as follows:
y＝(ymax-ymin)(x-xmin)/(xmax-xmin)+ymin；
Wherein y is the normalized SOFC prototype operation data, and y max＝1,ymin =0; x is the running data of the original SOFC prototype to be normalized currently; x max and x min are the maximum value and the minimum value of the original SOFC prototype operation data to be processed in the SOFC prototype operation data set respectively.
3. The method for predicting the performance degradation of an SOFC stack based on GA-BP according to claim 1, wherein in the step 2, the BP neural network structure is designed as follows: the number of the input layer nodes is 7, the number of the hidden layer nodes is 3, and the number of the output layer nodes is 1.
4. The method for predicting the performance degradation of a SOFC stack based on GA-BP according to claim 1, wherein after the GA-BP solid oxide fuel cell voltage prediction model is obtained in step 3, a model evaluation is further performed on the GA-BP solid oxide fuel cell voltage prediction model, and the parameter indexes used for the evaluation include: mean absolute error MAE, mean square error MSE, root mean square error RMSE, decision coefficient R 2.
5. A GA-BP based SOFC stack performance degradation prediction system, comprising:
the data set acquisition module is used for acquiring the SOFC prototype operation data acquired from the SOFC system and preprocessing the data to form an SOFC prototype operation data set; the SOFC prototype operation data comprise voltage and voltage influence parameters; the voltage influence parameters are system operation parameters for influencing the voltage, and comprise combustion chamber temperature, stack temperature, current, load power, system operation time, abnormal shutdown times of the system and shutdown time data;
the neural network structure design module is used for designing the structure of the BP neural network according to the voltage and the voltage influence parameters of the SOFC prototype operation data set;
the model training and optimizing module is used for taking the voltage influence parameters in the SOFC prototype operation data set as input, taking the voltage as output, training, optimizing the initial weight and the threshold value of the BP neural network by using an improved genetic algorithm, and obtaining a GA-BP solid oxide fuel cell voltage prediction model;
The genetic algorithm comprises the following steps:
the traditional roulette selection operation is changed to an index selection operator;
designing a traditional fixed crossover probability as an adaptive crossover probability;
Designing the traditional fixed variation probability as the adaptive variation probability;
Specifically, the method for optimizing the initial weight and the threshold value of the BP neural network by using the improved genetic algorithm comprises the following steps:
step 3.1, generating N initial strings by adopting a random method as an initial group, wherein each initial string is called an individual;
Step 3.2, calculating the fitness value F i of each individual of the ith generation of population according to the fitness function;
Step 3.3, selecting: the index selection operator is adopted in the parent population to select the child population, and the probability of the selected individual is P i as follows:
wherein: k is an adjusting factor, N is the population size, P i is the probability that the ith individual is selected, and F i is the fitness value of the ith individual;
Step 3.4, cross operation: adopting an adaptive crossover operator of adaptive crossover probability to carry out crossover operation on two individuals selected from a population, wherein the fitness of the ith individual in the population is F i, the average fitness of the population is avg F,pc, and the crossover probability of the individual i can be expressed as:
Wherein, a is a nonlinear function self-adaptive regulating factor, and p c is 0.5;
Step 3.5, mutation operation: adopting an adaptive mutation operator with adaptive mutation probability to select individuals from a population to perform mutation operation, wherein the fitness of the ith individual in the population is F i, the average fitness of the population is avg F,pm, and the mutation probability of the individual i can be expressed as:
Wherein b is a nonlinear function adaptive adjustment factor, and p m is 0.45;
Step 3.6, judging whether the genetic algebra meets the termination condition, if so, stopping operation, and assigning the network initial weight and the threshold value corresponding to the optimal individual to the BP neural network; otherwise, returning to the step 3.2;
The voltage prediction module is used for acquiring the SOFC actual operation data acquired from the SOFC system, inputting the voltage influence parameters into the GA-BP solid oxide fuel cell voltage prediction model, and obtaining a voltage prediction value;
And the electric pile performance attenuation prediction module is used for obtaining the voltage change trend according to the voltage predicted value, and the electric pile performance attenuation trend is obtained under the same current condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410173816.8A CN117725446B (en) | 2024-02-07 | 2024-02-07 | SOFC (solid oxide fuel cell) stack performance decay prediction method and system based on GA-BP (GAs-liquid-phase reactor) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410173816.8A CN117725446B (en) | 2024-02-07 | 2024-02-07 | SOFC (solid oxide fuel cell) stack performance decay prediction method and system based on GA-BP (GAs-liquid-phase reactor) |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117725446A CN117725446A (en) | 2024-03-19 |
| CN117725446B true CN117725446B (en) | 2024-05-14 |
Family
ID=90209214
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410173816.8A Active CN117725446B (en) | 2024-02-07 | 2024-02-07 | SOFC (solid oxide fuel cell) stack performance decay prediction method and system based on GA-BP (GAs-liquid-phase reactor) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117725446B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201740595U (en) * | 2010-06-13 | 2011-02-09 | 宝山钢铁股份有限公司 | Comparison detecting device for hearth flame sensor |
| CN109165764A (en) * | 2018-06-26 | 2019-01-08 | 昆明理工大学 | A kind of line loss calculation method of genetic algorithm optimization BP neural network |
| CN109585884A (en) * | 2018-12-13 | 2019-04-05 | 电子科技大学 | A kind of online method for excluding fuel cell deficency and water logging |
| CN110190304A (en) * | 2019-07-23 | 2019-08-30 | 潍柴动力股份有限公司 | A kind of monitoring method and monitoring device |
| CN111190113A (en) * | 2020-04-15 | 2020-05-22 | 中国人民解放军国防科技大学 | Spacecraft storage battery performance degradation abnormity detection method |
| CN112503019A (en) * | 2020-11-30 | 2021-03-16 | 华能国际电力股份有限公司营口电厂 | Desulfurization oxidation fan control system |
| CN113064346A (en) * | 2021-02-03 | 2021-07-02 | 华能国际电力股份有限公司营口电厂 | Intelligent optimization control method for wet desulphurization oxidation fan system |
| CN114696332A (en) * | 2022-03-23 | 2022-07-01 | 无锡隆基氢能科技有限公司 | Control method, device and system of solar hydrogen production system and storage medium |
| CN116259797A (en) * | 2023-02-21 | 2023-06-13 | 华中科技大学 | Health assessment and prediction method and system for solid oxide fuel cell system |
| CN117094701A (en) * | 2023-08-18 | 2023-11-21 | 首钢股份公司迁安钢铁公司 | Equipment state management system method and system and electronic equipment |
| CN117111462A (en) * | 2023-07-28 | 2023-11-24 | 南昌大学 | Circulating control method for SOFC system load tracking |
| CN117312939A (en) * | 2023-09-05 | 2023-12-29 | 南昌大学 | SOFC system working condition identification method based on deep learning |
| CN117335425A (en) * | 2023-08-31 | 2024-01-02 | 三峡大学 | Tidal current calculation method based on GA-BP neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112149822A (en) * | 2020-07-21 | 2020-12-29 | 吉林建筑大学 | Drinking water disinfection byproduct prediction method and system |
-
2024
- 2024-02-07 CN CN202410173816.8A patent/CN117725446B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201740595U (en) * | 2010-06-13 | 2011-02-09 | 宝山钢铁股份有限公司 | Comparison detecting device for hearth flame sensor |
| CN109165764A (en) * | 2018-06-26 | 2019-01-08 | 昆明理工大学 | A kind of line loss calculation method of genetic algorithm optimization BP neural network |
| CN109585884A (en) * | 2018-12-13 | 2019-04-05 | 电子科技大学 | A kind of online method for excluding fuel cell deficency and water logging |
| CN110190304A (en) * | 2019-07-23 | 2019-08-30 | 潍柴动力股份有限公司 | A kind of monitoring method and monitoring device |
| CN111190113A (en) * | 2020-04-15 | 2020-05-22 | 中国人民解放军国防科技大学 | Spacecraft storage battery performance degradation abnormity detection method |
| CN112503019A (en) * | 2020-11-30 | 2021-03-16 | 华能国际电力股份有限公司营口电厂 | Desulfurization oxidation fan control system |
| CN113064346A (en) * | 2021-02-03 | 2021-07-02 | 华能国际电力股份有限公司营口电厂 | Intelligent optimization control method for wet desulphurization oxidation fan system |
| CN114696332A (en) * | 2022-03-23 | 2022-07-01 | 无锡隆基氢能科技有限公司 | Control method, device and system of solar hydrogen production system and storage medium |
| CN116259797A (en) * | 2023-02-21 | 2023-06-13 | 华中科技大学 | Health assessment and prediction method and system for solid oxide fuel cell system |
| CN117111462A (en) * | 2023-07-28 | 2023-11-24 | 南昌大学 | Circulating control method for SOFC system load tracking |
| CN117094701A (en) * | 2023-08-18 | 2023-11-21 | 首钢股份公司迁安钢铁公司 | Equipment state management system method and system and electronic equipment |
| CN117335425A (en) * | 2023-08-31 | 2024-01-02 | 三峡大学 | Tidal current calculation method based on GA-BP neural network |
| CN117312939A (en) * | 2023-09-05 | 2023-12-29 | 南昌大学 | SOFC system working condition identification method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| Gaopeng Han 等.Temperature Prediction Model in Transition Ladle based on Guided Mutation GA-BP Neural Network.《EITCE '21: Proceedings of the 2021 5th International Conference on Electronic Information Technology and Computer Engineering》.2021,第1303–1308页. * |
| 固体氧化物燃料电池的神经模糊控制策略研究;王倩如 等;《热能动力工程》;20221031;第198-206页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117725446A (en) | 2024-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chang | The optimal design for PEMFC modeling based on Taguchi method and genetic algorithm neural networks | |
| CN110399928A (en) | Voltage of solid oxide fuel cell prediction technique, terminal device and storage medium | |
| CN114447378B (en) | Parameter optimization method for proton exchange membrane fuel cell | |
| Zhang et al. | MOEA/D with many-stage dynamical resource allocation strategy to solution of many-objective OPF problems | |
| CN112733417A (en) | Abnormal load data detection and correction method and system based on model optimization | |
| Wang et al. | Prognostics of remaining useful life for lithium-ion batteries based on hybrid approach of linear pattern extraction and nonlinear relationship mining | |
| CN116542382A (en) | Sewage treatment dissolved oxygen concentration prediction method based on mixed optimization algorithm | |
| CN111311001B (en) | Bi-LSTM network short-term load prediction method based on DBSCAN algorithm and feature selection | |
| Wang et al. | Parameters exploration of SOFC for dynamic simulation using adaptive chaotic grey wolf optimization algorithm | |
| Li et al. | Performance degradation decomposition-ensemble prediction of PEMFC using CEEMDAN and dual data-driven model | |
| CN115796011A (en) | Hydrogen storage bed heat transfer performance optimization method based on neural network and genetic algorithm | |
| CN117725446B (en) | SOFC (solid oxide fuel cell) stack performance decay prediction method and system based on GA-BP (GAs-liquid-phase reactor) | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| CN117634737A (en) | Photovoltaic power generation prediction model construction method, photovoltaic energy supply regulation and control method and device | |
| CN116070522B (en) | Power distribution network fault positioning method based on quantum ant colony algorithm | |
| Xu et al. | A Multistrategy‐Based Multiobjective Differential Evolution for Optimal Control in Chemical Processes | |
| CN117034762A (en) | Composite model lithium battery life prediction method based on multi-algorithm weighted sum | |
| CN116227350A (en) | Multi-target optimization method and device for boiler | |
| CN115716469A (en) | Output power distribution control method of hybrid power system | |
| CN115759458A (en) | Load prediction method based on comprehensive energy data processing and multi-task deep learning | |
| CN115169428A (en) | Transformer fault diagnosis method driven by artificial intelligence | |
| Chen et al. | A coupled and interactive influence of operational parameters for optimizing power output of cleaner energy production systems under uncertain conditions | |
| Wu et al. | Data-driven approaches for predicting performance degradation of solid oxide fuel cells system considering prolonged operation and shutdown accumulation effect | |
| Hou et al. | Performance prediction of automotive fuel cell stack with genetic algorithm-BP neural network | |
| Jin et al. | Degradation prediction of PEMFC based on data-driven method with adaptive fuzzy sampling |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |