Numerical Simulation Of Flow Field Structure Influence Of Solid Oxide Fuel Cell And Performance Prediction Based On Machine Learning

At present,the global energy situation is becoming more and more serious,and the demand to promote energy transformation and upgrading,find clean energy and renewable energy with high energy efficiency conversion rate is becoming more and more urgent.Solid Oxide Fuel Cell(SOFC)is a kind of distributed energy generation device that converts chemical energy into electrical energy in medium and high temperature environment,with high energy conversion efficiency,high flexibility in fuel selection,no need for precious metals as catalysts and complete solid electrolyte advantages,has broad development potential in distributed energy,is currently a research hotspot at home and abroad.Due to the tightness of SOFC,the experimental study of SOFC measured by working parameters,high experimental cost and long cycle,this paper first establishes a rectangular flow channel,X-shaped flow channel SOFC and SOFC multi-physics coupling model of two electrolytes based on COMSOL finite element,and conducts parameter sensitivity analysis of SOFC based on Sobol sensitivity analysis method,and secondly based on Latin Hypercube Sampling(LHS)The method obtains 500 sets of input parameters,and uses the combination of MATLAB and COMSOL to prepare the dataset,and finally builds three algorithm models based on machine learning Support Vector Machine(SVM),Multilayer Perceptron(MLP),and Multilayer Perceptron and Random Forest(MLP-RF)to predict the power and maximum stress of SOFC and compared the prediction results of the three algorithmic models.This paper finds that the electrochemical reaction rate of the X-shaped flow channel SOFC is faster under the conditions of 0.3V and 0.7V,but at 0.3V,the stress concentration of the two is distributed at the cathode and anode inlet,and the maximum temperature and maximum stress of the X-shaped flow channel are higher.This indicates that the X-shaped flow channel SOFC has higher material requirements.Compared with the rectangular flow channel,the reverse flow mode of the X-shaped flow channel has a better effect on the improvement of temperature distribution.Second,SOFC-H cathode produces water vapor,which aggravates the polarization of the cathode concentration,the electrochemical reaction rate is limited,and the X-shaped flow channel has a small cathode concentration polarization compared with the rectangular flow channel.Third,the five parameters of applied voltage,fuel inlet temperature,fuel inlet hydrogen concentration,air inlet temperature,and air inlet oxygen concentration are high for SOFC output power and maximum stress sensitivity,and fuel inlet flow and air inlet flow sensitivity are very small.Fourth,the MLP-RF model has the best prediction effect on the data set of this model,the prediction errors MAE and MSE of power are 0.003 and 0.0042,respectively,the prediction errors of maximum stress MAE and MSE are 0.00088 and 0.0012,respectively,SVM prediction results are the worst.