Performance Prediction Of Multi-Stack SOFC System Under Dynamic Conditions

Solid Oxide Fuel Cell（SOFC）is an all-solid-state chemical power generation device that directly converts chemical energy stored in fuels and oxidants into electrical energy at moderate and high temperatures in an efficient and environmentally friendly manner.In order to meet the needs of high-power scenarios such as fuel cell power station,it is urgent to develop multi stack SOFC systems.Due to dynamic conditions such as working environment,start stop cycle and load changes,performance degradation of multi stack SOFC systems will occur.In order to avoid unnecessary shutdown or safety accidents caused by performance degradation,it is necessary to predict the performance degradation trend of multi stack SOFC system under dynamic conditions.Therefore,the main research contents of this thesis are as follows:（1）By analyzing the mechanism of anodic nickel particle coarsening and nickel oxidation in SOFC system,a multi-stack SOFC system simulation model considering degradation was built,and the operation data under two dynamic conditions of slow load change and fast load change were collected through the simulation model,which was used to verify the performance prediction strategy of multi-stack SOFC system proposed later.（2）Aiming at the situation that there are multiple operating conditions in the multi stack SOFC system,and the operating conditions can be distinguished,a performance prediction strategy of the multi stack SOFC system based on multiple working states is proposed.Firstly,the correlation between the target stack and other stacks in the multi stack system is calculated by mutual information method and Pearson correlation coefficient,and the stacks with high correlation with the target stack are selected.Then,through the identification of dynamic conditions（IDC）method,the data corresponding to different working conditions are selected in turn,the performance degradation prediction model of the system under different working conditions is established,and the current working state of the test sample is selected by using GRU algorithm.Experiments on the test data sets show that when the working conditions can be distinguished,the proposed strategy can screen out the corresponding data of different working conditions in turn,and can accurately predict the performance degradation trend of SOFC.（3）A performance prediction strategy of multi-stack SOFC system based on Sa Conv LSTM model is proposed when the multi-stack SOFC system works under dynamic conditions with fast load changes.Firstly,the health indicator of each stack in the multi stack SOFC system is extracted by singular spectrum analysis method to eliminate the influence of dynamic working conditions.Then the Sa Conv LSTM model is used to predict the health indicators in space and time.By using the convolution operation,the model can solve the mutual influence between the stacks and realize the performance degradation trend prediction of the multi-stack system.Experiments on the test data sets show that the singular spectrum analysis method can extract the health indicator that can reflect the degradation process of SOFC from the voltage data,and the proposed Sa Conv LSTM model can accurately predict the performance degradation trend of each stack in the multi stack system.