Open AccessApplication of PSO-BP neural network in methane chemical looping reforming reaction
Author(s) -
Yongbin Liu,
Xiaoteng Ma,
Long Xu
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1617/1/012079
Subject(s) - particle swarm optimization , artificial neural network , convergence (economics) , methane , experimental data , biological system , computer science , algorithm , chemistry , mathematics , artificial intelligence , statistics , organic chemistry , economics , biology , economic growth
In order to overcome the disadvantage of BP neural network easy falling into local convergence, a BP neural network model with particle swarm optimization (PSO) is established to simulate and predict chemical looping reforming (CLR) of methane. The prediction results of PSO-BP model are analyzed, the results showed that the fitting degree of PSO-BP model training set and test set are both above 0.979, and the error of the network is finally stable at 0.013509. It is proved that this model is a good prediction model. The experimental results under a certain reaction condition are predicted by the PSO-BP model, and the relative errors between the predicted date and the actual data are 1.581% at the maximum and 0.118% at the minimum. Then the PSO-BP model is used to simulate and screen all the experimental data, and the optimal experimental conditions are obtained as follows: the ceria-iron composite oxygen carrier with a molar ratio of 7/3 prepared by co-precipitation method, calcinated at 800 °C for 6h, and the proper reaction temperature is 850 °C. Finally, the predicted reaction performance under optimized operating conditions were consistent with the actual experimental.