文献类型：期刊文献

英文题名：Prediction of three-dimensional temperature and current density field distribution in solid oxide electrolysis cells by modified transformer

作者：Wang, Fangzhou[1];Yang, Fan[2];Wu, Lily[1];Wang, Hao[1];Cao, Jun[1]

机构：[1]East China Univ Sci & Technol, 130 Meilong Rd, Shanghai 200237, Peoples R China;[2]Chengdu Univ Informat Technol, 24 Block 1,Xuefu Rd, Chengdu 610225, Peoples R China

年份：2025

卷号：27

外文期刊名：ENERGY CONVERSION AND MANAGEMENT-X

收录：EI(收录号：20253218948328);WOS:【ESCI(收录号:WOS:001547674300001)】；

基金：This work was supported by key project of National Natural Science Foundation of China under Grant No. 22393954.

语种：英文

外文关键词：SOEC; CFD; Transformer; Physical field distribution prediction

摘要：The temperature distribution spatial contour of solid oxide electrolysis cells (SOECs) is a critical indicator for evaluating device performance, energy consumption, and maintenance safety. Conventional computational fluid dynamics (CFD) simulations, while accurate, suffer from excessive computational time and poor timeliness, making them unsuitable for real-time field monitoring. To address this, this paper proposes an improved Transformer neural network model driven by CFD data for real-time prediction of SOEC internal temperature field spatial contours. The model integrates a hybrid architecture of CNN and Transformer, where a multicoupled CNN extracts operational parameter features, and a positional encoding Transformer reconstructs the spatial distribution of physical fields. Experimental results show that the model completes temperature field reconstruction within seconds, reducing computational resource consumption by over 90 % compared to traditional CFD methods. In terms of accuracy, the mean absolute error (MAE) of temperature field prediction is controlled below 2 K, and the current density field prediction accuracy exceeds 95 %. This approach breaks through the timeliness bottleneck of conventional simulations, provides an efficient and reliable technical support for intelligent operation and maintenance of SOEC digital twins and demonstrates significant engineering value for enhancing the operational stability and intelligent management of SOEC systems.