Time-series prediction of electromotive force for dissolved oxygen in lead-bismuth eutectic based on hybrid neural network

The concentration of dissolved oxygen is pivotal to the corrosion resistance of structural materials in lead-cooled fast reactor (LFR), which in turn has a substantial influence on operational safety. This study focuses on achieving precise prediction of oxygen variation trends in the liquid lead-bismuth eutectic (LBE) represented by electromotive force, with temperature serving as an assisting parameter. A novel forecasting approach, named Feature-Enhanced Variant Attention Network (FEVA-Net), is proposed. This approach is based on a hybrid neural network of CNN-BiLSTM incorporated with a variant attention mechanism. In this paper, experimental data were acquired from the LBE loop facility, KYLIN-II-M. A comprehensive verification process, involving multi-dimensional comparative experiments, was carried out to demonstrate the validity of the proposed method. The results reveal the superior performance of FEVA-Net over the EMD-CNNBiLSTM model across several key metrics. Specifically, improvements in RMSE, MAE, MAPE, and R 2 are observed, with average enhancements of 31.40 %, 43.58 %, 2.05 %, and 4.17 %, respectively. This study illustrates the promising accuracy of the proposed model, especially for dissolved oxygen process in practice.