基于ConvLSTM的核电温度传感器数据重构研究OA

The Convolutional Long Short-Term Memory(ConvLSTM)network effectively integrates the spatial feature extraction capability of convolutional layers with the temporal dependency modeling advantages of Long Short-Term Memory(LSTM)networks.Consequently,it has been increasingly utilized for predicting critical parameters of systems or equipment.Data reconstruction refers to the process of reconstructing faulty sensor data using available operational data.The application of the ConvLSTM model for data reconstruction of nuclear power plant process sensors enables operators to obtain timely and accurate operational status information of systems or equipment,thereby preventing misoperation caused by erroneous measurements from failed sensors and ensuring the proper functioning of systems or equipment.In this paper,the Fuqing nuclear power plant simulator dataset was utilized as experimental data,and a sensor fault model was developed to simulate the reconstruction process and results under actual sensor failure scenarios.Comparative analysis of CNN,LSTM,and CNN-LSTM models demonstrated that the ConvLSTM model exhibits superior reconstruction accuracy and applicability.