摘要
为了准确、高效的预测包壳峰值温度,本文提出了一种卷积神经网络-长短期记忆网络的混合神经网络模型。通过混合神经网络模型,充分提取数据局部特征的同时对时间序列信息进行充分的学习,实现了包壳峰值温度的预测。数据结果表明:卷积神经网络-长短期记忆网络的混合神经网络模型单次事故分析时间降低为0.55 s的同时具备很高的准确性和稳定性。峰值预测精度、序列预测精度、超限概率预测精度、平均绝对百分比误差分别达到了99.527%,91.098%,95.371%,2.522%,均方根误差为49.065。相较于传统的BP神经网络和卷积神经网络方法,卷积神经网络-长短期记忆网络的混合神经网络模型也体现出了明显的优势。
This paper presents a hybrid neural network model of a convolutional neural network and a long short-term memory network,which can predict the peak temperature of cladding accurately and efficiently.Through the hybrid neural network model,the local features of data are fully extracted,the time-series information is fully learned,and the peak temperature of cladding is predicted.The data results show that the hybrid neural network model of the convolutional neural network and long short-term memory network has high accuracy and stability while reducing the single accident analysis time to 0.55 s.The peak prediction accuracy,sequence prediction accuracy,exceedance probability prediction accuracy,and mean absolute percentage error reached 99.527%,91.098%,95.371%,and 2.522%,respectively,and the root mean square error was 49.065.Compared with the traditional backpropagation neural network and convolutional neural network,the hybrid neural network model combining a convolutional neural network and a long short-term memory network also shows obvious advantages.
作者
孙大彬
李磊
田兆斐
王贺
SUN Dabin;LI Lei;TIAN Zhaofei;WANG He(Fundamental Science on Nuclear Safety and Simulation Technology Laboratory,Harbin Engineering University,Harbin 150001,China)
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2022年第12期1728-1735,共8页
Journal of Harbin Engineering University
基金
国家重点研发计划(2018YFB1900302).
关键词
包壳峰值温度
卷积神经网络
长短期记忆网络
混合神经网络
峰值预测精度
序列预测精度
超限概率预测精度
平均绝对百分比误差
peak cladding temperature
convolutional neural network
long short-term memory network
hybrid neural network
peak prediction accuracy
sequence prediction accuracy
exceedance probability prediction accuracy
mean absolute percentage error
