电子器件2025，Vol.48Issue(3)：531-540,10.DOI:10.3969/j.issn.1005-9490.2025.03.010
基于Autoencoder的火电锅炉管道位移在线监测与预警系统
Online Monitoring and Warning System of Pipeline Displacement in Thermal Power Plant Based on Autoencoder
摘要
Abstract
Thermal boiler pipeline works under the condition of changing temperature and pressure for a long time.At present,most power plants use mechanical pointer displacement indicator to monitor pipeline displacement,which has low efficiency,high lag,and can not ac-curately reflect the health state of the pipeline.An autoencoder convolutional displacement neural network(ACDNN)is designed to monitor the three-dimensional thermal displacement of boiler pipes in real time.ACDNN uses CAESAR Ⅱ to calculate the three-dimensional ther-mal displacement of each node of the pipeline under different temperature and pressure,the corresponding dataset is built,which is used as the dataset for training,and the trained model is applied to the online monitoring of the displacement of the boiler pipeline in the power plant.At the same time,a real-time stereo displacement measurement device(SDMD)based on stereo vision theory and artificial neural network is developed.The ACDNN is verified by real-time pipeline 3D thermal displacement measured by SDMD.The field application re-sults show that compared with the traditional mechanical pointer,the maximum absolute error of SDMD is 2 mm.The simulation results of CAESAR Ⅱ have good accuracy,compared with SDMD,the maximum relative error is 9.6％,and the maximum error is 3.5％during stable operation.The residual connection of ACDNN can effectively reduce the prediction error.Compared with the simulation results,the relative error of the predicted value of ACDNN is less than 4％,the absolute error is less than 2mm,and it is stable in a long time.Finally,the sys-tem proposed is used to judge the state of the constant force hanger in the power plant,it is found that the spring travel of the hangers at the displacement value close to the design displacement value is also close to the maximum threshold of the safety travel,and the safety load verification is needed,which proves that the system has good applicability in thermal power plants.关键词
管道位移/火力发电厂/立体位移测量/残差连接/深度学习Key words
pipeline displacement/thermal power plant/stereoscopic displacement measurement/residual connection/deep learning分类
能源科技引用本文复制引用
简彦辰,戴明露,周宾..基于Autoencoder的火电锅炉管道位移在线监测与预警系统[J].电子器件,2025,48(3):531-540,10.基金项目
江苏省碳达峰碳中和科技创新专项资金项目(BT2024013) （BT2024013）
