基于3DCNN的锅炉再热器管壁减薄预测  

Prediction of wall thickness loss of boiler reheater tube based on 3DCNN

作　　者：闫佳瑛 朱希安[1] YAN Jiaying;ZHU Xi′an(School of Information and Communication Engineering,Beijing Information Science&Technology University,Beijing 100101,China)

机构地区：[1]北京信息科技大学信息与通信工程学院,北京100101

出　　处：《北京信息科技大学学报（自然科学版）》2021年第1期75-78,共4页Journal of Beijing Information Science and Technology University

基　　金：北京市科技创新服务能力建设-基本科研业务费(市级)(科研类)(PXM2019_014224_000026)。

摘　　要：为及时掌握锅炉再热器管壁减薄情况,基于锅炉再热器历史运行数据,结合三维卷积神经网络(3D convolutional neural network,3DCNN),搭建了锅炉再热器厚度损失预测模型。利用某电厂超临界660 MW机组锅炉再热器的运行数据,通过数据预处理构建三维变量作为模型的输入;通过对比不同结构模型的预测效果,确定3DCNN模型的最优结构;该方法在测试集上的最大绝对误差为12.04%,平均绝对误差为6.77%。结果表明,该方法能够比较准确地预测再热器管壁损失,方便电力企业及时掌握管壁厚度变化情况。In order to monitor the thickness loss of the boiler reheater tube wall,the predictive model of the thickness loss of the boiler reheater was built by combining 3 D convolutional neural network(3 DCNN)based on the historical operation data of the boiler reheater.Using the operation data of boiler reheater of a supercritical 660 MW unit in a power plant,3 d variables were constructed as the input of the model through data preprocessing.By comparing the prediction effects of different structural models,the optimal structure of 3 DCNN model was determined.The maximum absolute error of the method on the test set is 12.04%and the average absolute error is 6.77%.Experimental results show that the method can accurately predict the wall loss of the reheater pipe,which is convenient for power enterprises to monitor the wall loss in time.

关 键 词：三维卷积神经网络(3DCNN) 锅炉再热器 厚度损失预测 

分 类 号：TK229.6[动力工程及工程热物理—动力机械及工程]