摘要
电站锅炉主蒸汽管道长期服役在高温高压环境下,面临严重的蠕变损伤问题。为预防主蒸汽管道发生蠕变失效,提出了电站锅炉主蒸汽管道蠕变损伤预测方法。首先,通过文献研究发现温度和压力是引起主蒸汽管道蠕变的主要影响因子。然后,采用多种循环神经网络算法构建了主蒸汽管道运行温度和压力预测模型,对比发现Informer算法构建的预测模型性能最佳,且通过数据治理有效提升了模型性能。最后,结合L-M参数法实现了主蒸汽管道蠕变损伤预测。
The main steam piping of power station boilers,which have been operating under high-temperature and highpressure conditions for extended periods,are confronted with severe creep damage problems.To prevent creep failure in these main steam piping,a creep damage prediction method for them has been proposed.Initially,through a literature review,it was revealed that temperature and pressure are identified as the primary influencing factors contributing to creep in main steam piping.Subsequently,multiple recurrent neural network algorithms were employed to construct predictive models for the operating temperature and pressure of main steam piping.Among these,the predictive model developed using the Informer algorithm demonstrated superior performance,and its efficacy was further enhanced through data governance measures.Finally,creep damage prediction for main steam piping was realized by integrating the L-M parameter method.
作者
郝维勋
王硕
张述敏
杨旭
HAO Weixun;WANG Shuo;ZHANG Shumin;YANG Xu(State Key Laboratory of Low-Carbon Thermal Power Generation Technology and Equipments(Harbin Boiler Company Limited),Harbin,Heilongjiang 150046,China;China Special Equipment Inspection&Research Institute,Beijing 100029,China)
出处
《自动化应用》
2025年第16期47-50,56,共5页
Automation Application
基金
国家重点研发计划(2023YFB4102304)。
关键词
电站锅炉
蠕变预测
循环神经网络
power station boiler
creep prediction
recurrent neural network
作者简介
郝维勋,男,1985年生,硕士研究生,高级工程师,从事电站锅炉用材料及相关检测方法的研究工作。
