摘要
为避免失修和过修,提高燃气轮机可靠性和可用性,本文提出了一种基于模型与数据混合驱动的燃气轮机气路故障诊断方法。首先,通过自适应热力建模策略构建待诊断对象的燃气轮机热力模型,通过设置不同的部件健康参数和入口边界条件,模拟得到部件健康参数向量与入口边界条件参数和气路可测参数向量一一对应的数据集;其次,利用深度学习进行回归建模,训练得到燃气轮机气路故障诊断模型;最后,根据实际燃气轮机入口边界条件参数和气路可测参数向量,通过已训练的诊断模型来实时诊断输出各气路部件的健康参数向量。仿真实验结果表明,通过本文所提出的方法,可以准确地得到各个通流部件量化的健康参数,其总体均方根误差不超过0.033%,最大相对误差不超过0.36%,表明该方法具有较大的应用潜力。
To avoid disrepair and over repair,and improve the reliability and availability of gas turbines,this paper proposes a gas turbine gas circuit fault diagnosis method based on model and data hybrid drive.Firstly,the gas turbine thermodynamic model of the object to be diagnosed is constructed based on adaptive thermal modeling strategy.By setting different component health parameters and inlet boundary conditions,a data set of one-to-one correspondence between component health parameter vectors and inlet boundary condition parameters and gas path measurable parameter vectors is obtained through simulation.Secondly,deep learning is used to carry out regression modeling,and train a gas turbine gas path fault diagnosis model.Finally,according to the actual gas turbine inlet boundary condition parameters and gas path measurable parameter vectors,the trained diagnostic model is used to diagnose and output the health parameter vectors of each gas path component in real time.The simulation experiment results show that,accurate and quantified health parameters of each gas path component can be obtained by the proposed method,the overall root mean square error does not exceed 0.033%,and the maximum relative error does not exceed 0.36%,indicating the proposed method has great application potential.
作者
靳尧飞
应雨龙
李靖超
周宏宇
JIN Yaofei;YING Yulong;LI Jingchao;ZHOU Hongyu(School of Energy and Mechanical Engineering,Shanghai University of Electric Power,Shanghai 200090,China;School of Electronic and Information Engineering,Shanghai Dianji University,Shanghai 201306,China)
出处
《热力发电》
CAS
CSCD
北大核心
2021年第9期66-71,93,共7页
Thermal Power Generation
基金
国家自然科学基金项目(51806135)。
关键词
燃气轮机
热力学模型
部件健康参数
气路诊断
深度学习
gas turbine
thermodynamic model
component health parameter
gas path diagnosis
deep learning
