融合热力学模型与人工智能的燃气轮机压气机典型故障预警方法研究  

Typical fault warning method of gas turbine compressor combining thermodynamic model with artificial neural network

作　　者：谢岳生 万震天 李俊昆 XIE Yuesheng;WAN Zhentian;LI Junkun(Shanghai Power Equipment Research Institute Co.,Ltd.,Shanghai 200240,China)

机构地区：[1]上海发电设备成套设计研究院有限责任公司,上海200240

出　　处：《热力发电》2024年第3期117-125,共9页Thermal Power Generation

基　　金：上海市青年科技启明星项目(20QB1401900)。

摘　　要：为实现压气机叶片积垢和喘振故障的提前预警,提出了一种融合热力学模型与人工智能的燃气轮机压气机典型故障预警方法。根据模块化思想搭建燃气轮机热力学性能仿真模型,并利用燃气轮机实际运行数据完成模型的动态标定,形成高精度燃气轮机性能分析模型,实现排气流量、透平进口温度、燃气轮机热耗率等关键指标的计算。在热力性能仿真模型的基础上,结合压气机典型故障专家经验及专业知识确定影响压气机故障的主要特征参数,抽象表征出压气机叶片积垢和喘振的预警模型。选取历史健康数据,采用人工神经网络算法对模型进行训练,获取偏差曲线,通过监测预警模型输出预测值与测量值之间的偏差变化,实现压气机典型故障的提前预警,给出了某GE 9F型燃气轮机压气机的实测数据的有效性验证实例。结果表明:该方法能精准捕捉压气机叶片积垢和喘振故障,相对于传统阈值报警方式,提高预警的时间窗口。该研究成果可直接在燃气轮机电厂侧部署,实时为运维人员的检修和维护决策提供指导。In order to realize compressor blade fouling and surge faults early warning,a typical fault warning method of gas turbine compressor combining thermodynamic model with artificial neural network was proposed.The simulation model of gas turbine thermodynamic performance was built according to the modularization idea,and the dynamic calibration of the model was completed by using the actual operation data of the gas turbine to form a high-precision gas turbine performance analysis model,and the key indicators such as exhaust flow rate,turbine front temperature and heat consumption can be calculated.Based on the thermal performance simulation model and combined with the compressor typical faults expert experience and professional knowledge,the main characteristic parameters affecting compressor faults were determined,and the compressor blade fouling and surge warning models were abstracted.The historical health data were selected to train the models using the artificial neural network algorithm to obtain the deviation curve,and the early warning of typical compressor faults can be realized by monitoring the deviation changes between the predicted value and the measured value of the early warning model,the example to verify the validity of the measured data of a GE 9F gas turbine compressor was given.The results showed that the method can accurately capture the compressor blade fouling and surge faults,and improve the warning time window compared with the traditional threshold alarm method.The research achievement can be directly deployed in the gas turbine power plant and provide real-time guidance for operation and maintenance personnel to make overhaul and maintenance.

关 键 词：燃气轮机 压气机 性能仿真 人工神经网络 故障预警 

分 类 号：TK478[动力工程及工程热物理—动力机械及工程] TP18[自动化与计算机技术—控制理论与控制工程]