随着新能源大规模接入电网,发电型燃气轮机常需频繁切换工作状态,导致故障风险上升,因此,异常检测对燃气轮机安全运行更加重要。针对燃气轮机异常检测问题,提出了一种基于NARX-Catboost算法的基线建模方法。采用NARX建立燃气轮机声压特...随着新能源大规模接入电网,发电型燃气轮机常需频繁切换工作状态,导致故障风险上升,因此,异常检测对燃气轮机安全运行更加重要。针对燃气轮机异常检测问题,提出了一种基于NARX-Catboost算法的基线建模方法。采用NARX建立燃气轮机声压特征信号的基线模型,引入CatBoost算法以增强NARX拟合能力,并运用贝叶斯优化对模型超参数进行寻优,最终通过实验数据验证了该融合方法在异常检测方面的有效性。另外,将所提NARX-Catboost与基于向前回归正交最小二乘法的NARX模型(NARX-FROLS)和集成深度随机向量函数链接网络(Ensemble Deep Random Vector Functional Link network, edRVFL)方法及性能进行对比。结果表明:NARX-CatBoost方法对正常声压均方根值的拟合均方根误差(RMSE)值为0.008 50,拟合准确度明显优于NARX-FROLS与edRVFL方法;NARX-CatBoost方法对异常声压均方根的异常检测准确率为96.94%,表明通过正常声压特征数据建立基线模型进行异常检测的可行性与准确性。展开更多
The critical components of gas turbines suffer from prolonged exposure to factors such as thermal oxidation,mechanical wear,and airflow disturbances during prolonged operation.These conditions can lead to a series of ...The critical components of gas turbines suffer from prolonged exposure to factors such as thermal oxidation,mechanical wear,and airflow disturbances during prolonged operation.These conditions can lead to a series of issues,including mechanical faults,air path malfunctions,and combustion irregularities.Traditional modelbased approaches face inherent limitations due to their inability to handle nonlinear problems,natural factors,measurement uncertainties,fault coupling,and implementation challenges.The development of artificial intelligence algorithms has provided an effective solution to these issues,sparking extensive research into data-driven fault diagnosis methodologies.The review mechanism involved searching IEEE Xplore,ScienceDirect,and Web of Science for peerreviewed articles published between 2019 and 2025,focusing on multi-fault diagnosis techniques.A total of 220 papers were identified,with 123 meeting the inclusion criteria.This paper provides a comprehensive review of diagnostic methodologies,detailing their operational principles and distinctive features.It analyzes current research hotspots and challenges while forecasting future trends.The study systematically evaluates the strengths and limitations of various fault diagnosis techniques,revealing their practical applicability and constraints through comparative analysis.Furthermore,this paper looks forward to the future development direction of this field and provides a valuable reference for the optimization and development of gas turbine fault diagnosis technology in the future.展开更多
文摘随着新能源大规模接入电网,发电型燃气轮机常需频繁切换工作状态,导致故障风险上升,因此,异常检测对燃气轮机安全运行更加重要。针对燃气轮机异常检测问题,提出了一种基于NARX-Catboost算法的基线建模方法。采用NARX建立燃气轮机声压特征信号的基线模型,引入CatBoost算法以增强NARX拟合能力,并运用贝叶斯优化对模型超参数进行寻优,最终通过实验数据验证了该融合方法在异常检测方面的有效性。另外,将所提NARX-Catboost与基于向前回归正交最小二乘法的NARX模型(NARX-FROLS)和集成深度随机向量函数链接网络(Ensemble Deep Random Vector Functional Link network, edRVFL)方法及性能进行对比。结果表明:NARX-CatBoost方法对正常声压均方根值的拟合均方根误差(RMSE)值为0.008 50,拟合准确度明显优于NARX-FROLS与edRVFL方法;NARX-CatBoost方法对异常声压均方根的异常检测准确率为96.94%,表明通过正常声压特征数据建立基线模型进行异常检测的可行性与准确性。
基金funded by the Science and Technology Vice President Project in Changping District,Beijing(Project Name:Research on multi-scale optimization and intelligent control technology of integrated energy systemProject number:202302007013).
文摘The critical components of gas turbines suffer from prolonged exposure to factors such as thermal oxidation,mechanical wear,and airflow disturbances during prolonged operation.These conditions can lead to a series of issues,including mechanical faults,air path malfunctions,and combustion irregularities.Traditional modelbased approaches face inherent limitations due to their inability to handle nonlinear problems,natural factors,measurement uncertainties,fault coupling,and implementation challenges.The development of artificial intelligence algorithms has provided an effective solution to these issues,sparking extensive research into data-driven fault diagnosis methodologies.The review mechanism involved searching IEEE Xplore,ScienceDirect,and Web of Science for peerreviewed articles published between 2019 and 2025,focusing on multi-fault diagnosis techniques.A total of 220 papers were identified,with 123 meeting the inclusion criteria.This paper provides a comprehensive review of diagnostic methodologies,detailing their operational principles and distinctive features.It analyzes current research hotspots and challenges while forecasting future trends.The study systematically evaluates the strengths and limitations of various fault diagnosis techniques,revealing their practical applicability and constraints through comparative analysis.Furthermore,this paper looks forward to the future development direction of this field and provides a valuable reference for the optimization and development of gas turbine fault diagnosis technology in the future.
