Research On The Operation Fault Early Warning Of Steam Turbine Under Peak Regulation And Frequency Modulation Working Condition

With the global issues such as climate change and fossil energy crisis becoming more and more prominent,China has put forward the national strategy of "carbon peaking and carbon neutralization ",and promoting clean energy development and energy consumption is one of the important measures to achieve this goal.In order to improve the wind power and solar energy consumption capacity of the power grid,the role of thermal power units will change from the original basic load bearer to the flexible regulation capacity provider.The participation of thermal power units in peak regulation and frequency modulation will lead to changes in the operating state of steam turbines.Deep,rapid and frequent load changes are bound to have an impact on the safety and economy operation of equipment.The research on the operation fault early warning of steam turbine is of great significance to avoid serious equipment failure,improve the operation level,realize energy conservation and emission reduction,and promote the transformation of thermal power units to regulated power supply.Therefore,this paper summarizes and analyzes the possible faults of steam turbine under the deep peak regulation and fast frequency modulation working conditions.On this basis,considering the particularity of different sections and the interference factors to realize fault early warning,this paper carries out the research on fault early warning of key equipment in steam turbine.Aiming at the complex fault performance caused by frequent,fast and largescale regulation,based on the fault mechanism and external performance of the valve in the steam inlet section,the linearity and dispersion of the comprehensive flow characteristic and the dispersion of the valve opening characteristic are determined as the fault characterization parameters,and the online fault early warning method of the governing valve is proposed.Experiments with different fault type data verify that the proposed method can accurately judge and locate the software/hardware fault.Then,considering the thermal fatigue damage of rotor caused by large-scale regulation,based on the rotor thermal stress estimation and life damage monitoring algorithm verified by actual data,the life damage characteristics under different steam temperature variation amplitude and rate are analyzed,and it is pointed out that the steam temperature variation amplitude is the main factor affecting the rotor life,and the influence of temperature variation rate will not change significantly after reaching a threshold.Considering the influence of working condition variation on performance monitoring,the concept of normal pattern which does not change with working condition and reflects equipment performance is defined.Based on this,a fault early warning method of feedwater heater based on normal pattern model is proposed.Through the analysis of data characteristics,the influence of commonmode information on pattern recognition and normal pattern characterization is revealed.The modeling idea of using sparse autoencoder to eliminate the interference of common-mode information is proposed,then the effectiveness of the proposed method is verified by actual data.Then,from the perspective of mechanism,the feature selection results of the comparison method and the reasons for the difference of fault early warning performance are analyzed.The influence of common-mode interference on pattern recognition and the importance of feature engineering for fault early warning of feedwater heater are verified.Considering the complex coupling dynamic characteristics in the relevant parameters,an early warning method which considering dynamic information for performance degradation in intermediate stage flow path is proposed.Through the fusion of data and mechanism analysis,considering the expression forms of static and dynamic information,the feature combination is selected,and then the longshort term memory network which can adaptively consider the time-series information is used for normal pattern modeling.Through the comparative experiments with the static modeling method and different input feature combinations,the importance of considering the time-series information for the fault early warning of intermediate stage flow path is verified,and the essential reason of this phenomenon is explained from the perspective of working machanism.Aiming at the problem that the last stage blade flutters under small volume flow and related parameters cannot be measured directly,a soft measurement method of last stage blade flutter is proposed.Considering the relationship between the last stage blade flutter and volume flow,combined with the working principle of the steam turbine last stage,the thermodynamic simulation model is established.Through the analysis of parameter sensitivity,this paper explores the essential factors affecting the working state of the last stage blade,and then puts forward the soft measurement method of volume flow and flutter of the last stage blade after considering the coupling influence of multiple factors,so as to realize the on-line continuous monitoring for dynamic stress and soft measurement of last stage blade flutter.The simulation data are used to compare the proposed method with the traditional estimation method,which verifies the superiority of the proposed method for multi factor comprehensive consideration.Based on the above researches,this paper makes an integrated verification on the fault early warning effect of key components in steam turbine.From the point of view of meeting the actual needs of power plants,integrating the fault early warning methods of steam turbine studied in this study,the operation faults early warning platform of steam turbine is designed and developed,and the effectiveness of the platform is verified by using the actual fault data and simulation data.The results show that the relevant research results carried out in this paper can realize the fault early warning for key components in steam turbine under deep peak regulation and fast frequency modulation working condition.