Research On Data Reconciliation Methods For Heavy Duty Gas Turbines

Data quality is crucial for gas turbine health management and fault diagnosis.Sensor state drift and degradation can lead to a decrease in the quality of measurement data,affecting the reliability of fault diagnosis results and control stability.In order to eliminate the impact of data quality degradation,heavy gas turbines were studied as the research object,and data coordination technology was studied.Data quality inspection and hybrid fault diagnosis algorithms for heavy gas turbines were designed,and simulation experiments were completed to verify the relevant algorithms.The main research contents are as follows:(1)Research on Static Simulation Modeling of Gas Turbines: For the MS9001-FA heavyduty gas turbine,considering the thermophysical properties of the working fluid,a compressor module,a combustion chamber module,and a turbine module are established,and the interior point method is used to optimize the model input parameters;Verified the accuracy of the model and achieved simulation of the steady-state operation of the gas turbine.Compared with the design point operating conditions of the gas turbine,the average error of performance parameters such as compressor pressure ratio and turbine inlet/outlet temperature does not exceed 3%.(2)Gas turbine dynamic simulation modeling: Based on the thermal and physical processes of gas turbine operation,as well as the characteristics of main components,considering cooling extraction and inertia factors,a dynamic model of gas turbine was established through modular modeling method;Analyze the main control system of the GE 9FA gas turbine and establish a control strategy module to simulate the fuel flow,speed/power,and exhaust temperature of the dynamic model of the gas turbine.Compare the simulation results of the model with the measured values of the actual operating conditions of the gas turbine.The maximum error under high and medium operating conditions shall not exceed 3%,and the maximum error under low operating conditions shall not exceed 5%.(3)Research on data quality based on steady-state data coordination: Addressing the issues of low quality and high uncertainty of measured data in offline data analysis of gas turbines.Introducing a steady-state coordination method based on Gaussian correction criterion,data coordination was carried out for typical operating conditions,achieving data quality inspection and improvement of steady-state operation data;The results indicate that the uncertainty of the coordinated value is reduced by 10-15%compared to the measured value,meeting the data quality inspection standards.(4)Research on gas turbine sensor fault diagnosis based on data coordination,establishing a data coordination model based on PCA-LSTM multi model fusion,achieving the diagnosis of single sensor soft/hard faults and the reconstruction of fault data.The reconstructed signal is used to replace sensor fault data,providing accurate data for the gas pipeline fault diagnosis system to achieve gas turbine health management.(5)Research on combination faults diagnosis of gas turbines based on deep learning.Based on Convolutional Neural Networks Gate Recurrent Unit(CNN-GRU),a mixed fault diagnosis model is established to address the problem of gas turbine gas path fault diagnosis.By combining data coordination methods,a combination faults diagnosis model is established for the simultaneous occurrence of sensor and gas path faults,Verified the feasibility of combination faults diagnosis and the improvement effect of data coordination sensor fault diagnosis algorithm on the accuracy of gas path fault diagnosis.