| In recent years,with the industrial transformation in the field of industrial production and the continuous emergence of cutting-edge technologies such as cloud computing,internet of things,artificial intelligence,big data,and virtual reality,digital twin has gradually become a research and application hotspot.At present,our country’s thermal power generation continues to develop towards large-capacity,ultra-supercritical units,and it has become common for these large units to participate in primary frequency regulation and automatic generation control(AGC)for deep peak regulation and frequent load changes.This makes thermal equipments subject to harsher working conditions and greatly increases the probability of fault.Using digital twin technology to realize state monitoring and fault detection of thermal system has important practical significance for improving the safety and economy of unit operation.Using the five-dimensional model of the digital twin as a framework,the paper builds a digital twin model of the high-pressure heater system aiming at fault detection by studying the functions and implementation methods of each dimension,and verifies the system in terms of high-pressure heater fault detection through experiments.The effectiveness of digital twin technology lays the foundation for the further promotion and application of digital twin technology in the field of power station thermal system performance monitoring and fault diagnosis.Firstly,the definition of the five-dimensional model of digital twin is introduced,the similarities and differences between digital twin technology and simulation technology are compared,and the five dimensions of digital twin system including specific definitions and implementation,namely physical entity,digital model,data,service system,and interactive connection.Secondly,taking the high-pressure feedwater heater as the research object,two methods for establishing the digital twin model,namely the mechanism method and the data-driven method,are studied.The former is obtained by reconstructing the existing mechanism-based dynamic real-time simulation model,only using the measurable parameters of the DCS system to drive the model,and solving the problem of parallel and synchronous operation of the model;the latter is based on the historical operation data of the unit,using the extreme random forest algorithm in machine learning.By reasonably determining the input and output parameters of the model,and using the random search algorithm to optimize the model hyperparameters,a data-driven model with higher accuracy is obtained.Using the existing simulation platform and Python environment,the above two high-pressure heater digital twin models are established respectively.The tests show that both models can meet the application requirements of high-pressure heater fault detection.Finally,on the basis of analyzing the common faults of the high-pressure heater system,the Python platform and algorithms are employed to construct the high-pressure heater condition monitoring and fault detection service system.In the service system,the9 output parameters of the model are selected for dimensionless processing by the maximum scaling method,and the reciprocal of the average absolute error of the 9outputs is used as the weight to integrate a fault detection performance index,and its threshold is determined by combining fault database with coefficient method.With the help of the full-scope simulator of a 600 MW supercritical unit,the high-pressure heater faults are simulated,and the two digital twin models are used to carry out fault detection simulation tests in both steady-state and dynamic variable-load processes.The results show that with the two digital twin models and the service system accurate detection of common faults,such as leakage of high-pressure heater pipelines,blockage of some tube bundles,and short-circuit in the inlet and outlet water chambers can be achieved,which verifies the effectiveness of the method in this paper. |
PDF Full Text Request