Research On Modeling And Model Predictive Control Of Heavy-Duty Gas Turbines

With the continuous introduction of the "two-machine" major special policy by the state,the technical research related to gas turbines has great strategic significance.The most outstanding feature of heavy-duty gas turbine is that its thermal power conversion efficiency ranks first among all power generation equipment.Nowadays,the world regards it as an important embodiment of a country’s heavy industry level.Heavy duty gas turbine is directly related to the development of national energy.However,heavy-duty gas turbines have many complex characteristics such as multivariable,strong coupling,and nonlinearity.Among them,the data of core components such as compressors and high-temperature turbines are lacking,and the dynamic characteristics are unstable with the change of operating power,thus,control design presents difficulties and obstacles.In view of this,the idea of modular mechanism modeling is adopted.According to the structural characteristics of the GE9 FA heavy-duty gas turbine unit,a suitable static and dynamic mechanism model is established and a dynamic simulation test is carried out.For the mechanism model of the heavy-duty gas turbine,the linear subspace identification method and the Bayesian identification method based on the nonlinear Wiener model are used to obtain the linear transfer function model and the Wiener model,respectively.For the Wiener model,the nonlinear predictive control system is designed;for the transfer function model,the predictive control strategy is used to design the system control,and the predictive model is an augmented state space model.Starting from the mechanism modeling research of heavy-duty gas turbine units,designing intelligent algorithms to identify and carry out predictive control provides a complete idea for the research of such engineering objects.The specific research contents are as follows:(1)The control objectives and characteristics of the operation process of the heavy-duty gas turbine after being connected to the grid are analyzed,and the control tasks and objects are clarified.The dynamic model of the operation process of the heavy-duty gas turbine after being connected to the grid is established,and the modular modeling and system construction are realized based on the Modelica language platform,among which the compressor,combustion chamber,turbine,and pipeline modules are self-built modules.Taking the fuel quantity and compressor inlet air flow as inputs and taking the turbine exhaust temperature and gas turbine load as outputs,the dynamic characteristics of the system under the step change of input variables were analyzed.The accuracy of the model is verified by using the variable load process operating data of the GE 9FA heavy-duty gas turbine in actual operation.(2)Through the analysis of the working characteristics,mechanism and thermal process characteristics of the heavy-duty gas turbine system,single-variable nonlinear and multivariable linear models are designed respectively.The system parameter identification is carried out by using the Bayesian estimation algorithm based on the Wiener model and the subspace identification method.Negative effects of nonlinear and latent variables on parameter estimates are counteracted by the Wiener model.The results show that the two identification methods can effectively reflect the operating characteristics of the heavy-duty gas turbine system and meet the needs of practical engineering applications.(3)For the nonlinear model,the nonlinear part model is approximated by the generalized regression network,the nonlinear part in the control system is extracted and transformed into a linear predictive control problem,and the linearization idea is adopted to reduce the complexity of the nonlinear optimization problem.For the multi input multi output transfer function model,a predictive controller based on the augmented state space model is designed.Aiming at the multi-input and multi-output transfer function model,the augmented state space prediction model is derived,and the predictive control strategy is designed.Compared with traditional PID control,the predictive controller has faster control speed,smaller overshoot,and better anti-interference and stability.