A method to identify the parametric dynamic model of a single-shaft gas turbine engine.

摘要

This dissertation presents the development of a method to identify the dynamic model of a single-shaft industrial gas turbine engine operating in closed loop control. Besides providing a safe and robust identification procedure for a fully operational turbine system, the method alleviates the difficult issues of data identifiability and model accuracy that exist when estimating the open-loop dynamics from closed-loop experimental data.; A majority of previous work on gas turbine dynamics identification focuses primarily on open-loop estimation where the control feedback is disconnected. All of the work assumes a certain amount of knowledge about the turbine dynamics in the formulation of the excitation signal and the turbine model structure. The method presented here does not assume a priori knowledge about the turbine design, but relies on data from a step response to estimate the turbine characteristics. The on-line step response provides the necessary system parameters to design the appropriate excitation signal and to aid in formulating the model structure.; The ultimate objective is to arrive at an accurate turbine model that is easy to use for control synthesis and at the same time does not require excessive computational effort to obtain. It will be shown that this objective is well met using the linear discrete models, ARX and ARMAX, in conjunction with the multi-sine excitation signal and the prediction error estimation algorithm. The multi-sine excitation signal is selected based on the fact that it can reduce the risk of upsetting the turbine operation, and at the same time, contribute to maximizing the accuracy of identification test results. The prediction error algorithm is chosen for its computational robustness and applicability to the problem of estimating the open-loop dynamic parameters using closed-loop data.; The contribution of this research is a robust method that can be applied effectively to characterize the dynamics of the single-shaft gas turbine engine operating in a full-production mode. With minor improvements, the method can be utilized for identification of other turbine dynamic processes as well as other physical systems operating in closed loop control.