Skip to main content
Open Access Publications from the University of California

Diagnostics and Control of Gas Turbines Through System Identification

  • Author(s): Holcomb, Chad M.
  • et al.

In this dissertation, system identification methods are applied to characterize the dynamic system behavior of gas turbines for use in diagnostics, condition monitoring, and control design. The application examples are motivated by real-world engineering challenges with industrial gas turbines. Firstly, a method is proposed to represent the nonlinear interconnection of the fuel control system and gas turbine using a block Hammerstein model. A new iterative gradient descent prediction error minimization method is proposed to identify a low order dynamic model of the gas turbine and a parametrized static nonlinear mapping. The identified nonlinear mapping is demonstrated to capture fuel valve contamination. Secondly, a modification to existing subspace system identification methods is formulated to permit their application to multiple concatenated, but non-contiguous data sets, which jointly can be sufficiently rich for identification while individual records are not. This is a break from traditional system identification, where one uses a single contiguous synchronous input-output record. In addition, a set of simple sufficient conditions for identifiability are derived. In tandem, these developments increase the intrinsic value of archival data as it enables the construction of a sufficiently informative data set for dynamic modeling. These conditions represent the primary theoretical contribution of this dissertation. The identified models are suitable for use in condition monitoring methods that are based on characterizations of transient behavior. Finally, subspace system identification is applied for identification of control- relevant MIMO generalized plant models of the closed-loop gas turbine system in feedback control at a series of discrete operating points. An operating point dependent scaling method is proposed to enable a single nominal linear plant model to capture the dynamics of the closed- loop system. When combined, the procedure enables design of an outer loop controller to improve the disturbance rejection performance of the system. The method is demonstrated to significantly improve the transient load rejection of fuel and airflow control of an existing low emissions industrial gas turbine

Main Content
Current View