Design, modelling and control of a gas turbine air compressor

© 2014 Dr. Ashley P. WieseThe production of compressed air constitutes a considerable portion of industrial electrical consumption. An alternative to electrically driven air compression systems is a gas turbine cycle converted to the production of compressed air. Such devices possess characteristics that may be attractive relative to electrically driven air compressors. Features such as oil-free operation can be incorporated, while high power density and low vibration are intrinsic qualities of gas turbines. Further advantage is gained by potentially reducing greenhouse gas emissions and reducing load on electricity distribution infrastructure. This thesis presents a study of the design, modelling and closed-loop control of a prototype gas turbine air compressor (GTAC). In addition to the design and construction of the prototype, the key research objectives for this investigation are the development of steady-state and dynamic models, the development of a closed-loop control system, and the experimental evaluation of the prototype's performance at both steady-state, and in tracking strongly transient compressed air demand. In meeting these objectives, several contributions are made to the fields of gas turbine modelling, system identification and control. First, the steady-state performance of the GTAC and its constituent components are determined experimentally. To identify component characteristics that are difficult or impossible to measure in an operating gas turbine, an `inverse-cycle' model is developed. This model allows simultaneous inference of combustor efficiency, combustor pressure loss, hot-end heat transfer, turbine inlet temperature and turbine performance. This inverse-cycle technique is applied to the GTAC prototype, and the resulting performance characteristics are utilised in the next stage of modelling. To facilitate model-based control, a physics-based, transient model of the GTAC is first presented and validated experimentally. This model is an extension to an existing non-dimensional framework that includes a more comprehensive set of physical phenomena relevant to gas turbines. The extended framework includes new models of several gas turbine components, in particular the turbine and compressor, and also applies a well known method for prescribing boundary conditions to the gas path. This extended framework incorporates the previously determined steady-state characteristics to implement the source terms for the model's conservation equations. A least-squares optimisation is undertaken to estimate the shaft inertia and the significant thermal inertias. The complete gas turbine model is then validated against additional transient experimental data, and the overall system dynamics appear to be captured well. This appears to be the first experimental validation of this framework for a complete gas turbine. Control of the GTAC prototype is complicated by the nonlinear, MIMO nature of gas turbines, and their associated operating constraints related to compressor stability and turbine inlet temperature. Nonlinear model predictive control (NMPC) is identified as a promising approach to gas turbine control, particularly due to its inherent handling of operating constraints. While some authors have previously proposed MPC approaches for gas turbines, computational complexity has necessitated some degree of linearisation. This thesis presents an NMPC for the GTAC prototype, without the linearisation characteristic of previous approaches. This is achieved through the development of a nonlinear, reduced order modelling approach for gas turbines, a formal nonlinear stability analysis, and extensions to an existing solution algorithm for nonlinear optimisation problems. While the specific results of the reduced order modelling, controller development and stability analysis are all demonstrated for the GTAC prototype, the approach is equally applicable to gas turbines in general. The thesis concludes with an experimental demonstration of the NMPC on the GTAC prototype, and an assessment of its closed-loop performance