Implementation of self-tuning control for turbine generators
This thesis documents the work that has been done towards the development of a 'practical' self-tuning controller for turbine generator plant. It has been shown by simulation studies and practical investigations using a micro-alternator system that a significant enhancement in the overall performance in terms of control and stability can be achieved by improving the primary controls of a turbine generator using self-tuning control. The self-tuning AVR is based on the Generalised Predictive Control strategy. The design of the controller has been done using standard off-the-shelf microprocessor hardware and structured software design techniques. The proposed design is thus flexible, cost-effective, and readily applicable to 'real' generating plant. Several practical issues have been tackled during the design of the self-tuning controller and techniques to improve the robustness of the measurement system, controller, and parameter estimator have been proposed and evaluated. A simple and robust measurement system for plant variables based on software techniques has been developed and its suitability for use in the self-tuning controller has been practically verified. The convergence, adaptability, and robustness aspects of the parameter estimator have been evaluated and shown to be suitable for long-term operation in 'real' self-tuning controllers. The self-tuning AVR has been extensively evaluated under normal and fault conditions of the turbine generator. It has been shown that this new controller is superior in performance when compared with a conventional lag-lead type of fixed-parameter digital AVR. The use of electrical power as a supplementary feedback signal in the new AVR is shown to further improve the dynamic stability of the system. The self-tuning AVR has been extended to a multivariable integrated self-tuning controller which combines the AVR and EHG functions. The flexibility of the new AVR to enable its expansion for more complex control applications has thus been demonstrated. Simple techniques to incorporate constraints on control inputs without upsetting the loop decoupling property of the multivariable controller have been proposed and evaluated. It is shown that a further improvement in control performance and stability can be achieved by the integrated controller.