Multi-model self-learning control for turbine valving control

As turbine valving control of synchronous generator faced practical challenges as nonlinear characteristics, large-scale operating ranges and changing operation points, this paper proposed a multi-model self-learning controller (MMSC). Firstly fuzzy logic controller (FLC) rules for plant at various operation points were derived from operation samples. Then fuzzy clustering algorithm was employed to remove redundant operating models to N kinds of typical models, this reached N sub-model FLC (SFLC). Then control output of MMSC was just the output of SFLC multiplying their respective weights, which were decided by their matching degree. For the self-learning of SFLC, support vector machines was employed. Simulations show the effective and damping ability of the proposed MMSC.