A continually online trained neurocontroller for excitation and turbine control of a turbogenerator [PowerPoint Presentation]

A collection of slides from the author's PowerPoint conference presentation is provided. The presentation covers the following: (1) The design and implementation of a continually online trained (COT) neurocontroller and a COT neuroidentifier on a practical turbogenerator system test bed, using multi-layer perceptron (MLP) neural network. (2) Both simulation and practical implementation results are presented in order to prove that the neurocontroller is stable when using deviation signals for online training. (3) The drawback related to MLP neural networks compared to the RBF neural network, reported in references in the paper, is overcome with the use of deviation signals instead of using the actual signals. (4) It compares the performance of the neurocontroller with the conventional AVR and governor controllers but emphasizes on how to do neurocontrol with online training on a practical turbogenerator system.