On-line identification of synchronous generator using neural networks

Due to their approximation capabilities and inherent adaptivity features, neural networks have been employed in modelling of complex nonlinear systems. This paper presents an approach to effectively identify a synchronous generator when only the inputs and outputs are accessible for measurement. Such an identifier (or model) can be used in an indirect adaptive control or internal model control schemes. A one-hidden-layer feedforward neural network is proposed to identify the synchronous generator. Training is done in an on-line mode to allow the identifier to track the variable-parameter nonlinear plant, giving an adaptive attribute to the identifier. Simulation results are presented to complement the theoretical discussion.