Adaptive power system control with neural networks

The authors present a design of a new adaptive control system and demonstrate its performance in a computer simulation of a synchronous machine control task. The design utilizes the self-organization and predictive estimation capabilities of neural-net computing. The task of real-time adaptation is carried out using an error-based online learning scheme which is implemented on a cluster-wise segmented associative memory system. The systems ability to improve its own performance is demonstrated, and the possibility of using the existing control device in a supporting mode during a preliminary phase of the controller design is presented