ANN based load-frequency internal model control of multi-area power systems

摘要

An artificial neural network (ANN) based load-frequency self-tuning control system of multi-area power networks is designed in this paper. The control system for each subarea comprises chiefly two BP neural network models. One is for system identification and called internal model (IM). The other is for power regulation of generators and called reverse system model (RSM). Instead of a unique global optimization objective in a general discrete optimal control, rolling type limited time-domain optimization scheme is employed in the control. The error between subarea `imaginary frequency' (defined in section 3) and the output frequency of its internal model is used continuously for feed-back to RSM and tuning IM during the process of optimization. The control system can overcome the impacts of internal model error and uncertain disturbances, and is therefore robust and accurate. The effectiveness of the control system has been verified through numerical studies on a two-area power system.