Neural-based Predictive Control Applied to FACTS Devices

In this paper applications are presented employing the neural-based predictive control (NPC) for controlling flexible AC transmission systems (FACTS) devices with the purpose of regulating: bus voltage magnitudes through the use of a static compensator (STATCOM), and the active power flow on a transmission line where a static synchronous series compensator (SSSC) is embedded. Contrary to a proportional-integral (PI) conventional control with its trial-and error tuning, once trained a neural net able to predict the firing angle, simple calculations are required to achieve the wanted regulation. Thus, the NPC is a convenient tool to execute the power system adaptive control, with the possibility of carrying out such tasks considering non-linearities. The controller´s design simplicity and its performance compared with the conventional PI controller are shown, especially those related with overshoots and control signals quality, impacting directly into the controlled variable response and having softer behavior than that of the PI control. The applicability of the proposition is studied by digital simulation.