Skip to Main Content
In this paper, theoretical formulation and implementation results of an intelligent approach to develop a power system stabilizer are proposed. In this approach, a novel methodology that designs a hybrid controller with two algorithms, namely, a neural-network (NN)-based controller with explicit neuroidentifier and an adaptive controller that evolved from a model reference adaptive controller, is designed. This design performs as a novel system-centric controller (where the controller adapts based on system changes). The NN is trained offline with extensive data and is also adjusted online. The main advantage and uniqueness of the proposed scheme is the controllers' ability to complement each other in the case of parametric and functional uncertainties. Moreover, the online NN identifier, models and predicts the plant states/output in the event of functional change during abnormal operating conditions. The theoretical results are validated by conducting simulation studies on a fully nonlinear multimachine power system model consisting of five two-area equivalent generators and eight equivalent buses with varying generator schedules. The results confirm the theory, indicating that the proposed architecture damps interarea low-frequency oscillations faster than other conventional controllers, thus increasing generator stability margin as well as power transfer capability.