A Novel System-Centric Intelligent Adaptive Control Architecture for Power System Stabilizer Based on Adaptive Neural Networks

In this paper, an intelligent power system stabilizer based on a novel system-centric controller is proposed. The proposed method uses hybrid architecture with two algorithms: one, a neural network-based controller with explicit neuro-identifier, and the other, an adaptive controller implemented as a model reference adaptive controller (MRAC). The neuro-controller-identifier combination is used to approximate the nonlinear functional dynamics of the power system, and the MRAC controller adapts when power system (plant) parametric set changes. Additionally, a feed forward neural network (FFNN) identifier is used to predict system responses, and the control signals are adjusted in real-time to obtain improved system response. The FFNN is trained offline with extensive test data and is also adjusted online. The main advantage and uniqueness of the proposed scheme is the controllers' ability to complement each other in case of parametric and functional uncertainty and evolve in the presence of changing system dynamics. The theoretical results are validated by conducting simulation studies for electric-generator stabilization on a single-machine infinite-bus system and a two-area equivalent five-generator eight-bus multimachine power system with varying generator schedules that show fundamental subsynchronous oscillations.