NARX Neural Network based UPFC for Enhancement of Power System Security in Large Power Network

In modern power systems security challenges primarily arise from demand and supply imbalances placing excessive stress on transmission infrastructure and pushing the system beyond its thermal limits. To mitigate these issues advanced solutions such as Flexible AC Transmission Systems (FACTS) have been developed. FACTS technology enhances the power transfer capability of long distance transmission lines while improving overall system stability. These devices play a vital role to improve power grid stability, maintaining desired bus voltage levels and strengthening network reliability. This study examines the effectiveness of NARX (Nonlinear Autoregressive Exogenous) neural networks in optimizing large scale power networks. By dynamically regulating power flow the NARX neural network improves both operational efficiency and system stability. A comparative analysis is conducted between a NARX neural network based Unified Power Flow Controller (UPFC) and a conventional Proportional Integral (PI) controller based UPFC using MATLAB/SIMULINK for system modelling and performance evaluation. The results demonstrate that the NARX neural network based UPFC significantly reduces rotor speed deviations ($\Delta \omega$), stabilizes rotor speed ($\omega$) and effectively dampens oscillations caused by system faults. Additionally, it substantially enhances system real power, reactive power and node voltage profiles compared to conventional systems without NARX neural networks. These findings underscore the superiority of the NARX neural network based UPFC in improving oscillation damping and overall power system performance positioning it as a more efficient alternative to the PI based approach.