Machine Learning-Based Severity of Critical Line for Power System Security Enhancement with Zip Loads

This research proposes an advanced, data-driven framework to enhance power system security and operational reliability in the face of unforeseen faults and contingencies. A composite polynomial load model incorporating constant impedance (Z), constant current (I), and constant power (P) characteristics is employed to capture realistic load behavior under varying conditions. The methodology begins with modeling based on standard IEEE test systems, integrating this load model into power flow analysis using the Newton-Raphson algorithm. For contingency assessment, the focus is placed on single-line outage scenarios-critical events that may significantly impact system stability. A novel Hybrid Line Stability Ranking Index (HLSRI) is introduced to prioritise vulnerability, offering a more accurate ranking of transmission line criticality under stress conditions. Additionally, machine learning algorithms, including Gradient Boosting and Random Forest classifiers, are trained on system operational data to categorize the severity of line contingencies with high precision. To enhance control and stability, Flexible AC Transmission System (FACTS) devices such as the Unified Power Flow Controller (UPFC) and Interline Power Flow Controller (IPFC) are strategically deployed. Their optimal placements are determined through the metaheuristic Sparrow Search Algorithm (SSA), ensuring minimal power losses and improved dynamic performance. Simulation results validate the superiority of the proposed framework in terms of accuracy, adaptability, and system-wide resilience, making it a promising solution for real-time power grid reliability enhancement.