Adaptive Load Shedding Based on Voltage Stability in the Tambora System Using the XGBoost Algorithm

Modern power systems require adaptive defense mechanisms to maintain stability during disturbances. This paper proposes an Adaptive Load Shedding (ALS) strategy based on voltage and frequency stability for the Tambora system, utilizing the XGBoost machine learning algorithm. The proposed method detects the need for load shedding in real time by analyzing key system conditions and classifying events into three operational modes: No Load Shedding, Eco Mode (minimal curtailment), and Aggressive Mode (emergency response). When the system frequency drops due to a generator trip, the ALS mechanism is activated through the XGBoost classifier. To ensure both frequency and voltage stability, the Center of Inertia (COI) is used to allocate load shedding proportionally between the Bima and Sumbawa subsystems. Additionally, the Fast Voltage Stability Index (FVSI) is applied to rank substations based on voltage vulnerability, allowing the system to prioritize curtailment at the most critical locations. This coordinated approach helps prevent voltage collapse and enhances overall system resilience.