AI-Based Fault Diagnostic Framework for Load Frequency Control in Modern Power Systems

• AI-driven fault diagnostics and control framework for LFC under sensor faults • LSTM-based controller reconstructs accurate control during frequency sensor faults • The framework enables fault-tolerant LFC in a two-area interconnected power system • Robust performance is validated under step and variable load variations • Simulations show improved transient response compared to conventional PID-based LFC Modern interconnected power systems require load frequency control (LFC) mechanisms to maintain frequency stability under dynamic operating conditions and faults. Conventional LFC strategies, such as PID controllers, often suffer from limited adaptability when sensor faults and nonlinear disturbances occur. This research developed an AI-driven fault diagnostics and control framework for modern power system. By AI techniques, the study provided a deep learning based Long Short Term Memory (LSTM) controller to give an accurate control signal so that it can regulate the power system’s response. The proposed approach utilized sensor data collected from the power system, including frequency regulation, power deviation, load variations, and control signals, and trained deep learning LSTM models that recognize patterns, anomalies, and potential system failures. Unlike traditional fault detection approaches, the proposed method not only identifies abnormal conditions but also generates corrective control signals to maintain system stability during sensor faults. The proposed framework employed for the power system of two area networks and compared the control signal response of the conventional PID controller and the AI-based controller during a fault. System response is analyzed under both scenarios of step load variation and variable load disturbances to validate the proposed framework. Simulation results demonstrate that the proposed approach improves transient response, and provides fault-tolerant control faulty conditions compared with conventional PID control and other AI-based models. Furthermore, fault detection capability is validated using ROC-AUC and threshold sensitivity analysis, confirming reliable fault detection with controlled false alarm rates. The proposed framework offers a promising solution for intelligent and resilient LFC operation in modern smart grid environments.