Enhanced Cyber Resilient Load Frequency Control Scheme

Abstract

This study develops an integrated framework for improving load frequency control (LFC) performance and cyber resilience in modern interconnected power systems. A two-area multi-source power system is modelled in MATLAB/Simulink, where each area includes reheat thermal, hydro, and gas generation units with governor deadband and generation rate constraints. Renewable uncertainty is incorporated through a 50 MW photovoltaic plant in Area 1 and a 70 MW wind plant in Area 2. For controller tuning, a hybrid metaheuristic algorithm, Particle Swarm Optimization Grey Wolf Op-timizer with Adaptive Artificial Bee Colony (PGWA), is proposed to optimize PID gains by minimizing the integral time absolute error (ITAE). PGWA combines the global search capability of PSO, the hierarchical exploitation mechanism of GWO, and the stagnation recovery ability of ABC. Its performance is validated using benchmark functions and five LFC scenarios involving step load perturbations, random load varia-tions, asymmetric dual-area disturbances, renewable penetration, and communication delay. The results show that PGWA consistently achieves lower error indices, smaller frequency and tie-line deviations, and faster settling than PSO, GWO, ABC, and Modi-fied Hybrid ABCPSO (MHABCPSO). To address cybersecurity, an unsupervised Dis-turbance-Aware Contrastive GRU Autoencoder (DA-CGAE) is developed for false da-ta injection attack (FDI) detection and mitigation. The DA-CGAE model learns the normal operating patterns of the LFC system and detects abnormal behaviour through attention-guided BiGRU encoding, disturbance-sensitive reconstruction, and contras-tive feature learning. It achieves 97.61% accuracy, 99.99% recall, 97.47% F1 score, and 99.93% AUC, outperforming LSTM, BiLSTM, and CNN baselines. The findings confirm the effectiveness of the proposed control and detection framework under complex and uncertain grid conditions.