Simultaneous Faults and Cyber-Attacks Diagnosis in Wind Turbines; An LMI Approach using Memory-based Dynamic Residual Field

This paper presents a unified diagnostic framework for detecting actuator faults, sensor faults, and cyber-attacks in wind turbines operating under uncertainty, nonlinear perturbation and external disturbance. The core innovation is the Memory-based Dynamic Residual Field (MDRF), a dynamic structure that tracks both estimation errors and their temporal evolution, ensuring high sensitivity to anomalies with robust disturbance rejection. Using a state observer for augmented system and minimizing \(\:{H}_{\infty\:}\) gain of disturbances via Linear Matrix Inequality (LMI), an analytical thresholds that guarantee the separation of normal operations from cyber-attacks was derived. Crucially, a cross-correlation logic distinguishes legitimate set-point changes from malicious intrusions. Simulations on a benchmark \(\:4.8MW\) wind turbine confirm accurate, real-time isolation of all scenarios. Notably, a comprehensive comparative study demonstrates that the proposed framework outperforms standard Unknown Input Observers (UIO) in isolating cyber-attacks from legitimate command changes. Also, this model-based approach matches the detection accuracy of Neural Networks while reducing computational load by over \(\:90\%\), proving its suitability for resource-constrained industrial controllers.