Dynamic Event-Triggered RBFNN-Enabled Resilient Control of Cyber Physical Systems Under Sensor/Actuator Deception and DoS Attacks

This work focuses on the design and implementation of a cyber-defensive framework that integrates a recurrent neural network (RNN) to achieve power-frequency synchronization in an interconnected power system under malicious attacks. First, a feedforward compensator leveraging a double-hidden-layer RNN (DHLRNN) is developed to approximate aggregated plant uncertainties, encompassing exogenous disturbances, unmodelled dynamics, parametric perturbations, and sensor deception attacks (SDAs), thereby strengthening disturbance rejection ability and cyber-resilience. Subsequently, to ensure fast finite-time convergence while addressing the intrinsic singularity issue, an integral terminal sliding mode control (ITSMC) framework enhanced with DHLRNN estimation is developed. Furthermore, an adaptive super-twisting algorithm (STA) is integrated with ITSMC, forming an STA-ITSMC scheme that effectively mitigates chattering and enables smooth control effort. To optimize network resource utilization and prevent excessive data transmission, a dynamic event-triggered strategy is introduced in the controller-to-actuator channel, thereby refining the control law without compromising robustness. Lyapunov-based stability analysis is carried out to guarantee the ultimate boundedness of the states and estimation errors. In addition, sufficient triggering conditions and a minimum inter-event time are analytically derived to ensure input-to-state stability (ISS) and to preclude the Zeno phenomenon under malicious cyberattacks. Extensive numerical simulations and real-time experiments conducted on OPAL-RT 4610XG platform demonstrate the superior performance of the proposed controller in countering cyberattacks and enhancing overall system resilience.