Resilient Cyber-Attack Detection and Mitigation in Grid-Tied PEC9 Inverter Using A3C-Based Adaptive Control

In this paper, the cyber security problem of multi-level inverters (MLIs) in the grid-connected mode is addressed. The cyber-attack can falsify the measurement signals that are transmitted during insecure communication networks. In addition to targeting the data integrity., the cyber-attack can compromise the data availability transmitted during insecure communication networks. For this purpose., a two-stage resilient scheme is developed for identification and mitigation of denial-of-service (DoS) attacks in the measurement signals. i) a detection mechanism is designed by a convolutional neural network observer (CNNO) to estimate the signal transmitted during the insecure communication network., compare the estimated signal with the observed signal., and identify DoS theaters in the system., and ii) the mitigation is realized by an adaptive controller. In the mitigation mechanism., the asynchronous advantage actor-critic (A3C) is utilized to adaptively adjust coefficients embedded in the control structure. By interacting with the A3C learning with the multi-level inverter., the destructive effects of DoS threats are eliminated. The performance of the proposed two-stage resilient technique is evaluated on the MLI compromised by DoS attacks. The simulation results reveal that despite the cyber-attacks being able to impose a high level of instability on the multi-level inverter., the proposed scheme can ensure the ideal operation system from the stability and security point of view.