Event-Triggered Entropy Learning for Encountering of FDI Attack in Grid-Connected Packed E-Cell Inverter

With the high penetration of cyber-physical systems, the power electronic interfaces in smart grids (SGs) are threatened by cyber-attacks. False data injection (FDI) attacks are one of the most repetitive cyber threats that can adversely affect the performance of grid-connected multilevel inverters by manipulating the sensor data in the communication links. In particular, this brief focuses on the design of an event-triggering security control mechanism against cyber-attacks in a grid-connected nine-level packed e-cell (PEC9) inverter in two stages. (I) An adaptive detection is designed by incorporating based on high-order extended state observer (HOESO) and entropy learning to predict the system output. (II) An event trigger mechanism is established to block the false data injected into the measurement signals and eliminate it using the feedback controller. By training the capability of deep neural networks (DNNs), the coefficients embedded in the HOESO are designed to obtain an accurate estimation. By constructing a laboratory prototype of the grid-connected PEC9 inverter, experimental examinations under various types of FDI attacks, including manipulating the current signal using pulse and sinusoidal false data are carried out to verify the resilience of the suggested defense mechanism. Experimental outcomes of PEC9 reveal that the suggested scheme can effectively recognize and mitigate the effect of cyber threats, ensuring the security and reliability of power inverters in SG applications.