A Novel Standalone Implementation of MDNN Controller for DC–DC Converter Resilient to Sensor Attacks—A Design Approach
M. Prasad, Sriranga Suprabhath Koduru, Sreedhar Madichetty, Sukumar Mishra
Abstract
Power electronic converters have become an integral part of the direct current microgrid (DCMG) systems. The efficient control of these converters will decide the performance of the DCMG. With the evolution of cyber physical systems (CPSs), all these power converters are integrated into the communication networks to achieve intelligent and smart control. Data in the communication networks are highly vulnerable toward cyber-attacks, and leaving it unaddressed will lead to substantial economic losses and disasters. This article proposes a standalone implementation of multideep neural network (MDNN)-based dc–dc converter and its application to detect false data injection (FDI) attacks at the sensor level. An MDNN is developed with a combination of deep neural network (DNN) and error detection network (EDN). DNN is used as the controller to achieve closed-loop operation of the dc–dc converter and EDN is used to detect and mitigate the FDI attack. Initially, the proposed scheme is executed in the MATLAB simulink platform with various disturbances at controller level and several attack scenarios are verified with its experimental results.