AI-Based Self-Driving Grid-Following Inverters with Compromised Supervisory Layer Controller

This paper presents a Bayesian regularization-based artificial neural network (BRANN) for power setpoint correction of grid-following (GFL) inverter with compromised supervisory controller layer. Coordinated control schemes with multi-layered communication infrastructure makes the upcoming power electronics dominated grid (PEDG) susceptible to cyber threats. In particular, malicious power setpoints introduced by a compromised secondary controller can jeopardize the normal operation of PEDG. The main goal of this study is to develop an AI-scheme for the primary control layer of GFL inverter to correct the manipulated power setpoints in real-time independent of the compromised secondary control. A neural network (NN) module is trained via Bayesian regularization algorithm to address the challenges of AI-based schemes deficiencies. The provided case studies validate the performance and effectiveness of the proposed approach.