We propose a multi-agent self-learning control architecture of fleet of smart devices to enhance the grid resiliency and stabilize voltage/frequency in response to cyber-physical attack and other anomalies. The inverters hardware prototyped by the team, funded by previous EP-NPRP project, will be used as the groundwork to develop the smart inverters and P-HIL setup in this project. We will develop a coordinated fault-ride through capability of multiple smart inverters and meters via semi-supervised learning analytics. In case of detection of intrusion in smart inverter or other smart devices, the faulty unit will be detected and isolated, if required, while the operation set points of the healthy smart inverters can be changed autonomously to ensure energy services particularly to the critical infrastructure in the country.