This paper proposes a fast and reliable two-level islanding detection method (IDM) for grid-connected photovoltaic systems (GCPVSs) based microgrid. In the first level of the proposed IDM, the magnitude of rate of change of output voltage (ROCOV) is computed. If this variable exceeds a predefined threshold, a disturbance is injected into the duty cycle of DC/DC converter after a given time delay to deviate the system operating point away of its maximum power point (MPP) condition. This leads to a substantial active power output and voltage reduction in islanded mode. Therefore, the ROCOV and the rate of change of active power output (ROCOP) indices, measured in the second stage, pose great negative sets at the same time in islanding states. However, the variation of at least one of these variables is nearzero in non-islanding switching events. The assessment of the presented algorithm has been conducted under extensive islanding and non-islanding scenarios for a case study system with two PV power plants using hardware-in-the-loop (HiL) simulation tests. The provided results remark precise islanding classification with eminently small non-detection zone (NDZ) within 510 ms. The presented IDM has the advantages of self-standing thresholds determination, no improper effect on the output power quality, and simple and inexpensive structure. Moreover, the fast MPP restoration of the proposed scheme after islanding identification boosts the chance of seamless reconnection and DG autonomous operation in microgrid. Index Terms-Grid-connected photovoltaic system (GCPVS), Islanding detection method (IDM), Maximum power point (MPP), Microgrid, Non-detection zone (NDZ), Rate of change of active power output (ROCOP), Rate of change of output voltage (ROCOV).