The techno-socio economic considerations have triggered the installation of more renewable energy sources to the grid. Most of the renewable energy systems are connected to the grid in a distributed manner. The output of the renewable energy sources like wind, solar and tidal energies are highly unpredictable in nature as the sources of energy considerably depend on natural conditions. The add of effect of such these disturbances will affect the quality of the grid and is more serious in microgrids. The fluctuating demand and the uncertainty in the generation will affect the direction of the power flow from node to node. This paper focus on a multilayer machine learning procedure for identifying dissociative behaviour of a grid at various nodes, methods to effectively schedule and predict the power flow between the nodes, thus by identify the islanding. © 2018 IEEE.