Microgrid mandates a highly accurate and reliable protection measure for safe and secure operation. Reliability of the protecting system becomes a serious concern when the microgrid operates with various types of DG units which includes synchronous, induction and inverter based DGs with different topology and modes of operation. This paper demonstrates an intelligent protection scheme using deep neural network (DNN) to perform the task of fault detection and classification. Fault currents retrieved at the relaying points are given to the proposed protection scheme to perform the desired task. The performance of the scheme is measured with wide variations in system parameters such as modes of operation and network topologies of the microgrid. The scheme is extensively tested on a large dataset even by varying types of inputs and different deep neural network structures. The test results and response time show that the proposed algorithm can be used in real time for protective measures of microgrid. © 2019 IEEE.