Multiple Regression-Based Adaptive Protection Scheme for Microgrids

The dynamic behavior of microgrid on the protection point of views such as modes of operation (grid-connected/islanded) and generation levels of renewable energy sources makes the fault current vary randomly. Subsequently, the relay operating conditions are to be changed based on the above factors so that relay can differentiate between fault condition and normal operating condition. The conventional protection schemes will be unable to differentiate between modes of operations. Further, the fault current is dynamic in nature. Hence, electromechanical or static relays may be replaced with processor-based relays. This paper proposes an adaptive protection scheme for the microgrid using machine learning algorithm. Multiple regression machine learning algorithm is used to develop adaptive protection scheme. The proposed model will be able to differentiate between normal operating condition and fault condition and trip the relay if needed. The hardware prototype developed for the adaptive relaying scheme has been tested successfully.