Integrating fault detection and classification in microgrids using supervised machine learning considering fault resistance uncertainty

Microgrids (MGs) can enhance the consumers' reliability. Nevertheless, besides significant outcomes, some challenges arise. Regarding the intermittent nature of Renewable Energy Resources (RESs), MGs are not operated radially. Accordingly, the reliable protection of MGs considering uncertainty in RESs is crucial for planners and operators. This paper uses data analysis to extract knowledge from locally available measurements using RMS values of symmetrical components. The learning-based characteristic of the suggested technique with a low computational burden exempts the need for an available communication infrastructure in the MG. The Support Vector Machine (SVM) technique is applied to train the Intelligent Electronic Devices to have a reliable MG protection scheme. The proposed method, which performs fault detection and classification together, just requires local information and functions effectively to discriminate faulty from normal conditions considering different uncertainty of resistance faults. Digital simulations on an MV test network were conducted to construct an appropriate database to consider some aspects of uncertainty in the network. The various faults considering their uncertainty, the different modes of operation, the uncertainty of RESs generation, and the load levels are combined to produce myriad scenarios. The simulation results confirm the effectiveness of the proposed adaptive protection approach in accurately distinguishing different system modes and consistently protecting the MG, achieving an accuracy rate of 99.75%. Furthermore, it offers the MG an optimal protection scheme that is not limited by selectivity constraints across diverse conditions.