An Advanced Fault Detection Technique for DG Integrated Microgrid Using Fast Fourier Discrete Orthonormal Stockwell Transform-Based Hybrid Optimized Kernel Extreme Learning Machine

The proposed technique develops an intelligent protection scheme for multiple distributed generations (DGs) integrated microgrid systems. In this work, the retrieved fault current signal samples from the nearer faulty feeder are preprocessed through the fast Fourier discrete orthonormal Stockwell transform to obtain the spectral energy and differential energy. The normalized and prominent statistical differential energy features like maximum and minimum differential energy, entropy, standard deviation, mean, and median are calculated for analysis. The maximum and minimum values of differential energy are assessed to determine the severity of faults that can help in real-time applications. Further, these factors characterizing the fault type are considered as the input parameter to the hybrid approach as grey wolf optimization and particle swarm optimization-based kernel extreme learning machine algorithm for accurate fault detection. To justify the adaptability and usability of this method, this strategy is simulated under different types of faults like asymmetrical and symmetrical including different distance, fault inception angles, and fault resistance with different topologies like radial and looped configurations in both grid-connected and islanded modes of operation on a generalized IEC microgrid test system. The reliability and robustness of this methodology are examined by evaluating three parameters as accuracy, security, and dependability. Further, the comparative analysis and outcome of this work confirm the adaptability and superiority of this proposed method for the effective detection of a fault in the DGs-connected microgrid system.