Effective fault diagnosis and distance calculation for photovoltaic‐based DC microgrid using adaptive EWT and kernel random vector functional link network

Photovoltaic (PV) based DC microgrids are being introduced prominently in the recent energy market, especially towards local uninterrupted power supply. These low voltage DC distribution networks are facing implementation challenges due to the protection (monitoring and coordination) standards’ limitation; especially under multi‐distributed generations operation subjected to DC cable faults as well PV side arc events. Thus, a new DC fault diagnosis (DCFD) model is proposed here to monitor the fault events with accurately estimated distance, and to provide relay coordination podium. A new adaptive empirical wavelet transform (AEWT) based signal decomposition is incorporated to this model, where accurate fault detection is obtained in terms of sensitivity index (maximum weighted kurtosis index/ M‐WKCI) of different intrinsic mode functions towards less computational classification. The accuracy is ensured by an improved firefly algorithm to avoid false detection, sympathetic tripping, erroneous distance estimation etc. The fast, robust decision making is proposed by non‐iterative Kernelised Random Vector Functional Link Network, where higher dimensional mapping is obtained to cope with noise profile volatility of DC faults by Gaussian Kernel function. The effectiveness of the proposed DCFD model‐based classification and distance estimation is established through rigorous case studies in MATLAB environment.