A Multifractal Detrended Fluctuation Analysis-Based Framework for Fault Diagnosis in Autonomous Microgrids

The time series obtained during different fault events in an inverter-based microgrid are known to be inherently nonlinear, non-stationary and exhibits multifractal, chaotic behavior. This paper proposes a novel feature extraction and fault detection methodology based on multifractal detrended fluctuation analysis (MFDFA). The limitations of single-scale detrended fluctuation analysis and its susceptibility to interfere with the background noises are overcome in MFDFA which characterizes the multi-scaling nonlinear behavior of load signals during faults. The shape and distribution of the multifractal spectrum along with Hurst exponent are extracted from MFDFA analysis for pattern recognition and classification of different fault events. The efficacy of multifractal features in fault detection and localization with artificial neural network (ANN)-based classifier validates the adequacy of the proposed model.