Detection of islanding and non-islanding fault disturbances in microgrid using LMD and deep stacked RVFLN based auto-encoder

Abstract

This paper presents a hybrid AC/DC microgrid consisting of multiple distributed energy sources, modeled in MATLAB/Simulink environment that undergoes islanding and non-islanding (symmetrical and unsymmetrical Faults) disturbances. These disturbances result in various changes in circuit parameters (voltage, current, and frequency) and this paper focuses only on disturbed current signals that are extracted from the solar and wind-connected buses. Further to extract rich and useful fault features from the extracted current signals, an adaptive detection technique, local mean decomposition (LMD) is proposed that results in series of product functions (PFs). Various feature extraction indices are used to extract the underlying features of the decomposed PFs that are processed through a deep-stacked random vector functional link network (RVFLN) based auto-encoder (AE) technique for classifying the faults. The effectiveness of the proposed RVFLN based AE technique is evidenced in terms of classification accuracy (CA) and confusion matrix (CM). The performance of the proposed technique has been evaluated through reliability analysis (MAPE, MAE, RMSE, and CM) by comparison with various artificial neural networks under both islanding and non-islanding mode of operation.