COATI-OPTIMIZED BI-LSTM FOR ZERO NDZ ISLANDING DETECTION FOR HYBRID SOLAR-WIND SYSTEMS IN PUBLIC SECTOR

Abstract

The biggest issues prevailing in managing microgrids is the potential for unintentional islanding. This can result in critical risk exposure and technological problem. In these situations, locating and recognizing the defect is essential to equipment maintenance and protection. There are drawbacks to many of the works used in the literature, such as longer tripping times and larger non-detection zones (NDZ). To address these problems, a classifier based on green anaconda optimization with bi-directional long short-term memory (Bi-LTSM) and Fourier Bessel series expansion (FBSE) with empirical wavelet transform (EWT) is proposed. A modified Institute of Electrical and Electronics Engineers (IEEE) 33 bus distribution network with solar and wind integration is used to validate this work on the MATLAB/Simulink platform. Several distribution network defects, such as motor starting, nonlinear load, capacitor bank switching, islanding conditions, etc., are been used to identify islanding. The output gives that proposed method correctly identifies islanding activities and acts correctly in every other situation. According to simulation data, the proposed approach may identify the islanding process in less than 75 milliseconds. The original dataset with zero NDZ delivers 100% accuracy, the proposed approach with the noisy dataset produces an average accuracy of 99.95%.