Enhanced Microgrid Islanding Detection Using Optimized Deep Convolutional Neural Networks and Markov Transition Fields

Islanding detection is vital for maintaining the stability and safety of microgrid operations. This paper introduces an enhanced method using a Deep Convolutional Neural Network (CNN) optimized through the Grey Wolf Optimizer (GWO), combined with Markov Transition Fields (MTF) to detect islanding events with high accuracy. The GWO algorithm automatically tunes key CNN parameters, such as kernel size and number, enabling the model to efficiently learn and extract features from microgrid signals. Meanwhile, MTFs transform time-series signals into 2D representations, enriching the model with spatial-temporal context. The proposed GWO-CNN + MTF framework is trained and tested on microgrid datasets covering both islanding and non-islanding events. Experimental results confirm that the model achieves substantial accuracy and robustness even under noisy conditions. The proposed GWO-CNN model is evaluated against current state-of-the-art techniques, and results indicate its significant superiority. This research contributes to the advancement of islanding detection techniques, offering a promising solution for ensuring the reliability and resilience of microgrid systems.