Energy-Efficient Islanding Detection Using CEEMDAN and Neural Network Integration in Photovoltaic Distribution System

This paper proposes an enhanced islanding detection method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and a pattern recognition neural network (PANN). Negative sequence voltage data from both islanding and non-islanding scenarios were acquired through MATLAB Simulink simulations, with samples collected at a frequency of 3.84 kHz over a 2.5-second period. The voltage signals were decomposed into intrinsic mode functions (IMFs) using CEEMDAN, after which key features namely normalized max value, standard deviation, and entropy of the IMFs were extracted. The extracted features were used to train the PANN. The model was evaluated using cross-validation and several performance metrics, including accuracy, precision, recall, and F1 score. The proposed model achieved an overall accuracy of 98.6%, with a precision of 100%, a recall of 97%, and an F1 score of 98%. The detection time was found to be 0.2381 seconds, indicating the method's suitability for real-time applications. Furthermore, feature permutation importance analysis highlighted the critical role of certain features in the model's performance. The results demonstrate that the proposed method provides a reliable and efficient solution for islanding detection in grid-connected PV systems, significantly reducing the non-detection zone and ensuring high detection accuracy. This study contributes to developing of advanced detection techniques, enhancing the safety and reliability of modern power systems.