Islanding detection in distributed generation system based on optimized KNN utilizing S-transform based features

• Published in: 2023 International Conference in Advances in Power, Signal, and Information Technology (APSIT) pp. 41 - 47
• Main Authors: Mishra, Manohar, Patra, Chinmoy Kumar, Muni, Pratyush Kumar, Gadanayak, Debadatta Amaresh, Parida, Tanmoy
• Format: Conference Proceeding
• Language: English
• Published: IEEE 09.06.2023
• Subjects: classifier; DER; fault; islanding; KNN ensemble learning; neural network; transmission lines
• DOI: 10.1109/APSIT58554.2023.10201758

This paper presents an islanding detection approach for integrated distribution systems that incorporate distributed energy resources (DERs). The approach utilizes the S-transform and an ensemble K-Nearest Neighbor model (KNN). Initially, the S-transform is employed to extract the characteristic features of the system signals, effectively capturing the transient power variations that occur during islanding events. Subsequently, a KNN model is developed to classify the system states as either islanding or non-islanding. To achieve high accuracy and generalization performance, the KNN model is optimized using a Bayesian optimization algorithm. The proposed approach is evaluated on a simulated DER-integrated distribution system, considering various scenarios, and the results demonstrate its effectiveness in accurately detecting islanding events. This approach provides a reliable and efficient solution for islanding detection in integrated distribution systems (IDS), playing a crucial role in ensuring the stability and reliability of power systems. The modeling and simulation are conducted using MATLAB software.