Research on Islanding Detection Method of Distributed Photovoltaic Power Supply Based on Improved Adaboost Algorithm

摘要

The traditional islanding detection method based on Adaboost algorithm usually uses oversampling method to deal with the classification problem of unbalanced datasets, so as to achieve the balance of datasets. However, such processing usually introduces singular samples that are difficult to classify, resulting in the classification performance of classifiers to deteriorate. Aiming at these problems, an intelligent islanding detection method for distributed photovoltaic generator based on improved Adaboost algorithm is proposed. This method uses the Adaboost and K-means++ intelligent algorithms to classify the detected data twice, which can achieve accurate islanding operational detection. The simulation results show that the method of islanding detection can detect islanding state quickly and effectively without affecting the output power quality of the inverter.