Distributed AC Islanding Model Detection Method Based on Integrated Learning

With the increasing demand for DC transmission from new energy sources and the increasing DC load, the original AC distribution network system cannot meet the demand for power transmission, and AC-DC distribution network has become one of the development directions of future distribution network. When photovoltaic system is connected to AC/DC distribution system, unplanned island operation may occur on both AC and DC sides of the system. When unplanned islanding occurs in the system, maintenance personnel and system equipment fail to detect islanding in time and operate normally, which may pose a great threat to the safety of equipment and personnel. Therefore, it is necessary to accurately identify the operating state of the system, and then provide accurate judgment information for the action of the micro-switch to operate the photovoltaic system with islanding phenomenon through the micro-switch integrated with islanding detection algorithm, so as to ensure the safety and stability of the distribution system and load. In order to identify the running state of the system reliably, this paper analyzes the changes of the system electrical quantity before and after the islanding, and based on this, puts forward an islanding detection method. First of all, this paper uses a variety of data preprocessing techniques to clean and extract the features of the original data, and selects six island characteristic indicators such as voltage, current and output active power as detection features to generate a feature vector set to improve the quality and accuracy of the data. Secondly, this paper adopts three classification algorithms, including KNN, random forest and XGBoost, to integrate and learn the model, so as to improve the accuracy and robustness of the islanding detection task. Finally, it is verified by experiments that the islanding detection method based on multi-classification fusion model proposed in this paper has high fault detection accuracy and robustness, and has a wide application prospect and popularization value in practical application, which can provide performance improvement and stability guarantee for micro-cluster switches.