Research on automatic detection method of Microgrid bad data considering manifold sorting

Abstract

In order to improve the quality of Microgrid data, an automatic detection method of Microgrid bad data considering manifold ordering is proposed. The ELM network design and training model was constructed to extract the characteristic quantities of Microgrid data. Then, after analyzing the generation conditions of bad data, the quartile distance method is used to divide the identification interval of bad data, so as to improve the detection accuracy fundamentally. Then the Microgrid data are mapped to corresponding points in multi-dimensional vectors to form a weighted graph model. Considering the approximate manifold structure of data, a high-dimensional data automatic detection method based on manifold ordering is designed. The sorting scores of data nodes are calculated by confidence propagation, and the automatic detection results of bad data are obtained. Experimental results show that the root-mean-square error of monitoring results decreases obviously when the proposed method is used to detect bad data of Microgrid, indicating that the proposed method improves the accuracy of data detection.