Topology Recognition of Courts Based on the Improved PCA

With the further development of smart grid, the accurate recognition of each user in each transformer has become the basis of many high-end applications. In order to solve the problem of courts’ topology recognition, this paper proposes a new improved method based on the principle of Principal Component Analysis. First, the eigenvectors with larger eigenvalues from divergence matrix are selected, and these eigenvectors contain most of the information of the original time series. After analysis, this paper indicates that the essence of PCA for dimensionality reduction of time series is to compare the correlation between the original sequence and the characteristic waveform sequence. So, the similarity curves can be obtained from characteristic sequence waveforms. By comparing the Cosine similarity of the similarity curves, the non-local users can be excluded accurately. With the k–means, the relationship between local courts users can be accurately obtained. This method avoids the error caused by noise and abnormal data after directly using PCA to reduce dimension, and improves the accuracy of resolution. Subsequently, the normalized distance is used as the standard to match the user and the transformer voltage, and the identification of the courts’ topology is realized. In the end, the algorithm is validated with simulation model data. At the same time, the data are used for other algorithms, and get the conclusion that the topological recognition algorithm based on improved PCA is more effective.