M.S. AAI Capstone Chronicles 2024

ELECTRICITY DISTRIBUTION TOPOLOGY CLASSIFICATION

6

cannot account for distance relative to elevation as required. Cook also considered the BIRCH method to bind extreme points, perform hierarchical clustering, and accelerate K-means clustering. Still, this method must account for distribution systems with varied geographical radii. Another approach introduced by Cook was to focus on a subgroup or points and analyze the trend, in which the DBSCAN (Density-based spatial clustering of applications with noise) would form the clusters based on the neighbourhood region and the minimum number of data points in the neighbourhood. The DBSCAN approach can identify clusters of any irregular shape, which accounts for the issues that may arise from the varied geographical data of the streets. There is also a supervised learning approach that is shown in Tennakoon, R. N. et al. research that employs the Euclidean distance measure to calculate similarity values between each smart meter's voltage profile and a reference meter from each transformer, which are then used to train a cubic SVM classifier (Tennakoon, R. N. et al., 2020). Tennakoon, R. N. et al.’s research tested Euclidean distance, Manhattan distance, Jaccard Similarity, and Cosine similarity combined with a support vector machine, logistic regression, and random forest to compare the result where Euclidean plus support vector machine provided the highest accuracy result. However, the research and testing of this method shows that each similarity between each meter and each transformer reference needs to be calculated. Tennakoon, R. N. et al.’s research was based on 710 smart meters which connect to 5 transformers. A possible limitation has been observed that the Euclidean distance is calculated as 710 X 5 = 3550 in the research; however, in real life, meter numbers are generally at more than a million, and transformers normally at tens of thousands numbers; the process involves computing the Euclidean distance between each

97