Análisis de técnicas de clusterización en sistemas eléctricos de distribución

Abstract

The design of new distribution power lines is a task that engineers frequently encounter; this paper proposes to design medium and low-voltage power networks with the help of Clustering. Clustering methods have been used since ancient times for data analysis because of their ease of implementation and quality in delivering results. There are several clustering methods, but a comparison will be made between the most used clustering methods, K-Means and K-Medoids. One of the big problems of clustering methods is that it is impossible to precisely determine the number of K clusters that must be generated per case study. Still, several principles give us an idea of how many clusters can be generated; we will analyze several of these principles to have our system as optimized as possible. With the help of minimun spanning tree MSP, we will obtain the shortest route the driver has to travel along the entire projected electrical network, saving costs and reducing analysis times. Finally, with the results obtained, we will conclude which method gives more optimal results, analyzing implementation costs and losses along the network.