A Short Review on Data Mining Techniques for Electricity Customers Characterization

Abstract

This paper presents a short review on Data Mining (DM), with a focus on the characterization of electricity customers supported on knowledge discovery in database (KDD) process. The study includes several steps: first, few concepts of the KDD process are presented, including data selection, pre-processing, DM phase, data evaluation and data knowledge; following, a short review of clustering algorithms is presented including partitional, hierarchical, fuzzy, evolutionary methods, and Self-Organizing Maps; finally, the main concepts and methods for load classification are summarized. The main purpose of this work is to present a short review of DM techniques applied to the characterization of typical load profiles in electrical systems and new customers' classification.

An empirical analysis of domestic electricity load profiles: Who consumes how much and when?

“…The method of K-means clustering was used to segment households in relation to their hourly electricity load patterns. The goal of K-means clustering, which is one of the most popular unsupervised machine learning algorithms [23], [33], [37], is to partition n data points into homogeneous k clusters in such a way that households exhibiting similar magnitudes and time of use of electricity demand are grouped. In particular, K-means clustering minimizes the sum of the squared error over all clusters as follows:…”

“…Generally, the choice of the clustering algorithm depends on the nature of the research and available data. Clustering methods include K-means, K-medoids, hierarchical, self-organized maps, or other clustering approaches, with K-means being one of the most widely used due to its stability, efficiency, and ability to work well with a large dataset [3], [23], [33].…”

Heterogeneity of Electricity Consumption Patterns in Vulnerable Households

“…The method of K-means clustering was used to group vulnerable households by their electricity consumption profile. The K-means method is one of the most-used clustering techniques due to its stability, efficiency, and empirical success [47,48], and because it typically presents better results compared to other algorithms [49]. Unlike (agglomerative) hierarchical clustering algorithms (e.g., single linkage, complete linkage, and Ward linkage), which work in a bottom-up manner and start with clusters including a single element and proceed to merge pairs of clusters until there is only one, K-means clustering is a partitional algorithm that simultaneously finds all the clusters as a partition of the data and does not impose a hierarchical structure [47,[49][50][51].…”

Análise de algoritmos de agrupamento aplicados aos problemas de reconfiguração de rede de distribuição

“…Para a análise, são estudados os seguintes algoritmos: a) algoritmo K-means, b) algoritmos hierárquicos, c) algoritmo baseado em rede neural Self-Organizing Map (SOM), e d) algoritmo não paramétrico baseado em densidade DBSCAN, além explorar combinações entre os algoritmos (Granell, Axon and Wallom, 2015;Wang et al, 2017;Cembranel et al, 2019;Pan and Tan, 2019;Panapakidis and Moschakis, 2019).…”

“…Clássica técnica de agrupamento por similaridade, o algoritmo k-means está baseado em um método iterativo para partição de um conjunto de n-número de observações em k clusters (Sharma and Singh, 2015;Cembranel et al, 2019;Panapakidis and Moschakis, 2019).…”