Autor(es)

da Veiga, Carlos Ernani ; Ramos, Carlos ; Corchado, Juan Manuel ; Fernandes, Piara ; Soares, Joao

• Tipo de documento: Artigo Científico
• Fonte: DISTRIBUTED COMPUTING AND ARTIFICIAL INTELLIGENCE, SPECIAL SESSIONS I, 20TH INTERNATIONAL CONFERENCE; v. 741, p. 10-pg., 2023-01-01.

Resumo

Electric vehicle (EV) owners often have electric vehicle charging stations (EVCS) in their homes and leave the EV connected when it remains in the garage. This procedure does not guarantee thatEVcharging is adequate, and monitoring this period is necessary to identify anomalies. Astudywas developed where the EV charging curves was analyzed by clustering with K-means. The advantage presented in this work is that the necessary data for the charging analysis come from the EVCS power supply circuit. This proposal allows monitoring and rapidly classifying the data according to the form of the EV charging curves, with no need for the supervision of the input data. The unsupervised machine learning method can classify any situation by collecting data on the electrical power supplied to the EVCS. In this study, the classification was applied with learning performed in the history, identifying the clusters representing abnormal situations to identify anomalies in future charging curves. The results of the proposed approach presented around 27.8% as an anomaly in the EV charging, allowing to identify the moment in which it occurs and to make the necessary adjustments to avoid them. (AU)

• Processo FAPESP: 18/20355-1 - Otimização do planejamento da expansão e da operação de sistemas de distribuição de energia elétrica considerando restauração da carga
• Beneficiário: Leonardo Henrique Faria Macedo Possagnolo
• Modalidade de apoio: Bolsas no Brasil - Pós-Doutorado
• Processo FAPESP: 15/21972-6 - Otimização do planejamento e da operação de sistemas de transmissão e de distribuição de energia elétrica
• Beneficiário: Rubén Augusto Romero Lázaro
• Modalidade de apoio: Auxílio à Pesquisa - Temático