Data-driven analysis and machine learning for energy prediction in distributed photovoltaic generation plants: A case study in Queensland, Australia

Autor(es)

Ramos, Lucas ^[1] ; Colnago, Marilaine ^[1] ; Casaca, Wallace ^[1]

Afiliação do(s) autor(es)

^[1] Sao Paulo State Univ UNESP, Dept Energy Engn, BR-19274000 Rosana - Brazil

• Tipo de documento: Artigo Científico
• Fonte: ENERGY REPORTS; v. 8, n. -1, p. 745-751, APR 2022.
• Citações Web of Science: 0

Resumo

Understanding patterns and energy-related data in photovoltaic systems is one of the key tasks in energy generation and distribution. In fact, the use of data-driven tools and predictive learning models can support the government, power regulatory agencies, and the energy industry in improving their decision-making and operational activities. Bering this in mind, this paper presents a case study of data-driven analysis and machine learning to forecast the energy charge in the distributed photovoltaic power grid of Queensland, in Australia. Our analysis relies on a freely, open energy tracking platform and the design of three Machine Learning approaches built on the basis of Random Forest, Support Vector Machines, and Gradient Boosting methods. Experimental results with real data showed that the trained models allow for very consistent predictions while reaching a high forecasting accuracy (around 95%-93% in Generated-Exported prediction, respectively). Moreover, it was found that the Gradient Boosting-based model ensures robust behavior and low prediction errors, as endorsed by quality validation metrics. Another technical aspect observed is that the variables artificially created to boost the models substantially improve the post-analysis and overall accuracy of the results. (C) 2021 The Author(s). Published by Elsevier Ltd. (AU)

• Processo FAPESP: 13/07375-0 - CeMEAI - Centro de Ciências Matemáticas Aplicadas à Indústria
• Beneficiário: Francisco Louzada Neto
• Modalidade de apoio: Auxílio à Pesquisa - Centros de Pesquisa, Inovação e Difusão - CEPIDs
• Processo FAPESP: 19/18857-1 - Análise preditiva da geração e exportação distribuída de energia fotovoltaica: modelagem e aplicações via algoritmos de inteligência computacional
• Beneficiário: Lucas Gomes dos Ramos
• Modalidade de apoio: Bolsas no Brasil - Iniciação Científica