Solar Energy Forecasting: Case Study of the UNICAMP Gymnasium

Author(s)

do Nascimento, Gleyson Roberto ; Guillardi Junior, Hildo ; Attux, Romis

• Document type: Journal article
• Source: ENERGY INFORMATICS, EI.A 2023, PT II; v. 14468, p. 16-pg., 2024-01-01.

Abstract

Within the spectrum of studies conducted by the Sao Paulo Center for Energy Transition Studies (CPTEn), time series from the Photovoltaic Energy Plant of the UNICAMP Multidisciplinary Gymnasium (GMU-PV) were analyzed. This plant is associated with the first implementation of a photovoltaic system in the context of the Sustainable Campus Project (PCS) at UNICAMP - as a consequence, it originated the most extensive and robust time series in the project. The research, structured according to the Cross Industry Standard Process for Data Mining (CRISP-DM) methodology, aimed to identify the patterns and parameters associated with the energy production of the aforementioned photovoltaic system. Based on Machine and Deep Learning techniques, forecasting models were developed to maximize the use of available resources and promote the sustainability of this energy system at UNICAMP. In evaluating the results, it was observed that the most effective model was the Orthogonal Matching Pursuit (OMP) built from the Python low-code library, PyCaret. This regression machine learning model led to a coefficient of determination (R-2) of 0.935 494 and a root mean square error (RMSE) of 8.561 679. (AU)

• FAPESP's process: 20/16635-9 - Energy management in distribution networks through the concept of virtual power plants
• Grantee: Hildo Guillardi Júnior
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• FAPESP's process: 21/11380-5 - CPTEn - São Paulo Center for the Study of Energy Transition
• Grantee: Luiz Carlos Pereira da Silva
• Support Opportunities: Research Grants - Science Centers for Development