Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Predicting solar radiation in Minas Gerais using artificial intelligence techniques
Submission Author:
João Nazareth , MG
Co-Authors:
João Nazareth, Ricardo Furiati , Silvio GUIMARAES, Felipe Domingos Cunha, Cristiana Brasil Maia
Presenter: Ricardo Furiati
doi://10.26678/ABCM.COBEM2023.COB2023-1621
Abstract
One of the goals of the Ministério de Minas e Energia, as a part of the 2030 energy plan, is to diversify Brazil's main energy source towards renewables. Solar energy is one of the most promising green power generations due to its scalability and applicability. Although investment in solar generation equipment - solar panels and distribution infrastructure- is expected to grow in the next few years, it is heavily dependent on analyzing where those resources will yield better results. Moreover, installing new solar equipment is directly tied to analyzing the predicted solar radiation in a given area, which is currently made on a case-to-case basis. A reliable way to predict solar radiation accurately, at scale, in a given area could hold back further investments. The present study aims to create a Machine Learning model capable of predicting solar radiation in the Brazilian state of Minas Gerais. One of the study's goals is to validate theoretical values of solar incidence calculated from the clearness index in a Typical Meteorological Year (TMY) proposed in an analytical model. One of the key aspects of the study is the use of the meteorological data provided by NASA in 20 years used in the training of the machine learning model. A similar study was conducted using solar radiation data from the National Institute of Meteorology's (INMET) ground stations for 5 years. Whereas the past study could predict solar radiation somewhat accurately, the expansion of the time frame could improve the precision of the model. The principal methodology used in the study's development consists of calculating the theoretical values of solar radiation in Minas Gerais, using the clearness index obtained from the analytical model, and comparing it to the actual radiation incidence observed by NASA. Additionally, the calculated and observed solar incidence are used as a baseline to compare the predicted values generated by the machine learning model. Some correlation between the theoretical and observed values is expected due to its experimental nature. However, the differences in the time frame (theoretical meteorological year and actual yearly observations) could produce some uncertainty. The expected results are related to improving the machine learning model and prediction capabilities and generating heat maps from the predicted solar radiation values.
Keywords
solar radiation, machine learning, Clearness index, maps, prediction
