Modelagem numérica e experimental de uma "parede solar" utilizando nanofluidos de prata e dióxido de titânio

Abstract

The world population is increasingly aware of the environmental issues surrounding our planet. In the search for solutions that will contribute to the preservation of the environment is the use of renewable energies, which are in a great expansion process. The use of solar energy is a solution for environmental preservation, which is the subject of this work. The work aimed to evaluate the energy absorption capacity of silver and titanium dioxide nanofluids and the hybrid nanofluid composed by the agglomeration of the two nanoparticles used for direct solar absorption in a "Solar Wall". In addition, this work proposes the verification of the application of the learning models of the nearest K-neighbor machine, Support Vector Machine and Random Forests to predict the water temperature in the solar wall. The experiments were exposed for a period of 16 hours over several days, where different concentrations were tested for analysis. One of the analyzes of this work consisted of monitoring the temperature profile, where silver obtained the best results, achieving a gain of 26% against 5% of titanium dioxide. The stored energy rate (SER) was also analyzed, where the increase in this rate occurred in the early hours of the day, for both silver and titanium. This rate was also analyzed as a function of concentration, where a higher concentration represented higher SER values. The specific absorption rate (SAR) was also evaluated, where a behavior inversely proportional to that of SER was obtained, requiring a joint analysis to obtain the ideal concentration of each nanofluid. The silver obtained continuous improvements with the increase of the concentrations, being necessary more experiments to reach the ideal concentration. Titanium, on the other hand, showed a stability for SER and SAR at a concentration of 1.45 ppm, the concentration being considered ideal. The hybrid nanofluid was analyzed in a similar way, obtaining an ideal molar fraction of 6%, which corresponds to a silver concentration of 0.8125 ppm and 23.2 ppm of titanium. In numerical modeling the data were separated into training and test sets, where the purpose was to identify the smallest errors in the test phase, and thus the Random Forest achieved the best results, obtaining an RMSE of 0.562°C. The SVR and KNN methods obtained slightly higher error rates, being 0.818°C and 0.857°C respectively. The numerical modeling was able to establish an approach that was able to meet the objective of forecasting the temperature of the Solar Wall, predicting values very close to those measured, with error rates below 1°C