Estimating energy efficiency of the aeration process of stored grains through machine learning /Estimativa da eficiencia energetica do processo aeracao de graos armazenados atraves de machine learning

Aeration is carried out by blowing external air into the silo, with the aim to keep the temperature in the mass of stored grains at safe levels. In the present study, the energy efficiency of aeration of stored sunflower grains was estimated, and a model was proposed and tested to estimate the energy efficiency of aeration, using different algorithms in supervised and unsupervised machine learning. The objective of the work was to develop a Web application based on data mining and modeling with machine learning. The database was composed of information on the average temperature at the height of the sensors, average temperature of the silo, external ambient temperature, occurrence of aeration, if there was cooling, heating and direct heating during aeration, and the energy efficiency of the aeration process. The model for estimating the energy efficiency of the aeration process proved to be efficient, identifying that the energy efficiency was 97.78% during the aeration of stored sunflower grains. Among the classifier algorithms tested, Support Vector Machine (SVM-Poly) showed the best metrics and indicators, hence being recommended for implementation in system development networks capable of predicting the aeration status of stored grains. Key words: Weka, support vector machine, K-means A aeracao e realizada por meio da insuflacao do ar externo para dentro do silo, tendo como objetivo manter a temperatura da massa de graos armazenados em niveis seguros. No presente estudo foi estimada a eficiencia energetica da aeracao de graos de girassol armazenados, assim como proposto e testado um modelo de estimativa da eficiencia energetica da aeracao, utilizando diferentes algoritmos no aprendizado de maquinas supervisionado e nao supervisionado. O objetivo no trabalho foi desenvolver uma aplicacao Web a partir da mineracao e modelagem dos dados com o aprendizado de maquinas. O banco de dados foi composto pelas informacoes da temperatura media do nivel dos sensores, temperatura media do silo, temperatura ambiente externa, ocorrencia de aeracao, se houve resfriamento, aquecimento e aquecimento direto durante aeracao, e a eficiencia energetica do processo de aeracao. O modelo de estimativa da eficiencia energetica do processo de aeracao demonstrou-se eficiente, identificando que durante a aeracao de graos de girassol armazenados a eficiencia energetica foi de 97,78%. Dentre os algoritmos classificadores testados na Maquina de Vetores de Suporte (SVM-Poly) apresentou