Etanol de biomassa de milho: utilização de aprendizagem de máquina no estudo de casos de caldeiras

Abstract

In this dissertation, we work with a steam generator (boiler) operating within an industrial plant that produces biofuels (in this case, corn ethanol) using supervised machine learning as a tool. Sugarcane ethanol is extensively studied, but in the case of corn, in Brazil, it is still embryonic - with scarce records, of a fundamental equipment that is the boiler, operating with specific burning of eucalyptus chip, seeking to obtain optimized thermal efficiencies and understanding its behavior, mainly of a “full” corn plant. Data from process variables were obtained in the field through the DCS- Digital Control System (60 points from two boilers for 30 days), worked with Python and ML (Machine Learning) code that we use as a methodology. The results were positive, demonstrating that this important industrial equipment can be studied and its behavior predicted with the methodology used, helping and enabling us to achieve better efficiencies and better understand its technical behavior.