METODOLOGIA DE PREVISÃO EM TEMPO REAL DO PCI (PODER CALORÍFICO INFERIOR) BASEADO NA REGRESSÃO DE POISSON PARA APLICAÇÃO EM SISTEMAS DE GÁS COMBUSTÍVEL DE PLANTAS PETROQUÍMICAS

LHV (LOWER HEATING VALUE) REAL-TIME PREDICTION METHODOLOGY BASED ON POISSON REGRESSION FOR APPLICATION IN FUEL GAS SYSTEMS OF PETROCHEMICAL PLANTS. This study presents an innovative methodology for real-time prediction of the lower heating value (LHV) of fuel gases in petrochemical plants using Poisson regression. The approach combines statistical tools and phenomenological concepts to optimize energy consumption in thermal cracking furnaces. Thirty-two laboratory samples were used to analyze the gas composition, density, and LHV. Statistical analysis, through Pearson correlation coefficients, identified the key components affecting the LHV. The predictive model developed, based on Poisson regression with a logarithmic link function, showed an average error of only 0.32%, standing out in accuracy compared to other methodologies such as simple linear regression and principal component analysis (PCA). The application of this methodology can result in significant financial and operational benefits, providing a better understanding of the process conditions that affect combustion efficiency. Additionally, the methodology allows for precise adjustments in cracking furnaces, ensuring safe and efficient operation. This work contributes to reducing energy consumption and greenhouse gas emissions in the petrochemical industry, promoting more sustainable and efficient practices.