Using machine learning to predict the production and quality of bio-oil from pyrolysis biomass

Reducing the reserves of fossil energy sources serves as a warning sign for humanity. On the other hand, the increasing consumption of fossil fuels has led to significant environmental problems, such as global warming. These issues make the replacement of renewable energy sources with fossil fuels inevitable. Among various renewable energy sources, biomass is a reliable and sustainable resource. Thermochemical conversions of biomass are a promising method for converting raw biomass into liquid (bio-oil), solid (bio-char), and gas (biogas) fuels suitable for modern life. As one of the most important thermochemical conversions for efficient bio-oil production, pyrolysis has received significant attention. However, pyrolysis requires advanced equipment, precise product quantity, and quality measurement, which can be challenging and costly. Therefore, modeling has been extensively researched to enhance the performance and efficiency of pyrolysis. In recent years, machine learning has gained considerable attention in pyrolysis modeling, particularly for yield optimization, real-time monitoring, and process control. In addition to conventional techniques like artificial neural networks that capture nonlinear correlations between input and output values, combined machine learning models have been of particular interest for modeling and optimizing complex problems more effectively. This study provides a comprehensive overview of the research conducted on the application of machine learning in pyrolysis process modeling and assesses the prospects of this technology. These machine learning models have provided R2 between 0.26 in the weakest case and 0.99 in the best case for predicting bio-oil production. These values have been presented between 0.6 and 0.93 to predict the improvement of bio-oil quality modeling.