Application of Machine Learning in Optimizing Thermochemical Conversion Processes with Pre-treatment to Get Higher Bio-oil Yield from Biomass Waste

Improving the bio-oil yield was a challenging part in the thermochemical conversion processes. Implementing suitable pre-treatment technology to improve the biomass characteristics is an effective technique to increase the yield. In this study, a multi variate random forest algorithm was used to optimize the pre-treatment method in order to improve the biomass characteristics. The data collected from many previous studies were analysed to identify the importance of biomass characteristics in bio-oil yield. The correlation between biomass characteristics and bio-oil yield, was analysed using pearson method and the important influencing parameters %C and %H have a very good positive correlation with a coefficient value range 0.455 to 0.818. Among the six pre-treatment methods analysed, thermochemical pre-treatment method was found effective with more than 95% improvement of many biomass characteristics. The range of voting given to the parameters identify %H be the important characteristic optimized first. The suggested method was validated by laboratory experiments and %accuracy between predicted and calculated biomass characteristic values showed more than 90% accuracy for all the biomass characteristic parameters tested in this study.