Study on prediction models of oxygenated components content in biomass pyrolysis oil based on neural networks and random forests

Biomass pyrolysis oil, as a renewable energy source, has significant application value, with the content of its oxygenated components being a critical parameter affecting its properties and utilization methods. This study investigates the prediction of oxygenated component content in biomass pyrolysis oil using two machine learning methods: neural networks and random forests. A large dataset of biomass pyrolysis oil samples was collected and analyzed for their oxygenated component content. Neural network and random forest techniques were used for model training and validation, and the dataset was split into training and testing sets (90 %) and (10 %), respectively. The experimental results indicate that both algorithms can accurately predict the oxygenated component content in biomass pyrolysis oil (R² > 0.81, RMSE <3.46). Additionally, the models' performance was assessed and contrasted, providing effective methods and references for predicting the oxygenated component content in biomass pyrolysis oil.