Fourteen multivariate regression models were applied to model the bio-oil yield obtained by pyrolysis using different combinations of predictor variables. The data modeling was separated into the reactor regime: batch and continuous. For batch reactor, the Cubist model with the radial base function provided the best bio-oil prediction result with RMSEP of 0.92%, R² of 0.99, and MAE of 0.73%. This better result was obtained using the process’s modeling variables, proximate composition, elemental composition, and lignocellulose biomass concentration. For continuous reactor, the best result was obtained with the Extremely Randomized Tree model applied to the complete set of predictors with RMSEP of 2.15%, R² of 0.96, and MAE of 1.74%. Both models showed an outstanding performance for bio-oil yield prediction for batch and continuous reactors widely used in the chemical industry. The optimization analysis of the models showed that the batch reactor achieves a bio-oil yield as high as the fluidized bed reactor if operated under the right conditions. The PSO method used for the optimization found the global optimum for the defined analysis ranges.

应用十四个多元回归模型来使用预测变量的不同组合对热解获得的生物油产量进行建模。数据建模分为反应器状态：间歇式和连续式。对于间歇式反应器，具有径向基函数的Cubist模型提供了最好的生物油预测结果，RMSEP为0.92%，R ²为0.99，MAE为0.73%。这个更好的结果是使用过程的建模变量、近似成分、元素成分和木质纤维素生物质浓度获得的。对于连续反应器，将极端随机树模型应用于完整的预测变量集，获得了最佳结果，RMSEP 为 2.15%，R ²为 0.96，MAE 为 1.74%。这两个模型在化学工业广泛使用的间歇式和连续式反应器的生物油产量预测方面都表现出了出色的性能。模型的优化分析表明，如果在适当的条件下运行，间歇式反应器可以实现与流化床反应器一样高的生物油产率。用于优化的 PSO 方法找到了定义的分析范围的全局最优值。