Product prediction of fixed-bed coal pyrolysis using a fusion model

Abstract

Here, a proactively optimized fusion model (FM) for predicting the product yield of coal pyrolysis was developed. Eight coal characteristics (including pyrolysis temperature and proximate and ultimate analyses) were chosen as input parameters. Multiple linear regression (MLR), support vector machine (SVM), and random forest (RF) models were applied as the base models to form the FM. Sixty sets of experimental data from the literature were used for training and testing the base models. Different learning weights are assigned to the base models according to their predictive performance. The FM proactively improve the model outputs by means of the dynamical learning weight results. The coefficient of determination (R2) and the root-mean-squared error (RMSE) derived from the FM model were better than those of the base models. Moreover, the maximum relative error between the experimental data and model outputs was just 0.37%. These results suggest that FMs can be used to develop better predictive models for the yields of co-pyrolysis products. The FM proactively optimized the outputs base on learning weight algorithm and had better predicted performance than base models with less data.