Prediction of combustion activation energy of NaOH/KOH catalyzed straw pyrolytic carbon based on machine learning

Wheat straw, corn straw and sorghum straw were used as raw materials. KOH and NaOH were used as catalysts to prepare straw pyrolytic carbon (SPC) and the characteristics of combustion activation energy (AE) were analyzed by thermogravimetric analysis. The distributed modified Coats -Redfern integration method was used to compute the distributed AE. The predictive models of combustion AE based on Linear Regression (ER), Support Vector Regression (SVR) and Random Forest Regression (RFR) were proposed and compared. The results showed the AE variation trend of three kinds of SPCNaOH, SPCKOH and SPCNa-KOH were basically the same and obviously decreased. In the LR model, degree value was 2 and R-2 reached 0.8531. In the SVR model, the kernel function was Polynomial, C=3000, degree = 4, coef0 = 0.3 and R-2 reached 0.9048. In the RFR model, the n_estimators value was 400 and R-2 reached 0.9834. Compared with the LR and SVR model, the RFR model was more suitable for the AE prediction of alkali catalyzed SPC. (C) 2018 Elsevier Ltd. All rights reserved.