Machine learning in clarifying complex relationships: Biochar preparation procedures and capacitance characteristics

Abstract

Supercapacitors as a new type of energy storage device play an important role in the development of renewable energy. Activated biochar-based electrode materials has good energy storage potential because of useful physicochemical properties and renewability. The optimal preparation method for biochar-based electrodes for supercapacitors is difficult to determine experimentally due to the complexity of influencing factors. Therefore, three machine learning models were developed to describe the relationship between the preparation process and the energy storage characteristics of activated biochar. The Gradient Boosting Regression (GBR) model provided the best prediction, with an R2 value of 0.93. Interestingly, the heating rate was the most important factor affecting the specific capacitance of activated biochar and showed a significant negative correlation, with a feature importance of 35.08 %. In addition, Micropore volume proportion and specific surface area showed a significant positive correlation with specific capacitance, with feature importance of 15.39 % and 9.23 %, respectively. Significantly, the specific capacitance could be improved by increasing the added amount of N source and activator and a higher activation temperature, as well as using a shorter activation time and slower heating rate during preparation. This study provides new insights into the applications of activated biochar for energy storage through data analysis.