Data driven studies on Heating value of biomass and next generation carbon biobased Materials for energy storage

Abstract

Energy production and its storage has been the main thrust area for many scientific newlineresearchers globally. High power density supercapacitor has emerged as the ultimate newlinesolution for energy storage. Literature reports numerous computational attempts to newlinesynthesize ideal electrode material that improves the electrochemical performance of newlinesupercapacitors. Today, machine learning (ML) models have successfully evolved as newlinea solution to a number of chemical engineering challenges. Bulks of researchers have newlinegathered small dataset based on aqueous electrolyte to forecast the electrochemical performance newlineof supercapacitors. Moreover, some crucial electrolyte, structural and compositional newlinefactors are not thoroughly studied. Therefore, electrochemical performance newlineprediction requires careful selection of electrode material and electrolyte parameters newlinealong with optimization of machine learning models. newlineThis thesis deals with the prediction of specific capacitance based on literature newlinedriven data using six different physics-informed ML models namely: Linear Regression newline(LR), Random Tree (RT), Multilayer Perceptron (MLP), Support Vector Regression newline(SVR), Random Forest (RF) and Artificial Neural Network (ANN). Different electrolyte newlinetypes and its parameters along with structural and compositional features are newlineselected as input features. For determining the degree of correlation, Pearson Correlation newlineCoefficient (PCC) matrix was examined. ANN and RF offered lowest root mean newlinesquare (RMSE - 0.98 and 1.24 respectively) and highest correlation coefficient (R2 - newline0.852 and 0.851 respectively) amongst all the models taken into consideration. Further, newlineenergy density is predicted using RF and ANN model. In order to establish correlation newlinebetween surface area and specific capacitance, threshold pore size was also established. newlineIn recent years, materials derived from biomass have shown promise as sustainable newlineprecursors for the production of carbon nanomaterials for energy storage. Therefore, newlinefive ML models are designed to study the impact of surface c