Machine learning models for prediction of electrochemical properties in supercapacitor electrodes using MXene and graphene nanoplatelets

Herein, machine learning (ML) models using multiple linear regression (MLR), support vector regression (SVR), random forest (RF) and artificial neural network (ANN) are developed and compared to predict the output features viz. specific capacitance (Csp), electrical conductivity (σ) and sheet resistance (Rs) for MXene/Graphene Nanoplatelets (GNPs) based energy storage devices. These output features are modeled as a function of wt.% in different weight ratios of GNPs in MXene, optimum potential window and scan rates. The datasets are obtained by the real time output measurements through the experimental runs of these composites in different weight ratios. Among these models, ANN had achieved the highest performance followed by MLR, SVR and RF. From both the experimental and ANN results, the electrode with 20 wt% of GNPs in MXene (MG-80) exhibited the highest Csp of 226.6F/g at 5 mV/s with a long cycle life having 84.2 % retention even after 5000 cycles of charging-discharging. ANN model is further utilized to predict the cyclic stability of MG-80 electrode upto 10,000 cycles and the results show the accuracy of the ML model to fabricate storage devices. Furthermore, the structure of MXene/GNPs composites are investigated by different characterization techniques. XRD spectra confirmed the successful synthesis of MXene and the successful intercalation of GNPs into MXene sheets. The morphology of the embedded GNPs in layered MXenes are determined through FE-SEM/EDX and HR-TEM analysis. The increase in the surface area and pore volume in the MXene/GNPs composite are revealed by BET measurements. XPS is utilized to find out the chemical element states of the bare MXene as well as its composite with GNPs.