Enhanced supercapacitive performance of lead vanadate hybrid architect on nickel foam with machine learning-driven capacitance prediction

Abstract

The structural and chemical architecture of the active electrode material must provide an optimum environment for charge storage as well as stability to provide crucial supercapacitance properties. In this study, we developed a binder-free protocol to obtain an optimized polymorph of lead vanadate (PbV) with a unique hybrid architecture on nickel foam for hybrid supercapacitor applications. Furthermore, a detailed capacitance power prediction model is proposed based on several machine learning and recurrent neural network methods to improve the reliability of the fabricated hybrid supercapacitor. Specifically, recursive feature elimination is implemented to alleviate the problems of overfitting and the curse of dimensionality by identifying and eliminating extraneous characteristics while determining the appropriate features of the capacitance power. The experimental observations are validated by DFT analysis concerning the role of oxygen vacancies for the improvement of electrical characteristics. Ultimately, the fabricated energy storage device exhibits impressive energy and power densities of 26.18 Whkg−1 and 6−1, respectively. Importantly, the predicted results confirm that the long-short-term memory network provides excellent results for capacitance prediction of hybrid supercapacitors with R2 = 0.83 in the test set and synthesized sample prediction with a mean absolute error of 20.37.