DeepBET: Estimating the Surface Area of Plant Carbon using SEM Image

The importance of alternative and clean energy sources increases as the world faces global warming and energy shortages. Renewable energy sources like solar and wind energy require storage devices to store energy without sunlight or wind. Supercapacitors are high-capacity electrical charge storage devices with a higher power density and are safe for users. Carbon from natural sources is an exciting material for producing electrical charge storage because of its good electrical properties and natural resource conservation. However, a high specific surface area, Brunauer-Emmett-Teller (BET) surface area, is necessary for practical electrical charge storage. Standard-specific surface area calculation requires high resources such as time and cost using the conventional BET method. This research presents a machine learning model developed for estimating the BET surface area of carbon from plants using SEM images through a deep learning model, DeepBET. DeepBET predicts the BET surface area value with 76% accuracy, reducing the time and cost of calculating the specific surface area. This research explored the possibility to train a computer vision model through scientific publication databases.

I. Introduction

Recently, the popularity of electric vehicles (EVs) has grown due to their efficiency and concerns about global warming. While the battery is the vehicle's main electricity storage, supercapacitors have drawn significant attention from many sectors due to their high-power density and fast charging capabilities. Thus, discovering new supercapacitor materials has become one of physics and material science's most popular research topics.