Data Classification in High Throughput Screening

Abstract

Redox flow batteries offer an economical, low vulnerability means to store electrical energy at grid scale. Most commercial flow batteries use concentrated sulphuric acid with vanadium as the active redox material. While they have many inherent benefits, the use of vanadium drives up their cost. Deep eutectic solvents (DESs) are promising electrolyte solvents for less expensive and more sustainable organic redox species due to their biodegradability, low cost, high solvent quality and molecular versatility. High throughput screening (HTS) was used to synthesize and measure thermal and electrochemical properties of DESs to rapidly find an optimum DES for use in redox flow batteries. As a vast amount of data with varying characteristics based on the samples being tested is produced, data classification is essential to automate data analysis. Thus, convolutional neural networks and long short term memory networks were successfully used to classify HTS data into predefined categories and eliminate noisy data. Suitable data analysis techniques were then developed for each category of classified data to obtain accurate results.