Kornnum Sukanlaya, Chomkhuntod Praeploy, Schwaiger Nick, Limcharoen Kanwara, Deshsorn Krittapong, Jitapunkul Kulpavee, Iamprasertkun Pawin
School of Bio-Chemical Engineering and Technology, Sirindhorn International Institute of Technology, Thammasat University, Pathum Thani 12120, Thailand.
Research Unit in Sustainable Electrochemical Intelligent, Thammasat University, Pathum Thani 12120, Thailand.
Anal Chem. 2025 Feb 25;97(7):3881-3891. doi: 10.1021/acs.analchem.4c04764. Epub 2025 Jan 31.
Cyclic voltammetry (CV) is a standard method for assessing electrochemical properties in the electrochemical cells, typically in conventional aqueous contexts like 1 solutions ("salt-in-water"). However, recent advancements have extended electrochemistry into superconcentrated regimes, such as "water-in-salt" solutions with concentrations above 10 to 20 , which require large amounts of salt for experiments. To address this, machine learning (ML) has been applied, coupled with in-house data collection using lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) electrolytes. This work demonstrates the electrochemistry of YEC-8B in LiTFSI, given their broad potential window of up to 3.0 V across concentrations from 1 to 20 . The CV profiles were divided into two models: the upper curve for charging and the lower curve for discharging. Data were normalized and segmented by percentiles, and a decision tree model was developed to predict outputs based on input parameters like LiTFSI concentration, scan rates, and potential window. The model predicted nine target variables with a mean absolute percentage error of approximately 2% for both the upper and the lower CV profile curves. Trapezoidal rule was then used to calculate the system's capacitance. Additionally, tests showed a 75% accuracy in predicting the potential window and a suitable scan rate. Overall, the model effectively demonstrated the relationship between "water-in-salt" electrolytes and CV profiles in an electrochemical context using a simple machine learning (ML) algorithm, which continues to expand the integration of data science and electrochemistry.
循环伏安法（CV）是评估电化学电池中电化学性质的标准方法，通常用于传统的水性环境，如1 溶液（“水包盐”）。然而，最近的进展已将电化学扩展到超浓缩体系，例如浓度高于10至20 的“盐包水”溶液，这种溶液需要大量盐来进行实验。为了解决这个问题，机器学习（ML）已被应用，并结合使用双（三氟甲磺酰）亚胺锂（LiTFSI）电解质进行内部数据收集。这项工作展示了YEC - 8B在LiTFSI中的电化学性质，因为它们在1至20 的浓度范围内具有高达3.0 V的宽电位窗口。CV曲线分为两个模型：上部曲线用于充电，下部曲线用于放电。数据通过百分位数进行归一化和分段，并开发了一个决策树模型，根据LiTFSI浓度、扫描速率和电位窗口等输入参数预测输出。该模型预测了九个目标变量，上部和下部CV曲线的平均绝对百分比误差约为2%。然后使用梯形法则计算系统的电容。此外，测试表明在预测电位窗口和合适的扫描速率方面准确率为75%。总体而言，该模型使用简单的机器学习（ML）算法有效地展示了“盐包水”电解质与电化学环境中CV曲线之间的关系，这继续扩展了数据科学与电化学的整合。
