Thermo-electro-rheological properties of graphene oxide and MXene hybrid nanofluid for vanadium redox flow battery: Application of explainable ensemble machine learning with hyperparameter optimization
Praveen Kumar K , K. Deepthi Jayan , Prabhakar Sharma , Mansoor Alruqi
{"title":"Thermo-electro-rheological properties of graphene oxide and MXene hybrid nanofluid for vanadium redox flow battery: Application of explainable ensemble machine learning with hyperparameter optimization","authors":"Praveen Kumar K , K. Deepthi Jayan , Prabhakar Sharma , Mansoor Alruqi","doi":"10.1016/j.flatc.2023.100606","DOIUrl":null,"url":null,"abstract":"<div><p>Recent research has extensively focused on 2D materials such as graphene oxide (GO) and MXene due to their intriguing properties, significantly advancing nanotechnology and materials research. This experimental study explores the use of a vanadium electrolyte-based hybrid nanofluid (HNF) composed of GO and MXene (90:10) to enhance vanadium redox flow batteries (VRFBs). The synthesis and characterization of GO and Mxene nanoparticles (NPs) were conducted using various techniques. The HNF, produced at different weight concentrations, underwent analysis for stability, rheology, thermal conductivity (TC), and electrical conductivity (EC) within a temperature range of 10–45 °C. The results indicate that the HNF exhibits favorable stability and Newtonian behavior in the specified temperature range. At 45 °C, the HNF achieves a maximum enhancement of 20.5 % in EC and 6.81 % in TC for 0.1 wt% compared to the vanadium electrolyte. Subsequently, a prognostic model was developed using an explainable ensemble LSBoost-based machine learning approach, employing a test dataset and applying 5-fold cross-validation to prevent overfitting. Hyperparameter optimization was achieved using the Bayesian technique. The LSBoost-based prognostic models created for TC, EC, and viscosity (VST) demonstrated high effectiveness, with R<sup>2</sup> values of 0.9981, 0.99, and 0.9954, respectively. The prediction errors were minimal, with RMSE values of 0.00089255, 5.553, and 0.09391 for the TC, EC, and VST models, respectively. Similarly, the MAE values were low, at 0.00068948, 4.0919, and 0.06129.</p></div>","PeriodicalId":316,"journal":{"name":"FlatChem","volume":null,"pages":null},"PeriodicalIF":5.9000,"publicationDate":"2024-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"FlatChem","FirstCategoryId":"88","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2452262723001381","RegionNum":3,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"CHEMISTRY, PHYSICAL","Score":null,"Total":0}
引用次数: 0
Abstract
Recent research has extensively focused on 2D materials such as graphene oxide (GO) and MXene due to their intriguing properties, significantly advancing nanotechnology and materials research. This experimental study explores the use of a vanadium electrolyte-based hybrid nanofluid (HNF) composed of GO and MXene (90:10) to enhance vanadium redox flow batteries (VRFBs). The synthesis and characterization of GO and Mxene nanoparticles (NPs) were conducted using various techniques. The HNF, produced at different weight concentrations, underwent analysis for stability, rheology, thermal conductivity (TC), and electrical conductivity (EC) within a temperature range of 10–45 °C. The results indicate that the HNF exhibits favorable stability and Newtonian behavior in the specified temperature range. At 45 °C, the HNF achieves a maximum enhancement of 20.5 % in EC and 6.81 % in TC for 0.1 wt% compared to the vanadium electrolyte. Subsequently, a prognostic model was developed using an explainable ensemble LSBoost-based machine learning approach, employing a test dataset and applying 5-fold cross-validation to prevent overfitting. Hyperparameter optimization was achieved using the Bayesian technique. The LSBoost-based prognostic models created for TC, EC, and viscosity (VST) demonstrated high effectiveness, with R2 values of 0.9981, 0.99, and 0.9954, respectively. The prediction errors were minimal, with RMSE values of 0.00089255, 5.553, and 0.09391 for the TC, EC, and VST models, respectively. Similarly, the MAE values were low, at 0.00068948, 4.0919, and 0.06129.
期刊介绍:
FlatChem - Chemistry of Flat Materials, a new voice in the community, publishes original and significant, cutting-edge research related to the chemistry of graphene and related 2D & layered materials. The overall aim of the journal is to combine the chemistry and applications of these materials, where the submission of communications, full papers, and concepts should contain chemistry in a materials context, which can be both experimental and/or theoretical. In addition to original research articles, FlatChem also offers reviews, minireviews, highlights and perspectives on the future of this research area with the scientific leaders in fields related to Flat Materials. Topics of interest include, but are not limited to, the following: -Design, synthesis, applications and investigation of graphene, graphene related materials and other 2D & layered materials (for example Silicene, Germanene, Phosphorene, MXenes, Boron nitride, Transition metal dichalcogenides) -Characterization of these materials using all forms of spectroscopy and microscopy techniques -Chemical modification or functionalization and dispersion of these materials, as well as interactions with other materials -Exploring the surface chemistry of these materials for applications in: Sensors or detectors in electrochemical/Lab on a Chip devices, Composite materials, Membranes, Environment technology, Catalysis for energy storage and conversion (for example fuel cells, supercapacitors, batteries, hydrogen storage), Biomedical technology (drug delivery, biosensing, bioimaging)