New Material Design Of Vanadium-based Hydrogen Storage Alloys Based On Machine Learning

Hydrogen energy is widely recognized as a promising alternative to traditional fossil fuels.However,the widespread implementation of hydrogen energy requires advancements in hydrogen storage materials.Vanadium-based hydrogen storage alloys are especially promising due to their high theoretical hydrogen storage capacity and favorable thermodynamic and kinetic properties.Among these alloys,the V-Ti-Cr-Fe alloy stands out for its low cost and structural stability,but it has not achieved the expected performance only by means of experimental research.To address this issue,this paper aims to establish a composition-properties relationship and quantitative prediction model for the hydrogen storage properties of the V-Ti-Cr-Fe alloy using machine learning algorithms and first-principle calculation methods.Intelligent component design and experimental verification were also conducted to validate the accuracy of the machine learning model.Additionally,the research looks into the hydrogen storage mechanism of the alloy at a microstructure level to obtain the model interpretability analysis.Based on a large number of literature and experimental data collection,this paper establishes a quantitative prediction model for the maximum hydrogen storage capacity and hydrogen release enthalpy of V-Ti-Cr-Fe hydrogen storage alloy.After feature selection,this paper adopts various features such as element composition,atomic size,electronegativity,lattice constant,and elastic modulus.The prediction models of the maximum hydrogen storage capacity and the dehydrogenation enthalpy are constructed by using linear models,support vector machine,tree model and ensemble learning.The results show that reasonable feature selection plays a positive role in improving the accuracy of prediction of hydrogen release enthalpy,and the application of ensemble learning algorithm effectively reduces the prediction error of the maximum hydrogen storage capacity model.The prediction model of maximum hydrogen storage and hydrogen release enthalpy has achieved a relative error of less than 15%on average and a correlation coefficient of more than 0.6.Based on the ranking of features importance based on the gini impurity,it is found that the important features affecting the maximum hydrogen storage capacity of the alloy include atomic size,lattice size and valence electron concentration,and the factors that have an important impact on the enthalpy of hydrogen release are bulk modulus,electronegativity and lattice constant.Based on the quantitative prediction model of the maximum hydrogen storage capacity and hydrogen release enthalpy of the alloy,this paper further uses the genetic algorithm to optimize the composition of V-Ti-Cr-Fe alloy,with good comprehensive hydrogen storage performance.In this paper,V-Ti-Cr-Fe alloy was prepared according to the results of composition optimization,and the PCT test was carried out.The experimental verification results were in good agreement for the calculated results with the relative errors between calculated values and experimental values are less than 1.5%and 5%respectively.Last but not the least,hydrogen storage mechanism of V-Ti-Cr-Fe alloy are discussed as to the important factors affecting the hydrogen storage performance using the first principle calculation.Using the special quasi-random structure method,the lattice cell of V-Ti-Cr-Fe alloy is built which has a lattice constant close to the theoretical value with relative error of 0.4%.The variation trend of binding energy and crystal structure parameters was calculated using first principle calculation.The results show that the gap effect is obvious in hydrogen storage alloys,that is,larger gaps volume is more conducive to the occupation of hydrogen atoms.And larger electronegativity difference of metals and hydrogen benefits to the bonding between hydrogen atoms and metal atoms.The greater the modulus of the alloy,the greater the strain energy corresponding to the lattice distortion caused by hydrogen absorption and desorption,which means it is harder for the release of hydrogen atoms.Machine learning and first principles all show that volume,valence electron and modulus are the determinate structural factors that affect the hydrogen storage performance of hydrogen storage alloys.