CN112347697A - Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning - Google Patents

Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning Download PDF

Info

Publication number
CN112347697A
CN112347697A CN202011248936.8A CN202011248936A CN112347697A CN 112347697 A CN112347697 A CN 112347697A CN 202011248936 A CN202011248936 A CN 202011248936A CN 112347697 A CN112347697 A CN 112347697A
Authority
CN
China
Prior art keywords
neural network
polysulfide
deep neural
dimensional layered
carrier material
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011248936.8A
Other languages
Chinese (zh)
Inventor
李金金
汪志龙
张海阔
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Jiaotong University
Original Assignee
Shanghai Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Jiaotong University filed Critical Shanghai Jiaotong University
Priority to CN202011248936.8A priority Critical patent/CN112347697A/en
Publication of CN112347697A publication Critical patent/CN112347697A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • H—ELECTRICITY
    • H01—ELECTRIC ELEMENTS
    • H01M—PROCESSES OR MEANS, e.g. BATTERIES, FOR THE DIRECT CONVERSION OF CHEMICAL ENERGY INTO ELECTRICAL ENERGY
    • H01M10/00—Secondary cells; Manufacture thereof
    • H01M10/05—Accumulators with non-aqueous electrolyte
    • H01M10/052—Li-accumulators
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02E60/10—Energy storage using batteries

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Strategic Management (AREA)
  • Software Systems (AREA)
  • Economics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Computational Linguistics (AREA)
  • Development Economics (AREA)
  • Evolutionary Biology (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Game Theory and Decision Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Manufacturing & Machinery (AREA)
  • Chemical & Material Sciences (AREA)
  • Chemical Kinetics & Catalysis (AREA)
  • Electrochemistry (AREA)
  • General Chemical & Material Sciences (AREA)
  • Medical Informatics (AREA)

Abstract

The invention provides a method and a system for screening an optimal carrier material in a lithium-sulfur battery based on machine learning, which comprises the following steps: to AB2Optimizing the structure of the two-dimensional layered carrier material; setting a plurality of polysulfides, calculating adsorption energy by using a density functional theory, and constructing a data set of an adsorption structure corresponding to the adsorption energy; performing atomic local chemical environment calculation on the structure in the data set to obtain the structural characteristics in the data set; training the deep neural network based on the structural features in the data set, and predicting any configuration according to the trained deep neural networkAdsorption energy of any site; using a transfer learning algorithm to align the preset AB2Correspondingly adjusting the parameters of the deep neural network for adsorbing polysulfide by the type two-dimensional layered carrier material; for preset AB2Training a deep neural network for polysulfide adsorption by the type two-dimensional layered carrier material, predicting the adsorption energy of any configuration and site according to the trained deep neural network, and measuring the adsorption capacity according to the average value.

Description

Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning
Technical Field
The invention relates to the technical field of lithium-sulfur batteries, in particular to a method and a system for screening an optimal carrier material in a lithium-sulfur battery based on machine learning.
Background
Lithium sulfur batteries have been extensively studied in recent years as an energy storage system with high theoretical energy density (2600Wh kg-1) and capacity density (1675mAh g-1). However, the positive electrode sulfur has many problems, such as shuttle effect of discharge products, soluble polysulfide generated by discharge can migrate to the negative electrode through electrolyte and react with metal lithium, which leads to low coulombic efficiency and low cycle life of the lithium-sulfur battery, and further development and commercial application of the lithium-sulfur battery are greatly restricted, so that the shuttle effect can be obviously improved by selecting a material with high adsorption capacity to polysulfide as a carrier to inhibit the shuttle effect. However, in the face of many potential lithium-sulfur battery carrier materials, the research cannot depend on complicated experimental conditions. Although the adsorption energy of the support material to polysulfide can be obtained by means of the calculation of DFT theory, the calculation resource is expensive, the time cost is high, and the possible adsorption configuration and adsorption sites of polysulfide on the support material cannot be comprehensively calculated. It is therefore necessary to find a method for rapidly and comprehensively evaluating the adsorption energy of a support material for polysulfides.
The existing research technology is mainly divided into experimental measurement and theoretical calculation. In order to find the optimal lithium sulfur battery carrier material, the conventional experimental method is to synthesize various possible materials to serve as the positive electrode carrier of the lithium sulfur battery, assemble the lithium sulfur battery by loading active sulfur, and perform a series of electrochemical performance tests to investigate the influence of the carrier material on the improvement of the performance of the lithium sulfur battery. The technology can indirectly evaluate the carrier material, and has the defects of long experimental period, high cost, complex synthesis conditions of some carrier materials, high requirements on experimental equipment and difficult operation. The other technical means of theoretical calculation is also widely applied to screening of lithium sulfur battery carrier materials, the adsorption energy of the carrier materials to polysulfide is investigated by establishing a theoretical crystal model of the carrier materials and putting a polysulfide molecular model on a specific crystal face of the carrier materials for optimization calculation, the capability of the carrier materials for inhibiting the shuttle effect can be investigated on a theoretical layer, the technology is also the closest prior art to the invention, and the suitable lithium sulfur battery carrier materials can be searched through theoretical calculation before expensive experiments are found.
The technical problem to be solved by the invention is that the chemical formula of the material is AB2Type two-dimensional layered material and polysulfide (Li)2S4,Li2S6,Li2S8) The adsorption energy can be predicted, and then the optimal adsorption carrier material in the lithium-sulfur battery can be rapidly screened. In the experimental research of lithium sulfur batteries, the adsorption capacity of a carrier material to a discharged polysulfide of the carrier material cannot be directly observed and evaluated, and only tedious experimental steps are required to try to test the influence of different materials on the electrochemical performance of the carrier material so as to screen a proper sulfur carrier. In the aspect of theoretical calculation, researchers often investigate the adsorption energy of a carrier material on polysulfide by a Density Functional Theory (DFT) method at a theoretical level, but the traditional DFT method is time-consuming and expensive in calculation resources, and cannot perform calculation screening on a large amount of carrier materials. In addition, the DFT method cannot calculate all possible adsorption sites on the carrier material, so that it is difficult to evaluate the overall adsorption capacity, which has certain limitations.
The invention uses AB2The two-dimensional layered lithium-sulfur battery carrier material is a research object, on the basis of DFT theoretical calculation, the adsorption energy of any polysulfide adsorption configuration at different sites of the carrier is investigated, and AB is established by combining a deep neural network2The model is a high-precision and high-efficiency prediction model of polysulfide adsorption energy by using a two-dimensional layered material, and migration learning is introduced into the prediction of the adsorption energy to obtain any other AB2And (3) predicting the adsorption of the two-dimensional layered material to polysulfide. Comprehensive evaluation AB combined with big data statistical analysis2Adsorption capacity of the type two-dimensional layered material to polysulfide. Therefore, the high cost of experiment and theoretical calculation is greatly reduced, and the materials are rapidly and accurately screened. The invention is not only suitable for lithium sulfurThe screening of the carrier material in the battery is also suitable for the application field of other materials related to molecular adsorption.
Drawbacks of the prior art and objects of the present invention:
(1) the experiment cost is high. The synthesis of specific carrier materials of the prior art requires expensive laboratory drugs and instruments.
(2) The operation difficulty is high. The experimental synthesis conditions are complex, multi-step chemical synthesis is often needed, and the operation difficulty of some experimental instruments is higher.
(3) The experimental period is long. Experimental material synthesis requires a long time and requires assembly into a lithium sulfur battery for electrochemical testing, requiring a long test cycle.
(4) Not all configurations and sites can be investigated. Theoretical calculations cannot investigate various adsorption configurations of polysulfides and arbitrary sites on the support.
(5) Theoretical study of adsorption energy requires expensive computational resources and is time-consuming and costly.
Disclosure of Invention
In view of the deficiencies in the prior art, it is an object of the present invention to provide a method and system for machine learning based screening of optimal support materials in lithium sulfur batteries.
The invention provides a method for screening an optimal carrier material in a lithium-sulfur battery based on machine learning, which comprises the following steps:
step M1: AB pair by adopting density functional theory method2Optimizing the structure of the two-dimensional layered carrier material;
step M2: providing a plurality of polysulfides, and placing each polysulfide in the optimized AB2Different adsorption sites of the two-dimensional layered carrier material are formed, the spatial configuration of polysulfide is simultaneously changed, the adsorption energy is calculated by utilizing a density functional theory, and a data set corresponding to the adsorption structure and the adsorption energy is constructed;
step M3: performing atomic local chemical environment calculation on the structure in the data set to obtain the structural characteristics in the data set;
step M4: construction of AB2Deep neural net with polysulfide adsorbed by two-dimensional layered carrier materialTraining the deep neural network based on the structural features in the data set to obtain the trained deep neural network, and predicting the adsorption energy of any configuration and any site according to the trained deep neural network;
step M5: initializing a preset AB by using a transfer learning algorithm and using trained deep neural network model parameters2Adsorbing polysulfide deep neural network parameters by using a two-dimensional layered carrier material, and aiming at ensuring that the parameter oscillation amplitude reaches a preset value, carrying out on a preset AB2Correspondingly adjusting the parameters of the deep neural network for adsorbing polysulfide by the type two-dimensional layered carrier material;
step M6: using AB2The adsorption energy of a preset amount obtained by the theoretical calculation of the density functional of polysulfide adsorbed by the type two-dimensional layered carrier material is used for the preset AB2Training a deep neural network of polysulfide adsorbed by a type two-dimensional layered carrier material to obtain a trained preset AB2The type two-dimensional layered carrier material adsorbs a deep neural network of polysulfide;
step M7: according to preset AB after training2The deep neural network of polysulfide adsorbed by the type two-dimensional layered carrier material is used for further predicting the adsorption energy of any configuration and any site, and AB is calculated and analyzed based on statistics2Shaped two-dimensional layered support material and pre-defined AB2The adsorption capacity of the type two-dimensional layered carrier material is measured according to the minimum value, the maximum value, the standard deviation and the average value of the type two-dimensional layered carrier material for three polysulfides.
Preferably, the step M1 includes applying the method of PBE functional to AB2The structure of the two-dimensional layered material is optimized.
Preferably, the AB is2The deep neural network for adsorbing polysulfide by the two-dimensional layered material comprises preset layer hidden layers, wherein each hidden layer comprises a preset number of neurons.
Preferably, the step M2 includes:
the polysulfides contain n atoms, each atom having the spatial coordinate (x)i,yi,zi) The central atomic coordinate is
Figure BDA0002770955280000031
When any configuration is to be created, moving to any site, the polysulfide rotation operation involves: x'i＝xi-cx，y′i＝yi-cy，z′i＝zi-cz；
Figure BDA0002770955280000041
The molecule was then moved to the original point, x'i＝x″i+cx，y″′i＝y″i+cy，z″′i＝z″i+cz
Wherein γ represents an angle of rotation about the Z axis, β represents an angle of rotation about the Y axis, and α represents an angle of rotation about the Z axis;
calculating adsorption energy by using a density functional theory to construct a machine learning data set, wherein the data set is a structure and corresponding energy; calculating a formula according to the adsorption energy:
ΔEb＝Et-Es-Ep
wherein Et is the energy of the carrier material and the polysulfide as a whole, Es is the energy of the carrier, and Ep is the energy of the polysulfide.
Preferably, the step M3 includes:
local chemical environment G of ith atom and its neighbor atom jijExpressed as:
Figure BDA0002770955280000042
Figure BDA0002770955280000043
Figure BDA0002770955280000044
e(x)＝x/||x||
wherein x isijDenotes the difference between the X coordinates of the ith atom and the jth atom, a (i) and b (i) denote the two atoms closest to the ith atom, Ria denotes the difference between the coordinate of the ith atom and the coordinate vector of the a atom;
all atomic local chemical environments in a structure are combined into a large matrix to be used as the input of a deep neural network.
Preferably, the step M4 includes: using Adam optimizer, AB is aligned according to preset learning rate2Carrying out iterative training on the deep neural network of polysulfide adsorbed by the type two-dimensional layered material until the error reaches a preset value, and finishing the training to obtain the trained deep neural network;
the error includes:
Figure BDA0002770955280000045
wherein M represents the number of structures; eNN,iRepresenting the predicted energy of the neural network, EDFT,iRepresenting the energy calculated using the density functional theory.
According to the invention, the system for screening the optimal carrier material in the lithium-sulfur battery based on machine learning comprises:
module M1: AB pair by adopting density functional theory method2Optimizing the structure of the two-dimensional layered carrier material;
module M2: providing a plurality of polysulfides, and placing each polysulfide in the optimized AB2Different adsorption sites of the two-dimensional layered carrier material are formed, the spatial configuration of polysulfide is simultaneously changed, the adsorption energy is calculated by utilizing a density functional theory, and a data set corresponding to the adsorption structure and the adsorption energy is constructed;
module M3: performing atomic local chemical environment calculation on the structure in the data set to obtain the structural characteristics in the data set;
module M4: construction of AB2Type twoThe deep neural network is trained based on the structural features in the data set to obtain the trained deep neural network, and the adsorption energy of any configuration and any position point is predicted according to the trained deep neural network;
module M5: initializing a preset AB by using a transfer learning algorithm and using trained deep neural network model parameters2Adsorbing polysulfide deep neural network parameters by using a two-dimensional layered carrier material, and aiming at ensuring that the parameter oscillation amplitude reaches a preset value, carrying out on a preset AB2Correspondingly adjusting the parameters of the deep neural network for adsorbing polysulfide by the type two-dimensional layered carrier material;
module M6: using AB2The adsorption energy of a preset amount obtained by the theoretical calculation of the density functional of polysulfide adsorbed by the type two-dimensional layered carrier material is used for the preset AB2Training a deep neural network of polysulfide adsorbed by a type two-dimensional layered carrier material to obtain a trained preset AB2The type two-dimensional layered carrier material adsorbs a deep neural network of polysulfide;
module M7: according to preset AB after training2The deep neural network of polysulfide adsorbed by the type two-dimensional layered carrier material is used for further predicting the adsorption energy of any configuration and any site, and AB is calculated and analyzed based on statistics2Shaped two-dimensional layered support material and pre-defined AB2The adsorption capacity of the type two-dimensional layered carrier material is measured according to the minimum value, the maximum value, the standard deviation and the average value of the type two-dimensional layered carrier material for three polysulfides.
Preferably, the module M1 includes a method for processing AB by using PBE functional2Optimizing the structure of the two-dimensional layered material;
the AB is2The deep neural network for adsorbing polysulfide by the two-dimensional layered material comprises preset layer hidden layers, wherein each hidden layer comprises a preset number of neurons.
Preferably, said module M2 comprises:
the polysulfides contain n atoms, each atom having the spatial coordinate (x)i,yi,zi) The central atomic coordinate is
Figure BDA0002770955280000051
When any configuration is to be created, moving to any site, the polysulfide rotation operation involves: x'i＝xi-cx，y′i＝yi-cy，z′i＝zi-cz；
Figure BDA0002770955280000061
The molecule was then moved to the original point, x'i＝x″i+cx，y″′i＝y″i+cy，z″′i＝z″i+cz
Wherein γ represents an angle of rotation about the Z axis, β represents an angle of rotation about the Y axis, and α represents an angle of rotation about the Z axis;
calculating adsorption energy by using a density functional theory to construct a machine learning data set, wherein the data set is a structure and corresponding energy; calculating a formula according to the adsorption energy:
ΔEb＝Et-Es-Ep
wherein Et is the energy of the carrier material and the polysulfide as a whole, Es is the energy of the carrier, and Ep is the energy of the polysulfide.
Preferably, said module M3 comprises:
local chemical environment G of ith atom and its neighbor atom jijExpressed as:
Figure BDA0002770955280000062
Figure BDA0002770955280000063
Figure BDA0002770955280000064
e(x)＝x/||x||
wherein x isijDenotes the difference between the X coordinates of the ith atom and the jth atom, a (i) and b (i) denote the two atoms closest to the ith atom, Ria denotes the difference between the coordinate of the ith atom and the coordinate vector of the a atom;
forming a large matrix by using local chemical environments of all atoms in a structure as input of a deep neural network;
the module M4 includes: using Adam optimizer, AB is aligned according to preset learning rate2Carrying out iterative training on the deep neural network of polysulfide adsorbed by the type two-dimensional layered material until the error reaches a preset value, and finishing the training to obtain the trained deep neural network;
the error includes:
Figure BDA0002770955280000065
wherein M represents the number of structures; eNN,iRepresenting the predicted energy of the neural network, EDFT,iRepresenting the energy calculated using the density functional theory.
Compared with the prior art, the invention has the following beneficial effects:
1. the method theoretically predicts the adsorption energy between the carrier material and polysulfide, and avoids complex experimental preparation exploration;
2. according to the invention, a high-precision and high-efficiency adsorption energy prediction model is established by a machine learning method, so that the adsorption energy of polysulfide on any adsorption configuration and any adsorption site on a carrier material substrate can be predicted, and the action mechanism between the polysulfide and the carrier material can be more comprehensively analyzed;
3. the invention establishes an adsorption energy prediction model of polysulfide on any adsorption configuration and any adsorption site on a carrier material substrate, can comprehensively evaluate the adsorption energy of the carrier material on the polysulfide, and is used for searching the optimal carrier material for inhibiting the shuttle effect of the lithium-sulfur battery;
4. the method is combined with a transfer learning method, a large amount of data sets are not needed, the calculation cost is greatly reduced, and the model has strong generalization capability;
5. the invention selects the structural characteristics based on the atom local chemical environment, and the characteristics can well meet the requirements of translational invariance, rotational invariance and displacement invariance in periodic materials.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is AB calculated by DFT2A crystal structure of a two-dimensional layered material; (a) MoSe2Crystal structure of (b) WSe2The crystal structure of (a);
FIG. 2 shows AB2 type two-dimensional layered material and Li calculated by DFT2S4Adsorbed crystal structure, taking into account different adsorption configurations and adsorption sites (a) MoSe2/Li2S4Side and top views of (b) WSe2/Li2S4Side and top views of;
FIG. 3 is a machine learning model framework.
FIG. 4 is a neural network adsorption energy prediction model based on atomic local chemical environment
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
Example 1
The invention provides a method for screening an optimal carrier material in a lithium-sulfur battery based on machine learning, which comprises the following steps:
step M1: AB pair by adopting density functional theory method2Optimizing the structure of the two-dimensional layered carrier material;
step M2: providing a plurality of polysulfides, and placing each polysulfide in the optimized AB2Different adsorption sites of the two-dimensional layered carrier material are formed, the spatial configuration of polysulfide is simultaneously changed, the adsorption energy is calculated by utilizing a density functional theory, and a data set corresponding to the adsorption structure and the adsorption energy is constructed;
step M3: performing atomic local chemical environment calculation on the structure in the data set to obtain the structural characteristics in the data set;
step M4: construction of AB2The method comprises the steps that a deep neural network of polysulfide is adsorbed by a two-dimensional layered carrier material, the deep neural network is trained based on structural features in a data set to obtain the trained deep neural network, and the adsorption energy of any configuration and any site is predicted according to the trained deep neural network;
step M5: initializing a preset AB by using a transfer learning algorithm and using trained deep neural network model parameters2Adsorbing polysulfide deep neural network parameters by using a two-dimensional layered carrier material, and aiming at ensuring that the parameter oscillation amplitude reaches a preset value, carrying out on a preset AB2Correspondingly adjusting the parameters of the deep neural network for adsorbing polysulfide by the type two-dimensional layered carrier material;
step M6: using AB2The adsorption energy of a preset amount obtained by the theoretical calculation of the density functional of polysulfide adsorbed by the type two-dimensional layered carrier material is used for the preset AB2Training a deep neural network of polysulfide adsorbed by a type two-dimensional layered carrier material to obtain a trained preset AB2The type two-dimensional layered carrier material adsorbs a deep neural network of polysulfide;
step M7: according to preset AB after training2The deep neural network of polysulfide adsorbed by the type two-dimensional layered carrier material is used for further predicting the adsorption energy of any configuration and any site, and AB is calculated and analyzed based on statistics2Shaped two-dimensional layered support material and pre-defined AB2Shaped two-dimensional layered carrier materialAnd measuring the adsorption capacity of the three polysulfides according to the average value of the minimum value, the maximum value, the standard deviation and the average value of the adsorption.
Specifically, the step M1 includes applying the method of PBE functional to AB2The structure of the two-dimensional layered material is optimized.
In particular, the AB2The deep neural network for adsorbing polysulfide by the two-dimensional layered material comprises preset layer hidden layers, wherein each hidden layer comprises a preset number of neurons.
Specifically, the step M2 includes:
the polysulfides contain n atoms, each atom having the spatial coordinate (x)i,yi,zi) The central atomic coordinate is
Figure BDA0002770955280000091
When any configuration is to be created, moving to any site, the polysulfide rotation operation involves: x'i＝xi-cx，y′i＝yi-cy，z′i＝zi-cz；
Figure BDA0002770955280000092
The molecule was then moved to the original point, x'i＝x″i+cx，y″′i＝y″i+cy，z″′i＝z″i+cz
Wherein γ represents an angle of rotation about the Z axis, β represents an angle of rotation about the Y axis, and α represents an angle of rotation about the Z axis;
generating different polysulfide structures by translation and rotation;
calculating adsorption energy by using a density functional theory to construct a machine learning data set, wherein the data set is a structure and corresponding energy; calculating a formula according to the adsorption energy:
ΔEb＝Et-Es-Ep
wherein Et is the energy of the carrier material and the polysulfide as a whole, Es is the energy of the carrier, and Ep is the energy of the polysulfide.
Specifically, the step M3 includes:
local chemical environment G of ith atom and its neighbor atom jijExpressed as:
Figure BDA0002770955280000093
Figure BDA0002770955280000094
Figure BDA0002770955280000095
e(x)＝x/||x||
wherein x isijDenotes the difference between the X coordinates of the ith atom and the jth atom, a (i) and b (i) denote the two atoms closest to the ith atom, Ria denotes the difference between the coordinate of the ith atom and the coordinate vector of the a atom;
all atomic local chemical environments in a structure are combined into a large matrix to be used as the input of a deep neural network.
Specifically, the step M4 includes: using Adam optimizer, AB is aligned according to preset learning rate2Carrying out iterative training on the deep neural network of polysulfide adsorbed by the type two-dimensional layered material until the error reaches a preset value, and finishing the training to obtain the trained deep neural network;
the error includes:
Figure BDA0002770955280000101
wherein M represents the number of structures; eNN,iRepresenting the predicted energy of the neural network, EDFT,iRepresenting the energy calculated using the density functional theory.
According to the invention, the system for screening the optimal carrier material in the lithium-sulfur battery based on machine learning comprises:
module M1: AB pair by adopting density functional theory method2Optimizing the structure of the two-dimensional layered carrier material;
module M2: providing a plurality of polysulfides, and placing each polysulfide in the optimized AB2Different adsorption sites of the two-dimensional layered carrier material are formed, the spatial configuration of polysulfide is simultaneously changed, the adsorption energy is calculated by utilizing a density functional theory, and a data set corresponding to the adsorption structure and the adsorption energy is constructed;
module M3: performing atomic local chemical environment calculation on the structure in the data set to obtain the structural characteristics in the data set;
module M4: construction of AB2The method comprises the steps that a deep neural network of polysulfide is adsorbed by a two-dimensional layered carrier material, the deep neural network is trained based on structural features in a data set to obtain the trained deep neural network, and the adsorption energy of any configuration and any site is predicted according to the trained deep neural network;
module M5: initializing a preset AB by using a transfer learning algorithm and using trained deep neural network model parameters2Adsorbing polysulfide deep neural network parameters by using a two-dimensional layered carrier material, and aiming at ensuring that the parameter oscillation amplitude reaches a preset value, carrying out on a preset AB2Correspondingly adjusting the parameters of the deep neural network for adsorbing polysulfide by the type two-dimensional layered carrier material;
module M6: using AB2The adsorption energy of a preset amount obtained by the theoretical calculation of the density functional of polysulfide adsorbed by the type two-dimensional layered carrier material is used for the preset AB2Training a deep neural network of polysulfide adsorbed by a type two-dimensional layered carrier material to obtain a trained preset AB2The type two-dimensional layered carrier material adsorbs a deep neural network of polysulfide;
module M7: according to preset AB after training2Any configuration and any position obtained by further prediction of deep neural network for adsorbing polysulfide by type two-dimensional layered carrier materialAdsorption energy of the spots, based on statistical calculation and analysis of AB2Shaped two-dimensional layered support material and pre-defined AB2The adsorption capacity of the type two-dimensional layered carrier material is measured according to the minimum value, the maximum value, the standard deviation and the average value of the type two-dimensional layered carrier material for three polysulfides.
Specifically, the module M1 includes a method for processing AB by using PBE functional2Optimizing the structure of the two-dimensional layered material;
the AB is2The deep neural network for adsorbing polysulfide by the two-dimensional layered material comprises preset layer hidden layers, wherein each hidden layer comprises a preset number of neurons.
Specifically, the module M2 includes:
the polysulfides contain n atoms, each atom having the spatial coordinate (x)i,yi,zi) The central atomic coordinate is
Figure BDA0002770955280000111
When any configuration is to be created, moving to any site, the polysulfide rotation operation involves: x'i＝xi-cx，y′i＝yi-cy，z′i＝zi-cz；
Figure BDA0002770955280000112
The molecule was then moved to the original point, x'i＝x″i+cx，y″′i＝y″i+cy，z″′i＝z″i+cz
Wherein γ represents an angle of rotation about the Z axis, β represents an angle of rotation about the Y axis, and α represents an angle of rotation about the Z axis;
different polysulfide structures are generated by translation and rotation;
calculating adsorption energy by using a density functional theory to construct a machine learning data set, wherein the data set is a structure and corresponding energy; calculating a formula according to the adsorption energy:
ΔEb＝Et-Es-Ep
wherein Et is the energy of the carrier material and the polysulfide as a whole, Es is the energy of the carrier, and Ep is the energy of the polysulfide.
Specifically, the module M3 includes:
local chemical environment G of ith atom and its neighbor atom jijExpressed as:
Figure BDA0002770955280000113
Figure BDA0002770955280000114
Figure BDA0002770955280000115
e(x)＝x/||x||
wherein x isijDenotes the difference between the X coordinates of the ith atom and the jth atom, a (i) and b (i) denote the two atoms closest to the ith atom, Ria denotes the difference between the coordinate of the ith atom and the coordinate vector of the a atom;
forming a large matrix by using local chemical environments of all atoms in a structure as input of a deep neural network;
the module M4 includes: using Adam optimizer, AB is aligned according to preset learning rate2Carrying out iterative training on the deep neural network of polysulfide adsorbed by the type two-dimensional layered material until the error reaches a preset value, and finishing the training to obtain the trained deep neural network;
the error includes:
Figure BDA0002770955280000121
wherein M represents the number of structures; eNN,iRepresenting a neural netEnergy of the collaterals prediction, EDFT,iRepresenting the energy calculated using the density functional theory.
Example 2
Example 2 is a modification of example 1
The invention aims to establish a high-precision and high-efficiency adsorption energy prediction model by a machine learning method, avoid complicated and expensive experimental links, reduce the dependence on time-consuming and expensive theoretical calculation and reduce the dependence on AB2The adsorption configuration of any polysulfide on the two-dimensional layered carrier material and the adsorption energy of any site can be accurately predicted, so that the adsorption capacity of the carrier material on the polysulfide can be comprehensively considered, and then the adsorption capacity of the carrier material on the polysulfide can be evaluated through big data statistical analysis. The prediction model is suitable for other AB by further combining with a transfer learning method2The two-dimensional layered material improves the practicability of the model and further reduces the time cost.
(1) And (4) performing density functional theory calculation.
The density functional theory builds on the mathematical theorem demonstrated by Kohn and Hohenberg, which indicates that there is a one-to-one correspondence between the ground state wave function and the ground state charge density. That is, the ground state charge density uniquely determines all properties of the ground state, including wave function and energy. Instead of solving a wave function with 3N variables, we can solve schrodinger's equation by finding a charge density function with three spatial variables. This greatly simplifies the computational complexity and transforms the problem into a three-dimensional problem, which is the origin of the density functional theory. However, solving the ground state energy of schrodinger equation is very difficult because of the multibody problem, for which Kohn and Sham proposed Kohn-Sham equations in 1965 that can be used to determine the specific form of the energy functional. In which the ground state problem of a multi-particle interaction system can be transformed into a one-electron problem in an effective potential field, but the solution of this equation must be given an exchange relation. To solve this problem, the LDA (local density approximation) method expresses the exchange correlation energy approximation as a functional form residing in the charge density. This approximation is to see the electron density somewhere as a uniform electron gas. The calculation result of LDA is relatively close to the experimental result in the calculation of some non-strong correlation systems, but for some strong correlation systems, the calculation result of LDA has larger error, and the combination energy is often overestimated, and the band gap is underestimated. Therefore, GGA (generalized gradient approximation) is developed, the GGA develops the charge density in a gradient manner on the basis of LDA, the introduced charge obtains an uneven distribution form, and common GGA functional methods include PW91, PBE and the like, and can obtain a result closer to an experiment to a certain extent.
For AB2The optimization of the type two-dimensional layered material is completed by using Vienna Ab-initio Simulation Package (VASP) density functional theory calculation software. With MoSe2And WSe2Taking two-dimensional layered material as an example, firstly, the method of PBE functional is adopted to carry out alignment on AB2The structure of the two-dimensional layered material is optimized, the truncation energy is set to 520eV, a 3X 1K-point grid is adopted, and the force convergence standard is
Figure BDA0002770955280000135
The convergence criterion for the energy was 1X 10-6 eVatom-1. For accurate calculation of the adsorption energy, dipole moment corrections and van der waals force corrections are also considered.
(2) MoSe after optimization2And WSe2Three polysulfide Li are respectively placed on the surface2S4,Li2S6,Li2S8Different polysulfide adsorption configurations and adsorption sites are considered. With Li2S4For example, the molecule contains six atoms, each atom having a spatial coordinate of (x)i，yi，zi) Then its central atomic coordinate is
Figure BDA0002770955280000131
When any configuration is to be created, moved to any site, Li2S4The rotation operation of (2) is as follows.
x′i＝xi-cx，y′i＝yi-cy，z′i＝zi-cz (1)
Figure BDA0002770955280000132
The molecule was then moved to the original point, x ″'i＝x″i+cx，y″′i＝y″i+cy，z″′i＝z″i+cz (3)
The adsorption energy is calculated by using a density functional theory to construct a machine-learned data set, namely a structure and corresponding energy. Calculating a formula according to the adsorption energy:
ΔEb＝Et-Es-Ep
wherein Et is the energy of the carrier material and the polysulfide as a whole, Es is the energy of the carrier, and Ep is the energy of the polysulfide. It can be seen that the more negative the adsorption energy, the stronger the adsorption of polysulfides by the support material and vice versa. That is, the more negative the adsorption energy, the stronger the inhibition effect of the carrier material on the shuttle effect of the lithium sulfur battery, and the more suitable it is as a sulfur carrier material for the lithium sulfur battery.
Wherein for MoSe2Three polysulfides were adsorbed, and 3000-5000 adsorption energies were calculated. The 3000-5000 structures can ensure a huge machine learning data set, so that the final prediction precision is higher. For WSe2Three polysulfides are adsorbed, and only 300-500 adsorption energies are calculated, so that the calculation cost is reduced and the efficiency is improved.
FIG. 3 machine learning model framework. Prediction of adsorption energy of any configuration and any site based on data set calculated by DFT
(3) Construction of MoSe2/Li2S4,MoSe2/Li2S6,MoSe2/Li2S8The deep Neural Network (NN) of (1) to predict the adsorption energy of any configuration, any site, the data set of DFT computation is divided into training set and test set by 6: 1: using the calculation result based on the atom local chemical environment as the characteristic (making different material structures satisfy translation invariance, rotation invariance and displacement invariance), namely neural networkIs input. The calculation process is as follows:
local chemical environment G of ith atom and its neighbor atom jijExpressed as:
Figure BDA0002770955280000133
Figure BDA0002770955280000134
Figure BDA0002770955280000141
e(x)＝x/||x||
the subscripts in the formula indicate the difference between the two, such as xij indicating the difference between the X coordinates of the ith atom and the jth atom, a (i) and b (i) indicating the two atoms nearest to the ith atom, and Ria indicating the difference between the coordinate of the ith atom and the coordinate vector of the a atom.
The calculated cutoff distance between neighboring atoms is set to
Figure BDA0002770955280000142
Two atoms are at a distance of less than
Figure BDA0002770955280000143
Is considered a neighbor atom. The local chemical environment of each atom can be calculated according to the formula, the mathematical form of the local chemical environment is a vector, and the local chemical environments of all atoms form a large matrix which is used as the input of the neural network. A neural network consists of several sub-networks, with identical chemical elements corresponding to the same sub-networks.
According to the above steps, with MoSe2/Li2S4For example, the local chemical environment of all Mo, Se, Li, S atoms is first calculated to form a matrix as the input of the neural network, then the matrix is respectively corresponding to different sub-networks according to the difference of elements, and finally the matrix is classified as an outputI.e. the energy corresponding to this structure.
The neural network has a structure of 3-5 hidden layers, each layer comprises 10-50 neurons, iteration is carried out by an Adam optimizer (an optimization algorithm used for solving an optimization problem) in the training process, 32-128 are selected as batch training units, the initial learning rate is 0.002-0.003 in the training process, and namely the step length of each parameter updating is realized. Before training, the parameter values represented by the neurons in the neural network are initialized randomly, and then the parameters of the neurons are adjusted according to the output-input correspondence, namely the training process. After a neural network is trained, a model parameter is obtained, for a new structure, after a matrix of a local chemical environment is calculated, matrix operation is carried out by utilizing the matrix and the model parameter, and then the adsorption energy of any configuration and any site can be predicted, and the prediction is very fast and is six orders of magnitude faster than that of the adsorption energy calculated by a DFT theory.
(4) For WSe2/Li2S4,WSe2/Li2S6,WSe2/Li2S8The amount of DFT data of (a) is small, which is to reduce the calculation cost. Using the above MoSe using a transfer learning algorithm2/Li2S4,MoSe2/Li2S6,MoSe2/Li2S8Trained model parameters to initialize WSe2/Li2S4,WSe2/Li2S6,WSe2/Li2S8The model parameters (instead of random initialization) are subjected to parameter fine tuning, in order to ensure that the parameter oscillation amplitude is very small and achieve the purpose of fine tuning, the learning rate is adjusted to be 0.001-0.0015, and the parameters and MoSe of other training models2/Li2S4,MoSe2/Li2S6,MoSe2/Li2S8The same, such as the number of neurons, etc., is followed by training the three models.
(5) Repeating the steps 3-4, so that the predicted root mean square error RMSE of the adsorption energy is less than 0.1eV, namely the root mean square error of the predicted adsorption energy and the actual adsorption energy:
Figure BDA0002770955280000151
wherein M represents, ENN,iIs represented byDFT,iThe expression M indicates the number of structures and also the number of energies corresponding to the structures. ENN, i represents the energy predicted by the neural network, and EDFT, i represents the energy calculated using the density functional theory. The whole equation represents the root mean square error of the energy corresponding to all structures, and is also the error of the neural network.
(6) The adsorption energy of 3000-5000 corresponding structures is further predicted by using a trained model, and the minimum value, the maximum value, the standard deviation and the average value of the adsorption of the MoSe2 and WSe2 on the three polysulfides are calculated and analyzed by using statistical knowledge. The average value is stable and reliable, all energy data can be comprehensively considered, and the contained information is the most. Therefore, the average value of the energy is selected to measure the adsorption capacity.
(7) Based on the method, a prediction model of the adsorption energy of polysulfide on any adsorption configuration and any adsorption site on a carrier material substrate is established, the adsorption energy can be rapidly predicted, then the adsorption capacity of the AB2 type two-dimensional layered material to polysulfide is comprehensively evaluated according to statistical analysis, and the optimal carrier material for inhibiting the shuttle effect of the lithium-sulfur battery is searched.
Those skilled in the art will appreciate that, in addition to implementing the systems, apparatus, and various modules thereof provided by the present invention in purely computer readable program code, the same procedures can be implemented entirely by logically programming method steps such that the systems, apparatus, and various modules thereof are provided in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system, the device and the modules thereof provided by the present invention can be considered as a hardware component, and the modules included in the system, the device and the modules thereof for implementing various programs can also be considered as structures in the hardware component; modules for performing various functions may also be considered to be both software programs for performing the methods and structures within hardware components.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.

Claims (10)

1. A method for screening an optimal carrier material in a lithium-sulfur battery based on machine learning, comprising:
step M1: AB pair by adopting density functional theory method2Optimizing the structure of the two-dimensional layered carrier material;
step M2: providing a plurality of polysulfides, and placing each polysulfide in the optimized AB2Different adsorption sites of the two-dimensional layered carrier material are formed, the spatial configuration of polysulfide is simultaneously changed, the adsorption energy is calculated by utilizing a density functional theory, and a data set corresponding to the adsorption structure and the adsorption energy is constructed;
step M3: performing atomic local chemical environment calculation on the structure in the data set to obtain the structural characteristics in the data set;
step M4: construction of AB2The method comprises the steps that a deep neural network of polysulfide is adsorbed by a two-dimensional layered carrier material, the deep neural network is trained based on structural features in a data set to obtain the trained deep neural network, and the adsorption energy of any configuration and any site is predicted according to the trained deep neural network;
step M5: initializing a preset AB by using a transfer learning algorithm and using trained deep neural network model parameters2Adsorbing polysulfide deep neural network parameters by using a two-dimensional layered carrier material, and aiming at ensuring that the parameter oscillation amplitude reaches a preset value, carrying out on a preset AB2Correspondingly adjusting the parameters of the deep neural network for adsorbing polysulfide by the type two-dimensional layered carrier material;
step M6: using AB2The density functional theory of polysulfide adsorbed by the type two-dimensional layered carrier material is calculatedA predetermined amount of adsorption energy of, for a predetermined AB2Training a deep neural network of polysulfide adsorbed by a type two-dimensional layered carrier material to obtain a trained preset AB2The type two-dimensional layered carrier material adsorbs a deep neural network of polysulfide;
step M7: according to preset AB after training2The deep neural network of polysulfide adsorbed by the type two-dimensional layered carrier material is used for further predicting the adsorption energy of any configuration and any site, and AB is calculated and analyzed based on statistics2Shaped two-dimensional layered support material and pre-defined AB2The adsorption capacity of the type two-dimensional layered carrier material is measured according to the minimum value, the maximum value, the standard deviation and the average value of the type two-dimensional layered carrier material for three polysulfides.
2. The method for machine learning-based screening of optimal support materials in lithium-sulfur batteries according to claim 1, wherein the step M1 comprises applying the method of PBE functional to AB2The structure of the two-dimensional layered material is optimized.
3. The method for machine learning-based screening of optimal support materials in lithium sulfur batteries according to claim 1, wherein the AB2The deep neural network for adsorbing polysulfide by the two-dimensional layered material comprises preset layer hidden layers, wherein each hidden layer comprises a preset number of neurons.
4. The method for screening optimal carrier materials in lithium sulfur batteries based on machine learning according to claim 1, wherein the step M2 comprises:
the polysulfides contain n atoms, each atom having the spatial coordinate (x)i,yi,zi) The central atomic coordinates are:
Figure FDA0002770955270000021
when any configuration is to be created, moving to any site, the polysulfide rotation operation involves:
xi′＝xi-cx，yi′＝yi-cy，zi′＝zi-cz； (2)
Figure FDA0002770955270000022
the molecule was then moved to the original point, x'i＝x″i+cx，y″′i＝y″i+cy，z″′i＝z″i+cz； (4)
Wherein γ represents an angle of rotation about the Z axis, β represents an angle of rotation about the Y axis, and α represents an angle of rotation about the Z axis;
calculating adsorption energy by using a density functional theory to construct a machine learning data set, wherein the data set is a structure and corresponding energy; calculating a formula according to the adsorption energy:
ΔEb＝Et-Es-Ep (5)
wherein Et is the energy of the carrier material and the polysulfide as a whole, Es is the energy of the carrier, and Ep is the energy of the polysulfide.
5. The method for screening optimal carrier materials in lithium sulfur batteries based on machine learning according to claim 1, wherein the step M3 comprises:
local chemical environment G of ith atom and its neighbor atom jijExpressed as:
Figure FDA0002770955270000023
Figure FDA0002770955270000024
Figure FDA0002770955270000025
e(x)＝x/||x|| (9)
wherein x isijDenotes the difference between the X coordinates of the ith atom and the jth atom, a (i) and b (i) denote the two atoms nearest to the ith atom, RiaRepresenting the difference between the coordinates of the ith atom and the coordinate vector of the a atom;
all atomic local chemical environments in a structure are combined into a large matrix to be used as the input of a deep neural network.
6. The method for screening optimal carrier materials in lithium sulfur batteries based on machine learning according to claim 1, wherein the step M4 comprises: using Adam optimizer, AB is aligned according to preset learning rate2Carrying out iterative training on the deep neural network of polysulfide adsorbed by the type two-dimensional layered material until the error reaches a preset value, and finishing the training to obtain the trained deep neural network;
the error includes:
Figure FDA0002770955270000031
wherein M represents the number of structures; eNN,iRepresenting the predicted energy of the neural network, EDFT,iRepresenting the energy calculated using the density functional theory.
7. A system for screening optimal carrier materials in lithium sulfur batteries based on machine learning, comprising:
module M1: AB pair by adopting density functional theory method2Optimizing the structure of the two-dimensional layered carrier material;
module M2: providing a plurality of polysulfides, and placing each polysulfide in the optimized AB2Different adsorption sites of the two-dimensional layered carrier material can be simultaneously changed to change the spatial configuration of polysulfide, thereby facilitating the modification of the polysulfideCalculating adsorption energy by using a density functional theory, and constructing a data set of an adsorption structure corresponding to the adsorption energy;
module M3: performing atomic local chemical environment calculation on the structure in the data set to obtain the structural characteristics in the data set;
module M4: construction of AB2The method comprises the steps that a deep neural network of polysulfide is adsorbed by a two-dimensional layered carrier material, the deep neural network is trained based on structural features in a data set to obtain the trained deep neural network, and the adsorption energy of any configuration and any site is predicted according to the trained deep neural network;
module M5: initializing a preset AB by using a transfer learning algorithm and using trained deep neural network model parameters2Adsorbing polysulfide deep neural network parameters by using a two-dimensional layered carrier material, and aiming at ensuring that the parameter oscillation amplitude reaches a preset value, carrying out on a preset AB2Correspondingly adjusting the parameters of the deep neural network for adsorbing polysulfide by the type two-dimensional layered carrier material;
module M6: using AB2The adsorption energy of a preset amount obtained by the theoretical calculation of the density functional of polysulfide adsorbed by the type two-dimensional layered carrier material is used for the preset AB2Training a deep neural network of polysulfide adsorbed by a type two-dimensional layered carrier material to obtain a trained preset AB2The type two-dimensional layered carrier material adsorbs a deep neural network of polysulfide;
module M7: according to preset AB after training2The deep neural network of polysulfide adsorbed by the type two-dimensional layered carrier material is used for further predicting the adsorption energy of any configuration and any site, and AB is calculated and analyzed based on statistics2Shaped two-dimensional layered support material and pre-defined AB2The adsorption capacity of the type two-dimensional layered carrier material is measured according to the minimum value, the maximum value, the standard deviation and the average value of the type two-dimensional layered carrier material for three polysulfides.
8. The system for machine learning-based screening of optimal support materials in lithium-sulfur batteries according to claim 7, wherein module M1 comprises using PBE functionalMethod pair AB2Optimizing the structure of the two-dimensional layered material;
the AB is2The deep neural network for adsorbing polysulfide by the two-dimensional layered material comprises preset layer hidden layers, wherein each hidden layer comprises a preset number of neurons.
9. The system for machine learning based screening of optimal carrier materials in lithium sulfur batteries according to claim 7, wherein the module M2 comprises:
the polysulfides contain n atoms, each atom having the spatial coordinate (x)i,yi,zi) The central atomic coordinates are:
Figure FDA0002770955270000041
when any configuration is to be created, moving to any site, the polysulfide rotation operation involves:
xi′＝xi-cx，yi′＝yi-cy，zi′＝zi-cz； (2)
Figure FDA0002770955270000042
the molecule was then moved to the original point, x'i＝x″i+cx，y″′i＝y″i+cy，z″′i＝z″i+cz； (4)
Wherein γ represents an angle of rotation about the Z axis, β represents an angle of rotation about the Y axis, and α represents an angle of rotation about the Z axis;
calculating adsorption energy by using a density functional theory to construct a machine learning data set, wherein the data set is a structure and corresponding energy; calculating a formula according to the adsorption energy:
ΔEb＝Et-Es-Ep
wherein Et is the energy of the carrier material and the polysulfide as a whole, Es is the energy of the carrier, and Ep is the energy of the polysulfide.
10. The system for machine learning based screening of optimal carrier materials in lithium sulfur batteries according to claim 7, wherein the module M3 comprises:
local chemical environment G of ith atom and its neighbor atom jijExpressed as:
Figure FDA0002770955270000043
Figure FDA0002770955270000044
Figure FDA0002770955270000045
e(x)＝x/||x|| (9)
wherein x isijDenotes the difference between the X coordinates of the ith atom and the jth atom, a (i) and b (i) denote the two atoms nearest to the ith atom, RiaRepresenting the difference between the coordinates of the ith atom and the coordinate vector of the a atom;
forming a large matrix by using local chemical environments of all atoms in a structure as input of a deep neural network;
the module M4 includes: using Adam optimizer, AB is aligned according to preset learning rate2Carrying out iterative training on the deep neural network of polysulfide adsorbed by the type two-dimensional layered material until the error reaches a preset value, and finishing the training to obtain the trained deep neural network;
the error includes:
Figure FDA0002770955270000051
wherein M represents the number of structures; eNN,iRepresenting the predicted energy of the neural network, EDFT,iRepresenting the energy calculated using the density functional theory.
CN202011248936.8A 2020-11-10 2020-11-10 Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning Pending CN112347697A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011248936.8A CN112347697A (en) 2020-11-10 2020-11-10 Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011248936.8A CN112347697A (en) 2020-11-10 2020-11-10 Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning

Publications (1)

Publication Number Publication Date
CN112347697A true CN112347697A (en) 2021-02-09

Family

ID=74363100

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011248936.8A Pending CN112347697A (en) 2020-11-10 2020-11-10 Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning

Country Status (1)

Country Link
CN (1) CN112347697A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113536676A (en) * 2021-07-15 2021-10-22 重庆邮电大学 Lithium battery health condition monitoring method based on feature transfer learning
CN114371145A (en) * 2022-03-21 2022-04-19 武汉工程大学 Detection method and device for milk oil mixed pigment, electronic equipment and storage medium
CN114613445A (en) * 2022-02-18 2022-06-10 北京工业大学 Method for predicting carbon dioxide electroreduction copper alloy catalyst
CN114997066A (en) * 2022-06-24 2022-09-02 集美大学 Method and device for improving density functional theory expansion term and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018065158A1 (en) * 2016-10-06 2018-04-12 Siemens Aktiengesellschaft Computer device for training a deep neural network
CN109376933A (en) * 2018-10-30 2019-02-22 成都云材智慧数据科技有限公司 Lithium ion battery negative material energy density prediction technique neural network based
CN110443364A (en) * 2019-06-21 2019-11-12 深圳大学 A kind of deep neural network multitask hyperparameter optimization method and device

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018065158A1 (en) * 2016-10-06 2018-04-12 Siemens Aktiengesellschaft Computer device for training a deep neural network
CN109376933A (en) * 2018-10-30 2019-02-22 成都云材智慧数据科技有限公司 Lithium ion battery negative material energy density prediction technique neural network based
CN110443364A (en) * 2019-06-21 2019-11-12 深圳大学 A kind of deep neural network multitask hyperparameter optimization method and device

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
HAIKUO ZHANG ET AL.: "Ultra-fast and accurate binding energy prediction of shuttle effect-suppressive sulfur hosts for lithium-sulfur batteries using machine learning", 《ENERGY STORAGE MATERIALS》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113536676A (en) * 2021-07-15 2021-10-22 重庆邮电大学 Lithium battery health condition monitoring method based on feature transfer learning
CN114613445A (en) * 2022-02-18 2022-06-10 北京工业大学 Method for predicting carbon dioxide electroreduction copper alloy catalyst
CN114371145A (en) * 2022-03-21 2022-04-19 武汉工程大学 Detection method and device for milk oil mixed pigment, electronic equipment and storage medium
CN114997066A (en) * 2022-06-24 2022-09-02 集美大学 Method and device for improving density functional theory expansion term and storage medium

Similar Documents

Publication Publication Date Title
CN112347697A (en) Method and system for screening optimal carrier material in lithium-sulfur battery based on machine learning
Bhowmik et al. A perspective on inverse design of battery interphases using multi-scale modelling, experiments and generative deep learning
Turetskyy et al. Battery production design using multi-output machine learning models
CN108629147A (en) A kind of polycrystal Geometric Modeling Method
CN104715096B (en) BP neural network predicts dipeptides model multipole expansion attribute computing method
CN104836256B (en) The method and system that a kind of power distribution network photovoltaic digestion capability calculates
CN105574231A (en) Storage battery surplus capacity detection method
CN107562992B (en) Photovoltaic array maximum power tracking method based on SVM and particle swarm algorithm
US11817184B2 (en) Graph neural network force field computational algorithms for molecular dynamics computer simulations
Hong et al. Open-sourcing phase-field simulations for accelerating energy materials design and optimization
Giordano et al. Parallel execution of cellular automata through space partitioning: the landslide simulation sciddicas3-hex case study
Ma et al. Application of deep learning for informatics aided design of electrode materials in metal-ion batteries
Zhu et al. State of health prediction for li-ion batteries with end-to-end deep learning
Sandhu et al. Using neurocomputing techniques to determine microstructural properties in a Li-ion battery
Xu et al. Automated construction of neural network potential energy surface: The enhanced self-organizing incremental neural network deep potential method
Allen et al. A segregated approach for modeling the electrochemistry in the 3-D microstructure of Li-ion batteries and its acceleration using block preconditioners
Maphanga et al. Estimating DFT calculated voltage using machine learning regression models
Fang et al. Analysis of Performance Degradation in Lithium-Ion Batteries Based on a Lumped Particle Diffusion Model
CN116256639A (en) Estimation method and device for state of charge of vehicle battery and electronic device
Yan et al. Predicting 2D normal grain growth using a physics-regularized interpretable machine learning model
CN115099066A (en) Kriging-based power transmission tower structure wind disaster vulnerability modeling method
Rende et al. Fine-tuning Neural Network Quantum States
Li et al. Identifying geomechanical parameters of high cut rock slopes by an improved multi-output support vector machine method
Gutiérrez et al. Simple reservoirs with chain topology based on a single time-delay nonlinear node.
Galvez‐Aranda et al. Time‐Dependent Deep Learning Manufacturing Process Model for Battery Electrode Microstructure Prediction

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210209