Ionic liquid molecule design and carbon dioxide solubility verification method and system

Abstract

The invention discloses an ionic liquid molecule design and carbon dioxide solubility verification method and system. The method comprises the following steps: firstly, constructing a to-be-verified initial data set; executing preprocessing on the to-be-verified initial data set to obtain a target data set; the method comprises the following steps of: constructing a dual Transform decoder model; on the basis of the target data set and the dual Transform decoder model, an ionic liquid candidate is determined; and determining that the current ionic liquid candidate is the optimal target ionic liquid in response to verification that the ionic liquid candidate meets a preset condition through a density functional algorithm and a COSMO-RS method. According to the method, efficient and accurate ionic liquid molecule design and verification are realized by combining a dual Transform decoder model and a quantum chemistry verification method, and the development efficiency of the ionic liquid with high carbon dioxide solubility is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of quantum chemistry, and more specifically, to a method and system for ionic liquid molecular design and carbon dioxide solubility verification. Background Technology

[0002] Ionic liquids (ILs) exhibit immense application potential in fields such as carbon dioxide capture and drug delivery due to their extremely low volatility, tunable structure, and excellent chemical stability. However, the cation-anion combination space of ionic liquids is extremely vast. Traditional experimental trial-and-error screening methods are not only time-consuming, labor-intensive, and costly, but also struggle to cover this broad chemical space, resulting in low development efficiency for highly efficient ionic liquids for carbon dioxide capture. Furthermore, existing data-driven molecular generation models generally suffer from poor physical applicability, hindering the efficient and reliable development of ionic liquids that meet the requirements for carbon dioxide capture. This has become a pressing technical problem that needs to be solved. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for ionic liquid molecular design and carbon dioxide solubility verification, in order to overcome the shortcomings of the prior art. It is a Transformer-based reinforcement learning molecular design method for ionic liquids and a quantum chemical verification method for carbon dioxide solubility. By constructing a closed-loop framework of generation, optimization and verification, it can achieve efficient generation, chemical feasibility and high carbon dioxide solubility of ionic liquids.

[0004] One embodiment of this application provides a method for ionic liquid molecule design and carbon dioxide solubility verification, the method comprising: Construct an initial dataset to be validated, wherein the initial dataset to be validated includes a cation SMILES dataset, an anion SMILES dataset, and an ionic liquid carbon dioxide solubility dataset; Perform preprocessing on the initial dataset to be verified to obtain the target dataset; Construct a dual Transformer decoder model; Based on the target dataset and the dual Transformer decoder model, ionic liquid candidates are determined. If the ionic liquid candidate is verified to meet the preset conditions by using density functional theory algorithm and COSMO-RS method, then the current ionic liquid candidate is determined to be the optimal target ionic liquid.

[0005] Optionally, the preprocessing of the initial dataset to be verified to obtain the target dataset includes: Perform normalization processing on the cation SMILES dataset and the anion SMILES dataset to obtain a standard dataset; The dataset of carbon dioxide solubility of ionic liquids is divided into training set, validation set and test set according to a preset ratio; The target dataset is determined based on the standard dataset, the training set, the validation set, and the test set.

[0006] Optionally, the standardization process performed on the cation SMILES dataset and the anion SMILES dataset to obtain a standard dataset includes: Invalid data in the cation SMILES dataset and the anion SMILES dataset are filtered and removed. By adding preset token instructions, a vocabulary containing preset characters corresponding to the cation SMILES dataset and the anion SMILES dataset is constructed. The vocabulary contains at least element symbols, chemical bonds, and charge symbols. Based on the vocabulary, a standard dataset is determined.

[0007] Optionally, constructing the dual Transformer decoder model includes: Based on the cations and anions in the ionic liquid, a first Transformer decoder model and a second Transformer decoder model with conditional embedding are constructed respectively; wherein, the embedding conditions include the embedding temperature and pressure parameters; after training, both the first Transformer decoder model and the second Transformer decoder model have the function of obtaining ion-specific generation.

[0008] Optionally, determining ionic liquid candidates based on the target dataset and the dual Transformer decoder model includes: Based on the target dataset and the dual Transformer decoder model, the parameters of the dual Transformer decoder model are optimized using a reinforcement learning algorithm, with solubility prediction as a positive reward and structural invalidity as a penalty, to generate ionic liquid candidates.

[0009] Optionally, the step of determining the current ionic liquid candidate as the optimal target ionic liquid in response to verification by density functional theory and the COSMO-RS method that the ionic liquid candidate meets the preset conditions includes: If the structural stability of the ionic liquid candidate is verified to meet the first preset condition by density functional theory algorithm, and the thermodynamic rationality of the carbon dioxide solubility of the ionic liquid candidate is verified to meet the second preset condition by COSMO-RS method, then the current ionic liquid candidate is determined to be the optimal target ionic liquid.

[0010] Optionally, the method further includes: If the structural stability of the ionic liquid candidate does not meet the first preset condition when verified by density functional theory algorithm and / or the thermodynamic rationality of the carbon dioxide solubility of the ionic liquid candidate does not meet the second preset condition when verified by COSMO-RS method, then the reward function of the dual Transformer decoder model is adjusted and the process returns to the step of generating ionic liquid candidates by optimizing the parameters of the dual Transformer decoder model through reinforcement learning algorithm based on the target dataset and the dual Transformer decoder model, with the solubility prediction value as a positive reward and structural invalidity as a penalty.

[0011] Another embodiment of this application provides a system for ionic liquid molecular design and carbon dioxide solubility verification, the system comprising: The first construction module is used to construct the initial dataset to be verified, wherein the initial dataset to be verified includes the cation SMILES dataset, the anion SMILES dataset, and the ionic liquid carbon dioxide solubility dataset. The execution module is used to perform preprocessing on the initial dataset to be verified to obtain the target dataset; The second building block is used to build the dual Transformer decoder model; The first determining module is used to determine ionic liquid candidates based on the target dataset and the dual Transformer decoder model; The second determining module is used to determine the current ionic liquid candidate as the optimal target ionic liquid in response to the verification that the ionic liquid candidate meets the preset conditions through density functional algorithm and COSMO-RS method.

[0012] Another embodiment of this application provides a storage medium storing a computer program, wherein the computer program is configured to implement the method described in any of the above-described embodiments when running.

[0013] Another embodiment of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to implement the method described in any of the above embodiments.

[0014] Compared with existing technologies, this invention first constructs an initial dataset to be verified; performs preprocessing on the initial dataset to obtain a target dataset; constructs a dual Transformer decoder model; determines ionic liquid candidates based on the target dataset and the dual Transformer decoder model; and, in response to verification that the ionic liquid candidates meet preset conditions using density functional theory and the COSMO-RS method, determines the current ionic liquid candidate as the optimal target ionic liquid. By combining a dual Transformer decoder model with quantum chemical verification methods, it achieves efficient and accurate ionic liquid molecular design and verification, significantly improving the development efficiency of ionic liquids with high carbon dioxide solubility. Attached Figure Description

[0015] Figure 1 A hardware structure block diagram of a computer terminal for an ionic liquid molecule design and carbon dioxide solubility verification method provided in an embodiment of the present invention; Figure 2 A flowchart illustrating a method for ionic liquid molecule design and carbon dioxide solubility verification provided in an embodiment of the present invention; Figure 3 An example diagram of a molecular generation framework provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a Transformer decoder architecture provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of a Transformer decoder model training provided in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the change in carbon dioxide solubility during reinforcement learning, provided as an embodiment of the present invention. Figure 7 A schematic diagram of the electrostatic potential of a first ionic liquid provided in an embodiment of the present invention; Figure 8 A schematic diagram illustrating the interaction between a first type of ionic liquid and carbon dioxide molecules provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of an ionic liquid structure generated by a molecular generation model provided in an embodiment of the present invention; Figure 10 A schematic diagram of the electrostatic potential of a second ionic liquid provided in an embodiment of the present invention; Figure 11 This is a schematic diagram illustrating the interaction between a second type of ionic liquid and carbon dioxide molecules, provided in an embodiment of the present invention. Figure 12 A comparative schematic diagram of the first and second ionic liquids provided in the embodiments of the present invention; Figure 13 This is a schematic diagram of the structure of an ionic liquid molecule design and carbon dioxide solubility verification system provided in an embodiment of the present invention. Detailed Implementation

[0016] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0017] The present invention first provides a method for ionic liquid molecular design and carbon dioxide solubility verification. This method can be applied to electronic devices, such as computer terminals, specifically ordinary computers, quantum computers, etc.

[0018] The following detailed explanation uses a computer terminal as an example. Figure 1 This is a hardware structure block diagram of a computer terminal for a method of ionic liquid molecular design and carbon dioxide solubility verification provided in an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, memory, GPU, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0019] The non-volatile storage medium can store an operating system and a computer program. This computer program includes program instructions that, when executed, cause the processor to perform any method for designing ionic liquid molecules and verifying carbon dioxide solubility.

[0020] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0021] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to perform any ionic liquid molecule design and carbon dioxide solubility verification method.

[0022] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0023] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0024] See Figure 2 , Figure 2 The flowchart illustrating a method for ionic liquid molecule design and carbon dioxide solubility verification provided in this embodiment of the invention may include the following steps: S201: Construct an initial dataset to be validated, wherein the initial dataset to be validated includes a cation SMILES dataset, an anion SMILES dataset, and an ionic liquid carbon dioxide solubility dataset.

[0025] For example, the initial dataset to be constructed includes a cation SMILES dataset, for example, collecting 792 unique structures; an anion SMILES dataset, for example, collecting 689 unique structures; and an ionic liquid carbon dioxide solubility dataset, for example, collecting 1099 records, each record containing parameters such as ionic liquid SMILES, carbon dioxide solubility, temperature, and pressure.

[0026] S202: Perform preprocessing on the initial dataset to be verified to obtain the target dataset.

[0027] Specifically, the preprocessing of the initial dataset to be verified to obtain the target dataset may include: 1. Perform standardization processing on the cation SMILES dataset and the anion SMILES dataset to obtain a standard dataset.

[0028] Specifically, the standardization process performed on the cation SMILES dataset and the anion SMILES dataset to obtain a standard dataset may include: Step 1: Filter and remove invalid data from the cation SMILES dataset and the anion SMILES dataset. Construct a vocabulary containing preset characters corresponding to the cation SMILES dataset and the anion SMILES dataset by adding preset token instructions. The vocabulary includes at least element symbols, chemical bonds, and charge symbols.

[0029] It should be noted that the preset token instruction can be a preset token instruction.

[0030] Step 2: Determine the standard dataset based on the vocabulary.

[0031] For example, the RDKit library can be used to check the validity of SMILES strings, removing entries that cannot be parsed or contain illegal characters. And add [something] to each SMILES sequence. <sos>(Start of Sequence)、 <eos>(End ofSequence) and <pad>The (Padding) tokens are then used to construct a 63-character vocabulary based on all valid SMILES, covering element symbols, chemical bonds, charge symbols (such as +, -), etc.

[0032] It should be noted that the standardization process performed on the cation SMILES dataset and the anion SMILES dataset to obtain a standard dataset can also be used to set up an additional SMILES dataset to ensure a rich variety of SMILES. A preliminary check is performed during the preprocessing stage to ensure that each cation SMILES contains at least one "+" and each anion SMILES contains at least one "-".

[0033] 2. Divide the ionic liquid carbon dioxide solubility dataset into a training set, a validation set, and a test set according to a preset ratio.

[0034] For example, the ionic liquid carbon dioxide solubility dataset can be divided into a training set, a validation set, and a test set in a 6:2:2 ratio. For key indicators such as solubility, SMOTE (Synthetic Minority Oversampling) or Bootstrapping methods can be used to address the imbalance in data distribution, thereby improving the model's ability to learn from high-solubility samples.

[0035] 3. Determine the target dataset based on the standard dataset, the training set, the validation set, and the test set.

[0036] S203: Construct a dual Transformer decoder model.

[0037] Specifically, constructing the dual Transformer decoder model may include: Based on the cations and anions in the ionic liquid, a first Transformer decoder model and a second Transformer decoder model with conditional embedding are constructed respectively; wherein, the embedding conditions include the embedding temperature and pressure parameters; after training, both the first Transformer decoder model and the second Transformer decoder model have the function of obtaining ion-specific generation.

[0038] For example, the first Transformer decoder can be used specifically for cation processing, and its structure consists of a 6-layer stack, each layer containing an 8-head masked self-attention mechanism and a feedforward network with 1024-dimensional hidden units. The second Transformer decoder can be used specifically for anion processing, and its structure is the same as the first Transformer decoder, but the parameters of the two decoders are independent.

[0039] It should be noted that during model training, the Transformer decoder can also perform multi-task learning on anion and additional SMILES datasets or cation and additional SMILES datasets to ensure that the Transformer learns the writing patterns of SMILES. Embedding layer enhancements can be achieved through character embeddings, such as mapping SMILES characters to 256-dimensional vectors; or positional encoding, specifically sinusoidal positional encoding. Conditional embeddings involve mapping environmental parameters such as temperature and pressure to 256-dimensional vectors using a small multilayer perceptron (MLP including activation functions), and then adding these vectors element-wise with the character embeddings and positional encodings, enabling the model to perceive environmental conditions during generation.

[0040] S204: Based on the target dataset and the dual Transformer decoder model, determine the ionic liquid candidates.

[0041] Specifically, determining ionic liquid candidates based on the target dataset and the dual Transformer decoder model may include: Based on the target dataset and the dual Transformer decoder model, the parameters of the dual Transformer decoder model are optimized using the reinforcement learning REINFORCE algorithm, with solubility prediction as a positive reward and structural invalidity as a penalty, to generate ionic liquid candidates.

[0042] For example, see Figure 3 , Figure 3 This invention provides an example diagram of a molecular generation framework, including modules for model pre-training, reinforcement learning, and theoretical calculation verification. In model pre-training, conventional organic data and ionic liquid data are used, and a Transformer-based decoder is trained to obtain a pre-trained molecular SMILES generator and a pre-trained carbon dioxide solubility prediction model. Through the reinforcement learning module, theoretical calculations verify the surface charge distribution related to carbon dioxide solubility. The process of generating ionic liquid candidates also requires pre-training and fine-tuning of the dual Transformer decoder model. Pre-training can be performed on a corpus containing 10,000 general organic molecular SMILES, allowing the model to learn basic chemical grammar. Fine-tuning, or ion-specific enhancement, can be performed on cation / anion-specific datasets, strengthening the ability to generate ion-specific structures by constraining the generation logic (e.g., cations must contain "+", anions must contain "-"). A chemical grammar penalty term is added to the loss function. For example, if the generated cation does not contain "+", a fixed penalty value, such as 10, is added to the cross-entropy loss. Alternatively, prior knowledge of functional groups can be introduced to encourage the model to generate specific functions by adding additional biases to the attention layer.

[0043] See Figure 4 , Figure 4 This is a schematic diagram of a Transformer decoder architecture provided in an embodiment of the present invention. The Transformer decoder may include multiple decoder layers, and each decoder layer includes a masked multi-head attention mechanism, multiple layer normalization, a multi-head attention mechanism, and a feedforward neural network.

[0044] See Figure 5 , Figure 5 This is a schematic diagram of a Transformer decoder model training provided in an embodiment of the present invention. The ionic liquid SMILES includes cation SMILES and anion SMILES. The model training operation is performed by verifying whether the structure meets the preset feasibility and charge balance requirements. If the structure meets the preset feasibility and charge balance, the predicted carbon dioxide solubility is output, and loss calculation and parameter optimization are performed through positive feedback. The Transformer decoder generates molecules. If the structure does not meet the preset feasibility and charge balance, a negative feedback operation is performed.

[0045] It should be noted that the dual Transformer decoder model can also perform reinforcement learning environment optimization. Specifically, it uses the Morgan fingerprint of cations or anions and temperature / pressure parameters as the environmental state, and the cation or anion SMILES sequences generated by the decoder as the action. The model is based on the solubility value (range 0-100 points) output by a pre-trained QSPR (quantitative structure-property relationship) solubility prediction model. The penalty mechanism is set to -20 points for invalid SMILES (e.g., syntax errors) and -10 points for missing charges (e.g., no "+" in cations). Simultaneously, a reinforcement learning policy gradient algorithm is used to update the Transformer parameters through the AdamW optimizer, while entropy regularization is introduced to maintain structural diversity and prevent mode collapse. See also Figure 6 , Figure 6 This is a schematic diagram illustrating the change in carbon dioxide solubility during reinforcement learning, as provided in an embodiment of the present invention.

[0046] For example, Transformer pre-training can be performed on a corpus of 10,000 general organic molecules SMILES for 100 epochs with a batch size of 256, using the cross-entropy loss function, and saving a checkpoint every epoch. Then, fine-tuning can be performed on a cation / anion dataset for 50 epochs, with the constraint that "+" and "-" must appear. After fine-tuning, the model achieves a 94% success rate in generating cation SMILES and a 91% success rate in generating anion SMILES.

[0047] Reinforcement learning training can begin by initializing the RL environment, using the QSPR model's predicted solubility as the reward, and training for 200 epochs. In the first 50 epochs, the average reward increases from 2.1 to 4.9, corresponding to a solubility increase from 2.1 mol / kg to 4.9 mol / kg. Finally, the optimal candidate is generated, for example: SMILES is represented as: CCCCN1C=C[N+](=C1)CC(F)(F)(F)S(=O)(=O)[N-]S(=O)(=O)C(F)(F)F).

[0048] S205: In response to the verification by density functional theory algorithm and COSMO-RS method that the ionic liquid candidate meets the preset conditions, the current ionic liquid candidate is determined to be the optimal target ionic liquid.

[0049] Specifically, the step of determining the current ionic liquid candidate as the optimal target ionic liquid by verifying that the ionic liquid candidate meets the preset conditions through density functional theory and the COSMO-RS method can include: If the structural stability of the ionic liquid candidate is verified to meet the first preset condition by density functional theory algorithm, and the thermodynamic rationality of the carbon dioxide solubility of the ionic liquid candidate is verified to meet the second preset condition by COSMO-RS method, then the current ionic liquid candidate is determined to be the optimal target ionic liquid.

[0050] The COSMO-RS method is a way to predict the thermodynamic equilibrium of fluid and liquid mixtures using a static thermodynamic approach based on quantum chemical calculations. The underlying quantum chemical model, the so-called conductor-like shielded mode (COSMO), is an efficient variant of the dielectric continuous solvation method in COSMO calculations. Solute molecules are calculated in a virtual conductor environment. In this environment, solute molecules induce a non-polarized charge density σ at the interface between the molecule and the conductor, i.e., on the molecular surface. These charges react on the solute, producing a higher charge-polarized electron density than in a vacuum. In the quantum density algorithm, the quantum density is optimally consistent with the energy of the quantum conductor. The molecular weight determination can be optimized using the same calculation methods in a vacuum. For each molecule of interest, the time-consuming QCO calculation only needs to be performed once. The polarized charge density (also known as the shielded charge density) in the COSMO calculations is a good local descriptor of the molecular surface polarity, used to extend the model to "ReaSolvents" (Cosm0 rsp).

[0051] For example, the structural stability and thermodynamic verification of carbon dioxide solubility of the generated ionic liquid candidate are performed through the following steps: Density functional theory (DFT) structure verification can be performed by converting SMILES into an initial 3D structure using the MMFF94 force field in RDKit. Full geometry optimization is then conducted in quantum chemical calculation software using the B3LYP functional + 6-31G(d,p) basis set, combined with Grimme's D3 dispersion correction to handle long-range interactions. Stability analysis verifies the ion pair structure stability by calculating the binding energy (E_binding = E_IL - E_cation - E_anion) and bond lengths (e.g., 1.3-1.4 Å is effective for CN bonds). Electrostatic property calculations derive the ESP (electrostatic potential) charge through CHELPG / RESP fitting, calculate the charge anisotropy (σ_surf) and extreme surface potentials (V_max / V_min) using Multiwfn, and visualize the molecular electrostatic potential (MEP) distribution using quantum chemical calculation software.

[0052] COSMO-RS solubility thermodynamic verification can be performed using quantum chemical calculation software to export the ionic structure, optimized by density functional theory verification, into a preset format, such as .cosmo, and extract the surface polarization charge density to construct a σ-profile. Solubility prediction can be performed using preset application software to calculate Henry's law constant, activity coefficient, excess enthalpy, and carbon dioxide capacity under target conditions, such as 298 K and 1 bar partial pressure of carbon dioxide, thus verifying solubility performance.

[0053] Following the example above, the density functional algorithm is used for verification. The binding energy is -85 kcal / mol, higher than the training set average of -70 kcal / mol. The CN bond length is 1.35 Å, within the effective range of 1.3-1.4 Å, indicating structural stability. MEP analysis shows... of The group is a region rich in negative charge and has a strong interaction with the C=O dipole of carbon dioxide; COSMO-RS validation was performed at 298K and 1 bar to determine... The carbon dioxide capacity is 5.2 mol / kg, which is higher than the maximum value of 4.1 mol / kg in the training set. The σ-profile shows that its overlap with the carbon dioxide in the σ=-0.02-0.02e / Ų range reaches 85%, verifying its solubility advantage.

[0054] Compared with existing technologies, the technology generated in this embodiment... The invention demonstrates a 26.8% improvement in carbon dioxide solubility compared to traditionally screened ILs, and its synthetic feasibility has been confirmed through quantum chemical analysis, thus validating the effectiveness of the invention.

[0055] In one alternative implementation, the method may further include: If the structural stability of the ionic liquid candidate does not meet the first preset condition when verified by density functional theory algorithm and / or the thermodynamic rationality of the carbon dioxide solubility of the ionic liquid candidate does not meet the second preset condition when verified by COSMO-RS method, then the reward function of the dual Transformer decoder model is adjusted and the process returns to the step of optimizing the parameters of the dual Transformer decoder model using the reinforcement learning REINFORCE algorithm based on the target dataset and the dual Transformer decoder model, with the solubility prediction value as a positive reward and structural invalidity as a penalty, to generate the ionic liquid candidate.

[0056] For example, the quantum chemical verification results can be fed back to the reinforcement learning module. If the stability or solubility of the candidate does not meet the requirements, the reward function weights are adjusted, such as by increasing the structural constraint penalty, and the generation-optimization process is restarted until an ionic liquid that meets the requirements is obtained.

[0057] As can be seen, this invention first constructs an initial dataset to be verified; performs preprocessing on the initial dataset to obtain a target dataset; constructs a dual Transformer decoder model; determines ionic liquid candidates based on the target dataset and the dual Transformer decoder model; and, in response to verification that the ionic liquid candidates meet preset conditions using density functional theory and the COSMO-RS method, determines the current ionic liquid candidate as the optimal target ionic liquid. By combining a dual Transformer decoder model with quantum chemical verification methods, it achieves efficient and accurate ionic liquid molecular design and verification, significantly improving the development efficiency of ionic liquids with high carbon dioxide solubility.

[0058] In another embodiment, the cation SMILES dataset contains 1500 cations with different structures, including common cation types such as imidazoles, pyridines, ammonium compounds, and phosphonium compounds, each represented by the string "SMILES". The anion SMILES dataset contains 800 anions with different structures, including common anions such as [BF4], [PF6]⁻, and [Tf2N]⁻, also represented by the string "SMILES". The ionic liquid carbon dioxide solubility dataset contains 20,000 sets of experimental data, each set including the cation structure, anion structure, temperature, pressure, and corresponding carbon dioxide solubility value.

[0059] This embodiment can introduce an adaptive invalid data detection mechanism, which automatically identifies and removes invalid data by constructing a SMILES syntax tree and a chemical rationality rule base. Specifically: Construct a vocabulary of 200 predefined characters, including element symbols (C, H, N, O, etc.), chemical bonds (-, =, #, etc.), and charge symbols (+, -, etc.). Introduce special tokens. <start> 、 <end>and <pad>These are used to identify the beginning, end, and filling of a molecular structure, respectively. Special additions were made. <unknown>The token processing rarely structured, combined with transfer learning technology, through the pre-training model to predict and complete the unknown structure. The ionic liquid carbon dioxide solubility dataset is divided into training set, validation set and test set in the ratio of 7:2:1. The hierarchical sampling method can be used in this embodiment to ensure that the data distribution under different temperature and pressure conditions in each set is consistent, avoiding the influence of data bias on model training.

[0060] In particular, the embodiment can also introduce an active learning strategy, periodically evaluate the difficult example samples in the validation set, that is, the samples with larger prediction errors, and preferentially add them to the training set, thereby continuously improving the generalization ability of the model.

[0061] The double-Transformer decoder model constructed in this embodiment has the following design: First, a first Transformer decoder model (cation generation model) and a second Transformer decoder model (anion generation model) are respectively constructed. Each decoder model includes 6 Transformer layers, each Transformer layer includes 8 self-attention mechanisms and a feedforward neural network. A conditional embedding module is designed to integrate environmental parameters such as temperature and pressure into the model. Specifically, the temperature parameter is converted into a vector representation through the sine-cosine position encoding method, and the pressure parameter is mapped to the same dimension as the word embedding through linear transformation. The environmental parameter embedding and the molecular structure embedding are fused through a gating mechanism, so that the model can learn the relationship between molecular structure and performance under different temperature and pressure conditions.

[0062] It should be noted that the total loss is the average of the cross-entropy losses of all valid positions (non-PAD), that is: Where the batch size is , the sequence length is , The loss for masking the PAD position (PAD position contribution is 0), and the denominator is the total number of valid positions, to ensure that the loss is correctly averaged. For the i-th sample in the batch and the t-th position, the cross-entropy loss is: Where the numerator is the exponentialization of the target token (y(i,t)), and the denominator is the sum of the exponentialization of all tokens (i.e., softmax normalization), which measures the negative logarithm of the predicted probability of the model for the true token.

[0063] Through the cation-anion matching loss function, it is ensured that the generated ion pair has good matching in chemical properties.

[0064] The dual-Transformer decoder model is trained with the target data set, the model parameters are initialized, and a multi-target reward function is designed, which is expressed as follows: wherein, is the predicted carbon dioxide solubility value, is the stability score predicted based on the molecular structure characteristics, is a penalty term based on the molecular mass and structural complexity, 、 、 is a weight parameter, which can be adjusted according to actual requirements.

[0065] In an embodiment, the improved reinforcement learning REINFORCE algorithm can be used to optimize the model parameters, a baseline function is introduced to reduce the gradient estimation variance, an adaptive learning rate strategy is used to dynamically adjust the learning rate according to the reward value, and an exploration-exploitation balance mechanism is added to ensure the generation of high-quality molecules while maintaining a certain structural diversity. For example, Top-K ionic liquid candidates are generated by beam search, and in this embodiment, the value of K can be 50.

[0066] The present embodiment can use a multi-scale verification method to verify the candidates, such as using density functional theory (DFT) to verify the structural stability of ionic liquid candidates. Geometric optimization is performed at the B3LYP / 6-31G(d,p) level, the highest occupied molecular orbital energy (HOMO) and the lowest unoccupied molecular orbital energy (LUMO) are calculated, the binding energy and vibration frequency of the molecule are calculated, and it is verified whether there is a virtual frequency. The first preset condition can be that the binding energy is less than -50 kcal / mol and there is no virtual frequency.

[0067] The COSMO-RS method is used to verify the thermodynamic reasonableness of the carbon dioxide solubility, the COSMO surface of the molecule is constructed, the interaction parameters of the ionic liquid and carbon dioxide are calculated, and the carbon dioxide solubility under different temperatures and pressures is predicted. The second preset condition can be that the carbon dioxide solubility is greater than 0.15 mol / mol under the condition of 313K, 1bar.

[0068] For candidates that do not meet the preset conditions, the present embodiment can use a feedback adjustment mechanism to analyze the reasons for the verification failure, such as structural instability, insufficient solubility, etc., adjust the reward function weight parameters accordingly, and introduce failed cases as negative samples to retrain the model. The simulated annealing strategy is used to adjust the generation strategy to avoid falling into local optimum.

[0069] The embodiment introduces key technologies such as reward mechanism, quantum chemical verification and dynamic closed-loop feedback, and significantly improves the efficiency, reliability and practicability of intelligent design of ionic liquids. These improvements not only solve the core pain points of poor physical landing and low generation efficiency in the prior art, but also provide a solid methodological foundation for applying artificial intelligence technology to a wider functional molecule design field.

[0070] For example, see Figure 7 , Figure 7 The first ionic liquid provided by the embodiment of the present application is shown in the electrostatic potential diagram, wherein the electrostatic potential of the ionic liquid [MPPL][PEIM] is -75.52 kcal / mol and 42.33 kcal / mol, respectively. See Figure 8 , Figure 8 The interaction between the first ionic liquid provided by the embodiment of the present application and the carbon dioxide molecule is shown in the diagram, wherein the interaction between the ionic liquid [MPPL][PEIM] and the carbon dioxide molecule can be represented as follows: wherein, represents the charge density at the bond critical point, represents the hydrogen bond energy between different atoms.

[0071] See Figure 9 , Figure 9 The ionic liquid structure diagram generated by the molecular generation model provided by the embodiment of the present application can be represented as (CCCCC[N+]1(C)CCCC1.N=C([O-])Cc1ccccc1; 1-Methyl-1-pentylpyrrolidin-1-ium 2-Phenylethanimidate [MPPL][PEIM]).

[0072] See Figure 10 , Figure 10 The electrostatic potential diagram of the second ionic liquid provided by the embodiment of the present application is shown, wherein the electrostatic potential of the ionic liquid [BMIM][PF6] is 48.08 kcal / mol and -48.61 kcal / mol, respectively. See Figure 11 , Figure 11 The interaction between the second ionic liquid provided by the embodiment of the present application and the carbon dioxide molecule is shown in the diagram, wherein the interaction between the ionic liquid [BMIM][PF6] and the carbon dioxide molecule can be represented as follows: wherein, This represents the charge density at the bond critical point. It represents the hydrogen bond energy between different atoms.

[0073] As can be seen from the comparison, the hydrogen bond energy of the solvent [MPPL][PEIM] to carbon dioxide obtained by the model is -25.569 kJ / mol, which is much greater than the hydrogen bond energy of the common [BMIM][PF6] to carbon dioxide, which is -15.544 kJ / mol.

[0074] See Figure 12 , Figure 12 This is a comparative schematic diagram of a first ionic liquid and a second ionic liquid provided in an embodiment of the present invention, wherein, from Figure 12 It can be seen that both the first ionic liquid [MPPL][PEIM] and the second ionic liquid [BMIM][PF6] have strong hydrogen bond acceptor capabilities, and the first ionic liquid [MPPL][PEIM] has a stronger hydrogen bond acceptor capability, with a peak value close to 0.03.

[0075] Compared with existing technologies, this invention first constructs an initial dataset to be verified; performs preprocessing on the initial dataset to obtain a target dataset; constructs a dual Transformer decoder model; determines ionic liquid candidates based on the target dataset and the dual Transformer decoder model; and, in response to verification that the ionic liquid candidates meet preset conditions using density functional theory and the COSMO-RS method, determines the current ionic liquid candidate as the optimal target ionic liquid. By combining a dual Transformer decoder model with quantum chemical verification methods, it achieves efficient and accurate ionic liquid molecular design and verification, significantly improving the development efficiency of ionic liquids with high carbon dioxide solubility.

[0076] Another embodiment of this application provides a system for ionic liquid molecular design and carbon dioxide solubility verification, such as... Figure 13 The diagram shows a structural schematic of an ionic liquid molecule design and carbon dioxide solubility verification system, the system comprising: The first construction module 1301 is used to construct an initial dataset to be verified, wherein the initial dataset to be verified includes a cation SMILES dataset, an anion SMILES dataset, and an ionic liquid carbon dioxide solubility dataset. Execution module 1302 is used to perform preprocessing on the initial dataset to be verified to obtain the target dataset; The second building module 1303 is used to build a dual Transformer decoder model; The first determining module 1304 is used to determine ionic liquid candidates based on the target dataset and the dual Transformer decoder model; The second determining module 1305 is used to determine the current ionic liquid candidate as the optimal target ionic liquid in response to the verification that the ionic liquid candidate meets the preset conditions through density functional algorithm and COSMO-RS method.

[0077] Compared with existing technologies, this invention first constructs an initial dataset to be verified; performs preprocessing on the initial dataset to obtain a target dataset; constructs a dual Transformer decoder model; determines ionic liquid candidates based on the target dataset and the dual Transformer decoder model; and, in response to verification that the ionic liquid candidates meet preset conditions using density functional theory and the COSMO-RS method, determines the current ionic liquid candidate as the optimal target ionic liquid. By combining a dual Transformer decoder model with quantum chemical verification methods, it achieves efficient and accurate ionic liquid molecular design and verification, significantly improving the development efficiency of ionic liquids with high carbon dioxide solubility.

[0078] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to implement the steps in the above method embodiments when running.

[0079] Specifically, in this embodiment, the storage medium can be configured to store a computer program for performing the following steps: S201: Construct an initial dataset to be validated, wherein the initial dataset to be validated includes a cation SMILES dataset, an anion SMILES dataset, and an ionic liquid carbon dioxide solubility dataset; S202: Perform preprocessing on the initial dataset to be verified to obtain the target dataset; S203: Construct a dual Transformer decoder model; S204: Based on the target dataset and the dual Transformer decoder model, determine the ionic liquid candidates; S205: In response to the verification by density functional theory algorithm and COSMO-RS method that the ionic liquid candidate meets the preset conditions, the current ionic liquid candidate is determined to be the optimal target ionic liquid.

[0080] Specifically, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0081] Compared with existing technologies, this invention first constructs an initial dataset to be verified; performs preprocessing on the initial dataset to obtain a target dataset; constructs a dual Transformer decoder model; determines ionic liquid candidates based on the target dataset and the dual Transformer decoder model; and, in response to verification that the ionic liquid candidates meet preset conditions using density functional theory and the COSMO-RS method, determines the current ionic liquid candidate as the optimal target ionic liquid. By combining a dual Transformer decoder model with quantum chemical verification methods, it achieves efficient and accurate ionic liquid molecular design and verification, significantly improving the development efficiency of ionic liquids with high carbon dioxide solubility.

[0082] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps described in the method embodiments above.

[0083] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0084] Specifically, in this embodiment, the processor can be configured to perform the following steps via a computer program: S201: Construct an initial dataset to be validated, wherein the initial dataset to be validated includes a cation SMILES dataset, an anion SMILES dataset, and an ionic liquid carbon dioxide solubility dataset; S202: Perform preprocessing on the initial dataset to be verified to obtain the target dataset; S203: Construct a dual Transformer decoder model; S204: Based on the target dataset and the dual Transformer decoder model, determine the ionic liquid candidates; S205: In response to the verification by density functional theory algorithm and COSMO-RS method that the ionic liquid candidate meets the preset conditions, the current ionic liquid candidate is determined to be the optimal target ionic liquid.

[0085] Compared with existing technologies, this invention first constructs an initial dataset to be verified; performs preprocessing on the initial dataset to obtain a target dataset; constructs a dual Transformer decoder model; determines ionic liquid candidates based on the target dataset and the dual Transformer decoder model; and, in response to verification that the ionic liquid candidates meet preset conditions using density functional theory and the COSMO-RS method, determines the current ionic liquid candidate as the optimal target ionic liquid. By combining a dual Transformer decoder model with quantum chemical verification methods, it achieves efficient and accurate ionic liquid molecular design and verification, significantly improving the development efficiency of ionic liquids with high carbon dioxide solubility.

[0086] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0087] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0088] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0089] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0090] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0091] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0092] The embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.< / unknown> < / pad> < / end> < / start> < / pad> < / eos> < / sos>

Claims

1. A method for ionic liquid molecular design and carbon dioxide solubility verification, characterized in that, The method includes: Construct an initial dataset to be validated, wherein the initial dataset to be validated includes a cation SMILES dataset, an anion SMILES dataset, and an ionic liquid carbon dioxide solubility dataset; Perform preprocessing on the initial dataset to be verified to obtain the target dataset; Construct a dual Transformer decoder model; Based on the target dataset and the dual Transformer decoder model, ionic liquid candidates are determined. If the ionic liquid candidate is verified to meet the preset conditions by using density functional theory algorithm and COSMO-RS method, then the current ionic liquid candidate is determined to be the optimal target ionic liquid.

2. The method according to claim 1, characterized in that, The preprocessing of the initial dataset to be verified to obtain the target dataset includes: Perform normalization processing on the cation SMILES dataset and the anion SMILES dataset to obtain a standard dataset; The dataset of carbon dioxide solubility of ionic liquids is divided into training set, validation set and test set according to a preset ratio; The target dataset is determined based on the standard dataset, the training set, the validation set, and the test set.

3. The method according to claim 2, characterized in that, The standardization process performed on the cation SMILES dataset and the anion SMILES dataset to obtain a standard dataset includes: Invalid data in the cation SMILES dataset and the anion SMILES dataset are filtered and removed. By adding preset token instructions, a vocabulary containing preset characters corresponding to the cation SMILES dataset and the anion SMILES dataset is constructed. The vocabulary contains at least element symbols, chemical bonds, and charge symbols. Based on the vocabulary, a standard dataset is determined.

4. The method according to claim 3, characterized in that, The construction of the dual Transformer decoder model includes: Based on the cations and anions in the ionic liquid, a first Transformer decoder model and a second Transformer decoder model with conditional embedding are constructed respectively; wherein, the embedding conditions include the embedding temperature and pressure parameters; after training, both the first Transformer decoder model and the second Transformer decoder model have the function of obtaining ion-specific generation.

5. The method according to claim 4, characterized in that, The process of determining ionic liquid candidates based on the target dataset and the dual Transformer decoder model includes: Based on the target dataset and the dual Transformer decoder model, the parameters of the dual Transformer decoder model are optimized using a reinforcement learning algorithm, with solubility prediction as a positive reward and structural invalidity as a penalty, to generate ionic liquid candidates.

6. The method according to claim 5, characterized in that, The step of determining the current ionic liquid candidate as the optimal target ionic liquid after verifying that the ionic liquid candidate meets the preset conditions through density functional theory and the COSMO-RS method includes: If the structural stability of the ionic liquid candidate is verified to meet the first preset condition by density functional theory algorithm, and the thermodynamic rationality of the carbon dioxide solubility of the ionic liquid candidate is verified to meet the second preset condition by COSMO-RS method, then the current ionic liquid candidate is determined to be the optimal target ionic liquid.

7. The method according to claim 6, characterized in that, The method further includes: If the structural stability of the ionic liquid candidate does not meet the first preset condition when verified by density functional theory algorithm and / or the thermodynamic rationality of the carbon dioxide solubility of the ionic liquid candidate does not meet the second preset condition when verified by COSMO-RS method, then the reward function of the dual Transformer decoder model is adjusted and the process returns to the step of generating ionic liquid candidates by optimizing the parameters of the dual Transformer decoder model through reinforcement learning algorithm based on the target dataset and the dual Transformer decoder model, with the solubility prediction value as a positive reward and structural invalidity as a penalty.

8. A system for designing ionic liquid molecules and verifying carbon dioxide solubility, characterized in that, The system includes: The first construction module is used to construct the initial dataset to be verified, wherein the initial dataset to be verified includes the cation SMILES dataset, the anion SMILES dataset, and the ionic liquid carbon dioxide solubility dataset. The execution module is used to perform preprocessing on the initial dataset to be verified to obtain the target dataset; The second building block is used to build the dual Transformer decoder model; The first determining module is used to determine ionic liquid candidates based on the target dataset and the dual Transformer decoder model; The second determining module is used to determine the current ionic liquid candidate as the optimal target ionic liquid in response to the verification that the ionic liquid candidate meets the preset conditions through density functional algorithm and COSMO-RS method.

9. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to implement the method of any one of claims 1 to 7 when it is run.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to implement the method of any one of claims 1 to 7.

Images