Carbon dioxide absorbent development method and device and terminal equipment

Abstract

The invention discloses a carbon dioxide absorbent development method and device and terminal equipment. According to the method and the device, the property prediction model or the working condition simplification model is iteratively updated according to the first experimental sample obtained by each iteration until the difference value between the first performance data and the second performance data corresponding to the second experimental sample with the first preset proportion is smaller than the first preset range; outputting a second experimental sample with optimal second performance data as an optimal carbon dioxide absorbent; wherein after the property prediction model or the working condition simplification model is updated each time, the current newest property prediction model and the current newest working condition simplification model are combined to serve as a prior function of Bayesian optimization, and a first experiment sample is updated through Bayesian optimization; and screening a plurality of second experimental samples from the first experimental samples subjected to current iteration updating, and obtaining second performance data of the second experimental samples through the absorption process model. The development efficiency of the carbon dioxide absorbent can be improved.

Description

Technical Field

[0001] This application relates to the field of absorbent research and development, and particularly to a method, device, and terminal device for developing a carbon dioxide absorbent. Background Art

[0002] With the rapid development of industry, the commercial demand for carbon dioxide capture has gradually increased. Among many carbon capture technologies, the chemical absorption method has become the most technically mature post-combustion capture process due to its advantages in treating low-concentration flue gas and has achieved extensive commercial applications. The carbon dioxide absorbent is exactly the necessary material for the chemical absorption method, and the selection of the carbon dioxide absorbent plays a key role in the efficiency and cost of the absorption process. Currently, widely studied and applied carbon dioxide absorbents include alkanolamines, amino acid salts, potassium carbonate solutions, ammonia water, ionic liquids, and phase change absorbents, etc.

[0003] However, in actual production, problems such as high cost, high energy consumption, and imperfect process supporting still exist. Therefore, it is necessary to continuously research and develop new carbon dioxide absorbents to meet new industrial demands. Currently, the development methods of traditional carbon dioxide absorbents are inefficient, time-consuming, and costly, and it is difficult to meet industrial demands. Therefore, how to help researchers more quickly identify potential new materials and improve the research and development efficiency of carbon dioxide absorbents is a technical problem that needs to be solved currently. Summary of the Invention

[0004] This application provides a method, device, and terminal device for developing a carbon dioxide absorbent to solve the technical problem of how to improve the research and development efficiency of carbon dioxide absorbents.

[0005] To solve the above technical problem, an embodiment of this application provides a method for developing a carbon dioxide absorbent, including: iteratively updating a preset property prediction model or a preset working condition simplification model according to the first experimental samples obtained in each iteration until more than a first preset ratio of the second experimental samples meet a first preset condition, and outputting the second experimental sample with the optimal second performance data as the optimal carbon dioxide absorbent; the first preset condition is specifically that the difference between the first performance data and the second performance data corresponding to the second experimental sample is less than a first preset range; Among them, the first experimental samples used in the first iteration are generated and obtained according to the property prediction model before update in combination with the working condition simplification model; After each update of the property prediction model or the working condition simplification model, the current latest property prediction model is combined with the current latest working condition simplification model as the prior function of Bayesian optimization, and the first experimental samples are updated through Bayesian optimization, and several of the second experimental samples are screened from the first experimental samples updated in the current iteration; Each of the first experimental samples corresponds to a first performance data, and the first performance data is obtained according to the latest simplified working condition model; The second performance data is the performance data obtained by the second experimental sample through a preset absorption process model.

[0006] Compared with the prior art, the embodiments of the present application have the following beneficial effects: By introducing a property prediction model and a simplified working condition model as prior functions for Bayesian optimization, and using the property prediction model as a constraint in the optimization process, it is ensured that the finally obtained first experimental samples are effective and feasible. Further, the simplified working condition model is used to optimize the performance of the finally obtained first experimental samples under actual working conditions, so as to automatically generate first experimental samples with high potential and improve the development efficiency of carbon dioxide absorbents. In addition, by further screening a number of second experimental samples from the first experimental samples, and by judging whether the first performance data of the second experimental samples is consistent with the second performance data, the property prediction model and the simplified working condition model are further corrected. Combining with the automatic cycle process, the accuracy of the property prediction model and the simplified working condition model is automatically improved until a carbon dioxide absorbent that can pass the verification and has better performance is screened out, thereby effectively improving the development efficiency of carbon dioxide absorbents.

[0007] In some embodiments of the first aspect of the present application, the steps of obtaining the property prediction model before update include: obtaining carbon dioxide absorbent materials, and extracting the structural data of each carbon dioxide absorbent from the carbon dioxide absorbent materials; Through statistical sampling technology, a number of third experimental samples are randomly generated, and the structural data of each of the third experimental samples is determined according to the extracted structural data; Through high-throughput experimental technology, the measured property data of each of the third experimental samples is measured; A data set is constructed according to the measured property data and structural data of the third experimental samples, and the property prediction model is constructed according to the data set in combination with machine learning technology.

[0008] Compared with the prior art, the above embodiments have the following beneficial effects: By combining statistical sampling with machine learning technology, a property prediction model that can reflect the relationship between absorbent property data and structural data is constructed with a small amount of real experimental data, avoiding the use of traditional experimental modes to measure the specific property data of each absorbent. When facing the large-scale measurement of absorbent properties, the acquisition efficiency of absorbent property data can be effectively improved, thereby improving the subsequent screening and research and development efficiency of absorbents.

[0009] In some embodiments of the first aspect of the present application, the iterative update of the preset property prediction model or the preset simplified working condition model according to the first experimental samples obtained in each iteration includes: Among them, each of the first experimental samples corresponds to a property prediction data and a structure data, and the property prediction data and the structure data are obtained according to the latest property prediction model; Measure the actual property data of each of the first experimental samples through high-throughput experimental techniques; When the first experimental samples with a proportion lower than the second preset proportion meet the second preset condition, add the structure data and property data of each of the first experimental samples to the data set, and update the property prediction model according to the data set; the second preset condition is specifically that the difference between the property prediction data and the actual property data corresponding to the first experimental sample is less than the second preset range.

[0010] Compared with the prior art, the above embodiments have the following beneficial effects: When the property prediction data of the property prediction model is different from the actual property data obtained by actual measurement, it indicates that the accuracy of the property prediction model still needs to be further improved. Through the incremental learning method, while continuously generating new samples in the iterative loop, the prediction accuracy of the property prediction model is continuously and dynamically improved, effectively enhancing the reliability of the subsequent generated first experimental samples.

[0011] In some embodiments of the first aspect of the present application, the iterative update of the preset property prediction model or the preset working condition simplification model according to the first experimental samples obtained in each iteration further includes: When the proportion of the second experimental samples that meet the first preset condition is lower than the first preset proportion, update the working condition simplification model according to the second performance data of each of the second experimental samples.

[0012] Compared with the prior art, the above embodiments have the following beneficial effects: The first performance data of each first experimental sample can be predicted through the working condition simplification model. After further screening the first experimental samples to obtain the second experimental samples, the second performance data of the second experimental samples is further measured actually, and based on this, the accuracy of the first performance data predicted by the working condition simplification model is judged. When the working condition simplification model is inaccurate, it is updated and optimized, so as to dynamically adjust the working condition simplification model in the loop process, ensuring the efficiency and accuracy of subsequent screening of the first experimental samples.

[0013] In some embodiments of the first aspect of the present application, the screening of several of the second experimental samples from the first experimental samples updated in the current iteration includes: Input the property prediction data of each of the first experimental samples into a preset absorption process model, and combine mathematical optimization techniques to screen several second experimental samples from the first experimental samples.

[0014] Compared with the prior art, the above embodiments have the following beneficial effects: By combining the absorption process model with mathematical optimization techniques, further optimizing and screening the first experimental samples, and obtaining several more potential second experimental samples, thereby reducing the testing cost of subsequent pilot-scale devices and improving the development efficiency of absorbents.

[0015] In some embodiments of the first aspect of the present application, the second performance data is the performance data obtained by the second experimental samples through a preset absorption process model, including: Determine the process process parameters of each of the second experimental samples through the absorption process model; According to the process process parameters, test the second performance data of the corresponding second experimental samples through an automated pilot-scale device.

[0016] Compared with the prior art, the above embodiments have the following beneficial effects: While screening and obtaining the second experimental samples through the absorption process model, determine the optimal process process parameters under the optimal performance, and measure the true second performance data of the second experimental samples by the automated pilot-scale device according to the process process parameters to verify the accuracy of the simplified working condition model, thereby continuously improving the accuracy of the simplified working condition model in the whole cycle process, and improving the efficiency and accuracy of screening the first experimental samples.

[0017] In some embodiments of the first aspect of the present application, whenever the simplified working condition model is updated, it further includes: According to the latest simplified working condition model, combined with the principal component analysis method, re-determine the first property type of the property data output by the property prediction model; When the property type of the measured property data of the second experimental sample is inconsistent with the first property type, update the measured property data in the data set according to the first property type; Update the property prediction model according to the updated data set.

[0018] Compared with the prior art, the above embodiments have the following beneficial effects: Since there are many physical and chemical properties of absorbents, through the principal component analysis method, select the relatively key property types (i.e., the first property type) from all the physical and chemical properties of the absorbents. When the simplified working condition model is updated, it means that the first property type evaluated by the simplified working condition model has changed. If the currently evaluated first property type is different from the property type of the measured property data used when updating or training the property prediction model, the property prediction model needs to be updated again according to the measured property data of the first property type to ensure the accuracy of the property prediction model.

[0019] In a second aspect, an embodiment of the present application further provides a carbon dioxide absorbent development device, including: an iterative module and an experimental sample generation module; Among them, the iteration module iteratively updates a preset property prediction model or a preset working condition simplification model according to the first experimental samples obtained in each iteration until the second experimental samples greater than a first preset ratio meet a first preset condition, and outputs the second experimental samples with the optimal second performance data as the optimal carbon dioxide absorbent; the first preset condition is specifically that the difference between the first performance data and the second performance data corresponding to the second experimental samples is less than a first preset range. Among them, the first experimental samples used in the first iteration are generated and obtained according to the property prediction model before update in combination with the working condition simplification model. The experimental sample generation module is configured to, after each update of the property prediction model or the working condition simplification model, use the current latest property prediction model in combination with the current latest working condition simplification model as the prior function of Bayesian optimization, update the first experimental samples through Bayesian optimization, and screen a plurality of the second experimental samples from the first experimental samples updated in the current iteration. Each of the first experimental samples corresponds to a first performance data, and the first performance data is obtained according to the latest working condition simplification model. The second performance data is the performance data obtained by the second experimental samples through a preset absorption process model.

[0020] In some embodiments of the second aspect, the steps of obtaining the property prediction model before update include: Obtain carbon dioxide absorbent materials, and extract the structural data of each carbon dioxide absorbent from the carbon dioxide absorbent materials; Randomly generate a plurality of third experimental samples through statistical sampling techniques, and determine the structural data of each of the third experimental samples according to the extracted structural data; Measure the actual property data of each of the third experimental samples through high-throughput experimental techniques; Construct a data set according to the actual property data and structural data of the third experimental samples, and construct the property prediction model according to the data set in combination with machine learning techniques.

[0021] In a third aspect, the present application further provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the above-mentioned method for developing a carbon dioxide absorbent is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic flowchart of a method for developing a carbon dioxide absorbent provided in some embodiments of the present application; Figure 2Another process schematic diagram of a method for developing a carbon dioxide absorbent provided in some embodiments of the present application; Figure 3 Structural schematic diagram of a carbon dioxide absorbent development device provided in some embodiments of the present application; Figure 4 Another structural schematic diagram of a carbon dioxide absorbent development device provided in some embodiments of the present application. Detailed implementation manners

[0023] Currently, the development methods of traditional carbon dioxide absorbents are inefficient, time-consuming, and costly, and it is difficult to meet industrial requirements. Therefore, how to help researchers identify potential new materials faster and improve the R & D efficiency of carbon dioxide absorbents is a technical problem that needs to be solved currently.

[0024] To solve the above technical problems, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0025] Any experimental samples mentioned in the embodiments of the present application refer to carbon dioxide absorbents. The property prediction data or property measured data mentioned in the embodiments of the present application refer to the physical and chemical properties of carbon dioxide absorbents.

[0026] Embodiment 1 Please refer to Figure 1 , a method for developing a carbon dioxide absorbent provided in an embodiment of the present application, including S10 to S20, specifically: S10: Iteratively update a preset property prediction model or a preset simplified working condition model according to the first experimental samples obtained in each iteration until the second experimental samples greater than a first preset ratio meet a first preset condition, and output the second experimental samples with the optimal second performance data as the optimal carbon dioxide absorbent; the first preset condition is specifically that the difference between the first performance data and the second performance data corresponding to the second experimental samples is less than a first preset range; wherein, the first experimental samples used in the first iteration are generated and obtained according to the property prediction model before update in combination with the simplified working condition model. Further, in some embodiments of the present application, the steps of obtaining the property prediction model before update include: Obtain carbon dioxide absorbent materials, and extract the structural data of each carbon dioxide absorbent from the carbon dioxide absorbent materials; By means of statistical sampling techniques, a number of third experimental samples are randomly generated, and the structural data of each of the third experimental samples is determined according to the extracted structural data; By means of high-throughput experimental techniques, the measured property data of each of the third experimental samples is measured; A data set is constructed according to the measured property data and the structural data of the third experimental samples, and the property prediction model is constructed by combining the data set with machine learning techniques.

[0027] Preferably, referring to Figure 2 Another flow schematic diagram of the carbon dioxide absorbent development method shown, in some embodiments of the present application, before performing the iterative loop, a property prediction model is first constructed, and the corresponding property prediction model construction steps include S11 to S14: S11: According to known carbon dioxide absorbent materials, such as extracting the structural data of carbon dioxide absorbents from existing Chinese and foreign literature; wherein, the structural data includes but is not limited to the molecular characteristic structure of the carbon dioxide absorbent and the component formulation information of each group; S12: Using statistical sampling techniques, according to the extracted structural data, a series of carbon dioxide absorbents with different molecular structures or carbon dioxide absorbents with different component contents in the formulation are generated as the third experimental samples; S13: Using high-throughput experimental techniques, testing the measured property data of the third experimental samples; the type of property of the measured property data obtained here is the key property determined from all the physical and chemical properties of the carbon dioxide absorbent, and the determination of this key property can be obtained by combining a preset working condition simplification model with principal component analysis, or specified in advance; S14: Using machine learning techniques, combining the measured property data and the structural data of the third experimental samples, constructing an association model between the key properties of the carbon dioxide absorbent and the characteristic molecular structure and / or component ratio as the property prediction model; and combining external literature books and other materials to further expand the training samples and / or provide cross-validation to improve the robustness of the property prediction model.

[0028] Preferably, in some embodiments of the present application, it should be noted that the above S11 to S14 are the processes for initially training and constructing a property prediction model. When the third experimental sample is incrementally updated (or when the dataset is incrementally updated), and then S13 to S14 are executed, it is the process steps for retraining the property prediction model. It can be seen from the above preferred embodiments that by combining statistical sampling with machine learning techniques, a property prediction model that can reflect the correlation between absorbent property data and structural data is constructed with a small amount of real experimental data, avoiding the use of traditional experimental modes to measure the specific property data of each absorbent. When facing large-scale absorbent property measurements, the acquisition efficiency of absorbent property data can be effectively improved, thereby improving the subsequent screening and R & D efficiency of absorbents.

[0029] Preferably, in some embodiments of the present application, known carbon dioxide absorbents include, but are not limited to, amine compounds, alcohol compounds, ether compounds, potassium carbonate solutions, hydroxide solutions, ammonia water, amino acid solutions, ionic liquids, etc.

[0030] Preferably, in some embodiments of the present application, there is a possibility of using multiple chemical substances in combination for the carbon dioxide absorbent, such as forming binary, ternary or even multi-component mixed absorbents and other forms.

[0031] Preferably, the statistical sampling techniques adopted in some embodiments of the present application include, but are not limited to, orthogonal experimental design, Latin hypercube sampling, and clustering sampling.

[0032] Preferably, the high-throughput experimental techniques adopted in some embodiments of the present application include, but are not limited to, high-throughput sampling, high-throughput weighing, high-throughput liquid preparation, high-throughput characterization testing, and corresponding material and vessel transfer techniques.

[0033] Preferably, in some embodiments of the present application, property data such as property prediction data and property measured data corresponding to each experimental sample include, but are not limited to, viscosity, specific heat capacity, CO2 absorption capacity, and CO2 absorption heat, etc.

[0034] Preferably, in some embodiments of the present application, the input of the property prediction model can be the structural data corresponding to each experimental sample (i.e., the molecular structure and / or mixing ratio of the carbon dioxide absorbent), and the output is the property data determined as the first property type currently (i.e., the key property).

[0035] Preferably, the machine learning methods adopted for constructing the property prediction model in some embodiments of the present application include, but are not limited to, neural networks, random forests, support vector machines, Kriging interpolation fitting, etc.

[0036] Preferably, in some embodiments of the present application, artificial intelligence technologies for text processing and reading are used to process foreign language literature books, and further extract the structural data and measured property data of newly added experimental samples.

[0037] Preferably, in some embodiments of the present application, during the process of generating the first experimental sample, the working condition simplification model is used to predict the first performance data of the carbon dioxide absorbent corresponding to randomly generated sample points, so as to perform a screening operation according to the first performance data, and the property prediction model is used to constrain the feasibility of the first experimental samples screened by the working condition simplification model. Therefore, when performing Bayesian optimization, the working condition simplification model and the property prediction model serve as the prior functions of Bayesian optimization, and the corresponding optimization objective is the first performance data, which includes but is not limited to the minimum desorption energy consumption.

[0038] Furthermore, in some embodiments of the present application, the iterative update of the preset property prediction model or the preset working condition simplification model according to the first experimental sample obtained in each iteration includes: Wherein, each of the first experimental samples corresponds to a property prediction data and a structural data, and the property prediction data and the structural data are obtained according to the latest property prediction model; Measure the measured property data of each of the first experimental samples through high-throughput experimental techniques; When less than a second preset ratio of the first experimental samples meet the second preset condition, add the structural data and property data of each of the first experimental samples to the data set, and update the property prediction model according to the data set; the second preset condition is specifically that the difference between the property prediction data and the measured property data corresponding to the first experimental sample is less than a second preset range.

[0039] Preferably, referring to Figure 2 Another flow schematic diagram of the carbon dioxide absorbent development method shown, in some embodiments of the present application, after obtaining the initially constructed property prediction model, the first experimental sample can be generated through the currently obtained property prediction model and the working condition simplification model, and the property prediction model is iteratively optimized through the first experimental sample generated in each iteration. Next, the iterative optimization process of the property prediction model is described through S15. S15 includes S151 and S152, where S151 is the process of generating the first experimental sample through the property prediction model and the working condition simplification model. After obtaining the first experimental sample required for each iteration through S151, it is judged through S152 whether it is necessary to update the property prediction model. S151 to S152 are specifically as follows: S151: Using the property prediction model obtained or updated by S14, combining the simplified models of the operating conditions of the rate-based absorption tower and the equilibrium-based desorption tower, and selecting a number of highly uncertain sample points and potential optimal sample points through Bayesian optimization technology, so as to generate three types of mixed absorbents as shown in Figure 2 , which are used as the first experimental samples obtained in the current iteration.

[0040] S152: Through high-throughput experimental technology, measure the actual measured data of the properties of each of the first experimental samples obtained in the current iteration, and compare the predicted property data of the first experimental samples with the actual measured data. When the proportion of the second experimental samples whose deviation between the predicted result and the experimental result is less than the second preset range does not reach 90% or other set second preset proportions, then use the first experimental samples obtained in the current iteration as new samples, and re-execute S13 to S14 to retrain the property prediction model until the consistency between the predicted result and the experimental result meets 90% or other set second preset proportions.

[0041] Preferably, in some embodiments of the present application, the simplified model of the operating conditions can be specifically constructed by only simulating the absorption tower and the desorption tower through GAMS simulation software to simplify the carbon dioxide absorption process.

[0042] Preferably, in some embodiments of the present application, any uncertainty quantification method can be used to evaluate the confidence or reliability of the model prediction for the selection of the highly uncertain sample points and potential optimal sample points in S15.

[0043] It can be seen from the above preferred embodiments that when the predicted property data of the property prediction model is different from the actual measured property data obtained by actual measurement, it indicates that the accuracy of the property prediction model still needs to be further improved. Through the way of incremental learning, while continuously generating new samples in the iterative loop, continuously and dynamically improve the prediction accuracy of the property prediction model, and effectively improve the reliability of the subsequent generated first experimental samples.

[0044] Preferably, in some embodiments of the present application, the input data of the simplified model of the operating conditions is the process parameters corresponding to the carbon dioxide absorbent and the predicted property data output by the property prediction model, and the output includes but is not limited to the performance data of the carbon dioxide absorbent.

[0045] S20: After each update of the property prediction model or the simplified working condition model, the current latest property prediction model is combined with the current latest simplified working condition model as the prior function of Bayesian optimization. The first experimental samples are updated through Bayesian optimization, and several second experimental samples are screened from the first experimental samples updated in the current iteration; each first experimental sample corresponds to a first performance data, and the first performance data is obtained according to the latest simplified working condition model; the second performance data is the performance data obtained by the second experimental samples through a preset absorption process model.

[0046] Preferably, referring to Figure 2 Another schematic flow chart of the carbon dioxide absorbent development method shown, in some embodiments of the present application, when any one of the property prediction model or the simplified working condition model is updated each time, the steps S151 to S152 included in S15 need to be re-executed to re-obtain the first experimental samples.

[0047] Furthermore, in some embodiments of the present application, the screening of several second experimental samples from the first experimental samples updated in the current iteration includes: Input the property prediction data of each first experimental sample into a preset absorption process model, and combine mathematical optimization techniques to screen several second experimental samples from the first experimental samples.

[0048] Preferably, referring to Figure 2 Another schematic flow chart of the carbon dioxide absorbent development method shown, in some embodiments of the present application, step S16 of screening several second experimental samples from the first experimental samples is specifically: S16: Through the overall process optimization based on the Aspen Plus process simulation software, a strict absorption process model is constructed. The absorption process model can adopt an equilibrium stage (equilibrium) or a rate-based (rate-based) model, and using the property prediction data of the first absorbent as the input, combined with mathematical optimization techniques, 2 to 3 second experimental samples and the corresponding process parameters of each second experimental sample are obtained.

[0049] It can be seen from the above preferred embodiments that by combining the absorption process model with mathematical optimization techniques, the first experimental samples are further optimized and screened to obtain several more potential second experimental samples, thereby reducing the test cost of the subsequent pilot plant and improving the absorbent development efficiency.

[0050] Furthermore, in some embodiments of the present application, the second performance data is the performance data obtained by the second experimental samples through a preset absorption process model, including: Determine the process parameters of each of the second experimental samples through the absorption process model; According to the process parameters, test the second performance data of the corresponding second experimental samples through the automated pilot plant.

[0051] While screening and obtaining the second experimental samples through the absorption process model, determine the optimal process parameters under the optimal performance, and measure the real second performance data of the second experimental samples according to the process parameters through the automated pilot plant to verify the accuracy of the simplified working condition model, so as to continuously improve the accuracy of the simplified working condition model in the whole cycle process, and improve the efficiency and accuracy of screening the first experimental samples.

[0052] Further, in some embodiments of the present application, the iterative update of the preset property prediction model or the preset simplified working condition model according to the first experimental samples obtained in each iteration further includes: When the proportion of the second experimental samples that meet the first preset condition is lower than the first preset proportion, update the simplified working condition model according to the second performance data of each of the second experimental samples.

[0053] The first performance data of each first experimental sample can be predicted through the simplified working condition model. After further screening the first experimental samples to obtain the second experimental samples, the second performance data of the second experimental samples are further measured actually, and based on this, the accuracy of the first performance data predicted by the simplified working condition model is judged. When the simplified working condition model is inaccurate, it is updated and optimized, so as to dynamically adjust the simplified working condition model in the cycle process and ensure the efficiency and accuracy of subsequent screening of the first experimental samples.

[0054] Further, in some embodiments of the present application, whenever the simplified working condition model is updated, it further includes: According to the latest simplified working condition model, combined with the principal component analysis method, re-determine the first property type of the property data output by the property prediction model; When the property type of the measured property data of the second experimental sample is inconsistent with the first property type, update the measured property data in the data set according to the first property type; Update the property prediction model according to the updated data set.

[0055] Due to the wide variety of physical and chemical properties of absorbents, the relatively key property types (i.e., the first property types) are selected from all the physical and chemical properties of the absorbents by the principal component analysis method. When the simplified operating condition model is updated, it means that the first property types evaluated by the simplified operating condition model change. If the currently evaluated first property types are different from the property types of the property measured data used when updating or training the property prediction model currently, the property prediction model needs to be updated again according to the property measured data of the first property types to ensure the accuracy of the property prediction model.

[0056] Preferably, referring to Figure 2 Another schematic flow diagram of the carbon dioxide absorbent development method shown in the figure, in some embodiments of the present application, when a number of second experimental samples are obtained, the following step S17 needs to be executed to evaluate whether to continue to iteratively update the property prediction model or the simplified operating condition model according to the obtained second experimental samples. S17 is specifically as follows: S17: Use a fully automated pilot plant to test the second performance data of the second experimental samples obtained in S16. At this time, the obtained second performance data is the performance data measured according to the experiment, and compare the second performance data with the first performance data predicted by the simplified operating condition model. If the difference between the two performance data of more than the first preset proportion of the second experimental samples is within the first preset range (for example, the difference between the two performance data of more than 80% of the second experimental samples is between 0 and 2%, and the present application does not strictly limit the first preset proportion and the first preset range), then select the carbon dioxide absorbent with the best performance according to the second performance data; if the gap is large (for example, the difference between the two performance data of less than 80% of the second experimental samples is between 0 and 2%, then it can be considered that the gap is large), then update the simplified operating condition model according to the second performance data. When the simplified operating condition model is updated, it is necessary to re-evaluate the key properties (i.e., the measured property data of the first property types) according to the simplified operating condition model again. At this time, two results are allowed: 1) Obtain the same key properties as the previous evaluation result, and repeat S15 to S17; 2) Obtain different key properties from the previous evaluation result, and repeat S13 to S17.

[0057] Preferably, in some embodiments of the present application, the tests of the automated pilot plant include but are not limited to determining the operating procedures, safety magnification factors, and basic feasibility of process design, etc. The true performance tests of the absorbent include but are not limited to endothermic heat consumption, absorbent loss rate, and CO2 capture rate, etc.

[0058] Preferably, in some embodiments of the present application, the updated content of the simplified operating condition model includes but is not limited to optimizing the internals and tray numbers of the absorption tower and / or desorption tower, the liquid holdup in the tower, the mass transfer rate, etc.

[0059] Preferably, in some embodiments of the present application, the process parameters include, but are not limited to, the operating temperature, operating pressure, etc. of the absorption tower or the desorption tower.

[0060] In summary, it can be seen that a method for developing a carbon dioxide absorbent provided by the embodiments of the present application has the following beneficial effects: by introducing a property prediction model and a working condition simplification model as the prior functions of Bayesian optimization, and using the property prediction model as the constraint in the optimization process, it is ensured that the finally obtained first experimental samples are effective and feasible. Further, the working condition simplification model is used to optimize the performance of the finally obtained first experimental samples under actual working conditions, so as to automatically generate first experimental samples with high potential and improve the development efficiency of the carbon dioxide absorbent. In addition, by further screening a number of second experimental samples from the first experimental samples, and by judging whether the first performance data of the second experimental samples is consistent with the second performance data, the property prediction model and the working condition simplification model are further corrected. Combining with the automatic cycle process, the accuracy of the property prediction model and the working condition simplification model is automatically improved until a carbon dioxide absorbent that can pass the verification and has better performance is screened out, thus effectively improving the development efficiency of the carbon dioxide absorbent.

[0061] Embodiment 2 Reference Figure 3 , a carbon dioxide absorbent development device provided in some embodiments of the present application, includes: an iteration module 11 and an experimental sample generation module 12.

[0062] Further, in some embodiments of the present application, the iteration module 11 iteratively updates a preset property prediction model or a preset working condition simplification model according to the first experimental samples obtained in each iteration until more than a first preset ratio of the second experimental samples meet the first preset condition, and outputs the second experimental sample with the best second performance data as the optimal carbon dioxide absorbent; the first preset condition is specifically that the difference between the first performance data and the second performance data corresponding to the second experimental sample is less than a first preset range; wherein, the first experimental samples used in the first iteration are generated and obtained according to the property prediction model before update in combination with the working condition simplification model; the experimental sample generation module 12 is used to, after each update of the property prediction model or the working condition simplification model, use the current latest property prediction model in combination with the current latest working condition simplification model as the prior function of Bayesian optimization, update the first experimental samples through Bayesian optimization, and screen a number of the second experimental samples from the first experimental samples updated in the current iteration; each of the first experimental samples corresponds to a first performance data, and the first performance data is obtained according to the latest working condition simplification model; the second performance data is the performance data obtained by the second experimental sample through a preset absorption process model.

[0063] Further, in some embodiments of the present application, the step of obtaining the property prediction model before the update includes: obtaining carbon dioxide absorbent materials, and extracting the structural data of each carbon dioxide absorbent from the carbon dioxide absorbent materials; randomly generating a number of third experimental samples through statistical sampling techniques, and determining the structural data of each of the third experimental samples according to the extracted structural data; measuring the measured property data of each of the third experimental samples through high-throughput experimental techniques; constructing a data set according to the measured property data and the structural data of the third experimental samples, and constructing the property prediction model according to the data set in combination with machine learning techniques.

[0064] Further, in some embodiments of the present application, the iterative update of the preset property prediction model or the preset working condition simplification model according to the first experimental samples obtained in each iteration includes: wherein, each of the first experimental samples corresponds to a property prediction data and a structural data, and the property prediction data and the structural data are obtained according to the latest property prediction model. Measuring the measured property data of each of the first experimental samples through high-throughput experimental techniques; when the proportion of the first experimental samples lower than the second preset ratio satisfies the second preset condition, adding the structural data and the property data of each of the first experimental samples to the data set, and updating the property prediction model according to the data set; the second preset condition is specifically that the difference between the property prediction data and the measured property data corresponding to the first experimental sample is less than the second preset range.

[0065] Further, in some embodiments of the present application, the iterative update of the preset property prediction model or the preset working condition simplification model according to the first experimental samples obtained in each iteration further includes: when the proportion of the second experimental samples satisfying the first preset condition is lower than the first preset ratio, updating the working condition simplification model according to the measured property data of each of the second experimental samples.

[0066] Further, in some embodiments of the present application, the screening of a number of the second experimental samples from the first experimental samples after the current iterative update includes: inputting the property prediction data of each of the first experimental samples into a preset absorption process model, and screening a number of second experimental samples from the first experimental samples in combination with mathematical optimization techniques.

[0067] Further, in some embodiments of the present application, the second performance data is the performance data obtained by the second experimental samples through a preset absorption process model, including: determining the process process parameters of each of the second experimental samples through the absorption process model; testing the second performance data of the corresponding second experimental samples through an automated pilot plant according to the process process parameters.

[0068] Further, in some embodiments of the present application, whenever the simplified operating condition model is updated, it further includes: according to the latest simplified operating condition model, combining with the principal component analysis method, re-determining the first property type of the property data output by the property prediction model; when the property type of the measured property data of the second experimental sample is inconsistent with the first property type, updating the measured property data in the data set according to the first property type; and updating the property prediction model according to the updated data set.

[0069] Preferably, referring to Figure 4 , which is another structural schematic diagram of a carbon dioxide absorbent development device provided in some embodiments of the present application, including: a data acquisition module 13, a data processing module 14, a high-throughput chemical experiment module 15, a construction and iteration module 16 of a property prediction model, a process modeling and numerical optimization module 17, and an automated pilot module 18.

[0070] Further, in some embodiments of the present application, the data acquisition module 13 is used to automatically extract the molecular structure characteristics and mixture formula composition information of known carbon dioxide absorbents.

[0071] Further, in some embodiments of the present application, the data processing module 14 is used for statistical sampling processing, new data updating, and resampling based on Bayesian optimization after the data acquisition module 13.

[0072] Further, in some embodiments of the present application, the high-throughput chemical experiment module 15 includes, but is not limited to, high-throughput sampling, weighing, liquid preparation integrated units, high-throughput characterization test units, and corresponding material and utensil automatic transfer units and other experimental operation units.

[0073] Further, in some embodiments of the present application, the construction and iteration module 16 of the property prediction model is used to set termination conditions and determine whether the termination conditions in the iteration process are met, such as whether the prediction results of each model are consistent with the experimental results.

[0074] Further, in some embodiments of the present application, the process modeling and numerical optimization module 17 is used to simulate and optimize the entire process of the second experimental sample screened after model iteration, and the software used includes, but is not limited to, AspenPlus and GAMS.

[0075] Further, in some embodiments of the present application, the control program and functions of the automated pilot module 18 include, but are not limited to, realizing feeding automation, storing and collecting temperature and pressure data in each instrument and pipeline at regular intervals, and being equipped with a mobile terminal remote operation control function.

[0076] Further, in some embodiments of the present application, in the data acquisition module 13, its data sources include, but are not limited to, known carbon dioxide absorbents, newly reported carbon dioxide absorbents in the literature, and carbon dioxide absorbents with excellent comprehensive performance generated by the high-throughput chemical experiment module 15, that is, the first experimental samples.

[0077] Further, in some embodiments of the present application, in the data acquisition module 13, the data accuracy and effectiveness can be guaranteed by artificial intelligence technologies for text reading and processing.

[0078] Further, in some embodiments of the present application, in the data processing module 14, the resampling after the new data update can use the method of adaptive sampling in active learning to improve the reliability of the data.

[0079] Further, in some embodiments of the present application, in the high-throughput chemical experiment module 15, its operation connection can be realized by an automatic transfer module for materials and utensils.

[0080] Further, in some embodiments of the present application, in the module 16 for constructing and iterating the property prediction model, two related contents of the property prediction model are the key physical and chemical properties of the carbon dioxide absorbent and its molecular structure and / or mixing ratio.

[0081] It can be understood that the above device item embodiments correspond to the method item embodiments of the present invention. A carbon dioxide absorbent development device provided by the embodiments of the present invention can implement any one of the method item embodiments of the present invention, that is, the carbon dioxide absorbent development method provided in Embodiment 1.

[0082] In summary, it can be seen that a carbon dioxide absorbent development device provided by the embodiments of the present application has the following beneficial effects: by introducing a property prediction model and a working condition simplification model as prior functions of Bayesian optimization, and using the property prediction model as a constraint in the optimization process, it is ensured that the finally obtained first experimental samples are effective and feasible. Further, the working condition simplification model is used to optimize the performance of the finally obtained first experimental samples in the actual working conditions, so as to automatically generate first experimental samples with high potential and improve the development efficiency of carbon dioxide absorbents. In addition, by further screening out several second experimental samples from the first experimental samples, and by judging whether the first performance data and the second performance data of the second experimental samples are consistent, the property prediction model and the working condition simplification model are further corrected. Combining with the automatic cycle process, the accuracy of the property prediction model and the working condition simplification model is automatically improved until carbon dioxide absorbents that can pass the verification and have better performance are screened out, thereby effectively improving the development efficiency of carbon dioxide absorbents.

[0083] Embodiment 3 Based on the embodiments of the above carbon dioxide absorbent development method, another embodiment of the present application provides a carbon dioxide absorbent development terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the carbon dioxide absorbent development method of any embodiment of the present application is implemented.

[0084] Exemplarily, in this embodiment, the computer program may be divided into one or more modules. The one or more modules are stored in the memory and executed by the processor to complete the present application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the carbon dioxide absorbent development device.

[0085] The carbon dioxide absorbent development device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The carbon dioxide absorbent development terminal device may include, but is not limited to, a processor and a memory.

[0086] The processor may be a Central Processing Unit (CPU), or may 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. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the carbon dioxide absorbent development device, and connects various parts of the entire carbon dioxide absorbent development device through various interfaces and lines. The memory can be used to store the computer programs and / or modules. By running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory, the processor realizes various functions of the carbon dioxide absorbent development device. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0087] Embodiment 4 Based on the embodiments of the above carbon dioxide absorbent development method, another embodiment of the present application provides a storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the storage medium is located to execute the carbon dioxide absorbent development method of any embodiment of the present application.

[0088] In this embodiment, the above storage medium is a computer-readable storage medium, and the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice within the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0089] The specific embodiments described above further elaborate on the purpose, technical solution, and beneficial effects of the present application. It should be understood that the above description is only specific embodiments of the present application and is not used to limit the protection scope of the present application. In particular, for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for developing a carbon dioxide absorbent, characterized in that: include: Iteratively update the preset property prediction model or the preset simplified operating condition model according to the first experimental sample obtained in each iteration until the second experimental sample greater than the first preset ratio meets the first preset condition, and output the second experimental sample with the best second performance data as the optimal carbon dioxide absorbent; the first preset condition is specifically that the difference between the first performance data and the second performance data corresponding to the second experimental sample is less than a first preset range; The first experimental sample used in the first iteration is generated and obtained based on the property prediction model before the update and the simplified working condition model; Each time the property prediction model or the simplified operating condition model is updated, the latest property prediction model is combined with the latest simplified operating condition model as a priori function of Bayesian optimization, the first experimental sample is updated through Bayesian optimization, and a number of the second experimental samples are selected from the first experimental samples after the current iterative update; Each of the first experimental samples corresponds to a first performance data, and the first performance data is obtained according to the latest simplified model of the operating condition; The second performance data is performance data obtained by the second experimental sample through a preset absorption process model.

2. A method for developing a carbon dioxide absorbent according to claim 1, characterized in that: The step of obtaining the property prediction model before updating comprises: Acquiring carbon dioxide absorbent data, and extracting structural data of each carbon dioxide absorbent from the carbon dioxide absorbent data; Randomly generate a number of third experimental samples by statistical sampling technology, and determine the structural data of each of the third experimental samples according to the extracted structural data; Measuring the measured data of the properties of each of the third experimental samples by high-throughput experimental technology; A data set is constructed based on the measured property data and structural data of the third experimental sample, and the property prediction model is constructed based on the data set in combination with machine learning technology.

3. A method for developing a carbon dioxide absorbent according to claim 2, characterized in that: The iterative updating of the preset property prediction model or the preset simplified working condition model according to the first experimental samples obtained in each iteration includes: wherein each of the first experimental samples corresponds to a property prediction data and a structure data, and the property prediction data and the structure data are obtained according to the latest property prediction model; Measuring the measured data of the properties of each of the first experimental samples by high-throughput experimental technology; When the first experimental samples below the second preset ratio meet the second preset condition, the structural data and property data of each of the first experimental samples are added to the data set, and the property prediction model is updated according to the data set; the second preset condition is specifically that the difference between the property prediction data and the property measured data corresponding to the first experimental sample is less than a second preset range.

4. A method for developing a carbon dioxide absorbent according to claim 3, characterized in that: The method of iteratively updating a preset property prediction model or a preset simplified operating condition model based on the first experimental sample obtained in each iteration also includes: when the proportion of second experimental samples that meet the first preset condition is lower than the first preset proportion, updating the simplified operating condition model based on the second performance data of each of the second experimental samples.

5. A method for developing a carbon dioxide absorbent according to claim 3, characterized in that: The selecting a plurality of the second experimental samples from the first experimental samples updated in the current iteration includes: The property prediction data of each of the first experimental samples is input into a preset absorption process model, and a number of second experimental samples are selected from the first experimental samples in combination with mathematical optimization technology.

6. A method for developing a carbon dioxide absorbent according to claim 3, characterized in that: The second performance data is performance data obtained by the second experimental sample through a preset absorption process model, including: Determining the process parameters of each of the second experimental samples by using the absorption process model; According to the process parameters, second performance data corresponding to the second experimental sample is tested by the automated pilot plant.

7. A method for developing a carbon dioxide absorbent according to any one of claims 2 to 6, characterized in that: Whenever the simplified operating condition model is updated, it also includes: According to the latest simplified working condition model, combined with the principal component analysis method, the first property type of the property data output by the property prediction model is re-determined; When the property type of the measured property data of the second experimental sample is inconsistent with the first property type, updating the measured property data in the data set according to the first property type; The property prediction model is updated according to the updated data set.

8. A carbon dioxide absorbent development device, characterized in that: include: An iteration module and an experimental sample generation module; wherein the iteration module iteratively updates a preset property prediction model or a preset simplified operating condition model according to a first experimental sample obtained in each iteration, until a second experimental sample greater than a first preset ratio satisfies a first preset condition, and outputs a second experimental sample with the best second performance data as the best carbon dioxide absorbent; the first preset condition is specifically that the difference between the first performance data and the second performance data corresponding to the second experimental sample is less than a first preset range; The first experimental sample used in the first iteration is generated and obtained based on the property prediction model before the update and the simplified working condition model; The experimental sample generation module is used for, after each update of the property prediction model or the simplified operating condition model, using the latest property prediction model combined with the latest simplified operating condition model as a priori function of Bayesian optimization, updating the first experimental sample through Bayesian optimization, and selecting a number of the second experimental samples from the first experimental sample after the current iterative update; Each of the first experimental samples corresponds to a first performance data, and the first performance data is obtained according to the latest simplified model of the operating condition; The second performance data is performance data obtained by the second experimental sample through a preset absorption process model.

9. A carbon dioxide absorbent development device as claimed in claim 8, characterized in that: The step of obtaining the property prediction model before updating comprises: Acquiring carbon dioxide absorbent data, and extracting structural data of each carbon dioxide absorbent from the carbon dioxide absorbent data; Randomly generate a number of third experimental samples by statistical sampling technology, and determine the structural data of each of the third experimental samples according to the extracted structural data; Measuring the measured data of the properties of each of the third experimental samples by high-throughput experimental technology; A data set is constructed based on the measured property data and structural data of the third experimental sample, and the property prediction model is constructed based on the data set in combination with machine learning technology.

10. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a method for developing a carbon dioxide absorbent according to any one of claims 1 to 7 is implemented.

Images