Quantum chemistry calculation result intelligent analysis and error correction system and method

Abstract

The invention belongs to the technical field of material information, and particularly relates to a quantum chemistry calculation result intelligent analysis and error correction system and method. The calculation result analysis module is in communication connection with the data preprocessing module and is used for analyzing the standardized data based on a preset rule; the machine learning analysis module is in communication connection with the calculation result analysis module, receives an analysis result sent by the calculation result analysis module, and performs deep analysis on the analysis result based on a machine learning algorithm; and the error correction module is in communication connection with the calculation result analysis module and the machine learning analysis module and is used for performing error correction on the quantum chemistry calculation result based on the analysis result and the analysis result, so that the overall analysis efficiency is improved, the analysis result is continuously optimized, and the practical value is remarkable.

Description

Technical Field

[0001] The present invention belongs to the field of materials information technology, in particular to an intelligent analysis and error correction system and method for quantum chemical calculation results. Background Art

[0002] With the rapid development of materials science and computational chemistry, quantum chemical calculations have become an important tool for designing new materials, predicting material properties, and understanding material behavior. However, despite the significant progress made in quantum chemical calculation methods in the past few decades, the accuracy and reliability of their calculation results still face many challenges.

[0003] Existing methods for analyzing quantum chemical calculation results mainly rely on the experience and intuition of researchers, and often have problems such as strong subjectivity, low efficiency, and difficulty in processing large-scale data. For example, when dealing with the electronic structure calculation results of complex material systems, researchers usually need to manually check band diagrams, density of states diagrams, etc., which is not only time-consuming and laborious, but also prone to overlooking some potential anomalies or errors.

[0004] In addition, the systematic errors commonly existing in quantum chemical calculations are also an urgent problem to be solved. Taking density functional theory (DFT) calculations as an example, different exchange-correlation functionals have different degrees of deviation in predicting physical quantities such as band gaps and binding energies. Most of the existing error correction methods are based on empirical formulas or simple linear regression, and it is difficult to adapt to the complex error patterns under different material systems and calculation conditions.

[0005] Existing quantum chemical calculation auxiliary tools generally lack the effective use of domain knowledge. This results in relatively limited generalization ability and interpretability of the system when dealing with the calculation results of new materials or under extreme conditions. Summary of the Invention

[0006] In view of the above problems, the present invention proposes an intelligent analysis and error correction system and method for quantum chemical calculation results. The system aims to achieve automatic and intelligent analysis and correction of quantum chemical calculation results through advanced data processing technologies, machine learning algorithms, and knowledge engineering methods, thereby significantly improving the accuracy, reliability, and interpretability of the calculation results.

[0007] The present invention proposes an intelligent analysis and error correction system for quantum chemical calculation results, including:

[0008] A data preprocessing module, configured to collect the original data obtained from quantum chemical calculations and perform normalization processing on the original data;

[0009] A calculation result analysis module, configured to receive the data after normalization processing sent by the data preprocessing module and analyze the data after normalization processing based on preset rules;

[0010] A machine learning analysis module, configured to receive the analysis result sent by the calculation result analysis module and deeply analyze the analysis result based on a machine learning algorithm;

[0011] An error correction module, which receives the analysis result sent by the calculation result analysis module and the analysis result sent by the machine learning analysis module, and corrects the error of the quantum chemistry calculation result based on the analysis result and the analysis result.

[0012] Preferably, the data preprocessing module includes:

[0013] A data acquisition unit, configured to obtain original calculation data from quantum chemistry calculation software;

[0014] A data standardization unit, configured to receive the original calculation data sent by the data acquisition unit and perform normalization processing on the original calculation data;

[0015] A data cleaning unit, configured to receive the data after normalization processing sent by the data standardization unit and remove noise and outliers in the data after normalization processing.

[0016] Preferably, the calculation result analysis module includes:

[0017] A rule library unit, configured to store preset analysis rules, where the analysis rules include: an outlier identification rule based on data distribution; an accuracy verification rule based on the comparison between the calculation result and the real experimental result; a cleaning and normalization rule based on data characteristics;

[0018] A rule application unit, which obtains the analysis rules in the rule library unit, selects and applies the analysis rules corresponding to the data after standardization processing;

[0019] A result analysis unit, configured to process the data after standardization processing by applying the corresponding analysis rules, so as to delete the following problematic data from the data after standardization processing: data with outliers, data with low accuracy, and data features with low correlation with the prediction target, and obtain the data after standardization processing with problematic data removed.

[0020] Preferably, the machine learning analysis module includes:

[0021] A feature extraction unit, configured to extract key features from the data after standardization processing with problematic data removed, and perform vectorization processing on the key features;

[0022] A feature selection unit, configured to receive the vectorized features sent by the feature extraction unit, obtain the feature importance scores of the key features, and select the optimal feature subset based on the feature importance scores;

[0023] A model training unit for constructing a prediction model for quantum chemistry calculation results, receiving the optimal feature subset sent by the feature selection unit as the input of the prediction model for quantum chemistry calculation results.

[0024] The output of the prediction model for quantum chemistry calculation results is the properties of the material, where the material is composed of periodic molecules, and the periodic molecules are the research objects of the quantum chemistry calculation results.

[0025] Use a supervised learning method to train the prediction model for quantum chemistry calculation results.

[0026] A prediction analysis unit for using a new result input unit to obtain the quantum chemistry calculation results to be predicted. After being processed by a data preprocessing module and a calculation result parsing module, it is input into the trained prediction model for quantum chemistry calculation results to obtain the predicted value and confidence level of the material properties.

[0027] Preferably, the error correction module includes:

[0028] An error pattern recognition unit for obtaining a systematic error pattern library of the predicted value of the material properties and analyzing the systematic error pattern of the predicted value of the material properties.

[0029] A correction parameter generation unit for obtaining corresponding correction parameters from the systematic error pattern library of the predicted value of the material properties based on the identified error pattern.

[0030] A correction execution unit that receives the correction parameters sent by the correction parameter generation unit and applies the correction parameters to correct the predicted value of the material properties.

[0031] A correction verification unit for cross - validating the predicted value of the corrected material properties to evaluate the effectiveness and reliability of the correction.

[0032] Preferably, the system further includes a knowledge base module, which is communicatively connected to the feature selection unit in the machine learning parsing module and the error pattern recognition unit in the error correction module. The knowledge base module stores professional knowledge in the field of materials science, including the feature importance scores of key features and the systematic error pattern library of the predicted value of the material properties, and is used to provide prior knowledge for machine learning model training and assist in decision - making during the error correction process.

[0033] Preferably, the prediction model for quantum chemistry calculation results adopts a deep neural network with the following features:

[0034] The input layer adopts a convolutional structure to adapt to the periodicity of the molecules composing the material; the hidden layer contains gated recurrent units to capture the long - range correlations of the material properties; the output layer uses an attention mechanism to highlight the influence of key material features.

[0035] Preferably, the acquisition of the correction parameters in the systematic error pattern library of the predicted values of the material properties includes the following steps:

[0036] Set the objective function based on minimizing the deviation between the corrected result and the true experimental data, and use the Bayesian optimization algorithm to find the optimal correction parameters through an iterative method; the constraint conditions include the material physical laws and thermodynamic limitations.

[0037] The present invention also provides an intelligent analysis and error correction method for quantum chemistry calculation results, including the following steps:

[0038] S1. Data preprocessing: Collect the original data of quantum chemistry calculation and perform standardization processing;

[0039] S2. Calculation result analysis: Analyze the standardized data based on preset rules, identify abnormal points and evaluate the calculation accuracy;

[0040] S3. Machine learning analysis: Use deep learning algorithms to deeply analyze the analysis results, extract key features and establish a prediction model;

[0041] S4. Error correction: Based on the analysis results and machine learning prediction results, identify error patterns and generate correction parameters to correct the predicted values of material properties.

[0042] Provide professional knowledge support in the field of materials science through the knowledge base module to improve the accuracy of analysis and correction.

[0043] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0044] The intelligent analysis and error correction system for quantum chemistry calculation results of the present invention is realized through innovative system architecture design and advanced algorithms, and significant technological breakthroughs and performance improvements have been achieved in many aspects.

[0045] First, from a macroscopic perspective, the present invention establishes a complete intelligent analysis process, which organically integrates key links such as data preprocessing, result analysis, machine learning analysis and error correction. This systematic method not only improves the overall analysis efficiency, but also can form a positive interaction and information feedback among various links, so as to continuously optimize the analysis results.

[0046] Secondly, a series of innovative algorithms and technologies are adopted in each functional module of the present invention, achieving a leapfrog improvement in performance. In the data preprocessing stage, a GPU cluster is used for parallel processing, significantly improving the processing speed of large-scale material data. In practical applications, for a complex material system containing millions of atoms, this system can shorten the data preprocessing time from several hours of traditional methods to a few minutes, greatly improving the research efficiency.

[0047] In terms of machine learning analysis, innovative model structures such as the periodic graph convolutional neural network (P-GCN) and the material property gated unit (MP-GRU) applied in the present invention can effectively capture key features such as periodicity and long-range interactions in the material system. In multiple benchmark tests, the average error of these models in the material property prediction task is reduced by more than 30% compared with traditional methods, providing more reliable theoretical guidance for material design and performance optimization.

[0048] In terms of error correction, the method of combining the Bayesian optimization algorithm with physical constraints adopted in the present invention not only improves the accuracy of correction but also ensures the physical rationality of the correction result.

[0049] In addition, the knowledge base module of the present invention significantly improves the intelligence level and interpretability of the system by integrating domain expert knowledge and data-driven methods. This method not only improves the analysis accuracy but also provides valuable knowledge discovery for researchers.

[0050] In summary, through the innovative design of the system architecture, the application of advanced algorithms, and the deep integration of knowledge and data, the present invention comprehensively improves the analysis efficiency, accuracy, and interpretability of quantum chemistry calculation results. This not only provides a powerful computational auxiliary tool for material researchers but also is expected to accelerate the discovery and development process of new materials, promoting innovative development in the fields of materials science and computational chemistry. Brief Description of the Drawings

[0051] Figure 1 It is the logical block diagram of the overall system of the present invention.

[0052] Figure 2 It is the logical block diagram of the data preprocessing module of the present invention.

[0053] Figure 3 It is the logical block diagram of the calculation result analysis module of the present invention.

[0054] Figure 4 It is the logical block diagram of the machine learning analysis module of the present invention.

[0055] Figure 5 It is the logical block diagram of the error correction module of the present invention. Detailed Description of the Invention

[0056] Refer to Figures 1-5 , the present invention relates to an intelligent analysis and error correction system and method for quantum chemical calculation results, and is particularly suitable for the analysis and optimization of quantum chemical calculation results in the field of material chemistry. The present invention will be described in detail below in conjunction with specific embodiments.

[0057] As Figure 1 shown, the intelligent analysis and error correction system for quantum chemical calculation results of the present invention includes a data preprocessing module 1, a calculation result analysis module 2, a machine learning analysis module 3, and an error correction module 4. These modules are interconnected through a data bus and cooperate with each other to complete the intelligent analysis and error correction of quantum chemical calculation results.

[0058] The data preprocessing module 1 is used to collect the original data obtained from quantum chemical calculations and perform standardization processing on the original data. For example, for the quantum chemical calculation results of a large material system containing 10 6 atoms, a cluster of 100 GPU nodes can reduce the data preprocessing time from several hours of traditional CPU processing to a few minutes, greatly improving the overall efficiency of the system.

[0059] The calculation result analysis module 2 is communicatively connected to the data preprocessing module 1 and is used to receive the standardized data sent by the data preprocessing module 1 and analyze the standardized data based on preset rules.

[0060] The machine learning analysis module 3 is communicatively connected to the calculation result analysis module 2 and is used to receive the analysis results sent by the calculation result analysis module 2 and perform in-depth analysis on the analysis results based on machine learning algorithms. The present invention adopts a deep learning algorithm, which can automatically learn the complex patterns and features in the quantum chemical calculation results, thereby improving the accuracy and efficiency of the analysis.

[0061] The error correction module 4 is communicatively connected to the calculation result analysis module 2 and the machine learning analysis module 3 respectively, and is used to receive the analysis results sent by the calculation result analysis module 2 and the analysis results sent by the machine learning analysis module 3, and perform error correction on the quantum chemical calculation results based on the analysis results and the analysis results. The error correction module 4 of the present invention adopts an innovative adaptive correction algorithm, which can automatically adjust the correction parameters according to the characteristics of different material systems, significantly improving the accuracy and applicable range of the correction.

[0062] Furthermore, as Figure 2As shown in the figure, the data preprocessing module 1 of the present invention includes a data acquisition unit 11, a data normalization unit 12, and a data cleaning unit 13. The data acquisition unit 11 is used to obtain original calculation data from quantum chemistry calculation software. In the present invention, the data acquisition unit 11 supports data formats of multiple mainstream quantum chemistry calculation software, such as Gaussian, VASP, etc., greatly improving the compatibility and application scope of the system. The data normalization unit 12 is communicatively connected to the data acquisition unit 11, receives the original calculation data sent by the data acquisition unit 11, and performs normalization processing on the original calculation data. The present invention innovatively adopts an adaptive normalization algorithm, which can automatically select the optimal normalization method according to different types of quantum chemistry calculation data. For example, for band structure data, the min-max normalization method is adopted; for electron density data, the Z-score normalization method is adopted. This flexible normalization strategy significantly improves the accuracy and comparability of subsequent analysis. The data cleaning unit 13 is communicatively connected to the data normalization unit 12, receives the data after normalization processing sent by the data normalization unit 12, and removes noise and outliers in the data after normalization processing. The data cleaning unit 13 of the present invention adopts a hybrid cleaning strategy based on statistics and machine learning, which can effectively identify and process various types of data anomalies. For example, for band structure data, the 3σ principle is used to identify outliers; for electron density data, the isolation forest algorithm is used to detect abnormal regions.

[0063] As Figure 3 shown in the figure, the calculation result parsing module 2 of the present invention includes a rule library unit 21, a rule application unit 22, and a result analysis unit 23. The rule library unit 21 is used to store preset parsing rules, including the four rules mentioned above. These rules are summarized based on a large amount of quantum chemistry calculation experience and materials science knowledge, and can effectively guide the parsing process of calculation results. The rule application unit 22 is communicatively connected to the rule library unit 21, obtains the parsing rules in the rule library unit 21, and applies the parsing rules to the data after normalization processing. The present invention innovatively adopts a rule priority dynamic adjustment algorithm, which can automatically adjust the application order and weight of the rules according to the characteristics of different material systems and calculation tasks, thereby improving the accuracy and efficiency of parsing. The result analysis unit 23 is communicatively connected to the rule application unit 22, receives the application results of the rule application unit 22, and generates a parsing report including outlier marking, accuracy evaluation, and data characteristics.

[0064] As Figure 4As shown in the figure, the machine learning parsing module 3 of the present invention includes a feature extraction unit 31, a feature selection unit 32, a model training unit 33, a new result input unit 34, and a prediction analysis unit 35. The feature extraction unit 31 extracts key features from the parsing results and vectorizes the key features. For example, for the electronic structure calculation results of materials, the system automatically extracts key features such as Fermi level, band gap width, and density of states, and converts these features into numerical vectors that can be processed by machine learning algorithms. The feature selection unit 32 is communicatively connected to the feature extraction unit 31, receives the vectorized features sent by the feature extraction unit 31, and selects an optimal feature subset based on feature importance scores. The present invention innovatively adopts a multi-criterion feature selection algorithm, comprehensively considering the correlation, redundancy, and physical meaning of features, so as to select the most representative and predictive feature subset. This method not only improves the performance of subsequent machine learning models but also ensures the interpretability of model results. The model training unit 33 is communicatively connected to the feature selection unit 32, receives the optimal feature subset sent by the feature selection unit 32, and trains a quantum chemistry calculation result analysis model using a deep learning algorithm. The model training unit 33 of the present invention adopts an advanced periodic graph convolutional neural network (P-GCN) algorithm, which can effectively capture the interactions and long-range correlations between atoms in the material structure. The new result input unit 34 is communicatively connected to the result parsing module and is used to receive the data obtained after data preprocessing and parsing of the quantum chemistry calculation results. The prediction analysis unit 35 is communicatively connected to the new result input unit 34 and the model training unit 33, uses the trained model to perform prediction analysis on the new quantum chemistry calculation results, and generates an analysis report containing prediction results and confidence levels. The prediction analysis unit 35 of the present invention not only outputs point estimation results but also provides uncertainty quantification based on Bayesian deep learning, which enables researchers to better evaluate the reliability of prediction results, especially when dealing with calculation results of new materials or under extreme conditions.

[0065] As Figure 5As shown, the error correction module 4 of the present invention includes an error pattern recognition unit 41, a correction parameter generation unit 42, a correction execution unit 43, and a correction verification unit 44. The error pattern recognition unit 41 analyzes the systematic error patterns in the historical quantum chemistry calculation results and establishes an error pattern library. This unit adopts a hybrid algorithm based on clustering and anomaly detection, which can automatically discover and classify different types of calculation errors. The correction parameter generation unit 42 is communicatively connected to the error pattern recognition unit 41. Based on the recognized error patterns, it generates corresponding correction parameters and optimizes the correction parameters to minimize the errors. The present invention innovatively adopts a meta-learning algorithm, which can quickly generate initial correction parameters for a new material system and fine-tune them through a small amount of high-precision calculation or experimental data. This method greatly improves the efficiency and application scope of the correction. The correction execution unit 43 is communicatively connected to the correction parameter generation unit 42, receives the optimized correction parameters sent by the correction parameter generation unit 42, and applies the correction parameters to the quantum chemistry calculation results. The correction execution unit 43 of the present invention adopts an adaptive multi-scale correction strategy, which can handle both local and global calculation errors simultaneously. The correction verification unit 44 is communicatively connected to the correction execution unit 43, cross-verifies the corrected results, and evaluates the effectiveness and reliability of the correction. The present invention uses a verification method based on statistical learning theory, which not only evaluates the average error of the correction results but also considers the variance and skewness of the error distribution. This comprehensive verification method can provide researchers with a more reliable evaluation of the correction results.

[0066] In addition, the present invention further includes a knowledge base module 5, which is communicatively connected to the machine learning analysis module 3 and the error correction module 4. The knowledge base module 5 is used to store the professional knowledge and experience rules in the field of materials science, provide prior knowledge for the training of machine learning models, and assist in decision-making during the error correction process.

[0067] As mentioned above, the quantum chemistry calculation result prediction model in the model training unit 33 of the present invention adopts an advanced periodic graph convolutional neural network (P-GCN). This neural network has the following characteristics: the input layer adopts a convolutional structure adapted to the periodicity of materials, the hidden layer contains gated recurrent units to capture the long-range correlations of material properties, and the output layer uses an attention mechanism to highlight the influence of key material features.

[0068] Specifically, the convolution operation of P-GCN is defined as follows:

[0069]

[0070] where, represents the feature vector of node v at the l-th layer, is the set of edge types, is the set of neighbors of node v under edge type r, and b(l) They are the weight matrix and bias vector of the l-th layer respectively, and σ is the activation function. This structure can effectively handle the periodic boundary conditions of the material crystal and improve the model's ability to express the material structure.

[0071] In the hidden layer, the present invention uses a variant of the long short-term memory (LSTM) unit, called the material property gated unit (MP-GRU). The update rule of MP-GRU is as follows:

[0072] z t = σ(W z [h t-1 , x t ),

[0073] r t = σ(W r [h t-1 , x t ),

[0074]

[0075] where z t is the update gate, r t is the reset gate, is the candidate hidden state, h t is the hidden state at the current time step, x t is the input, W z , W r and W are weight matrices, and ⊙ represents element-wise multiplication. This structure can capture the long-range correlations of material properties with changes in composition and structure, and is particularly suitable for handling complex material systems.

[0076] In the output layer, the present invention introduces a multi-head self-attention mechanism, which is calculated as follows:

[0077]

[0078] where Q, K, and V are the query, key, and value matrices respectively, and d k is the dimension of the key. The multi-head attention calculates multiple attention heads in parallel and then concatenates the results:

[0079] MultiHead(Q, K, V) = Concat(head1,..., head h )W o

[0080] This mechanism can adaptively focus on different scales and types of material features, improving the interpretability and prediction accuracy of the model.

[0081] In the systematic error pattern library of the predicted values of the material properties of the present invention, the Bayesian optimization algorithm is adopted, and the optimal correction parameters are found through an iterative method. The objective function of this algorithm is set to minimize the deviation between the corrected result and the experimental data, and the constraints include material physical laws and thermodynamic limitations. Specifically, the present invention uses Gaussian process regression (GPR) as the surrogate model of Bayesian optimization. Given the observed data

[0082]

[0083] The GPR model is defined as follows:

[0084]

[0085] where m(x) is the mean function and k(x, x′) is the kernel function. The present invention adopts the Matérn kernel function:

[0086]

[0087] where ν and are hyperparameters, and K ν is the modified Bessel function.

[0088] During the optimization process, the present invention uses the expected improvement (EI) as the acquisition function:

[0089] EI(x) = E[max(f(x) - f(x + ), 0)],

[0090] where f(x + ) is the best function value observed currently.

[0091] To ensure that the correction parameters satisfy the physical constraints, the present invention introduces a constraint handling mechanism based on the Lagrange multiplier method. For example, for bandgap correction, the following constraints are added:

[0092]

[0093] where is the corrected bandgap, is the bandgap calculated by DFT, and ΔE max is the maximum allowable correction amplitude. These constraints ensure the physical rationality of the correction result.

[0094] In summary, the intelligent analysis and error correction system and method for quantum chemistry calculation results provided by the present invention offer a powerful and intelligent computational assistance tool for materials researchers through an innovative deep learning architecture, Bayesian optimization algorithm, and knowledge-driven analysis process. This system can not only significantly improve the accuracy and reliability of quantum chemistry calculation results but also provide researchers with in-depth data insights and knowledge discovery, promising breakthroughs in new material design, performance optimization, and theoretical research.

[0095] It should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent analysis and error correction system for quantum chemistry calculation results, characterized in that, It includes: A data preprocessing module, which is used to collect the original data obtained from quantum chemistry calculations and perform standardization processing on the original data; A calculation result parsing module, which is used to receive the data after standardization processing sent by the data preprocessing module and parse the data after standardization processing based on preset rules; A machine learning parsing module, which is used to receive the parsing result sent by the calculation result parsing module and perform in-depth analysis on the parsing result based on machine learning algorithms; An error correction module, which receives the parsing result sent by the calculation result parsing module and the analysis result sent by the machine learning parsing module, and performs error correction on the quantum chemistry calculation result based on the parsing result and the analysis result.

2. The intelligent analysis and error correction system for quantum chemistry calculation results according to claim 1, characterized in that, The data preprocessing module includes: A data acquisition unit, which is used to obtain the original calculation data from the quantum chemistry calculation software; A data standardization unit, which is used to receive the original calculation data sent by the data acquisition unit and perform normalization processing on the original calculation data; A data cleaning unit, which is used to receive the data after normalization processing sent by the data standardization unit and remove the noise and outliers in the data after normalization processing.

3. The intelligent analysis and error correction system for quantum chemical calculation results according to claim 1, characterized in that The calculation result parsing module includes: A rule library unit, which is used to store preset parsing rules. The parsing rules include: an outlier identification rule based on data distribution; an accuracy verification rule based on the comparison between the calculation result and the real experimental result; a cleaning and normalization rule based on data characteristics; A rule application unit, which obtains the parsing rules in the rule library unit, selects and applies the parsing rules corresponding to the data after standardization processing; A result analysis unit, which is used to apply the corresponding parsing rules to process the data after standardization processing, so as to delete the following problematic data from the data after standardization processing: data with outliers, data with low accuracy, and data features with low correlation with the prediction target, and obtain the data after standardization processing with problematic data removed.

4. The intelligent analysis and error correction system for quantum chemical calculation results according to claim 1, wherein The machine learning parsing module includes: A feature extraction unit, which is used to extract key features from the data after standardization processing with problematic data removed and perform vectorization processing on the key features; A feature selection unit, which is used to receive the vectorized features sent by the feature extraction unit, obtain the feature importance scores of the key features, and select the optimal feature subset based on the feature importance scores; A model training unit, which is used to construct a quantum chemistry calculation result prediction model, receive the optimal feature subset sent by the feature selection unit as the input of the quantum chemistry calculation result prediction model. The output of the quantum chemistry calculation result prediction model is the properties of the material. The material is composed of periodic molecules, and the periodic molecules are the research objects of the quantum chemistry calculation results; use a supervised learning method to train the quantum chemistry calculation result prediction model; A prediction analysis unit, which is used to obtain the quantum chemistry calculation result to be predicted using a new result input unit, and after being processed by the data preprocessing module and the calculation result parsing module, input it into the trained quantum chemistry calculation result prediction model to obtain the predicted value and confidence level of the properties of the material.

5. The intelligent analysis and error correction system for quantum chemical calculation results according to claim 1, wherein The error correction module includes: An error pattern recognition unit, configured to obtain a systematic error pattern library of predicted values of material properties and analyze the systematic error patterns of the predicted values of material properties; A correction parameter generation unit, configured to obtain corresponding correction parameters from the systematic error pattern library of the predicted values of the material properties based on the recognized error patterns; A correction execution unit, configured to receive the correction parameters sent by the correction parameter generation unit and apply the correction parameters to correct the predicted values of the material properties.

6. The intelligent analysis and error correction system for quantum chemical calculation results according to claim 5, wherein It further includes a knowledge base module, which is communicatively connected to the feature selection unit in the machine learning analysis module and the error pattern recognition unit in the error correction module. The knowledge base module stores professional knowledge in the field of materials science, including the feature importance scores of key features and the systematic error pattern library of the predicted values of material properties.

7. The intelligent analysis and error correction system for quantum chemistry calculation results according to claim 4, characterized in that The quantum chemistry calculation result prediction model adopts a deep neural network including the following features: The input layer adopts a convolutional structure to adapt to the periodicity of the molecules constituting the material; the hidden layer includes gated recurrent units to capture the long-range correlations of material properties; the output layer uses an attention mechanism to highlight the influence of key material features.

8. The intelligent analysis and error correction system for quantum chemical calculation results according to claim 6, characterized in that The acquisition of the correction parameters in the systematic error pattern library of the predicted values of the material properties includes the following steps: Setting an objective function based on minimizing the deviation between the corrected result and the real experimental data, adopting a Bayesian optimization algorithm, and searching for the optimal correction parameters through an iterative manner; the constraint conditions include material physical laws and thermodynamic limitations.

9. The intelligent analysis and error correction method for quantum chemical calculation results of the quantum chemical calculation result intelligent analysis and error correction system according to any one of claims 6-8, characterized in that, It includes the following steps: S1. Data preprocessing: Collecting the original data of quantum chemistry calculations and performing normalization processing; S2. Calculation result analysis: Analyzing the data after normalization processing based on preset rules, identifying abnormal points and evaluating the calculation accuracy; S3. Machine learning analysis: Using a deep learning algorithm to analyze the analysis results, extracting key features and establishing a prediction model; S4. Error correction: Based on the analysis results and the machine learning prediction results, identifying error patterns and generating correction parameters to correct the predicted values of the material properties.

Images