Power transmission and transformation project electronic file verification method and system based on fuzzy reasoning mechanism

Abstract

The invention discloses a power transmission and transformation project electronic file verification method based on a fuzzy reasoning mechanism. The method comprises the following steps: carrying out consistency verification on metadata of an electronic file through a Transform model, an LSTM network and Hash verification; on the basis of a conditional weight dynamic adjustment mechanism, screening out features most relevant to the target task in the electronic archive management system, and removing redundant data; according to an input condition, adaptively adjusting a scoring standard, and according to a fuzzy reasoning mechanism and the scoring standard, scoring the screened features to obtain a target feature score; and generating a final verification result of the electronic archive based on the target feature score. The design is suitable for integrated processing and accurate verification of large-scale and multi-modal data, and the accuracy and flexibility of electronic file verification can be effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power engineering management, and particularly relates to a power transmission and transformation engineering electronic archive verification method and system based on a fuzzy reasoning mechanism. BACKGROUND

[0002] The completeness and accuracy of the electronic archives of power transmission and transformation engineering are crucial for ensuring the effectiveness and compliance of project management. However, existing electronic archive verification methods face complex data verification problems, such as information missing, inconsistency between different archive data (such as text contracts, image materials, operation records, etc.), and complex interaction between features, resulting in poor archive data verification effect. The current method lacks an effective fusion mechanism when dealing with multi-modal data, and cannot fully utilize the information association in the data, resulting in the accuracy and efficiency of data verification being affected. Therefore, how to verify incomplete or incorrect data in a large amount of archive data, filter out features closely related to the verification task, and optimize the scoring and evaluation model, has become a key problem in the current verification of power transmission and transformation engineering electronic archives.

[0003] After comparing the existing literature, it is found that the literature Application of neural networks for gross error detection (Gupta, G. and S. Narasimhan, Application of neural networks for gross error detection [J]. Ind. Eng. Chem. Res., 1993, 32(8):1651-1657.) first explored the application of artificial neural networks to solve the problem of gross error detection; the literature Data rectification and gross error detection in a steady-state process via artificial neural networks (Terry, P. A. and D. Himmelblau, M., Data rectification and gross error detection in a steady-state process via artificial neural networks [J]. Ind. Eng. Chem. Res., 1993, 32(12):3020-3028.) proposed using artificial neural networks for data checking in view of the shortcomings of repeated strategy steps in data checking using statistical methods; the above literatures prove that neural networks can achieve good results in terms of checking accuracy, but have limitations in dealing with fuzzy and uncertain data.The fully adaptive data reconciliation method proposed in the document "A framework for robust data reconciliation based on a generalized objective function" (Wang, D. and J. A. Romagnoli, A framework for robust data reconciliation based on a generalized objective function [J]. Industrial & Engineering Chemistry Research, 2003, 42(13):3076-3084.) uses kernel density estimation method to estimate the actual distribution of measurement data; the document "Rectification of plant measurements using a statistical framework. Computers & Chemical Engineering" (Morad, K., B. R. Young, and W. Y. Svrcek, Rectification of plant measurements using a statistical framework. Computers & Chemical Engineering, 2005, 29(5):919-940.) uses adaptive mixed distribution to update and estimate the actual measurement variables; the document "Research on theory and application of data reconciliation technology" uses robust least squares algorithm for linear and nonlinear data reconciliation. Although the above methods can detect noise and abnormal data, the effect is not ideal when checking the metadata in complex or multi-modal data. SUMMARY

[0004] The purpose of the present application is to overcome the problem in the prior art that there is a lack of effective fusion mechanism when checking multi-modal data in electronic archives, and the information correlation in the data cannot be fully utilized, resulting in the accuracy and efficiency of data checking being affected. The present application provides a power transmission and transformation engineering electronic archive checking method and system based on a fuzzy reasoning mechanism, which can effectively improve the accuracy and flexibility of power transmission and transformation engineering electronic archive checking, and is suitable for integrated processing and accurate checking of large-scale, multi-modal data.

[0005] The technical solution of the present application is:

[0006] In a first aspect, the present application provides a power transmission and transformation engineering electronic archive checking method based on a fuzzy reasoning mechanism, which comprises:

[0007] S1, sequentially processing the metadata of the electronic archives through a Transformer model, an LSTM network and a hash check to obtain feature vectors corresponding to each metadata field;

[0008] S2, based on a conditional weight dynamic adjustment mechanism, filtering out feature vectors most relevant to the check task in the electronic archives management system and removing redundant data to obtain target feature vectors most relevant to the check task;

[0009] S3, scoring the target feature vectors according to fuzzy reasoning to obtain target feature scores corresponding to the check task;

[0010] S4, generating a final check result of the electronic archives based on the target feature scores.

[0011] The S1 comprises:

[0012] S11, vectorizing all metadata fields to obtain first vectors corresponding to each metadata field;

[0013] S12, calculating attention weights between each first vector and other first vectors through a self-attention mechanism in the Transformer model, and generating second vectors corresponding to each first vector according to the attention weights between each first vector and other first vectors;

[0014] S13, generating a hidden state vector for each second vector through an LSTM network to obtain third vectors corresponding to each metadata field;

[0015] S14, generating a corresponding hash value for each third vector through a hash check to obtain feature vectors corresponding to each metadata field.

[0016] The S2 comprises:

[0017] S21, calculating mutual information between each feature vector and the check task;

[0018] S22, filtering out relevant feature vectors with mutual information greater than a mutual information threshold between the feature vectors and the check task;

[0019] S23, assigning initial weights to the relevant feature vectors;

[0020] S24, using gradient descent method to iteratively adjust the initial weights according to a loss function to obtain iteratively adjusted weights;

[0021] S25, adjusting the iteratively adjusted weights according to a conditional adjustment formula to obtain feature weights corresponding to each relevant feature vector;

[0022] S26, discard the related feature vector with feature weight less than weight threshold value, and obtain k target feature vectors most related to the verification task.

[0023] In the S21, the mutual information calculation formula is:

[0024] ;

[0025] In the formula, denotes the jth feature vector, denotes the verification target, denotes the feature vector and the mutual information between the verification target ; denotes a specific example, denotes a sample space belonging to ; denotes a specific example, denotes a sample space belonging to ; denotes the joint probability between the feature vector and the verification target ; and denote the marginal probability of the feature vector and the verification target , respectively.

[0026] In the S24, the initial weight is iteratively adjusted according to the loss function by using the gradient descent method, including the following steps:

[0027] The predicted value is calculated according to the current weight and the corresponding related feature vector, the value of the loss function is calculated by the predicted value and the true value, the current weight is updated according to the value of the loss function and the gradient descent formula, and the above steps are repeated until the value of the loss function is less than the preset threshold value or the number of weight updates reaches the preset upper limit;

[0028] The loss function is:

[0029] ;

[0030] In the formula, denotes the loss function; denotes the system predicted value; denotes the true value; is the current feature weight vector; is the sample number;

[0031] ​The gradient descent formula is trained by a standard file:

[0032] ;

[0033] In the formula, represents the weight of the i-th relevant feature vector when the weight is updated for the t-th time; represents the weight of the i-th relevant feature vector when the weight is updated for the t+1-th time; is the learning rate.

[0034] In the S25, the condition adjustment formula is:

[0035] ;

[0036] In the formula, is the feature weight of the i-th relevant feature vector ; is a condition adjustment coefficient; is a condition function, which reflects the weight adjustment requirement of the feature vector , and is determined by the verification task and the data type corresponding to the relevant feature vector; is the weight of the relevant feature vector after iteration.

[0037] S31, map the target feature vector to the fuzzy set, and convert each target feature vector to the corresponding membership degree through the membership function, wherein the membership function is:

[0038] ;

[0039] In the formula, is the m-th target feature vector; , , is the parameter of the triangular function corresponding to the p-th fuzzy set; is the membership degree of the target feature vector in the p-th fuzzy set ;

[0040] S32, adaptively adjust the membership degree according to the verification target and the data type corresponding to the target feature vector, and the adaptive adjustment formula is:

[0041] ;

[0042] In the formula, is the membership degree of the adjusted target feature vector in the p-th fuzzy set membership in the fuzzy set is the target feature vector before adjustment in the fuzzy set membership in the fuzzy set is the adjustment coefficient, the adjustment coefficient According to the data type corresponding to the check target and the target feature vector.

[0043] S33, according to the membership corresponding to each target feature vector, reasoning through the fuzzy rule base containing the preset fuzzy rule, obtaining the fuzzy output distribution, the fuzzy output distribution The expression is:

[0044] ;

[0045] ;

[0046] In the formula, is the membership corresponding to the ith rule of the fuzzy rule base; denotes the target feature vector in the fuzzy set ; denotes the target feature vector in the fuzzy set ; denotes the target feature vector in the fuzzy set ; denotes the first fuzzy set related to the ith rule; denotes the target feature vector in the fuzzy set ; denotes the target feature vector in the fuzzy set ; denotes the target feature vector in the fuzzy set ; denotes the second fuzzy set related to the ith rule; denotes the target feature vector in the fuzzy set ; denotes the target feature vector in the fuzzy set ; denotes the target feature vector in the fuzzy set ; denotes the third fuzzy set related to the ith rule a fuzzy set; denotes the total number of fuzzy sets related to the i-th rule; denotes the total number of rules;

[0047] S34, converting the fuzzy output into a target feature score by defuzzification, the calculation formula of the defuzzification being:

[0048]

[0049]

[0050] wherein, denotes the target feature score after defuzzification; denotes the m-th target feature vector in the file to be verified; denotes the feature vector corresponding to in the verification standard file, denotes a preset numerical value, > > .

[0051] In a second aspect, the present application provides a power transmission and transformation engineering electronic archive verification system based on a fuzzy reasoning mechanism, comprising:

[0052] a data preprocessing module, configured to sequentially process metadata of the electronic archive through a Transformer model, an LSTM network and a hash verification;

[0053] a screening module, configured to screen out features most relevant to the verification task in an electronic archive management system based on a conditional weight dynamic adjustment mechanism, and remove redundant data;

[0054] a scoring module, configured to adaptively adjust the verification rules according to the data attributes of the features, obtain a scoring standard according to the verification rules, and score the screened features according to the scoring standard and the fuzzy reasoning mechanism to obtain target feature scores;

[0055] a result generation module, configured to generate a final verification result of the electronic archive based on the target feature scores.

[0056] In a third aspect, the present application provides an electronic device, which comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, and the processor implements the steps in the power transmission and transformation engineering electronic archive verification method based on the fuzzy reasoning mechanism when executing the computer program.

[0057] ​​​​​​In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program, when executed by a processor, implements the steps of the power transmission and transformation engineering electronic archive verification method based on the fuzzy reasoning mechanism.

[0058] Compared with the prior art, the present application has the following beneficial effects:

[0059] 1. In the power transmission and transformation engineering electronic archive verification method based on the fuzzy reasoning mechanism, a data consistency verification model based on adaptive attention is constructed. The model combines the technologies of Transformer and LSTM to process global context information, time series and sequential dependency fields. Meanwhile, hash verification is introduced to lay a foundation for subsequent consistency, integrity and accuracy detection, and to realize intelligent verification. Therefore, based on the metadata missing or inconsistency in the electronic archive management of power transmission and transformation engineering, a data consistency verification model based on adaptive attention is constructed, and hash verification is introduced to lay a foundation for subsequent consistency, integrity and accuracy detection.

[0060] 2. In the power transmission and transformation engineering electronic archive verification method based on the fuzzy reasoning mechanism, the data after verification is filtered through a conditional weight dynamic adjustment mechanism. This feature selection mechanism can filter out the most relevant features for the verification task and remove redundant information to improve the efficiency and usability of data processing. Meanwhile, the feature selection process uses a dynamically adjusted weight model to adjust the importance of each feature in real time according to the actual needs of different archive verification tasks, ensuring that the most relevant features are fully utilized. Therefore, in this design, based on the dynamically adjusted weight model, the importance of each feature in the data is adjusted according to the actual needs of different verification tasks, the most relevant features for the verification task are filtered out, the efficiency and usability of data processing are improved, and the most relevant features are fully utilized.

[0061] 3. In the power transmission and transformation engineering electronic archive verification method based on the fuzzy reasoning mechanism, based on the improved fuzzy reasoning mechanism, multi-dimensional data is analyzed by fuzzy logic, fuzzy rules are dynamically optimized by adaptive adjustment mechanism, scoring standards are adjusted for different input conditions, high correlation features are quantitatively scored, and the final electronic archive evaluation result is generated. This can meet the needs of different tasks and ensure the accuracy and flexibility of the verification result in different task scenarios. Therefore, this design can meet the needs of different tasks and ensure the accuracy and flexibility of the verification result in different task scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1is a flowchart of the power transmission and transformation project electronic archive verification method based on the fuzzy reasoning mechanism provided by the embodiment of the present application.

[0063] Figure 2 is a flowchart of the consistency verification provided by the embodiment of the present application.

[0064] Figure 3 is a flowchart of the feature most relevant to the target task in the electronic archive management system screened by the embodiment of the present application.

[0065] Figure 4 is a flowchart of the scoring of the screened features provided by the embodiment of the present application.

[0066] Figure 5 is a structural diagram of the power transmission and transformation project electronic archive verification system based on the improved fuzzy reasoning mechanism provided by the embodiment of the present application.

[0067] Figure 6 is a structural diagram of the electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0068] The present application is further described in detail below in conjunction with the accompanying drawings, description and specific embodiments.

[0069] In the related art, the missing, inconsistency and potential abnormality in the power transmission and transformation project electronic archive data may exist, for example, the information missing, inconsistency between different archive data (such as text contract, image material, operation record, etc.) and the complex interaction relationship between features, which leads to poor archive data verification effect.

[0070] In order to solve the technical problems existing in the related art, the embodiment of the present application provides a power transmission and transformation project electronic archive verification method based on a fuzzy reasoning mechanism, please refer to Figure 1 , Figure 1 is a flowchart of the power transmission and transformation project electronic archive verification method based on the fuzzy reasoning mechanism provided by the embodiment of the present application, which includes steps S1 to S5.

[0071] S1, the metadata of the electronic archive is processed by the Transformer model, the LSTM network and the hash check in turn;

[0072] In this embodiment, based on the adaptive attention mechanism, the multi-modal data consistency verification method is improved, the multi-modal data consistency verification model is constructed, the global dependence (Transformer) and the sequential dependence (LSTM) are combined, and the hash check is introduced at the same time, so as to protect the integrity and authenticity of the electronic archive, and lay a foundation for the verification and scoring of the electronic archive.

[0073] In the electronic file management of power transmission and transformation projects, the situation of missing or inconsistent metadata occurs from time to time. In order to verify the integrity of the metadata, the present application constructs a data consistency verification model based on adaptive attention. Through the combination of two technologies of Transformer and LSTM, we use the Transformer model to process global context information, and the LSTM network to process time series and sequential dependent fields, so as to realize intelligent verification. For details, please refer to Figure 2 , Figure 2 is a flowchart of the consistency verification provided by the embodiment of the present application, as shown in Figure 2 , the consistency verification provided by the embodiment includes steps in S11 to S14.

[0074] S11, vectorizing all metadata fields to obtain the first vector corresponding to each metadata field.

[0075] In the present embodiment, Word2Vec model can be used to vectorize all input metadata fields. Word2Vec analyzes the usage patterns of metadata fields in context and maps each metadata field to a high-dimensional vector , which captures the semantic information of the metadata field. The formula of the process is as follows:

[0076] ;

[0077] In the formula, is the i-th metadata in the electronic file; is the word vector corresponding to the i-th metadata in the electronic file.

[0078] S12, calculate the attention weight between each first vector and other first vectors through the self-attention mechanism in the Transformer model, and generate the second vector corresponding to each first vector according to the attention weight between each first vector and other first vectors.

[0079] Self-attention mechanism is one of the core components of Transformer model. It allows the model to dynamically focus on information at different positions in the sequence when processing sequence data, and assigns different attention weights according to the importance of these information. This mechanism enables the model to better understand the internal structure and dependency of sequence data.

[0080] In this embodiment, after obtaining the vector representation corresponding to each metadata field, the self-attention mechanism in the Transformer model is used to capture the global dependency between the metadata fields, which can dynamically allocate attention weights according to the context of each field. The calculation formula of the attention mechanism is:

[0081] ;

[0082] wherein, is the vector representation of the missing field currently to be predicted; is the vector representation of other known fields; is the content of other known fields; represents the dimension of the key vector, which is used for scaling processing to avoid large numerical values in calculation.

[0083] The attention weight calculation formula is:

[0084] ;

[0085] wherein, is the similarity score between the fields and , is the attention weight allocated to each field. Through this formula, the model can capture the association between fields and ensure the global consistency of metadata.

[0086] S13, generate a hidden state vector for each second vector through an LSTM network, to obtain a third vector corresponding to each metadata field.

[0087] The LSTM network is a Long Short-Term Memory network, which is good at processing and predicting time dependencies and long-distance dependencies in sequence data. The LSTM network controls the storage, forgetting and output of information by introducing memory cells and input gates, forgetting gates and output gates, thereby solving the gradient vanishing or gradient explosion problem faced by traditional RNN when processing long sequence data. The memory cell is used to store long-term information, and the three gates are responsible for regulating the flow of information in the memory cell.

[0088] In this embodiment, the LSTM network is introduced to process metadata that depends on previous and subsequent fields, to ensure the sequential consistency of time series fields.

[0089] The LSTM verifies whether the time dependency of the field meets the requirements by processing the hidden states of the previous and subsequent time points. The forgetting gate formula of the LSTM network is:

[0090] ;

[0091] wherein, is the output of the forget gate, which determines whether to retain the memory state of the previous moment; is the weight matrix of the forget gate; is the hidden state of the previous moment, which saves the information of the previous moment; is the input vector of the current moment; is the bias term of the forget gate.

[0092] The memory state update formula of the LSTM network is:

[0093]

[0094] wherein, is the current memory state; is the memory state of the previous moment; is the output of the input gate, which determines the degree of update of the current input information; is the candidate memory state of the current moment;

[0095] The output gate of the LSTM network is used to generate the hidden state of the current moment, and the output gate formula of the LSTM network is:

[0096]

[0097] wherein, is the hidden state of the current moment; is the output of the output gate, which determines whether the hidden state of the current moment generates the final output.

[0098] Through the hidden states of the previous and subsequent moments, the LSTM can capture the time sequence dependence of the metadata, ensuring the correctness of the sequential field.

[0099] S14, generate a corresponding hash value for each third vector through hash verification, and obtain the feature vector corresponding to each metadata field.

[0100] The hash verification mechanism is a commonly used data integrity verification method, which calculates the hash value of the data and compares the value in the subsequent verification process to determine whether the data remains unchanged in the transmission or storage process. In this embodiment, the hash verification mechanism is introduced to further verify the authenticity of the metadata in the transmission and processing process. Specifically, we generate a unique identifier for the data through a hash function, and ensure that the data has not been tampered with by comparing the hash values. The hash function can be represented as:

[0101]

[0102] wherein, represents the metadata field; ​​​For metadata fields The corresponding hash value can effectively verify the authenticity and integrity of the data through hash check.

[0103] Through the adaptive attention mechanism and sequential dependency analysis described above, combined with the hash check model, a foundation is laid for subsequent consistency, integrity and accuracy detection.

[0104] S2, based on the condition weight dynamic adjustment mechanism, the most relevant features in the electronic archives management system are screened out with the target task, and redundant data is removed.

[0105] In the electronic archives management system, different metadata fields are not important for checking and analysis, so an effective feature selection mechanism is necessary, which can ensure that the model focuses on the most relevant data fields and reduces redundant information. Therefore, the embodiment of the application provides a feature selection method based on condition weight dynamic adjustment, which reduces redundant information, optimizes feature selection, and improves the efficiency of data processing and subsequent analysis of the system. For details, please see Figure 3 , Figure 3 is a flowchart for screening out the most relevant features in the electronic archives management system with the target task, as shown in Figure 3 The embodiment provided by the present application screens out the most relevant features in the electronic archives management system with the target task, which includes steps S21 to S26.

[0106] S21, calculate the mutual information between each feature vector and the verification task.

[0107] In the process of detecting the authenticity of electronic archives of power transmission and transformation projects, we set six verification tasks as shown in the verification task list, and score the electronic archives of power transmission and transformation projects according to the six verification tasks:

[0108]

[0109] For any one of the verification tasks in the verification task list, first, we preliminarily screen all the metadata fields. Based on the correlation analysis and initial feature evaluation, the importance of each field to the verification task is determined, and mutual information is used as the measurement standard for feature selection to calculate the information between each feature and the verification task.

[0110] Specifically, the mutual information calculation formula is:

[0111] ;

[0112] In the formula, Xj represents the jth feature vector, represents the verification target, represents the feature vector mutual information between the feature vector and the check target; represents one specific example, represents belongs to the sample space; represents one specific example, represents belongs to the sample space; represents the joint probability between the feature vector and the check target; and respectively represent the marginal probability of the feature vector and the check target.

[0113] The check target is set according to the check task, for example, if the check task is to check the integrity of the electronic file, the check target is the result required to check the integrity of the electronic file, that is, whether missing data is found; if the check task is to check the consistency of the electronic file, then the check target is whether the data fields of the electronic file meet the consistency requirement.

[0114] Through mutual information analysis, the system can filter out feature vectors with high correlation with the target task, which will be the focus of subsequent analysis.

[0115] S22, filter out relevant feature vectors with mutual information greater than the mutual information threshold between the feature vectors and the check task, specifically, the mutual information threshold is set to 0.2.

[0116] S23, according to the system prior knowledge or the established rule, assign an initial weight to the relevant feature vector.

[0117] After the mutual information analysis is completed, the system will assign an initial weight to the filtered relevant feature vector, which is used to represent the importance of each relevant feature vector in the subsequent task.

[0118] The initial weight is not directly determined by the mutual information, specifically, we assign an initial weight to the filtered relevant feature vector according to the system prior knowledge or the established rule. For example, the initial weight can be set according to the historical performance of the feature vector and the task requirements, and the expression of the initial weight is:

[0119] ;

[0120] In the formula, represents the i-th relevant feature vector​​​ initial weights, which are dynamically adjusted in subsequent tasks.

[0121] S24, using gradient descent method, dynamically adjusting the initial weights according to the loss function, to obtain the iterated weights.

[0122] After the initial weights are assigned, the system will update the weights of the relevant feature vectors according to the task requirements using a dynamic adjustment mechanism. In this embodiment, dynamic adjustment relies on the gradient descent algorithm, and as the loss function changes, the system optimizes the weight of each feature in real time. Specifically, updating the weight of a feature includes:

[0123] According to the current weight and the corresponding relevant feature vector, the predicted value is calculated, the value of the loss function is calculated by the predicted value and the true value, and the current weight is updated according to the value of the loss function and the gradient descent formula, then enter the next iteration, repeat the above steps until the value of the loss function is less than the preset threshold or the number of iterations reaches the preset upper limit, to obtain the iterated weights. The ground truth is the objective correct answer that the model tries to predict in supervised learning, representing the actual state or label of the data sample in the real world. The ground truth is the deterministic information obtained by observation, measurement or manual annotation, which is used to measure the deviation of the prediction result and drive weight adjustment.

[0124] The loss function is:

[0125] ;

[0126] In the formula, represents the loss function; represents the system prediction value; represents the true value; is the current feature weight vector; is the number of samples.

[0127] In this embodiment, each of the samples is a preset verification standard file, and each of the verification standard files corresponds to a standard verification result, i.e. the true value. The system prediction value is the actual verification result of the verification with the current weight value and the corresponding relevant feature vector. By calculating the difference between the true value and the system prediction value, the quality of the weight value can be judged.

[0128] The gradient descent formula is:

[0129] ;

[0130] In the formula, represents the weight of the ith relevant feature vector at the tth iteration; denotes the weight of the i-th relevant feature vector at the t+1-th iteration of the weight; . is a learning rate.

[0131] S25, dynamically adjusting the weight of the relevant feature vector after iteration according to the conditional adjustment formula to obtain the feature weight corresponding to each relevant feature vector.

[0132] In this embodiment, in order to meet the specific requirements of different tasks, the system introduces a conditional adjustment factor to adjust the weight of a specific feature through this factor. This allows the system to weight or reduce the weight of certain features according to the requirements of a specific task. The conditional adjustment formula is:

[0133] .

[0134] wherein, denotes the feature weight of the i-th relevant feature vector . is a conditional adjustment coefficient; is a conditional function, which reflects the weight adjustment requirement of the feature vector , and is determined by the verification task and the attribute of the corresponding relevant feature vector; is the weight of the relevant feature vector after iteration. In the data verification task, the conditional function is used to dynamically adjust the weight of the feature, and determines how to adjust the weight of each feature according to the specific task requirements, to ensure that the final verification result can better meet the task goal.

[0135] In this embodiment, the data types include numbers, texts, dates, etc. reflect the weight adjustment requirement of the relevant feature vector , is determined according to the data attribute of the relevant feature vector , for example, the attribute of the relevant feature vector is "start date", then the importance of this attribute in the verification task is judged and the weight is adjusted to ensure that the use of the relevant feature vector meets the task requirements.

[0136] Through the above dynamic adjustment method, the system can adaptively adjust the feature weight according to the task to ensure the flexibility and pertinence of the system.

[0137] S26, removing the relevant feature vector whose feature weight is less than the weight threshold to obtain k target feature vectors most relevant to the target task, and generating a set of target feature vectors .

[0138] ​In the process of dynamic adjustment, the weights of some features may gradually decrease, and when they reach below the set weight threshold, these features are considered redundant features. The system will automatically eliminate these features that contribute less to the task, reducing noise in data processing.

[0139] In this embodiment, the condition for eliminating redundant features is defined as:

[0140] ;

[0141] In the formula, is a preset weight threshold, and features with weights below this threshold will be eliminated by the system to improve system efficiency.

[0142] After multiple iterations and weight adjustments, the k target feature vectors most relevant to the target task are obtained. The system generates an optimized feature vector set, which is obtained through the processes of mutual information screening, dynamic weight adjustment, and redundancy elimination. The set of target feature vectors is generated , .

[0143] The system continuously optimizes the feature selection process through feedback and iteration. When new data is input, the system reevaluates the weights of the features based on the latest data and makes necessary adjustments to ensure that the system remains sensitive to the most relevant features. Through the steps in S21 to S26, the system filters out the most relevant features to the task under the premise of ensuring data quality, reducing unnecessary processing burden, and thus improving the efficiency and accuracy of the scoring process.

[0144] S3, according to the input conditions, the scoring criteria are adjusted adaptively, and the filtered features are scored according to the fuzzy reasoning mechanism and the scoring criteria to obtain target feature scores.

[0145] To handle the scoring needs of complex multi-dimensional data, the present application introduces an adaptive scoring model based on fuzzy reasoning, which can handle uncertainty and fuzziness. By dynamically adjusting the scoring criteria, the highly relevant features selected are quantitatively scored to generate the final electronic file evaluation results, meeting the needs of different tasks and improving the flexibility and accuracy of scoring. For details, please refer to Figure 4 , Figure 4 is a flowchart of scoring the filtered features provided by the embodiment of the present application, as shown in Figure 4 The scoring of the filtered features provided by the embodiment includes steps S31 to S33.

[0146] S31, map the target feature vectors to the preset fuzzy set, and convert each target feature vector to the corresponding membership degree through the membership function.

[0147] The first step of fuzzy reasoning is to map the input into fuzzy sets, and the input value is converted into membership degree by fuzzy function; for each input target feature vector, the membership degree function is used to map it into a fuzzy set.

[0148] In this embodiment, triangular membership function is used for mapping calculation, and the membership degree function is:

[0149] ;

[0150] In the formula, is the mth target feature vector; , , is the parameter of the triangular function corresponding to the pth fuzzy set; is the target feature vector in the pth fuzzy set .

[0151] S32, the membership degree is adaptively adjusted according to the data attribute corresponding to the verification target and the target feature vector, and the adaptive adjustment formula is:

[0152] ;

[0153] In the formula, is the target feature vector in the pth fuzzy set after adjustment; is the target feature vector in the pth fuzzy set before adjustment; is the adjustment coefficient, and the adjustment coefficient is set according to the data type corresponding to the verification target and the target feature vector.

[0154] S33, reasoning is carried out through the fuzzy rule base, and the fuzzy output of the membership degree is calculated.

[0155] After obtaining the membership degree of the input value, the system carries out reasoning through the fuzzy rule base containing the preset fuzzy rules, wherein the core of the reasoning is to combine the membership degree corresponding to each target feature vector through the rules, and the minimum membership degree rule is used in the reasoning process to calculate the fuzzy output distribution, and the expression of the fuzzy output distribution is:

[0156] ;

[0157] ;

[0158] In the formula, a membership degree corresponding to the i-th rule of the fuzzy rule base; a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a first fuzzy set related to the i-th rule a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a second fuzzy set related to the i-th rule a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a target feature vector a membership degree in a fuzzy set a membership degree in a fuzzy set a first fuzzy set related to the i-th rule a total number of fuzzy sets related to the i-th rule a total number of rules a total number of rules

[0159] In this embodiment, after the membership degrees of each target feature vector in each fuzzy set are calculated, the fuzzy sets corresponding to each rule in the rule base are obtained, and the membership degrees corresponding to the rules are calculated. A series of fuzzy rules in the fuzzy rule base define the fuzzy relationship between the input and the output. The fuzzy rules are defined based on practical operation experience or expert experience knowledge, or obtained based on learning algorithms of experimental data.

[0160] S34, converting the fuzzy output into a specific score value by defuzzification.

[0161] The output result of the fuzzy inference is a fuzzy membership distribution, in order to obtain a specific score value, the system needs to convert the fuzzy output into a specific numerical value through defuzzification, the application adopts a weighted average method to calculate the score value, and the defuzzification calculation formula is:

[0162] ;

[0163] ;

[0164] In the formula, is the target feature score after defuzzification; represents the mth target feature vector in the file to be verified; represents the feature vector corresponding to in the verification standard file, 、 、 is a preset fuzzy output numerical value, the fuzzy output numerical value is set as a group of determined numerical values, representing the output result of each rule; is the total number of rules;k is the total number of target feature vectors.

[0165] In the embodiment, > > We can set the values of 、 、 to be 90, 70 and 50 respectively, if the field data is“in line with the rule”for the ith rule, that is , then takes the value 90, if the field data is“basically in line with the rule”for the ith rule, that is , then corresponding value is 70, and if the field data is“not in line with the rule”for the ith rule, that is , then corresponding value is 50.

[0166] In order to enable the system to adapt to different tasks and conditions, the application introduces an adaptive adjustment mechanism, by adjusting the membership function and fuzzy rules, the system can dynamically optimize the scoring standard according to the task requirements.

[0167] Through the above steps, we can obtain the target feature score corresponding to each verification task shown in the verification task list, and evaluate the data authenticity of the electronic file of the power transmission project according to the target feature score corresponding to each verification task. Similarly, we can also set other verification tasks according to requirements to evaluate the integrity, availability and security of the file.

[0168] Through the adaptive adjustment, the system can optimize the scoring rules under different tasks, and ensure that the scoring results are consistent with the task objectives. The system optimizes the fuzzy rule base and membership function through multiple iterations, and the accuracy of the scoring results is evaluated through a feedback mechanism, so that the fuzzy reasoning model is continuously optimized, and the scoring performance of the system is improved.

[0169] S4, generating an electronic file final verification result based on the target feature score.

[0170] After the target feature score of the electronic file is calculated, the quality of the electronic file can be known according to the height of the target feature score, and it is judged whether the electronic file meets the “four nature” requirements.

[0171] To sum up, the application discloses a power transmission and transformation project electronic file verification method based on an improved fuzzy reasoning mechanism. The method combines global dependency information (Transformer) and time sequence dependency information (LSTM), and lays a foundation for subsequent consistency, integrity and accuracy detection through hash verification and global dependency analysis. Then, the system selects features from the verified data through a conditional weight dynamic adjustment mechanism. The feature selection mechanism can filter out the most relevant features for the verification task, remove redundant information, improve data processing efficiency and usability, and use a dynamically adjusted weight model to adjust the importance of each feature in real time according to the actual needs of different file verification tasks, so that the most relevant features can be fully utilized. Next, a comprehensive scoring model based on an improved fuzzy reasoning mechanism is proposed. The model analyzes multi-dimensional data through fuzzy logic and dynamically optimizes fuzzy rules through an adaptive adjustment mechanism. The scoring standard is adjusted for different input conditions to ensure the accuracy and flexibility of the verification results in different task scenarios.

[0172] According to the method described in the above embodiment, the present embodiment will be further described from the perspective of a power transmission and transformation project electronic file verification system based on a fuzzy reasoning mechanism. The system can be implemented as an independent entity or integrated in an electronic device such as a terminal, which can include a mobile phone, a tablet computer, etc.

[0173] Please refer to Figure 5 , Figure 5 is a structural schematic diagram of the power transmission and transformation project electronic file verification system based on the fuzzy reasoning mechanism provided by the embodiment of the application, as Figure 5 shown, the power transmission and transformation project electronic file verification system based on the fuzzy reasoning mechanism provided by the embodiment of the application comprises:

[0174] The data preprocessing module is configured to sequentially process the metadata of the electronic archives by a Transformer model, an LSTM network and hash verification to obtain feature vectors corresponding to each metadata field.

[0175] The screening module is configured to screen the feature vectors most relevant to the verification task from the electronic archives management system based on a conditional weight dynamic adjustment mechanism, remove redundant data, and obtain target feature vectors most relevant to the verification task.

[0176] The scoring module is configured to score the target feature vectors according to fuzzy reasoning to obtain target feature scores corresponding to the verification task.

[0177] The result generation module is configured to generate a final verification result of the electronic archives based on the target feature scores.

[0178] In addition, please refer to Figure 6 , Figure 6 is a structural schematic diagram of an electronic device provided by the embodiment of the present application, as Figure 6 shown, the electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps in the power transmission project electronic archives verification method based on the fuzzy reasoning mechanism.

[0179] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or controlled by related hardware by instructions, which can be stored in a computer readable storage medium and loaded and executed by a processor. For this purpose, the embodiment of the present application provides a storage medium, in which a plurality of instructions are stored, which can be executed by a processor to implement the steps in the power transmission project electronic archives verification method based on the fuzzy reasoning mechanism.

[0180] Generally, computer instructions used to implement the method of the present application can be carried by any combination of one or more computer readable storage media. Non-transitory computer readable storage media can include any computer readable medium except a signal propagating transitorily.

[0181] A computer readable storage medium, for example, can be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0182] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an

[0183] While the embodiments of the present application have been illustrated and described, it will be understood by those skilled in the art that various changes can be made and equivalents can be substituted for elements thereof without departing from the scope of the present application. In addition, many modifications can be made to adapt a particular situation or material to the teachings of the present application without departing from the central inventive concept described herein.

Claims

1. A method for verifying electronic records of power transmission and transformation projects based on fuzzy reasoning mechanism, characterized in that: The verification method includes: S1. The metadata of the electronic archive is processed sequentially through the Transformer model, LSTM network, and hash check to obtain the feature vector corresponding to each metadata field; S2. Based on the conditional weight dynamic adjustment mechanism, the feature vectors most relevant to the verification task in the electronic records management system are selected to obtain the target feature vectors most relevant to the verification task. S3. Score the target feature vector based on fuzzy inference to obtain the target feature score corresponding to the verification task; S4. Based on the target feature score, generate the final verification result of the electronic archive.

2. The method for verifying electronic archives of power transmission and transformation projects based on fuzzy reasoning mechanism according to claim 1, characterized in that, S1 includes: S11. Perform vectorization processing on all metadata fields to obtain the first vector corresponding to each metadata field; S12. Calculate the attention weights between each first vector and other first vectors using the self-attention mechanism in the Transformer model, and generate the second vectors corresponding to each first vector based on the attention weights between each first vector and other first vectors. S13. Generate a hidden state vector for each second vector using an LSTM network to obtain the third vector corresponding to each metadata field; S14. Generate a corresponding hash value for each third vector through hash verification to obtain the feature vector corresponding to each metadata field.

3. The method for verifying electronic archives of power transmission and transformation projects based on fuzzy reasoning mechanism according to claim 1, characterized in that, S2 includes: S21. Calculate the mutual information between each feature vector and the verification task; S22. Filter out relevant feature vectors from the feature vectors whose mutual information with the verification task is greater than the mutual information threshold; S23. Assign initial weights to the relevant feature vectors; S24. Using the gradient descent method, the initial weights are iteratively adjusted according to the loss function to obtain the iterative weights; S25. Adjust the weights after iteration according to the conditional adjustment formula to obtain the feature weights corresponding to each relevant feature vector; S26. Set the feature weights to be less than the weight threshold. By removing the relevant feature vectors, we obtain the k target feature vectors that are most relevant to the verification task.

4. The method for verifying electronic archives of power transmission and transformation projects based on fuzzy reasoning mechanism according to claim 3, characterized in that: In step S21, the mutual information calculation formula is as follows: ； In the formula, Let j represent the j-th eigenvector. Indicates the verification target. Representing the eigenvector With verification target Mutual information between them; express A specific example, express belong The sample space; express A specific example, express belong The sample space; Representing the eigenvector With verification target The joint probability between them; and They represent the eigenvectors respectively. With verification target The marginal probability.

5. The method for verifying electronic archives of power transmission and transformation projects based on fuzzy reasoning mechanism according to claim 3, characterized in that: In step S24, the gradient descent method is used to iteratively adjust the initial weights according to the loss function, which includes the following steps: The predicted value is calculated based on the current weights and the corresponding relevant feature vectors. The loss function is calculated by comparing the predicted value with the true value. The current weights are updated based on the loss function and the gradient descent formula. The above steps are repeated until the loss function value is less than the preset threshold or the number of weight updates reaches the preset upper limit. The loss function is: ； In the formula, Represents the loss function; Indicates the system's predicted value; Represents the actual value; This is the vector of current feature weights; The number of samples; The gradient descent formula is as follows: ； In the formula, This represents the i-th relevant feature vector when the weights are updated for the t-th time. The weights; This represents the i-th relevant feature vector when the weights are updated for the (t+1)th time. The weights; This is the learning rate.

6. The method for verifying electronic archives of power transmission and transformation projects based on fuzzy reasoning mechanism according to claim 3, characterized in that: In S25, the condition adjustment formula is: ； In the formula, For the i-th relevant feature vector Feature weights; Adjustment coefficients for conditions; The condition function reflects the eigenvector. The weight adjustment requirements are determined by the data type of the verification task and the relevant feature vectors; For relevant feature vectors The corresponding weights after iteration.

7. The method for verifying electronic archives of power transmission and transformation projects based on fuzzy reasoning mechanism according to claim 1, characterized in that, S3 includes: S31. Map the target feature vectors to a preset fuzzy set, and transform each target feature vector into a corresponding membership degree using a membership function, wherein the membership function is: ； In the formula, Let m be the feature vector of the target; , , The parameters of the trigonometric function corresponding to the p-th fuzzy set; For target feature vectors The p-th fuzzy set Membership degree in; S32. Adaptively adjust the membership degree according to the data type corresponding to the verification target and the target feature vector. The adaptive adjustment formula is as follows: ； In the formula, For the adjusted target feature vector The p-th fuzzy set Membership degree in; Adjustment factor, adjustment factor The settings are based on the data type corresponding to the verification target and the target feature vector; S33. Based on the membership degree corresponding to each target feature vector, inference is performed through a fuzzy rule base containing preset fuzzy rules to obtain a fuzzy output distribution. The expression is: ； ； In the formula, Let be the membership degree corresponding to the i-th rule in the fuzzy rule base; Represents the target feature vector In fuzzy sets Membership degree in; Represents the target feature vector In fuzzy sets Membership degree in; Represents the target feature vector In fuzzy sets Membership degree in; Let represent the first fuzzy set related to the i-th rule; Represents the target feature vector In fuzzy sets Membership degree in; Represents the target feature vector In fuzzy sets Membership degree in; Represents the target feature vector In fuzzy sets Membership degree in; Let represent the second fuzzy set related to the i-th rule; Represents the target feature vector In fuzzy sets Membership degree in; Represents the target feature vector In fuzzy sets Membership degree in; Represents the target feature vector In fuzzy sets Membership degree in; Indicates the i-th rule related to the 1st rule. A fuzzy set; This represents the total number of fuzzy sets associated with the i-th rule; The total number of rules; S34. The fuzzy output is converted into a target feature score through defuzzification. The calculation formula for defuzzification is as follows: ； ； In the formula, To score the target features after defuzzification; This represents the m-th target feature vector in the file to be verified; This indicates that the verification standard file is related to The corresponding feature vectors, , , The preset value, > > .

8. A verification system for electronic archives of power transmission and transformation projects based on fuzzy reasoning mechanism, including: The data preprocessing module is used to process the metadata of electronic archives sequentially through the Transformer model, LSTM network, and hash check to obtain the feature vectors corresponding to each metadata field. The filtering module is used to filter out the feature vectors most relevant to the verification task in the electronic records management system based on the condition weight dynamic adjustment mechanism, and remove redundant data to obtain the target feature vectors most relevant to the verification task. The scoring module is used to score the target feature vector based on fuzzy inference to obtain the target feature score corresponding to the verification task. The result generation module is used to generate the final verification result of the electronic archive based on the target feature score.

9. An electronic device, characterized in that, The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 7.

Images