Electric power engineering technical-economic review method based on deep learning

Abstract

The invention discloses an electric power engineering technical-economic review method based on deep learning. The method comprises the following steps: S1, automatically capturing project data such as engineering quantity, an equipment list and material cost, and carrying out preliminary detection; s2, analyzing data by using a deep learning model, and analyzing and classifying text information through natural language processing; s3, establishing a charging benchmark rule base, and automatically checking the cost data in the technical and economic files; s4, adopting a self-adaptive material semantic matching algorithm to examine equipment and material data, and carrying out market information price matching; s5, constructing a mapping relation between the project amount and the typical cost, and generating a deviation report; s6, establishing a data center to support multi-dimensional statistical analysis and a real-time cost report; and S7, combining historical and real-time data, performing early warning through a deep learning algorithm, and generating an intelligent review report. According to the invention, automation and intellectualization of electric power engineering technical-economic review are realized, and data processing efficiency and review accuracy are optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power engineering technical and economic review, and in particular to an electric power engineering technical and economic review method based on deep learning. Background Art

[0002] With the rapid development of power engineering projects, the technical and economic review of power engineering has become increasingly important. Technical and economic review involves multiple links such as engineering quantity verification, budget preparation, cost analysis and project progress control, and its purpose is to ensure the economic rationality of engineering projects and the effectiveness of resource utilization. However, the existing technical and economic review methods mainly rely on manual analysis and traditional data processing tools. This method is inefficient and inaccurate when facing complex data structures and large-scale engineering projects, and it is difficult to adapt to the needs of modern power engineering.

[0003] Traditional power engineering technical and economic review methods usually use rule bases and empirical formulas for data verification and analysis. These methods can play a certain role when the amount of data is small and the project structure is simple, but when dealing with complex engineering quantities, material lists and dynamic market information, the existing methods have obvious shortcomings:

[0004] The first is that data processing is complex and it is difficult to quickly analyze large-scale data. Traditional technical and economic review methods can work when the amount of data is small, but they are not efficient enough when faced with large-scale data in power engineering. The data of engineering projects include various types such as engineering quantities, equipment lists, and cost information, and the quantity is huge and the data structure is complex. Existing methods rely on manual data capture and processing, which is difficult to adapt to the needs of rapid review, resulting in a time-consuming and error-prone data verification process.

[0005] The second is the lack of intelligent cost verification and comparison mechanism; technical and economic reviews require accurate verification of various costs in engineering projects (such as construction costs, installation costs, and material costs). However, existing methods usually rely on manual experience and fixed rule comparisons, lack a real-time update mechanism for dynamic market prices, and are difficult to accurately reflect current market fluctuations, affecting the accuracy of cost control. The inefficiency and uncertainty of the cost verification process can easily lead to budget deviations.

[0006] The third is the lack of decision support and the lack of early warning and risk warning functions. In technical and economic review, the existing methods are difficult to conduct intelligent analysis and automatic early warning, and lack the integration and analysis mechanism of historical data and real-time data. Traditional methods cannot provide intelligent risk warnings and review opinions, and decision makers rely on manual analysis to judge budgets and resource allocation, which increases management risks. The shortcomings of existing technical and economic review methods in risk control and decision support limit the effectiveness of project management.

[0007] Therefore, how to provide a power engineering technical review method based on deep learning is an urgent problem that technicians in this field need to solve. Summary of the invention

[0008] One purpose of the present invention is to propose a method for technical review of electric power engineering based on deep learning. The invention makes full use of deep learning and data mining technology, and describes in detail the whole process from data capture, cost verification to intelligent review, and has the advantages of high data processing efficiency, accurate cost verification, and timely risk warning.

[0009] A method for evaluating power engineering technical qualifications based on deep learning according to an embodiment of the present invention comprises the following steps:

[0010] S1. Automatically capture project data from power engineering technical documents, including engineering quantities, equipment lists and material cost data, and conduct preliminary testing on the captured data;

[0011] S2. Analyze and process captured data based on deep learning models, parse text information through natural language processing algorithms, identify data items and classify and archive them;

[0012] S3. Establish a billing benchmark rule base, use preset rules and industry standards to automatically verify the cost data in technical and economic documents, including construction project fees, installation project fees and engineering inspection fees, and output standardized inspection results;

[0013] S4. Apply the adaptive material semantic analysis and matching algorithm to review the equipment and material data in the technical and economic documents, automatically match them with the market information price, and output the verification report;

[0014] S5. Construct a mapping relationship between engineering quantities and typical construction costs and control line indicators, perform comparative analysis of typical construction costs by calling historical project data in the database, generate a difference report, identify deviations, and provide audit suggestions;

[0015] S6. Establish an enterprise-level data center, integrate technical and economic data, support multi-dimensional statistical analysis and refinement of cost management indicators, dynamically adjust data models, and generate real-time cost analysis reports;

[0016] S7. Combine historical data and real-time monitoring data, conduct data analysis and early warning through deep learning algorithms, generate intelligent review opinions, and output technical and economic review reports containing predictions and risk warnings.

[0017] Optionally, the S2 specifically includes:

[0018] S21. Preprocess the captured project data and map different data items to multi-dimensional feature space based on the deep learning model to form an initial data vector set D={ d1,d2,...,d n} , where d i Characteristic vectors representing data items, including engineering quantities, equipment lists, and material costs;

[0019] S22. Apply natural language processing algorithms to parse text data in technical documents and use word embedding models to map text information into a semantic vector space T= { t1,t2,...,t m} , where t j A semantic vector representing each piece of text information;

[0020] S23. Calculate the text data vector t based on the semantic similarity algorithm j and the preprocessed initial data vector d i The similarity S(d i ,t j ), filter out highly relevant data items by setting a similarity threshold;

[0021] S24, classify and archive the selected data items according to the preset classification standard, build a hierarchical data structure, and form a data classification set C= { c1,c2...,,c k} , where c k Represents a subset of data archived by engineering category and material category.

[0022] Optionally, the S23 specifically includes:

[0023] S231, based on the preprocessed initial data vector set D= { d1,d2,...,d n} And text semantic vector set T = { t1,t2,...,t m } , calculate each text vector t j With each data vector d i The semantic similarity S(d i ,t j );

[0024] S232, using the improved cosine similarity formula with dynamic weight factor to calculate the semantic similarity value S(d i ,t j ):

[0025]

[0026] Among them, d i ·t j Represents vector d i and tj The dot product of ∥d i ∥ and ∥t j ∥ respectively represent vector d i and t j The norm of di According to the data item d i The weight factor of the dynamic adjustment of the characteristics;

[0027] S233, for each calculated similarity value S(d i ,t j ), set the similarity threshold θ, when S(d i ,t j )≥θ, the text data t j Identified as data item d i Highly relevant data and mark them as candidate data items;

[0028] S234, sort the selected candidate data items according to the similarity values ​​from high to low, and generate a high correlation data set R = {(d i ,t j )|S(d i ,t j )≥θ}, where every pair (d i ,t j 0 indicates a highly correlated match.

[0029] Optionally, the S3 specifically includes:

[0030] S31. According to the industry standards of power engineering budget and cost management, a billing benchmark rule base B = {b1, b2,, b k}, where each rule b k Includes standard values ​​and threshold ranges for construction project fees, installation project fees, and project inspection fees;

[0031] S32, extract the cost data in the technical and economic files to form a cost vector set F = {f1, f2, ..., f n}, where f i Indicates the actual value of the cost items, including construction project fees, installation project fees and project inspection fees;

[0032] S33. For each cost item f i The corresponding standard value b in the billing benchmark rule base k Perform deviation calculation and define the deviation value as:

[0033] Δf i =f i -b k ;

[0034] Where Δfi represents the cost item f i With benchmark rule b k The difference between

[0035] S34, judging the deviation value Δf i Is it within the threshold range? If |Δf is satisfied i |≤∈ k , then the cost item f i Marked as compliant if |Δf i |>∈ k , then the cost item f i Mark as an exception and record the exception type;

[0036] S35. Output the normative test results, including the list of cost items that meet the norms. 合规 and a list of expense items marked as abnormal F 异常 , where F 合规 = { f i ||Δf i |≤∈ k } , (F 异常 ={f i ||Δf i |>∈ k}.

[0037] Optionally, the S4 specifically includes:

[0038] S41. Extract the list data of equipment and materials from the technical and economic documents and generate the material data vector set M = m1, m2, ..., m p}, where m i The data vector representing the i-th material contains the material's name, quantity, specification, and unit cost information;

[0039] S42, through the material semantic analysis matching algorithm, the material data vector m i Match with the corresponding material information in the design drawing, and the similarity of material matching is calculated as:

[0040]

[0041] Among them, d i represents the reference vector of the i-th material in the design drawing, is the adaptive weight factor, m i ·d i Represents the dot product of material data and design drawing data, ∥m i ∥ and ∥d i ∥ represents the norm of the material and design drawing vectors respectively;

[0042] S43, calculate the similarity value S ( m i ,d i) After that, compare it with the preset similarity threshold θ m For comparison, if S ( m i ,d i) ≥θ m , then the material item m i Mark as a successful match, if S9m i ,d i) <θ m , then the material item is marked as a match failure and recorded as an item to be verified;

[0043] S44, for the successfully matched material item m i , obtain the corresponding market reference price p from the market database i , generate the market price vector set P = {p1, p2, ..., p p};

[0044] S45, for material item m i The actual cost of i and the market price p i , introducing a dynamic adjustment factor The formula used for price deviation calculation is to adaptively adjust the price deviation according to the characteristics of the material category and the purchase batch:

[0045]

[0046] in, is the dynamic adjustment factor, f i For material item m i Actual cost value, p i is the market reference price, Δp i Indicates the adjusted price deviation value;

[0047] S46. Set the price deviation Δp i With deviation threshold ∈ p For comparison, if |Δp i |≤∈ p , then the material item m i Marked as price compliant if |Δp i |>∈ p , it is marked as price abnormality and a verification report is generated. The verification report includes the material matching status and price deviation analysis results.

[0048] Optionally, the S45 specifically includes:

[0049] S451. Extract each material item m from the market databasei The market price p i , and set dynamic adjustment factors based on material item categories, purchase batches, and market fluctuations The initial value of forms a dynamic adjustment factor set in Reflects the material item m i Adaptability coefficient in the current market environment;

[0050] S452, for different material types, adjust the factor Classify and set materials according to their scarcity, frequency of use, and purchasing cycle, and set low Set a high value for scarce and customized materials. value;

[0051] S453, adjust according to the purchase batch For bulk purchases, reduce For small batch and urgent purchase items, increase The value of

[0052] S454, the final price deviation calculation is expressed as

[0053] Optionally, the 7 specifically includes:

[0054] S71, obtain technical and economic document related data from the historical database and real-time monitoring system, and construct a historical data vector set H = {h1,h2,...,h q} and real-time monitoring data vector set R = { r1,r2,...,r s } , where each vector h i Represents historical project data items, including engineering quantities, equipment lists, and material costs. Each vector r j Represents the real-time data items in the current monitoring system, including the current cost status and progress information;

[0055] S72. Through the deep learning model, the historical data vector set H and the real-time data vector set R are correlated and analyzed, and the time series prediction algorithm is applied to generate the prediction vector P= { p1,p2,...,p s] , where p j Indicates the real-time data within a specific time period. j forecasts to identify potential cost deviations and risk trends;

[0056] S73. Define risk assessment indicator λ risk , calculate each real-time data item rj Its corresponding predicted value p j The deviation Δr j , introduce the dynamic adjustment coefficient γ j For increased flexibility in deviations:

[0057] Δr j =γ j ·(r j -p j );

[0058] Among them, γ j is the real-time data item r j The dynamic adjustment coefficient is set dynamically according to the criticality of the project, the volatility of real-time data and the historical deviation trend;

[0059] S74, if |Δr j |>λ risk , then the data item is marked as high-risk data and assigned a risk level of L risk ; According to the deviation value Δr j The size of the data item is assigned a corresponding risk level, and the high-risk data item r j Classify by risk level to form a risk level set R L = { r L1 ,r L2 ,...,r Ln} ;

[0060] S75, combined with risk level set R L and the actual deviation value Δr j , generate intelligent technical and economic review reports, including risk level classification, potential deviation trends, and predicted cost change information;

[0061] S76. Output a technical and economic review report containing predictions and risk warnings. The report can be used to support real-time monitoring and adaptive adjustments. By integrating historical data and real-time monitoring information, dynamic management and intelligent early warning of the technical and economic review process can be achieved.

[0062] Optionally, the S73 specifically includes:

[0063] S731, extract the project-related deviation data and fluctuation information from the historical database, and generate a deviation trend vector V = {v1, v2, ..., v q}, where each vector v i Indicates the deviation mean, fluctuation range and criticality level of specific project data in the historical period;

[0064] S732, for each real-time data item r jAccording to the criticality level and fluctuation range of the project to which it belongs, the corresponding deviation mode is selected from the deviation trend vector V and the initial dynamic adjustment coefficient γ is set j,init ;

[0065] S733, based on the current data item r j Real-time monitoring of the initial adjustment coefficient γ j,init Further fine-tuning is performed to calculate the final dynamic adjustment coefficient γ j , whose value depends on the criticality of the project and the volatility of the data:

[0066] γ j =γ j,init ·(1+δ j );

[0067] Among them, δ j It is a real-time volatility correction factor, which is set according to the volatility of current monitoring data and risk warning status;

[0068] S734, the final dynamic adjustment coefficient γ j Applied to the calculation of real-time data deviation, the deviation value is expressed as Δr j =γ j ·(r j -p j );

[0069] S735, compare deviation value Δr j Compared with the preset risk assessment index λ risk , if |Δr j |>λ risk , then mark r j It is a high-risk data item, and generates high-risk warning information, recording the deviation size and adjustment coefficient γ of this data item j application situation.

[0070] The beneficial effects of the present invention are:

[0071] (1) The present invention combines deep learning models and natural language processing technology to realize the whole process of automatically capturing and parsing project data from power engineering technical documents, and quickly completes the preliminary detection and classification and archiving of data through semantic analysis and data mapping. This method has the characteristics of high efficiency and automation in multidimensional data processing, greatly reducing the necessity of manual intervention, improving the efficiency of large-scale data processing, and adapting to the technical and economic review needs of complex engineering projects.

[0072] (2) The present invention uses an adaptive material semantic matching algorithm and a dynamic adjustment factor to accurately verify and compare the equipment and material cost data in the technical and economic documents. By constructing a billing benchmark rule library and combining it with market reference prices, the present invention realizes the automatic verification of cost items and the real-time comparison of market prices, effectively ensuring the rationality and market adaptability of the project budget. The cost verification results are further optimized through deviation analysis, which significantly improves the accuracy of cost data and the reliability of review.

[0073] (3) The present invention establishes an enterprise-level data center, integrates historical data and real-time monitoring information, and uses deep learning algorithms to perform multi-dimensional analysis on review data to generate intelligent review opinions and risk warnings. The dynamic adjustment of historical data and the risk warning mechanism based on real-time data make project review more intelligent, identify potential risks in a timely manner, improve the decision-making support capabilities of technical and economic reviews, and provide a scientific basis for the economic control of power engineering projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0075] Figure 1 A flowchart of a power engineering technical review method based on deep learning proposed by the present invention;

[0076] Figure 2 A flowchart of cost verification and material matching in a power engineering technical review method based on deep learning proposed by the present invention; DETAILED DESCRIPTION

[0077] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0078] refer to Figure 1-2 , a power engineering technical review method based on deep learning, comprising the following steps:

[0079] S1. Automatically capture project data from power engineering technical documents, including engineering quantities, equipment lists and material cost data, and conduct preliminary testing on the captured data;

[0080] S2. Analyze and process captured data based on deep learning models, parse text information through natural language processing algorithms, identify data items and classify and archive them;

[0081] S3. Establish a billing benchmark rule base, use preset rules and industry standards to automatically verify the cost data in technical and economic documents, including construction project fees, installation project fees and engineering inspection fees, and output standardized inspection results;

[0082] S4. Apply the adaptive material semantic analysis and matching algorithm to review the equipment and material data in the technical and economic documents, automatically match them with the market information price, and output the verification report;

[0083] S5. Construct a mapping relationship between engineering quantities and typical construction costs and control line indicators, perform comparative analysis of typical construction costs by calling historical project data in the database, generate a difference report, identify deviations, and provide audit suggestions;

[0084] S6. Establish an enterprise-level data center, integrate technical and economic data, support multi-dimensional statistical analysis and refinement of cost management indicators, dynamically adjust data models, and generate real-time cost analysis reports;

[0085] S7. Combine historical data and real-time monitoring data, conduct data analysis and early warning through deep learning algorithms, generate intelligent review opinions, and output technical and economic review reports containing predictions and risk warnings.

[0086] In this implementation, S2 specifically includes:

[0087] S21. Preprocess the captured project data and map different data items to multi-dimensional feature space based on the deep learning model to form an initial data vector set D= { d1,d2,...,d n} , where d i Characteristic vectors representing data items, including engineering quantities, equipment lists, and material costs;

[0088] S22. Apply natural language processing algorithms to parse text data in technical documents and use word embedding models to map text information into a semantic vector space T= { t1,t2,...,t m} , where t j A semantic vector representing each piece of text information;

[0089] S23. Calculate the text data vector t based on the semantic similarity algorithm j and the preprocessed initial data vector d i The similarity S(d i ,t j ), filter out highly relevant data items by setting a similarity threshold;

[0090] S24, classify and archive the selected data items according to the preset classification standard, build a hierarchical data structure, and form a data classification set C = {c1, c2..., c k}, where c k Represents a subset of data archived by engineering category and material category.

[0091] In this implementation manner, the S23 specifically includes:

[0092] S231, based on the preprocessed initial data vector set D = {d1, d2, ..., d n} and text semantic vector set T = {t1, t2, ..., t m}, calculate each text vector t j With each data vector d i The semantic similarity S(d i ,t j );

[0093] S232, using the improved cosine similarity formula with dynamic weight factor to calculate the semantic similarity value S(d i ,t j ):

[0094]

[0095] Among them, d i ·t j Represents vector d i and t j The dot product of ∥d i ∥ and ∥t j ∥ respectively represent vector d i and t j The norm of di According to the data item d i The weight factor of the dynamic adjustment of the characteristics;

[0096] S233, for each calculated similarity value S(d i ,t j ), set the similarity threshold θ, when S(d i ,t j )≥θ, the text data t j Identified as data item d i Highly relevant data and mark them as candidate data items;

[0097] S234, sort the selected candidate data items according to the similarity values ​​from high to low, and generate a high correlation data set R = {(d i ,t j )|S(d i ,t j)≥θ}, where every pair (d i ,t j ) indicates a high correlation matching pair.

[0098] In this implementation, S3 specifically includes:

[0099] S31. According to the industry standards of power engineering budget and cost management, a billing benchmark rule base B = {b1, b2,, b k}, where each rule b k Includes standard values ​​and threshold ranges for construction project fees, installation project fees, and project inspection fees;

[0100] S32, extract the cost data in the technical and economic files to form a cost vector set F = {f1, f2, ..., f n}, where f i Indicates the actual value of the cost items, including construction project fees, installation project fees and project inspection fees;

[0101] S33. For each cost item f i The corresponding standard value b in the billing benchmark rule base k Perform deviation calculation and define the deviation value as:

[0102] Δf i =f i -b k ;

[0103] Where Δf i represents the cost item f i With benchmark rule b k The difference between

[0104] S34, judging the deviation value Δf i Is it within the threshold range? If |Δf is satisfied i |≤∈ k , then the cost item f i Marked as compliant if |Δf i |>∈ k , then the cost item f i Mark as an exception and record the exception type;

[0105] S35. Output the normative test results, including the list of cost items that meet the norms. 合规 and a list of expense items marked as abnormal F 异常 , where F 合规 ={f i ||Δf i |≤∈ k},(F 异常 ={f i ||Δfi |>∈ k}.

[0106] In this implementation manner, the S4 specifically includes:

[0107] S41. Extract the list data of equipment and materials from the technical and economic documents and generate the material data vector set M = m1, m2, ..., m p}, where m i The data vector representing the i-th material contains the material's name, quantity, specification, and unit cost information;

[0108] S42, through the material semantic analysis matching algorithm, the material data vector m i Match with the corresponding material information in the design drawing, and the similarity of material matching is calculated as:

[0109]

[0110] Among them, d i represents the reference vector of the i-th material in the design drawing, is the adaptive weight factor, m i ·d i Represents the dot product of material data and design drawing data, ∥m i ∥ and ∥d i ∥ represents the norm of the material and design drawing vectors respectively;

[0111] S43, calculate the similarity value S ( m i ,d i) After that, compare it with the preset similarity threshold θ m For comparison, if S ( m i ,d i) ≥θ m , then the material item m i Mark as a successful match, if S ( m i ,d i) <θ m , then the material item is marked as a match failure and recorded as an item to be verified;

[0112] S44, for the successfully matched material item m i , obtain the corresponding market reference price p from the market database i , generate the market price vector set P = {p1, p2, ..., p p};

[0113] S45, for material item m i The actual cost of i and the market price pi , introducing a dynamic adjustment factor The formula used for price deviation calculation is to adaptively adjust the price deviation according to the characteristics of the material category and the purchase batch:

[0114]

[0115] in, is the dynamic adjustment factor, f i For material item m i Actual cost value, p i is the market reference price, Δp i Indicates the adjusted price deviation value;

[0116] S46. Set the price deviation Δp i With deviation threshold ∈ p For comparison, if |Δp i |≤∈ p , then the material item m i Marked as price compliant if |Δp i |>∈ p , it is marked as price abnormality and a verification report is generated. The verification report includes the material matching status and price deviation analysis results.

[0117] In this implementation manner, the S45 specifically includes:

[0118] S451. Extract each material item m from the market database i The market price p i , and set dynamic adjustment factors based on material item categories, purchase batches, and market fluctuations The initial value of forms a dynamic adjustment factor set in Reflects the material item m i Adaptability coefficient in the current market environment;

[0119] S452, for different material types, adjust the factor Classify and set materials according to their scarcity, frequency of use, and purchasing cycle, and set low Set a high value for scarce and customized materials. value;

[0120] S453, adjust according to the purchase batch For bulk purchases, reduce For small batch and urgent purchase items, increase The value of

[0121] S454, the final price deviation calculation is expressed as

[0122] In this implementation manner, the 7 specifically includes:

[0123] S71, obtain technical and economic document related data from the historical database and real-time monitoring system, and construct a historical data vector set H = {h1,h2,...,h q} and real-time monitoring data vector set R = { r1,r2,...,r s } , where each vector h i Represents historical project data items, including engineering quantities, equipment lists, and material costs. Each vector r j Represents the real-time data items in the current monitoring system, including the current cost status and progress information;

[0124] S72. Through the deep learning model, the historical data vector set H and the real-time data vector set R are correlated and analyzed, and the time series prediction algorithm is applied to generate the prediction vector P= { p1,p2,...,p s] , where p j Indicates the real-time data within a specific time period. j forecasts to identify potential cost deviations and risk trends;

[0125] S73. Define risk assessment indicator λ risk , calculate each real-time data item r j Its corresponding predicted value p j The deviation Δr j , introduce the dynamic adjustment coefficient γ j For increased flexibility in deviations:

[0126] Δr j =γ j ·(r j -p j );

[0127] Among them, γ j is the real-time data item r j The dynamic adjustment coefficient is set dynamically according to the criticality of the project, the volatility of real-time data and the historical deviation trend;

[0128] S74, if |Δr j |>λ risk , then the data item is marked as high-risk data and assigned a risk level of L risk ; According to the deviation value Δr j The size of the data item is assigned a corresponding risk level, and the high-risk data item r j Classify by risk level to form a risk level set RL = { r L1 ,r L2 ,...,r Ln} ;

[0129] S75, combined with risk level set R L and the actual deviation value Δr j , generate intelligent technical and economic review reports, including risk level classification, potential deviation trends, and predicted cost change information;

[0130] S76. Output a technical and economic review report containing predictions and risk warnings. The report can be used to support real-time monitoring and adaptive adjustments. By integrating historical data and real-time monitoring information, dynamic management and intelligent early warning of the technical and economic review process can be achieved.

[0131] In this implementation manner, the S73 specifically includes:

[0132] S731, extract the project-related deviation data and fluctuation information from the historical database, and generate a deviation trend vector V = {v1, v2, ..., v q}, where each vector v i Indicates the deviation mean, fluctuation range and criticality level of specific project data in the historical period;

[0133] S732, for each real-time data item r j According to the criticality level and fluctuation range of the project to which it belongs, the corresponding deviation mode is selected from the deviation trend vector V and the initial dynamic adjustment coefficient γ is set j,init ;

[0134] S733, based on the current data item r j Real-time monitoring of the initial adjustment coefficient γ j,init Further fine-tuning is performed to calculate the final dynamic adjustment coefficient γ j , whose value depends on the criticality of the project and the volatility of the data:

[0135] γ j =γ j,init ·(1+δ j );

[0136] Among them, δ j It is a real-time volatility correction factor, which is set according to the volatility of current monitoring data and risk warning status;

[0137] S734, the final dynamic adjustment coefficient γ j Applied to the calculation of real-time data deviation, the deviation value is expressed as Δr j =γ j ·(r j-p j );

[0138] S735, compare deviation value Δr j Compared with the preset risk assessment index λ risk , if |Δr j |>λ risk , then mark r j It is a high-risk data item, and generates high-risk warning information, recording the deviation size and adjustment coefficient γ of this data item j application situation.

[0139] Embodiment 1:

[0140] The deep learning-based power engineering technical and economic review method of the present invention was implemented in Company B, a large power engineering company located in a province in southern China. Company B is responsible for the construction of multiple key power engineering projects in the province, mainly involving the installation of power facilities, equipment configuration, and management of engineering quantities. With the expansion of project scale and fluctuations in the market environment, Company B faces a series of challenges in the process of technical and economic review, including large amounts of data, cumbersome manual verification, and inaccurate cost control. In order to solve these problems, Company B decided to introduce the method of the present invention to improve the efficiency and accuracy of technical and economic review through deep learning models and intelligent data verification methods.

[0141] During the implementation process, Company B first used the method of the present invention to capture and automatically parse the technical and economic documents of multiple projects under construction. Through the deep learning model, the system can automatically extract important data items such as engineering quantities, equipment lists, and material costs, and conduct preliminary inspections on the format and standardization of the data to ensure the integrity of the files and the availability of the data. Compared with the traditional method that relies on manual sorting and proofreading, the time for data capture and preliminary inspection has been shortened from an average of 6 hours to 1 hour, greatly improving work efficiency.

[0142] Company B uses the adaptive material semantic matching algorithm of the present invention to accurately verify the equipment and material data in the technical and economic documents, and compares it with the material information in the design drawings to ensure the consistency of the engineering quantity and the bill of materials. In addition, through the market price database, the system of the present invention can obtain the latest market reference prices of equipment and materials in real time, and automatically complete the rationality verification of the cost. The data shows that through the automatic verification process, the material cost deviation is reduced from 3.8% of the traditional manual verification to 0.9%, and the accuracy of the material cost is significantly improved. In addition, Company B also uses the present invention to establish an enterprise-level data center in the technical and economic review, which is used to uniformly manage various types of engineering data and support multi-dimensional statistical analysis. The data center integrates the historical data of more than 50 projects completed by Company B in the past three years, and dynamically compares it with the real-time data of the current project, and analyzes potential cost deviations and risk factors through a deep learning model. For each project, the system generates a technical and economic review report containing cost forecasts, deviation analysis and risk prompts to help decision makers adjust budgets in a timely manner and optimize resource allocation. Data shows that in the process of dynamic management of the five pilot projects, an average of four potential cost overruns per project were discovered and adjusted through the early warning mechanism, effectively controlling budget overruns.

[0143] Table 1 Comparison of data before and after the implementation of technical and economic review

[0144]

[0145] As shown in Table 1 above, the implementation of the method of the present invention significantly improved the efficiency and accuracy of the technical and economic review of Company B. Specifically, the data capture and detection time was shortened by 6 times, the material cost deviation was reduced by 3 times, and the project cost control deviation was reduced from ±5.0% to ±1.5%, achieving precise control of the project budget. At the same time, through the intelligent early warning and dynamic management mechanism, Company B was able to identify and adjust potential risks in advance during the technical and economic review process, effectively reducing cost overruns.

[0146] After implementing the present invention, Company B significantly improved the efficiency and accuracy of technical and economic review. Compared with traditional methods, the present invention provides a more intelligent and accurate solution in data capture, cost verification and dynamic management, which not only significantly shortens the review time, but also improves the reliability of the project budget. This deep learning-based technical and economic review method helps Company B better control costs and manage resources in a complex engineering environment, and improves the economic benefits and management level of the project.

[0147] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A power engineering technical review method based on deep learning, characterized in that: The following steps are involved: S1. Automatically capture project data from power engineering technical documents, including engineering quantities, equipment lists and material cost data, and conduct preliminary testing on the captured data; S2. Analyze and process captured data based on deep learning models, parse text information through natural language processing algorithms, identify data items and classify and archive them; S3. Establish a billing benchmark rule base, use preset rules and industry standards to automatically check the cost data in technical and economic documents, including construction project fees, installation project fees and project inspection fees, and output standardized inspection results; S4. Apply the adaptive material semantic analysis and matching algorithm to review the equipment and material data in the technical and economic documents, automatically match them with the market information price, and output the verification report; S5. Construct a mapping relationship between engineering quantities and typical construction costs and control line indicators, perform comparative analysis of typical construction costs by calling historical project data in the database, generate a difference report, identify deviations, and provide audit suggestions; S6. Establish an enterprise-level data center, integrate technical and economic data, support multi-dimensional statistical analysis and refinement of cost management indicators, dynamically adjust data models, and generate real-time cost analysis reports; S7. Combine historical data and real-time monitoring data, conduct data analysis and early warning through deep learning algorithms, generate intelligent review opinions, and output technical and economic review reports containing predictions and risk warnings.

2. The method for evaluating the technical experience of electric power engineering based on deep learning according to claim 1, characterized in that: The S2 specifically includes: S21, pre-process the captured project data, map different data items to multi-dimensional feature space based on the deep learning model, and form an initial data vector set D = {d1, d2, ..., d n }, where d i Characteristic vectors representing data items, including engineering quantities, equipment lists, and material costs; S22, apply natural language processing algorithms to parse the text data in the technical documents, and use word embedding models to map the text information to the semantic vector space T = {t1, t2, ..., t m }, where t j A semantic vector representing each piece of text information; S23. Calculate the text data vector t based on the semantic similarity algorithm j and the preprocessed initial data vector d i The similarity S(d i ,t j ), filter out highly relevant data items by setting a similarity threshold; S24, classify and archive the selected data items according to the preset classification standard, build a hierarchical data structure, and form a data classification set C = {c1, c2..., c k }, where c k Represents a subset of data archived by engineering category and material category.

3. The method for evaluating the technical experience of electric power engineering based on deep learning according to claim 2, characterized in that: The S23 specifically includes: S231, based on the preprocessed initial data vector set D = {d1, d2, ..., d n } and text semantic vector set T = {t1, t2, ..., t m }, calculate each text vector t j With each data vector d i The semantic similarity S(d i ,t j ); S232, using the improved cosine similarity formula with dynamic weight factor to calculate the semantic similarity value S(d i ,t j ): Among them, d i ·t j Represents vector d i and t j The dot product of ∥d i ∥ and ∥t j ∥ respectively represent vector d i and t j The norm of According to the data item d i The weight factor of the dynamic adjustment of the characteristics; S233, for each calculated similarity value S(d i ,t j ), set the similarity threshold θ, when S(d i ,t j )≥θ, the text data t j Identified as data item d i Highly relevant data and mark them as candidate data items; S234, sort the selected candidate data items according to the similarity values ​​from high to low, and generate a high correlation data set R = {(d i ,t j )|S(d i ,t j )≥θ}, where every pair (d i ,t j ) indicates a high correlation matching pair.

4. The method for evaluating the technical experience of electric power engineering based on deep learning according to claim 1, characterized in that: The S3 specifically includes: S31. According to the industry standards of power engineering budget and cost management, build a billing benchmark rule base B = b1, b2,, b k }, where each rule b k Includes standard values ​​and threshold ranges for construction project fees, installation project fees, and project inspection fees; S32, extract the cost data in the technical and economic files to form a cost vector set F = {f1, f2, ..., f n }, where f i Indicates the actual value of the cost items, including construction project fees, installation project fees and project inspection fees; S33. For each cost item f i The corresponding standard value b in the billing benchmark rule base k Perform deviation calculation and define the deviation value as: Δf i =f i -b k ; Where Δf i represents the cost item f i With benchmark rule b k The difference between S34, judging the deviation value Δf i Is it within the threshold range? If |Δf is satisfied i |≤∈ k , then the cost item f i Marked as compliant if |Δf i |>∈ k , then the cost item f i Mark as an exception and record the exception type; S35. Output the normative test results, including the list of cost items that meet the norms. 合规 and a list of expense items marked as abnormal F 异常 , where F 合规 ={f i ||Δf i |≤∈ k },(F 异常 ={f i ||Δf i |>∈ k }.

5. The method for evaluating power engineering technical qualifications based on deep learning according to claim 1, characterized in that: The S4 specifically includes: S41. Extract the list data of equipment and materials from the technical and economic documents and generate the material data vector set M = m1, m2, ..., m p }, where m i The data vector representing the i-th material contains the material's name, quantity, specification, and unit cost information; S42, through the material semantic analysis matching algorithm, the material data vector m i Match with the corresponding material information in the design drawing, and the similarity of material matching is calculated as: Among them, d i represents the reference vector of the i-th material in the design drawing, is the adaptive weight factor, m i ·d i Represents the dot product of material data and design drawing data, ∥m i ∥ and ∥d i ∥ represents the norm of the material and design drawing vectors respectively; S43, calculate the similarity value S(m i ,d i ), and then compare it with the preset similarity threshold θ m For comparison, if S(m i ,d i )≥θ m , then the material item m i Mark as a successful match, if S(m i ,d i )<θ m , then the material item is marked as a match failure and recorded as an item to be verified; S44, for the successfully matched material item m i , obtain the corresponding market reference price p from the market database i , generate the market price vector set P = {p1, p2, ..., p p }; S45, for material item m i The actual cost of i and the market price p i , introduce dynamic adjustment factor β mi The formula used for price deviation calculation is to adaptively adjust the price deviation according to the characteristics of the material category and the purchase batch: in, is the dynamic adjustment factor, f i For material item m i Actual cost value, p i is the market reference price. Indicates the adjusted price deviation value; S46. Set the price deviation Δp i With deviation threshold ∈ p For comparison, if |Δp i |≤∈ p , then the material item m i Marked as price compliant if |Δp i |>∈ p , it is marked as price abnormality and a verification report is generated. The verification report includes the material matching status and price deviation analysis results.

6. The method for evaluating power engineering technical qualifications based on deep learning according to claim 5, characterized in that: The S45 specifically includes: S451. Extract each material item m from the market database i The market price p i , and set dynamic adjustment factors based on material item categories, purchase batches, and market fluctuations The initial value of forms a dynamic adjustment factor set in Reflects the material item m i Adaptability coefficient in the current market environment; S452, for different material types, adjust the factor Classify and set materials according to their scarcity, frequency of use, and purchasing cycle, and set low Set a high value for scarce and customized materials. value; S453, adjust according to the purchase batch For bulk purchases, reduce For small batch and urgent purchase items, increase The value of S454, the final price deviation calculation is expressed as 7. The method for evaluating power engineering technical qualifications based on deep learning according to claim 1, characterized in that: The 7 specifically include: S71, obtain technical and economic document related data from the historical database and real-time monitoring system, and construct a historical data vector set H = {h1,h2,...,h q } and real-time monitoring data vector set R = {r1, r2, ..., r s }, where each vector h i Represents historical project data items, including engineering quantities, equipment lists, and material costs. Each vector r j Represents the real-time data items in the current monitoring system, including the current cost status and progress information; S72, through the deep learning model, the historical data vector set H and the real-time data vector set R are correlated and analyzed, and the time series prediction algorithm is applied to generate the prediction vector P = (p1, p2, ..., p s ), where p j Indicates the real-time data within a specific time period. j forecasts to identify potential cost deviations and risk trends; S73. Define risk assessment indicator λ risk , calculate each real-time data item r j Its corresponding predicted value p j The deviation Δr j , introduce the dynamic adjustment coefficient γ j For increased flexibility in deviations: Δr j =c j ·(r j -p j ); Among them, γ j is the real-time data item r j The dynamic adjustment coefficient is set dynamically according to the criticality of the project, the volatility of real-time data and the historical deviation trend; S74, if |Δr j |>λ risk , then the data item is marked as high-risk data and assigned a risk level of L risk ; According to the deviation value Δr j The size of the data item is assigned a corresponding risk level, and the high-risk data item r j Classify by risk level to form a risk level set R L = {r L1 ,r L2 ,...,r Ln }; S75, combined with risk level set R L and the actual deviation value Δr j , generate intelligent technical and economic review reports, including risk level classification, potential deviation trends, and predicted cost change information; S76. Output a technical and economic review report containing predictions and risk warnings. The report can be used to support real-time monitoring and adaptive adjustments. By integrating historical data and real-time monitoring information, dynamic management and intelligent early warning of the technical and economic review process can be achieved.

8. The method for evaluating power engineering technical qualifications based on deep learning according to claim 7, characterized in that: The S73 specifically includes: S731, extract the project-related deviation data and fluctuation information from the historical database, and generate a deviation trend vector V = {v1, v2, ..., v q }, where each vector v i Indicates the deviation mean, fluctuation range and criticality level of specific project data in the historical period; S732, for each real-time data item r j According to the criticality level and fluctuation range of the project to which it belongs, the corresponding deviation mode is selected from the deviation trend vector V and the initial dynamic adjustment coefficient γ is set j,init ; S733, based on the current data item r j Real-time monitoring of the initial adjustment coefficient γ j,init Further fine-tuning is performed to calculate the final dynamic adjustment coefficient γ j , whose value depends on the criticality of the project and the volatility of the data: c j =c j,init ·(1+d j ); Among them, δ j It is a real-time volatility correction factor, which is set according to the volatility of current monitoring data and risk warning status; S734, the final dynamic adjustment coefficient γ j Applied to the calculation of real-time data deviation, the deviation value is expressed as Δr j =γ j ·(r j -p j ); S735, compare deviation value Δr j Compared with the preset risk assessment index λ risk , if |Δr j |>λ risk , then mark r j It is a high-risk data item, and generates high-risk warning information, recording the deviation size and adjustment coefficient γ of this data item j application situation.

Images