Power transmission and transformation project cost analysis model construction method based on quantitative indexes

Abstract

The invention discloses a power transmission and transformation project cost analysis model construction method based on quantitative indexes, and belongs to the technical field of electric power project cost, and the method comprises the steps: S1, carrying out the semantic recognition of first historical power transmission and transformation project data based on natural language processing, and constructing a final feature library; s2, performing quantitative index and quantitative index correlation extraction on the final feature library based on a convolutional neural network to establish a first initial cost analysis model; s3, constructing an adaptive simplification model at least comprising a weight calculation rule and a similar engineering screening rule based on a discriminant clustering algorithm and an AHP method; and S4, updating the first initial cost analysis model based on the self-adaptive simplified model to obtain a project cost analysis model. The technical problem that it is difficult to improve the discrimination efficiency while improving the discrimination accuracy of the same kind of engineering is solved, the model complexity is reduced, and the discrimination accuracy and efficiency of the same kind of engineering are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power project cost, and specifically to a method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators. Background Art

[0002] With the rapid development of big data, higher requirements are put forward for the timeliness and accuracy of construction cost information in the construction industry. The construction of power transmission and transformation projects has the characteristics of huge project investment, multiple involved fields, and complex influencing factors. Therefore, controlling and obtaining the cost of power transmission and transformation project construction has always been a difficult problem. The existing technologies mainly predict project costs by constructing prediction models. For example, a method, system, device, and medium for constructing a power grid project cost prediction model with the patent number CN113537614A. Among them, the construction method includes: obtaining historical data of construction costs of different project types in the power grid project, and dividing the historical data into several stage project training sets; selecting at least one target training set corresponding to a preset prediction requirement from the several stage project training sets and combining them to obtain a combined training set; using the combined training set to train an initial prediction model to obtain a target power grid project cost prediction model. The above method splits and combines the training data, making the trained model more in line with the power grid project cost prediction requirements, and improving the accuracy of the prediction results to a certain extent. However, due to the particularity of the construction industry, the data sources are rich and there are a large number of homogeneous data. It can be seen that the above cost prediction model is only constructed based on historical data. Therefore, the above cost prediction model has a high complexity. On this basis, it is also difficult to accurately identify similar projects. When analyzing the project costs of similar projects, it is difficult to ensure accuracy and analysis efficiency. Summary of the Invention

[0003] Aiming at the problem that it is difficult to improve the discrimination accuracy and efficiency of similar projects in the prior art, the present invention provides a method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators. By using the feature extraction and parameter sharing characteristics of the convolutional neural network, quantitative indicators and the correlation of quantitative indicators are extracted from the final feature library to establish a first initial cost analysis model, eliminating a large amount of homogeneous data and initially improving the discrimination efficiency. An adaptive refinement model including at least a weight calculation rule and a similar project screening project is constructed through a discrimination clustering algorithm and AHP to update the first initial cost analysis model to obtain a project cost analysis model. On the basis of realizing in-depth mining of project data, the discrimination accuracy of similar projects is improved. At the same time, direct discrimination is carried out through rules, and the discrimination efficiency is also improved. The technical problem of being difficult to improve the discrimination accuracy and efficiency of similar projects at the same time is solved, and the model complexity is reduced.

[0004] To solve the above technical problems, the present invention provides a method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators, including the following steps: S1: Based on natural language processing, perform semantic recognition on the first historical power transmission and transformation project data to construct a final feature library; S2: Based on a convolutional neural network, extract the correlation between quantitative indicators from the final feature library to establish a first initial cost analysis model; S3: Based on the discriminant clustering algorithm and the AHP method, construct an adaptive refinement model that at least includes a weight calculation rule and a similar project screening rule; S4: Based on the adaptive refinement model, update the first initial cost analysis model to obtain a project cost analysis model.

[0005] After adopting the above technical solutions, the present invention has the following advantages: By using the feature extraction and parameter sharing characteristics of the convolutional neural network to extract the correlation between quantitative indicators from the final feature library to establish a first initial cost analysis model, a large amount of homogeneous data is eliminated, the complexity of the model is reduced, and the discrimination efficiency is thereby improved; Considering that the discriminant clustering algorithm only has classification characteristics, but in the process of discriminating similar projects, the importance of different equipment in the discrimination of similar projects is different. Therefore, the accuracy of discriminating similar projects by only using the discriminant clustering algorithm is relatively low. And through AHP, the importance between different equipment can be obtained, but AHP does not have classification characteristics. Therefore, by combining the clustering algorithm and AHP, the limitations in the power project cost analysis are mutually compensated. On the basis of realizing in-depth mining of project data, the accuracy of discriminating similar projects is improved. At the same time, by directly discriminating through rules, the discrimination efficiency is also improved. It solves the technical problem of being difficult to improve the discrimination efficiency while improving the accuracy of discriminating similar projects, reduces the model complexity, and significantly improves the accuracy and efficiency of discriminating similar projects.

[0006] Preferably, the S1 includes: S11: Define word segmentation rules based on industry term habits, and perform division and part-of-speech tagging processing on the first historical power transmission and transformation project data according to the analysis rules to obtain tagged historical power transmission and transformation project data; S12: Based on the word vector model in natural language processing, perform vector conversion on the tagged historical power transmission and transformation project data to obtain power transmission and transformation project vectors; S13: Calculate the first semantic similarity between power transmission and transformation project vectors and the second semantic similarity between power transmission and transformation project vectors and the to-be-confirmed feature library, and obtain the final feature library according to the first semantic similarity and the second semantic similarity.

[0007] In this solution, the labeled historical power transmission and transformation project data is converted into power transmission and transformation project vectors suitable for calculating semantic similarity through a word vector model. Then, the final feature library is obtained by calculating the first semantic similarity between the power transmission and transformation project vectors, initially eliminating the homogenized data existing in the to-be-confirmed feature library and reducing the complexity of the first initial cost analysis model. By calculating the second semantic similarity between the power transmission and transformation project vectors and the to-be-confirmed feature library, accurate classification of the labeled historical power transmission and transformation project data is achieved, improving the accuracy of the final feature library.

[0008] Preferably, the S2 includes: S21: Input the final feature library into a convolutional neural network for convolution and pooling operations to obtain the first correlation between the features in the final feature library and cost calculation, and use the features with the first correlation greater than the preset correlation as quantization indicators; S22: Obtain the second correlation between the quantization indicators based on regression analysis, and construct a second initial cost analysis model based on the quantization indicators and the second correlation; S23: Perform information price matching based on the second historical power transmission and transformation project data, and update the second initial cost analysis model according to the information price matching result to obtain the first initial cost analysis model.

[0009] In this solution, the first correlation with cost calculation is obtained through the feature extraction and parameter sharing characteristics of the convolutional neural network, and then the quantization indicators are obtained, eliminating the redundant data in the final feature library. A model is established with clean quantization indicators, further reducing the complexity of the model. The second initial cost analysis model is constructed by obtaining the second correlation through regression analysis. When some equipment attributes are obtained, all attributes of the equipment can be obtained according to the second initial cost analysis model, improving the accuracy of cost analysis and simplifying the analysis process at the same time.

[0010] Preferably, the S23 includes: S231: Perform information price matching between the second historical power transmission and transformation project data and the data in the second initial cost analysis model. If there is data in the second initial cost analysis model with an information price matching result greater than or equal to the preset value, the second initial cost analysis model is the first initial cost analysis model; if not, execute S232; S232: Update the second initial cost analysis model by adding data attributes to the second initial cost analysis model based on the information price matching result to obtain the first initial cost analysis model.

[0011] Preferably, the S3 includes: S31: Construct a hierarchical framework based on project attributes and discriminant clustering algorithms; S32: Define weight calculation rules and similar project screening rules through the hierarchical framework, AHP, and entropy weight method.

[0012] In this solution, a hierarchical framework is constructed through a discriminant clustering algorithm to achieve the classification of indicators or features. Rules are obtained through AHP and the entropy weight method. Based on the classification, the weight calculation rules and the screening rules for similar projects are obtained through the subjectivity of experts in AHP and the objectivity of the entropy weight method, improving the scientificity and accuracy of the weight calculation rules and the screening rules for similar projects. Further, the accurate weight calculation rules are combined with the classification, and screening is carried out according to the screening rules for similar projects, further improving the accuracy of the determination of similar projects.

[0013] Preferably, in S32, the weight calculation rules include: Obtain the judgment matrix of the second indicator in the second layer of the hierarchical framework through expert judgment, and obtain the subjective weight of the second indicator based on the judgment matrix; Standardize the judgment matrix through the entropy weight method to obtain the objective weight of the second indicator, and obtain the final weight based on the subjective weight and the objective weight; The screening rules for similar projects include: Conduct a preliminary screening based on the first indicator in the first layer of the hierarchical framework; On the basis of the preliminary screening, obtain the project scores based on the final weight and the Pearson correlation coefficient, and screen out similar projects according to the project scores.

[0014] Preferably, the expression for obtaining the final weight based on the subjective weight and the objective weight is In the formula, final is the final weight, subj is the subjective weight, and obj is the objective weight.

[0015] Preferably, the S4 includes: S41: Merge the data in the first initial cost analysis model through the weight calculation rules and the screening rules for similar projects, and obtain the corresponding relationship between the data in the first initial cost analysis model; S42: Based on the regression analysis, mine the functional relationship between the data in the first initial cost analysis model to construct a correlation function, and obtain the cost analysis model for engineering projects based on the corresponding relationship and the correlation function.

[0016] In this solution, data is merged through the weight calculation rules and the screening rules for similar projects, further reducing the complexity of the model. By obtaining the corresponding relationship, it is convenient to obtain cost data based on the basic engineering data, further improving the efficiency of cost analysis. Through regression analysis, in-depth mining is carried out on the data without a corresponding relationship, and then a correlation function is constructed. The cost analysis model for engineering projects is obtained based on the corresponding relationship and the correlation function, further improving the accuracy and analysis efficiency of the cost analysis model for engineering projects.

[0017] Preferably, it further includes: Set corresponding accounting algorithms for the engineering quantities of different classifications based on the equipment types used in the project, and formulate an abnormal warning mechanism according to the results of the accounting algorithms.

[0018] Beneficial effects of this solution: Through the feature extraction and parameter sharing characteristics of the convolutional neural network, the quantization index and the correlation of the quantization index are extracted from the final feature library to establish the first initial cost analysis model, eliminating a large amount of homogeneous data, reducing the complexity of the model, and thus improving the discrimination efficiency; Considering that the discriminant clustering algorithm only has classification characteristics, but in the process of discriminating similar projects, the importance degrees of different equipment are also different. Therefore, the accuracy of discriminating similar projects only through the discriminant clustering algorithm is relatively low, while the importance between different equipment can be obtained through AHP, but AHP does not have classification characteristics. Therefore, by combining the clustering algorithm with AHP, the limitations in engineering cost analysis are mutually compensated. On the basis of realizing in-depth mining of engineering data, the accuracy of discriminating similar projects is improved. At the same time, through direct discrimination by rules, the discrimination efficiency is also improved. It solves the technical problem of being difficult to improve the discrimination efficiency while improving the accuracy of discriminating similar projects, reduces the model complexity, and significantly improves the accuracy and efficiency of discriminating similar projects; Specifically, a hierarchical framework is constructed through the discriminant clustering algorithm to realize the classification of indicators or features. Rules are obtained through AHP and the entropy weight method. On the basis of classification, the weight calculation rules and the screening rules for similar projects are obtained through the subjectivity of experts in AHP and the objectivity of the entropy weight method, improving the scientificity and accuracy of the weight calculation rules and the screening rules for similar projects. Further, the accurate weight calculation rules are combined with the classification, and screening is carried out according to the screening rules for similar projects, further improving the accuracy of determining similar projects.

[0019] The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the method for constructing an input and output power transmission and transformation project cost analysis model based on quantization indexes are implemented. Description of the Drawings

[0020] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, objectives, and advantages of the present invention will become more obvious. The drawings are only for the purpose of showing the preferred embodiments and are not considered as limiting the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components.

[0021] Figure 1 It is a flowchart of the method for constructing an input and output power transmission and transformation project cost analysis model based on quantization indexes of the present invention; Figure 2 This is the scatter plot of the single and double circuit lengths and the main body costs in the method for constructing the power transmission and transformation project cost analysis model based on quantitative indicators of the present invention; Figure 3 This is the visualization model verification result diagram in the method for constructing the power transmission and transformation project cost analysis model based on quantitative indicators of the present invention. Specific Embodiments

[0022] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only the best embodiments of the present invention, which are only used to explain the present invention and do not limit the protection scope of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations (or steps) can be implemented in parallel, concurrently or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but there can also be additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0024] Embodiment 1: As Figure 1 shown, the method for constructing the power transmission and transformation project cost analysis model based on quantitative indicators includes the following steps: S1: Based on natural language processing, perform semantic recognition on the first historical power transmission and transformation project data to construct the final feature library.

[0025] The S1 includes: S11: Define the word segmentation rules based on the industry term habits, and perform division and part-of-speech annotation processing on the first historical power transmission and transformation project data according to the word segmentation rules to obtain the annotated historical power transmission and transformation project data; S12: Based on the word vector model in natural language processing, perform vector conversion on the annotated historical power transmission and transformation project data to obtain the power transmission and transformation project vectors; S13: Calculate the first semantic similarity between the power transmission and transformation project vectors and the second semantic similarity between the power transmission and transformation project vectors and the to-be-confirmed feature library, and obtain the final feature library according to the first semantic similarity and the second semantic similarity.

[0026] In this embodiment, the first historical power transmission and transformation project data includes data in carriers such as tables, texts, drawings, and GIM files. In the process of identifying the text semantics and obtaining the power transmission and transformation project vectors by defining word segmentation rules, part-of-speech tagging, word vector calculation, etc., a corresponding correction algorithm is also formulated to correct the data with problems such as data missing, data errors, duplicate records, and data inconsistency to ensure data quality. Through the formula Calculate the first semantic similarity. In the formula, Similarity is the similarity, and F1 and V1 represent the power transmission and transformation project vectors or the vectors corresponding to the to-be-confirmed feature library. Merge the power transmission and transformation project vectors with the first semantic similarity greater than the first preset similarity, and classify the power transmission and transformation project vectors with the second similarity greater than the second preset similarity into the corresponding to-be-confirmed feature library. In addition, the first preset similarity is usually greater than the second preset similarity. The final feature library at least includes the equipment material information library and the equipment material information price library. In the present invention, the labeled historical power transmission and transformation project data is converted into power transmission and transformation project vectors suitable for calculating semantic similarity through a word vector model, and then the final feature library is obtained by calculating the first semantic similarity between the power transmission and transformation project vectors, initially eliminating the homogeneous data existing in the to-be-confirmed feature library and reducing the complexity of the first initial cost analysis model. By calculating the second semantic similarity between the power transmission and transformation project vectors and the to-be-confirmed feature library, accurate classification of the labeled historical power transmission and transformation project data is achieved, and the accuracy of the final feature library is improved.

[0027] Specifically, taking the new substation project and the new transmission line project as the research objects, typical power engineering equipment materials in specialties such as primary substation, line structure, and line electrical are selected. According to the eight major types of primary equipment, secondary equipment, secondary equipment of intelligent substations, installation materials, communication equipment, and information equipment, the relevant characteristic attributes of 73 kinds of equipment materials are sorted out and analyzed, and the characteristic items that can uniquely describe the equipment materials are identified to construct an equipment material feature library for part-of-speech tagging of equipment material description texts. Among them, the main equipment material information of the power transmission and transformation project is based on the historical project budget data of the power transmission and transformation project. A total of 8,983 basic data items are sorted out, forming 18 categories of primary equipment with 65 characteristics; 10 categories of secondary equipment with 11 characteristics; 13 categories of secondary equipment of intelligent substations with 14 characteristics; 28 categories of installation materials with 62 characteristics; 3 categories of communication equipment with 10 characteristics; 1 category of information equipment with 4 characteristics. The main equipment material information library is shown in Table 1: Table 1. Main Equipment Material Information Library Taking the electrical appliances and circuits in substation transformation as the research object, typical power engineering equipment materials in specialties such as primary substation, line structure, and line electrical are selected. Among them, the main equipment information database for power engineering and the installation material information database for power engineering are based on "Information Price of Grid Engineering Equipment Materials (2022 Edition)", identify the relevant attributes and price fields of the equipment materials, and make specific identifications to form an information price database for equipment materials. The main equipment material information price database is shown in Table 2: Table 2. Information Price Database of Equipment Materials

[0028] S2: Establish the first initial cost analysis model based on extracting the correlation between the quantization index and the quantization index from the final feature library using a convolutional neural network.

[0029] The above S2 includes: S21: Input the final feature library into the convolutional neural network for convolution and pooling operations to obtain the first correlation between the features in the final feature library and cost calculation, and use the features with the first correlation greater than the preset correlation as quantization indexes; S22: Obtain the second correlation between the quantization indexes based on regression analysis, and construct the second initial cost analysis model based on the quantization indexes and the second correlation; S23: Perform information price matching based on the second historical power transmission and transformation project data, and update the second initial cost analysis model according to the information price matching result to obtain the first initial cost analysis model.

[0030] The above S23 includes: S231: Perform information price matching between the second historical power transmission and transformation project data and the data in the second initial cost analysis model. If there is data in the second initial cost analysis model with an information price matching result greater than or equal to the preset value, then the second initial cost analysis model is the first initial cost analysis model; if not, execute S232; S232: Add data attributes to the second initial cost analysis model based on the information price matching result to update the second initial cost analysis model to obtain the first initial cost analysis model.

[0031] In this embodiment, the preset correlation is flexibly set according to user needs. The second initial cost analysis model is a model including a knowledge graph. The knowledge graph framework is constructed through quantization indexes, and the quantization indexes with correlations in the knowledge graph framework are associated through the second correlation, thereby obtaining a model including a knowledge graph. When some attributes of the equipment are obtained, all attributes of the equipment can be obtained according to this model, improving the accuracy of cost analysis and simplifying the analysis process at the same time.

[0032] In this embodiment, part-of-speech tagging is performed on the second historical power transmission and transformation project data to construct a power engineering equipment and material library. Feature recognition is carried out on the second historical power transmission and transformation project data through a model including a knowledge graph and the power engineering equipment and material library to obtain the specific features of the second historical power transmission and transformation project data, and the information price is obtained by comparing with the preliminary design budget estimate. The overall structure of the second initial cost analysis model is divided into three layers, namely the input layer, the model layer, and the output layer. The core part of the natural language processing task execution is the acquisition and processing of the training data. The data used in this training is mainly the information price of State Grid equipment and materials and the equipment and material list. The data format of the equipment and materials is as follows: {"major category": "primary equipment", "equipment and material name": "1000 kV transformer", "attribute combination": "1000 kV, 1000 MVA, single-phase, 1000 / 500 / 110, non-excited, three-winding autotransformer", "unit": "set", "information price": 38054264, "information price (excluding tax)": 32805400}; the data format of the equipment and material list is as follows: {"equipment name": "500 kV main transformer 334 MVA (no-load) (outdoor installation)", "type": "main equipment", "unit": "set"}; Before the input layer, the data needs to be preprocessed. First, the text data is roughly cleaned to filter out invalid characters, such as: outdoor installation, etc.; second, jieba is used to segment the text to filter out punctuation marks and stop words, so as to select the core vocabulary; finally, the jieba tool is used to identify the part of speech of the core vocabulary. At this time, the number of part-of-speech classifications in jieba is used as the total number of all part-of-speech categories in the experiment. The input layer integrates word embedding, part-of-speech encoding, and position encoding. The input is sequentially converted into a comprehensive vector integrating part-of-speech features and position features through a matrix, which contains the encoding of the original sequence information and the target sequence information. The input passes through each layer of the Transformer, and a fully connected layer and a softmax layer are added to the output of the last layer to convert the output into the information price. The model layer is similar to the Transformer decoder layer, and a normalization layer is added after each layer and then passed to the next layer. In the training stage, actually the model output result is a probability distribution, and the information price corresponding to the position with the highest probability is the required unit price, which is similar to a multi-classification problem, and the cross entropy is used to calculate the loss of the model. Taking the transformer as an example: A total of 799 types of equipment information prices and 8137 sample equipment and material data are collected through sorting. Select the information price data for the experiment, and divide the selected sample data into a training set and a test set according to a ratio of 4:1. Taking the transformer as an example, the input sample is: {"equipment name": "110 kV oil-immersed on-load transformer, 50 MVA, 110 / 10, horizontal split", "type": "main equipment", "unit": "set"}. The output result after model training is shown in Table 3 below: Table 3. Equipment and Material Matching Degree From the probability column, it can be seen that the information price of the device is most likely to be 1,799,224. To verify the accuracy of the algorithm, 678 power transmission and transformation sample projects were screened, including 1,162 transformers and 3,311 gas-insulated switchgear (GIS) for identification. After being identified by the algorithm, the matching accuracy of the information price is 92%, and the matching errors are mainly caused by too little device attribute information. Therefore, when the information price matching results are all less than the preset value, data attributes are added to the second initial cost analysis model. The preset value is flexibly set according to requirements. By matching the predicted information price with the actual information price, the second initial cost analysis model is adjusted accordingly according to the matching results, improving the accuracy and flexibility of the second initial cost analysis model.

[0033] S3: Construct an adaptive reduction model based on the discriminant clustering algorithm and the AHP method, which includes at least the weight calculation rule and the similar project screening rule.

[0034] The above S3 includes: S31: Construct a hierarchical framework based on project attributes and the discriminant clustering algorithm; S32: Define the weight calculation rule and the similar project screening rule through the hierarchical framework, AHP and entropy weight method.

[0035] In S32, the weight calculation rule includes: Obtain the judgment matrix of the second index in the second layer of the hierarchical framework through expert judgment, and obtain the subjective weight of the second index based on the judgment matrix; Standardize the judgment matrix through the entropy weight method to obtain the objective weight of the second index, and obtain the final weight based on the subjective weight and the objective weight; The similar project screening rule includes: Conduct a preliminary screening based on the first index in the first layer of the hierarchical framework; On the basis of the preliminary screening, obtain the project score based on the final weight and the Pearson correlation coefficient, and screen out similar projects according to the project score.

[0036] The expression for obtaining the final weight based on the subjective weight and the objective weight is In the formula, final is the final weight, subj is the subjective weight, and obj is the objective weight.

[0037] In this embodiment, the first indicators are attributes, structures, scales, functions, etc. The second indicators are obtained specifically according to the project type. For example, for substation projects, the second indicators, i.e., discrimination indicators, are sorted out mainly from aspects such as the main transformer system, the high-voltage side of the distribution device, the medium-voltage side of the distribution device, the low-voltage side of the distribution device, and reactive power compensation. For line projects, the second indicators, i.e., discrimination indicators, are sorted out mainly from aspects such as voltage level, number of circuits, number of splits, conductor type, meteorological conditions, terrain, altitude, etc. The subjective weight table is obtained through expert evaluation, and the relative importance judgment matrix of the second indicators is obtained through the subjective weight table. Taking the main transformer system as an example, as shown in Table 4: Table 4. Subjective weight table The matrix is normalized to obtain the normalized matrix, and the eigenvector W of the target layer matrix p =(0.333, 0.143, 0.238, 0.048, 0.238), that is, the required weight W p1 =0.333, W p2 =0.143, W p3 =0.238, W p4 =0.048, W p5 =0.238. Conduct a consistency test on the judgment matrix. When the consistency test passes, the obtained weight is the subjective weight; otherwise, recalculate the subjective weight, thereby effectively avoiding the situation where judgments between individuals conflict when making multi-factor judgments.

[0038] In this embodiment, the entropy weight method is used to standardize each element in the judgment matrix. Taking the main transformer system as an example, the obtained objective weight table is shown in Table 5 below: Table 5. Objective weight table The weights of each index of the number of main transformers (current period), the capacity of main transformers (current period), the number of main transformers (long-term), the capacity of main transformers (long-term), and the layout form of main transformers are calculated to be v1 = v2 = v3 = v4 = v5 = 0.2. The weight results of each index of the main transformer system are obtained through the combined weighting method as shown in Table 6 below: Table 6. Final weight table Index layer AHP relative weight Entropy weight method relative weight Combined weight Number of main transformers (current period) 0.333 0.2 0.267 Capacity of main transformers (current period) 0.143 0.2 0.171 Number of main transformers (long term) 0.238 0.2 0.219 Capacity of main transformers (long term) 0.048 0.2 0.124 Arrangement form of main transformers 0.238 0.2 0.219 After obtaining the final weights, first compare the first indicators, and then conduct a preliminary screening. The projects obtained from the preliminary screening are screened again. In statistics, the Pearson correlation coefficient, also known as the Pearson product-moment correlation coefficient, is used to measure the correlation (linear correlation) between two variables X and Y, and its value ranges from -1 to 1. Taking the overhead line as an example, to analyze the correlation between the project quantity data and the project basic data and technical data, relevant indicators are selected based on expert experience for correlation analysis to calculate the Pearson correlation coefficient. Where X are variables such as area tags, line length grades (single-folded), average span distances, tension tower ratios, line lengths and terrain distributions of lines with different designed icing wind speeds, and Y represents engineering quantity data such as tower material quantity per unit length and conductor quantity per unit length. According to the above formula, the correlation coefficient between these two variables can be calculated. What the neural network calls "learning" refers to the process of automatically obtaining the optimal weight parameters from the training data with a certain index as a clue. The index used in the learning of the neural network is called the loss function. This loss function can use any function, but generally the mean squared error and cross-entropy error, etc. The mean squared error is shown as follows: y k represents the output of the neural network, t k represents the supervised data, and k represents the dimension of the data. In addition to the mean squared error, the cross-entropy error is also often used as a loss function. The cross-entropy error is shown as follows: E = -∑ k t k log y k , log represents the natural logarithm with e as the base (log e ). y k is the output of the neural network, t kIt is the correct solution of labels, using cross entropy as the loss function. Taking the overhead line project as an example, the tower type, wind speed, ice cover, split number, material and other variables are the input layer, and the output layer is the project score. By training 161 sample projects and randomly inputting a project, the scores of 161 sample projects can be obtained. The higher the score, the higher the similarity between the projects. Taking a 110 kV overhead line project in Zhejiang Province as an example, the tower type of the project is an angle steel tower, the split number is one split, the conductor cross-section is 300 mm, the wind speed is 33 km / h, there is no ice zone, the unit total investment is 1.4871 million yuan / km, and the unit length body cost is 1.1543 million yuan / km. The technical parameters of the project were input into the model. After training, the scores of 161 sample projects were obtained, of which the lowest score was 12 points and the highest score was 93 points. It was a 110 kV overhead line project in Zhejiang Province. The tower type of the project was an angle steel tower, the number of splits was one split, the conductor cross-section was 300 mm, the wind speed was 35 km / h, and the ice area was light. The total unit investment of the project was 1.3899 million yuan / km, with a difference rate of 6.54%, and the unit length body cost was 1.0469 million yuan / km, with a difference rate of 9.3%. In this scheme, a hierarchical framework was constructed by discriminant clustering algorithm to achieve the classification of indicators or features, and rules were obtained by AHP and entropy weight method. On the basis of classification, the weight calculation rules and similar project screening rules were obtained by the expert subjectivity of AHP and the objectivity of entropy weight method, which improved the scientificity and accuracy of the weight calculation rules and similar project screening rules, and further combined the accurate weight calculation rules with classification, and screened according to the screening rules of similar projects, which further improved the accuracy of similar project judgment.

[0039] S4: updating the first initial cost analysis model based on the adaptive simplified model to obtain a project cost analysis model.

[0040] The S4 includes: S41: merging the data in the first initial cost analysis model through the weight calculation rule and the similar project screening rule and obtaining the corresponding relationship between the data in the first initial cost analysis model; S42: Based on the regression analysis, mining the functional relationship between the data in the first initial cost analysis model constructs a correlation function, and obtaining a project cost analysis model based on the corresponding relationship and the correlation function.

[0041] In this embodiment, the corresponding relationship is the strength and correspondence of the correlation between technical conditions and technical conditions, between technical conditions and engineering quantities, and between technical conditions and construction costs. By performing correlation analysis on cost data, technical parameters, etc., it can be found that there is a one-to-one correspondence between some engineering quantities and a multiple relationship between some engineering quantities. The corresponding method of each engineering quantity is clarified through statistics as follows. For engineering quantity objects with consistent engineering quantity and technical solution data, a one-to-one correspondence is adopted for determination, that is, the main technical conditions are determined, and the engineering quantity is determined, and no measurement and adjustment are required. For example, the engineering quantity indicator "main transformer", once the main technical conditions are determined, the number of main transformers is also determined, and the engineering quantity can be directly extracted from the corresponding data form. The example table of the one-to-one correspondence determination algorithm for standard engineering quantities is shown in Table 7: Table 7. Example of algorithm for determining standard engineering quantities in one-to-one correspondence The multiple formula is used for calculation when there is a multiple relationship between the engineering quantity and the technical solution data, that is, there is an obvious multiple relationship (proportional relationship) between the engineering quantity and the technical parameter, and between the engineering quantity and the engineering quantity. Under the premise of clarifying the multiple relationship, the standard engineering quantity is calculated by the formula: engineering quantity = multiple * technical parameter (or engineering quantity). The example table of the algorithm of the standard engineering quantity multiple formula is shown in Table 8: Table 8. Example table of standard engineering quantity multiples formula algorithm

[0042] In this example, in order to explore the relationship between some costs and technical parameters that are difficult to summarize, a self-built function is used through regression analysis to further explore the functional relationship between data. Taking overhead line engineering as an example, combined with expert experience, if we need to study the relationship between the main body cost and the length of a single circuit and double circuit, we can first visualize the data. The obtained scatter plot of the length of a single circuit and double circuit and the main body cost is as follows: Figure 2 As shown by Figure 2 It can be seen that there is a linear relationship between these three indicators. Therefore, if the engineering cost of the 110kV overhead line in the above sample is set as the dependent variable y, the single circuit length is set as the independent variable x1, and the double circuit length is set as the independent variable x2, further analysis is performed through regression analysis method, and the function expression between the three is y=126.01+69.75x1+103.94x2. To verify the fitting effect of the model, the data and model are visualized, as shown in Figure 3As shown, from the fitting effect, it can be seen that this function can accurately reflect the correlation between various data. By merging data through the weight calculation rule and the screening rule for similar projects, the complexity of the model is further reduced. By obtaining the corresponding relationship, it is convenient to obtain the cost data based on the basic engineering data, further improving the efficiency of cost analysis. Through regression analysis, the data without corresponding relationships in the data is deeply mined, and then relevant functions are constructed. Based on the corresponding relationship and relevant functions, a project cost analysis model is obtained, further improving the accuracy and analysis efficiency of the project cost analysis model.

[0043] It also includes: Set corresponding accounting algorithms for the engineering quantities of different classifications based on the equipment types used in the project, and formulate an abnormal warning mechanism according to the results of the accounting algorithms.

[0044] For example, for engineering quantities related to technical solutions such as the number of main transformers and the number of lightning arresters, direct verification algorithms are adopted. For those that cannot be directly judged by the technical solution, standard verification algorithms need to be formulated by combining standards such as pre-specifications and quotas, or an influence factor verification algorithm is constructed based on the confidence interval measured from historical projects in combination with the influencing factors of the corresponding engineering quantities. Different levels of abnormal warning mechanisms are formulated for the algorithm results.

[0045] Embodiment 2: This embodiment also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the method for constructing an analysis model of transmission and transformation project cost based on quantization indicators.

[0046] The above specific implementation manners are the preferred implementation manners of the method for constructing an analysis model of transmission and transformation project cost based on quantization indicators of the present invention, and do not limit the specific implementation scope of the present invention by this. The scope of the present invention includes but is not limited to this specific implementation manner. All equivalent changes made according to the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators, characterized in that It includes the following steps: S1: Based on natural language processing, perform semantic recognition on the first historical power transmission and transformation project data to construct a final feature library; S2: Based on a convolutional neural network, extract the correlation between quantization metrics and quantization metrics from the final feature library to establish a first initial cost analysis model; S3: Based on a discriminant clustering algorithm and the AHP method, construct an adaptive refinement model that includes at least a weight calculation rule and a similar project screening rule; S4: Update the first initial cost analysis model based on the adaptive refinement model to obtain a project cost analysis model.

2. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 1, wherein The S1 includes: S11: Define word segmentation rules based on industry terminology habits, and perform partitioning and part-of-speech tagging on the first historical power transmission and transformation project data according to the word segmentation rules to obtain tagged historical power transmission and transformation project data; S12: Perform vector conversion on the tagged historical power transmission and transformation project data based on the word vector model in natural language processing to obtain power transmission and transformation project vectors; S13: Calculate the first semantic similarity between power transmission and transformation project vectors and the second semantic similarity between power transmission and transformation project vectors and the to-be-confirmed feature library, and obtain the final feature library according to the first semantic similarity and the second semantic similarity.

3. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 1, characterized in that The S2 includes: S21: Input the final feature library into a convolutional neural network for convolution and pooling operations to obtain the first correlation between the features in the final feature library and cost calculation, and use the features with the first correlation greater than the preset correlation as quantization metrics; S22: Obtain the second correlation between quantization metrics based on regression analysis, and construct a second initial cost analysis model based on the quantization metrics and the second correlation; S23: Perform information price matching based on the second historical power transmission and transformation project data, and update the second initial cost analysis model according to the information price matching result to obtain the first initial cost analysis model.

4. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 3, characterized in that, The S23 includes: S231: Perform information price matching between the second historical power transmission and transformation project data and the data in the second initial cost analysis model. If there is data in the second initial cost analysis model with an information price matching result greater than or equal to the preset value, then the second initial cost analysis model is the first initial cost analysis model. If not, execute S232; S232: Update the second initial cost analysis model by adding data attributes to the second initial cost analysis model based on the information price matching result to obtain the first initial cost analysis model.

5. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 1, wherein The S3 includes: S31: Construct a hierarchical framework based on project attributes and a discriminant clustering algorithm; S32: Define a weight calculation rule and a similar project screening rule through the hierarchical framework, AHP, and entropy weight method.

6. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 5, characterized in that, In S32, the weight calculation rule includes: Obtain the judgment matrix of the second index in the second layer of the hierarchical framework through expert judgment, and obtain the subjective weight of the second index based on the judgment matrix; Perform standardization processing on the judgment matrix through the entropy weight method to obtain the objective weight of the second index, and obtain the final weight based on the subjective weight and the objective weight; The similar project screening rule includes: Perform preliminary screening based on the first index in the first layer of the hierarchical framework; On the basis of the preliminary screening, obtain the project score based on the final weight and the Pearson correlation coefficient, and screen out similar projects according to the project score.

7. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 6, characterized in that, The expression for obtaining the final weight based on subjective weight and objective weight is In the formula, final is the final weight, subj is the subjective weight, and obj is the objective weight.

8. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 3, characterized in that, The S4 includes: S41: Merging the data in the first initial cost analysis model through the weight calculation rule and the similar project screening rule, and obtaining the corresponding relationship between the data in the first initial cost analysis model; S42: Based on the regression analysis, mining the functional relationship between the data in the first initial cost analysis model to construct a correlation function, and obtaining a project cost analysis model based on the corresponding relationship and the correlation function.

9. The method for constructing a power transmission and transformation project cost analysis model based on quantitative indicators according to claim 1, wherein, It further includes: Setting corresponding accounting algorithms for the engineering quantities of different classifications based on the equipment types used in the project, and formulating an abnormal warning mechanism according to the results of the accounting algorithms.

10. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it realizes the steps of the method for constructing a transmission and transformation project cost analysis model based on quantitative indicators described in any one of claims 1 to 9.

Images