Electric power material price intelligent monitoring method and system based on same product identification

Abstract

The invention provides an electric power material price intelligent monitoring method and system based on same product identification, and relates to the technical field of electric power material management, and the method comprises the steps: obtaining the technical specification data, historical settlement data and actual bidding quotation of to-be-purchased non-standard electric power materials; the method comprises the following steps: encoding technical indexes in technical specification data and historical settlement data to construct a feature vector set, and iteratively screening parameter masks of the technical indexes representing core value attributes from the feature vector set by using a genetic algorithm; performing spatial reconstruction on the feature vector set according to the parameter mask to generate a similarity sequence; based on the similarity sequence, obtaining a theoretical estimation price of the to-be-purchased non-standard electric power materials; and calculating the difference ratio between the theoretical estimated price and the actual bidding quoted price so as to realize intelligent monitoring of the price of the electric power materials, and solving the problem that the historical price of the non-standard materials cannot be compared due to fine adjustment of parameters.

Description

Technical Field

[0001] This application relates to the field of power material management technology, and in particular to a method and system for intelligent monitoring of power material prices based on identical product identification. Background Technology

[0002] Large energy companies procure massive amounts of non-standard equipment, spare parts, and engineering services annually across their core business segments, including coal mining, thermal power generation, and railway transportation. These materials are diverse, have complex technical parameters, and often vary due to project-specific customization requirements. Reasonable management of procurement prices and risk control are crucial for reducing overall supply chain costs, ensuring stable production operations, and improving procurement efficiency within the group. With the deepening of enterprise digital transformation, utilizing intelligent methods to monitor, analyze, and provide early warnings about procurement prices has become a key trend in enhancing procurement compliance, economic efficiency, and strategic insight.

[0003] Existing technologies for monitoring the prices of purchased materials by enterprises typically rely on strict coding and matching of standard material master data, or on comparing prices by manually reviewing historical transaction records of the same type of materials.

[0004] However, in specialized fields such as coal mining, thermal power, and railway transportation, there are a large number of non-standard and customized materials lacking unified standard codes or models. Due to the numerous and frequently adjusted technical parameters, it is difficult to find completely identical reference products in historical procurement databases, rendering traditional price comparison methods based on precise matching ineffective. Therefore, existing technologies struggle to accurately match effective historical samples for non-standard materials across different business segments, leading to technical problems such as low accuracy and efficiency in price monitoring, and delayed risk detection. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for intelligent monitoring of power material prices based on same-product identification, in order to solve the technical problems in the existing technology, such as the difficulty in accurately matching effective historical samples for non-standard materials across business segments, which leads to low accuracy, low efficiency and delayed risk detection in price monitoring.

[0006] Firstly, this application provides a method for intelligent monitoring of electricity material prices based on identical product identification, including:

[0007] Obtain technical specifications, historical settlement data, and actual bid prices for non-standard power materials to be procured. The historical settlement data includes technical indicators and settlement price information from multiple historical settlement records.

[0008] By encoding the technical indicators in the technical specifications data and historical settlement data, a feature vector set is constructed, and a genetic algorithm is used to iteratively filter the parameter mask of the technical indicators representing the core value attributes from the feature vector set. The feature vector set includes the vector of materials to be procured and multiple historical procurement material vectors.

[0009] The feature vector set is spatially reconstructed based on the parameter mask, and the Euclidean distance between the reconstructed material to be procured vector in the reconstructed feature vector set and each reconstructed historical material to be procured vector is calculated to generate a similarity sequence.

[0010] Based on the similarity sequence, the N historical procurement material vectors with the highest similarity are selected as target samples, and the weighted K-nearest neighbor algorithm is used to calculate the weighted average of the settlement unit price of the historical settlement records corresponding to the target samples to obtain the theoretical estimated price of the non-standard power materials to be procured.

[0011] The system calculates the ratio of the theoretically estimated price to the actual bid price, and generates price anomaly information when the ratio exceeds a preset threshold, thereby enabling intelligent monitoring of power material prices.

[0012] Optionally, the weighted K-nearest neighbor algorithm is used to calculate the weighted average of the settlement unit prices of the historical settlement records corresponding to the target sample, thereby obtaining the theoretical estimated price of the non-standard power materials to be procured, including:

[0013] Obtain the transaction timestamp of the historical settlement record corresponding to each target sample;

[0014] The reciprocal of the sum of the Euclidean distance and the preset smoothing factor for each target sample is used as the spatial similarity weight for each target sample.

[0015] Based on the time difference between each transaction timestamp and the current time, the time decay weight of each target sample is calculated using the exponential decay function;

[0016] The spatial similarity weight and time decay weight of each target sample are multiplied to obtain the target weight of each target sample.

[0017] Based on the target weight, the settlement unit price of the historical settlement records corresponding to each target sample is calculated by weighted average to obtain the theoretical estimated price of the non-standard power materials to be purchased.

[0018] Optionally, the method further includes:

[0019] Calculate the standard deviation of the settlement unit price for all target samples, and use the standard deviation of the preset multiple as the price fluctuation tolerance value;

[0020] Based on theoretical estimates of prices and price volatility tolerance values, a dynamic confidence interval is constructed.

[0021] When the actual bid price is less than the lower limit of the dynamic confidence interval or greater than the upper limit of the dynamic confidence interval, price anomaly information is generated to achieve intelligent monitoring of power material prices.

[0022] Optionally, a feature vector set is constructed by encoding the technical indicators in the technical specifications data and historical settlement data, including:

[0023] Extract the text description field with technical indicators from the technical specification data as the first text information, and extract the text description field with technical indicators from each historical settlement record in the historical settlement data as the second text information;

[0024] Using preset word segmentation rules, the first text information and each second text information are segmented into words to obtain a first keyword set and multiple second keyword sets;

[0025] The first keyword set and all second keyword sets are deduplicated and merged to construct a feature index table;

[0026] Based on the index position and occurrence status of each keyword in the feature index table in the first keyword set, construct a vector of materials to be procured;

[0027] Based on the index position and occurrence status of each keyword in the feature index table in each set of secondary keywords, construct multiple historical procurement material vectors;

[0028] The vector of materials to be procured is combined with the vectors of all historical procured materials to obtain a set of feature vectors.

[0029] Optionally, a genetic algorithm is used to iteratively select a parameter mask representing the technical indicators of the core value attributes from the feature vector set, including:

[0030] Using the number of feature dimensions of the feature vector set as the bit length of the binary string, an initial population consisting of multiple binary strings is randomly generated.

[0031] Based on each binary string, the feature vector set is initially reconstructed, and based on the initially reconstructed feature vector set, a preset number of temporary historical samples that are closest to the Euclidean distance of the initially reconstructed vector of materials to be procured are selected.

[0032] The dispersion of the settlement price information set of temporary historical samples is calculated, and the dispersion is used as the fitness index of the corresponding binary string. The fitness index is used to measure whether the technical indicators reflect the core value attributes.

[0033] The initial population is sorted in ascending order according to the fitness index, and the first preset number of binary strings in the sorting result are used as the parent binary strings.

[0034] Perform crossover and mutation operations on the parent binary string to generate the child binary string;

[0035] The initial population is updated based on the binary string of the offspring until the updated population meets the preset iterative convergence condition. The binary string corresponding to the lowest fitness index in the updated population is used as the parameter mask.

[0036] Optionally, crossover and mutation operations are performed on the parent binary string to generate the child binary string, including:

[0037] The parent binary strings are randomly paired in pairs to obtain multiple parent pairing groups;

[0038] Generate a first random integer within the bit length range of the parent binary string, and use the first random integer as the intersection index of each parent pairing group;

[0039] Swap the binary bits after the intersection index of the two parent binary strings in each parent pairing group to obtain the first binary string and the second binary string.

[0040] Generate a second random integer within the bit length range of each first binary string and each second binary string, and use the second random integer as the mutation bit index of each parent pairing group;

[0041] The binary values ​​at the mutated bit indexes of each first binary string and each second binary string are flipped to obtain the child binary strings.

[0042] Optionally, the feature vector set is spatially reconstructed based on the parameter mask, and the Euclidean distance between the reconstructed material to be procured vector and each reconstructed historical material to be procured vector in the reconstructed feature vector set is calculated to generate a similarity sequence, including:

[0043] The binary bits marked as reserved in the parameter mask are used to determine the target location;

[0044] The feature dimensions that are not located at the target position in the feature vector set are removed to obtain the reconstructed feature vector set;

[0045] Calculate the Euclidean distance between the reconstructed material vector to be procured in the reconstructed feature vector set and each reconstructed historical material vector;

[0046] The historical procurement material vectors corresponding to the reconstructed historical material vectors are arranged in ascending order according to Euclidean distance to generate a similarity sequence.

[0047] Secondly, this application provides an intelligent monitoring system for electricity material prices based on identical product identification, comprising:

[0048] The acquisition module is used to acquire the technical specifications, historical settlement data, and actual bid prices of the non-standard power materials to be procured. The historical settlement data includes the technical indicators and settlement price information of multiple historical settlement records.

[0049] The encoding module is used to encode the technical indicators in the technical specifications data and historical settlement data, construct a feature vector set, and use a genetic algorithm to iteratively filter the parameter mask of the technical indicators representing the core value attributes from the feature vector set. The feature vector set includes the vector of the material to be purchased and multiple historical material purchase vectors.

[0050] The reconstruction module is used to spatially reconstruct the feature vector set according to the parameter mask, and calculate the Euclidean distance between the reconstructed material to be purchased vector and each reconstructed historical material to be purchased vector in the reconstructed feature vector set to generate a similarity sequence.

[0051] The calculation module is used to select the N historical procurement material vectors with the highest similarity based on the similarity sequence as target samples, and use the weighted K-nearest neighbor algorithm to calculate the weighted average of the settlement unit price of the historical settlement records corresponding to the target samples to obtain the theoretical estimated price of the non-standard power materials to be procured.

[0052] The calculation module is also used to calculate the difference ratio between the theoretical estimated price and the actual bid price, and to generate price anomaly information when the difference ratio is greater than a preset ratio threshold, so as to realize intelligent monitoring of the price of power materials.

[0053] Thirdly, this application provides an electronic device, comprising:

[0054] Memory, used to store computer programs;

[0055] A processor is used to execute computer programs to implement the steps of an intelligent monitoring method for electricity material prices based on same-item identification, as described in the first aspect above.

[0056] Fourthly, this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the steps of the intelligent monitoring method for power material prices based on same-item identification as described in the first aspect above.

[0057] This application provides an intelligent monitoring method for power material prices based on similarity identification. By collecting comprehensive data on technical indicators and price information of materials from multiple fields such as coal mining, thermal power, and transportation, it solves the problem of insufficient valuation basis caused by differences in business segments and fragmented information. It also addresses the problem of chaotic price comparison dimensions caused by the complex parameter system and mixed primary and secondary features of non-standard materials. Furthermore, it avoids misjudgments of similarity due to differences in non-critical parameters. It overcomes the limitations of traditional precise matching methods based on material codes, which suffer from the inability to value non-standard materials due to missing or changed codes. Finally, it solves the problems of low efficiency, strong subjectivity, and inconsistent standards associated with cross-category price comparisons based on manual experience, thus improving the accuracy, efficiency, and compliance of price monitoring in multi-business contexts.

[0058] Furthermore, a discrete-degree-oriented genetic algorithm screening strategy is adopted. Specifically, a population of randomly generated binary strings is used to initially reconstruct the feature vector set and find the nearest neighbor temporary historical samples of the materials to be purchased. The discreteness of the price sets of these samples is calculated as a fitness index to measure whether the combination of technical indicators reflects the core value attributes. Subsequently, the population is iteratively updated through sorting, crossover, and mutation operations based on the fitness index, and finally, the binary string that minimizes the price discreteness of similar samples is selected as the parameter mask. This improves the accuracy of identification and the reliability of valuation of non-standard materials in complex parameter environments such as coal mines, thermal power plants, and transportation. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 A flowchart illustrating an intelligent monitoring method for power material prices based on identical product identification, provided as an embodiment of this application;

[0061] Figure 2 A flowchart illustrating a method for obtaining a theoretically estimated price, provided as an embodiment of this application;

[0062] Figure 3 A flowchart illustrating a method for obtaining a parameter mask provided in an embodiment of this application;

[0063] Figure 4 A schematic diagram of the structure of an intelligent monitoring system for power material prices based on identical product identification provided in this application embodiment;

[0064] Figure 5This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0065] In the procurement of non-standard materials involving multiple sectors such as coal mining, thermal power, and railway transportation, the lack of unified standard coding or frequent model iterations makes it difficult to find effective historical references using existing technologies that rely on precise matching or manual experience. This leads to low accuracy in price monitoring. This application addresses this issue by encoding the technical specifications and historical settlement data of non-standard materials and constructing a feature vector set. A genetic algorithm is then introduced to iteratively select parameter masks that represent core value attributes, thereby eliminating redundant parameter interference in feature space reconstruction and accurately quantifying the essential similarity between non-standard materials and historical samples. Based on this, a weighted K-nearest neighbor algorithm is used to select the most similar target samples based on Euclidean distance for weighted pricing. By comparing the difference between theoretically estimated prices and actual bid prices, automated and high-precision intelligent early warning of price risks for cross-domain non-standard materials is achieved in the absence of completely consistent references.

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] The core of this application is to provide an intelligent monitoring method for electricity material prices based on identical product identification. A flowchart of one specific implementation method is shown below. Figure 1 As shown, the method includes:

[0068] Step 101: Obtain the technical specifications, historical settlement data, and actual bid prices of the non-standard power materials to be procured. The historical settlement data includes the technical indicators and settlement price information of multiple historical settlement records.

[0069] In this step, non-standard power materials refer to materials that are not listed in the national or industry standard general catalog in the power industry and require customized production according to specific project or equipment requirements. Technical specifications data refers to a set of structured or unstructured data describing the physical properties, performance parameters, and manufacturing requirements of the non-standard power materials. Historical settlement data refers to a set of transaction records for similar materials that have been settled within a certain period in the past. Actual bid price refers to the unit price of the non-standard power materials to be procured submitted by the supplier in the current procurement project. Historical settlement record refers to a single transaction entry in the historical settlement data. Technical indicators refer to the specific parameter items included in the technical specifications data or historical settlement records used to define the performance of the materials. Settlement price information refers to the finally confirmed transaction unit price in the historical settlement records.

[0070] In this embodiment, for obtaining technical specification data, the technical specifications of the non-standard power materials to be procured can be directly extracted from the enterprise's supply chain management platform or engineering design documents, from which key parameter fields such as height, power, and traction force can be parsed. Simultaneously, for obtaining historical settlement data, the enterprise's historical procurement database or financial settlement system can be accessed to retrieve transaction records of similar materials from the past three years, with each record including a complete description of the technical parameters and the final settlement unit price. For obtaining actual bid prices, during the current bidding or inquiry process, the bid documents submitted by each supplier are received from the electronic bidding platform, and the bid prices are extracted.

[0071] For example, in a coal mine equipment procurement scenario, the technical specifications T of a high-mining-height electro-hydraulic controlled coal mining machine to be procured is obtained, which includes [mining height range: 3.5-6.0m, installed power: 2000kW, traction force: 800kN]; at the same time, historical settlement data H is obtained, including multiple historical settlement records, such as record A as {technical specifications: [mining height range: 3.0-5.5m, installed power: 1800kW, traction force: 750kN], settlement price information: P}, record B as {technical specifications: [mining height range: 2.8-5.0m, installed power: 1500kW, traction force: 600kN], settlement price information: Q}, etc.; and the actual bid price R submitted by supplier C based on the coal mining machine is also obtained.

[0072] Step 102: By encoding the technical indicators in the technical specifications data and historical settlement data, a feature vector set is constructed, and a genetic algorithm is used to iteratively filter the parameter mask of the technical indicators representing the core value attributes from the feature vector set. The feature vector set includes the vector of the material to be procured and multiple historical material procured vectors.

[0073] In this step, the genetic algorithm refers to a stochastic search and optimization algorithm that simulates the biological evolution mechanism in nature. It iteratively searches for the optimal or near-optimal solution in the solution space by simulating processes such as natural selection, crossover, and mutation. The core value attribute refers to the combination of key indicators that decisively influence the price of a commodity among numerous technical indicators. The parameter mask is a binary filtering identifier used to indicate which technical indicators belong to the core value attribute.

[0074] Step 10201: Extract the text description field with technical indicators from the technical specification data as the first text information, and extract the text description field with technical indicators from each historical settlement record in the historical settlement data as the second text information.

[0075] In this step, the text description field with technical specifications refers to the unstructured or semi-structured data field in the data source specifically used to store information such as material specifications, models, materials, and process requirements.

[0076] In this embodiment, the extraction of the first text information can be achieved by reading the source file of the technical specification data, locating predefined key fields, such as the technical parameter section in the specification document or the material description field in the purchase requisition, and then merging and extracting the text content from these fields. For the extraction of the second text information, each historical settlement record in the historical settlement data can be traversed, and the corresponding material description field can be located, extracting the text content from each record separately. For example, the first text information of the coal mining machine to be purchased can be extracted as "high-seam electro-hydraulic control coal mining machine, 2000kW, 800kN," and the second text information of two historical records can be extracted as "high-seam electro-hydraulic control coal mining machine, 1800kW" and "medium-thick coal seam coal mining machine, 1500kW."

[0077] Step 10202: Using the preset word segmentation rules, perform word segmentation on the first text information and each second text information respectively to obtain the first keyword set and multiple second keyword sets.

[0078] In this step, the preset word segmentation rule refers to the logical strategy used to divide a continuous text sequence into independent semantic units.

[0079] In this embodiment, a natural language processing algorithm is used to scan the first text information and each second text information. The text information is segmented into independent lexical units according to word segmentation rules, as exemplarily shown in Table 1. Meaningless function words, punctuation marks, and common stop words are removed. Nouns, verbs, or adjectives with practical technical meaning are retained, thereby forming a first keyword set and a second keyword set corresponding to each historical settlement record.

[0080] Table 1: Preset word segmentation rules

[0081]

[0082] For example, after processing using the above word segmentation rules, for the first text information: "high-mining-height electro-hydraulic control coal mining machine, 2000kW, 800kN", the first step is to filter out meaningless characters containing "and" using stop word rules; if none are found, this step is skipped. The numerical features "2000kW" and "800kN" are extracted using regular expression matching rules. Then, industry dictionary rules are used for verification and matching, breaking it down into atomic features: "high-mining-height", "electro-hydraulic control", and "coal mining machine". Finally, the first keyword set is obtained as [high-mining-height, electro-hydraulic control, coal mining machine, 2000kW, 800kN]. Similarly, for the second text information, after the same processing, the second keyword sets are obtained as [high-mining-height, electro-hydraulic control, coal mining machine, 1800kW] and [medium-thick coal seam, coal mining machine, 1500kW].

[0083] Step 10203: Deduplicate and merge the first keyword set and all second keyword sets to construct a feature index table.

[0084] In this step, the feature index table refers to an ordered list or dictionary containing all relevant material and technical feature terms, used to establish a mapping relationship between keywords and vector dimension indexes.

[0085] In this embodiment, the first keyword set and all second keyword sets are aggregated into a temporary set. A deduplication operation is used to remove duplicate keywords, ensuring that each technical feature term appears only once in the set. The deduplicated keyword list is then sorted, for example, by alphabetical order or frequency of occurrence. Each keyword in the list is assigned a unique integer index value, starting from 0 and incrementing until all keywords are covered. The final generated mapping relationship, including all unique keywords and their corresponding index values, is the feature index table, as shown in Table 2.

[0086] Table 2: Feature Index Table

[0087]

[0088] For example, after deduplication and merging, according to the mapping relationship shown in the table above, the system maps the high mining height to the 0th dimension of the vector, the electro-hydraulic control to the 1st dimension, the medium-thick coal seam to the 2nd dimension, 2000kW to the 3rd dimension, the coal mining machine to the 4th dimension, and so on.

[0089] Step 10204: Construct a vector of materials to be procured based on the index position and occurrence status of each keyword in the feature index table in the first keyword set.

[0090] In this step, the index position refers to the sequence number of the keyword in the feature index table, which determines the dimensional coordinates of the keyword in the vector. The occurrence status is a numerical value used to identify whether the keyword exists in the text information or its importance; it can be a Boolean value or a word frequency value. The vector of materials to be procured refers to the mathematical representation of the first text information digitized based on the feature index table.

[0091] In this embodiment, a zero-based vector with a dimension length equal to the length of the feature index table is initialized. Each keyword in the first keyword set is traversed. The index position corresponding to the keyword is found in the feature index table. The value at the corresponding index position in the zero-based vector is modified to reflect the occurrence status of the keyword. For example, if one-hot encoding logic is used, the value at the corresponding position is set to 1. If word frequency statistics logic is used, the value at the corresponding position is set to the number of times the keyword appears in the first keyword set. The vector obtained after traversal is the vector of materials to be procured. For example, based on the index position and occurrence status, here it is set to 1 if it exists. Since the first keyword set is [high mining height, electro-hydraulic control, coal mining machine, 2000kW, 800kN], corresponding to indices 0, 1, 4, and 3 respectively, the vector of materials to be procured, V=[1,1,0,1,1], is constructed.

[0092] Step 10205: Construct multiple historical procurement material vectors based on the index position and occurrence status of each keyword in the feature index table in each second keyword set.

[0093] In this step, the historical procurement material vector refers to the mathematical representation of the second text information of a single historical settlement record after it has been digitized based on the feature index table.

[0094] In this embodiment, the vector construction process is repeated for each set of second keywords. A vector of all zeros with the same length as the feature index table is initialized. Keywords in the current set of second keywords are traversed. The index position of each keyword is determined according to the feature index table. The value at the corresponding position in the vector is updated to reflect its occurrence. This operation is repeated for all historical settlement records, ultimately generating multiple historical procurement material vectors equal to the number of historical settlement records.

[0095] For example, for the second keyword set [high mining height, electro-hydraulic control, coal mining machine, 1800kW], corresponding to indices 0, 1, 4, a historical procurement material vector U1=[1,1,0,0,1] is constructed; for the second keyword set [medium-thick coal seam, coal mining machine, 1500kW], corresponding to indices 2, 4, a historical procurement material vector U2=[0,0,1,0,1].

[0096] Step 10206: Combine the vector of materials to be procured with all historical material procured vectors to obtain a feature vector set.

[0097] In this embodiment, the constructed vector of materials to be procured is used as the first row of a matrix or a row with a specific identifier. Multiple constructed historical material procurement vectors are then used as subsequent rows of the matrix. These vectors are organized into a unified data structure in memory, ensuring that all vectors have the same dimension length and that each dimension corresponds to the same technical feature meaning, thus obtaining a feature vector set. For example, V, U1, and U2 are finally combined to obtain the feature vector set S.

[0098]

[0099] like Figure 3 As shown, Figure 3 A flowchart illustrating a method for obtaining a parameter mask provided in an embodiment of this application is shown.

[0100] Step 10211: Using the number of feature dimensions of the feature vector set as the bit length of the binary string, randomly generate an initial population including multiple binary strings.

[0101] In this step, the number of feature dimensions refers to the total number of data columns included in each vector in the feature vector set, corresponding to the total number of keywords in the feature index table. A binary string is a sequence of numbers consisting of 0s and 1s, where each bit corresponds to a feature dimension, used to hypothesize whether that dimension belongs to a core value attribute. The initial population refers to a set of binary strings randomly generated when the algorithm starts, representing the first generation of potential solution space.

[0102] In this embodiment, the number of columns in the feature vector set is determined as the length of the binary string. Set population size This refers to the number of individuals that need to be generated. This is achieved using a random number generator. A length of A binary sequence. Each bit of the sequence is randomly assigned a value of 0 or 1 with a preset probability. This... The binary sequences are collected and stored to form the initial population. For example, suppose the feature vector set... If there are 5 feature dimensions, the binary string length is 5 bits; the initial population is randomly generated and includes one binary string. .

[0103] Step 10212: Based on each binary string, perform preliminary reconstruction of the feature vector set, and based on the preliminary reconstruction of the feature vector set, select a preset number of temporary historical samples that are closest to the Euclidean distance of the preliminary reconstruction of the material to be procured vector.

[0104] In this step, the temporary historical sample refers to the historical material record that is geometrically closest to the material to be purchased in the temporary feature space defined by the current binary string, and is only used to verify the validity of the current feature combination.

[0105] In this embodiment, each binary string in the initial population is traversed. For the current binary string, the index position where the value is 1 is identified. Only the column data corresponding to these index positions in the feature vector set is retained to generate a temporary feature vector set. The Euclidean distance between the vector of the material to be procured in the temporary feature vector set and each historical material to be procured vector is calculated. All historical material to be procured vectors are sorted in ascending order according to this distance. The top-ranked vectors are then truncated. Each historical procurement material vector is used to determine the corresponding historical settlement record as a temporary historical sample.

[0106] For example, suppose the vector of materials to be procured Historical procurement material vector , . use The above vectors are initially reconstructed, retaining only the 0th, 1st, and 4th dimensions, resulting in the reconstructed vector. , , Calculating the Euclidean distance revealed... distance Recently, the corresponding historical records were selected as temporary historical samples.

[0107] Step 10213: Calculate the dispersion of the settlement price information set of the temporary historical sample, and use the dispersion as the fitness index of the corresponding binary string. The fitness index is used to measure whether the technical indicator reflects the core value attribute.

[0108] In this step, the settlement price information set refers to the list of transaction prices corresponding to the multiple temporary historical samples selected above. Dispersion is a statistical measure reflecting the degree of fluctuation in the prices, usually expressed as variance or standard deviation. Fitness index is a standard used in genetic algorithms to evaluate the quality of individuals. Here, lower dispersion means that the feature combinations selected by the binary string can cluster goods with similar prices together, thus better reflecting the core value attributes.

[0109] In this embodiment of the application, extraction Settlement price information corresponding to each temporary historical sample forms a price set. The variance or standard deviation of the set is calculated using statistical formulas. The calculated variance or standard deviation value is then directly assigned as the fitness index of the current binary string. If the price set is empty or the sample size is insufficient, a very large penalty value is assigned. This calculation process is repeated for each binary string in the population.

[0110] For example, extracting temporary historical samples The settlement price is 100,000 yuan. The settlement price is 102,000 yuan; calculate the variance or standard deviation of these two prices to obtain a small dispersion. take it as Fitness index.

[0111] Step 10214: Sort the initial population in ascending order according to the fitness index, and take the first preset number of binary strings in the sorting result as the parent binary strings.

[0112] In this step, the parent binary string refers to the high-performing binary string individuals selected from the current population that will be used to reproduce the next generation.

[0113] In this embodiment, the fitness index corresponding to each binary string in the initial population is read. All binary strings are sorted in ascending order of fitness index value, and a predetermined number of binary strings at the top of the sorted list are selected. These selected binary strings are marked as parent binary strings. For example, the top 50% of individuals with the lowest fitness index in the population are selected as parent binary strings, including those with excellent performance. .

[0114] Step 10215: Perform crossover and mutation operations on the parent binary string to generate the child binary string.

[0115] In this step, the offspring binary string refers to the binary string newly generated after the genetic operation that will be evaluated in the next generation.

[0116] Step 10221: Randomly pair the parent binary strings in pairs to obtain multiple parent pairing groups.

[0117] In this embodiment, the parent binary strings are shuffled using a random sampling algorithm or a shuffling algorithm. Two binary strings are selected sequentially as a combination unit. This process is repeated until all parent binary strings have been assigned, thus forming multiple parent pairing groups. If the number of parent strings is odd, one can be randomly discarded or a particular parent string can participate in two pairings.

[0118] For example, in a power grid project procurement scenario in region A, the feature vector set has a dimension of 5. After screening, the two parent binary strings that enter the reproduction stage are respectively... and ,Will and Form a parent pairing group.

[0119] Step 10222: Generate a first random integer within the bit length range of the parent binary string, and use the first random integer as the intersection index of each parent pairing group.

[0120] In this step, the first random integer refers to an integer value that is dynamically generated during algorithm execution and falls within the valid index range. The intersection index refers to the specific coordinates used to indicate the location of the binary string during splitting and swapping operations.

[0121] In this embodiment of the application, the bit length of the parent binary string is obtained. This length is equal to the number of feature dimensions in the feature vector set. Using a pseudo-random number generator, within the interval... Generate an integer. Use this integer as the crossover index for the current parent pairing group; this index determines the boundary between parent gene chain breaks and recombination. For example, randomly generate a first random integer 2 between 1 and 4 as the crossover index.

[0122] Step 10223: Swap the binary bits after the intersection index of the two parent binary strings in each parent pairing group to obtain the first binary string and the second binary string.

[0123] In this embodiment, for each parent pairing group, two parent binary strings are identified. Based on the determined crossover index, the two parent binary strings are respectively divided into head segments and tail segments. The head segment of the first parent is concatenated with the tail segment of the second parent to form a first binary string. Simultaneously, the head segment of the second parent is concatenated with the tail segment of the first parent to form a second binary string. This process simulates chromosome crossing over in biological evolution.

[0124] For example, exchange and The segment after index 1, i.e., the second position and subsequent segments. head and tail The first binary string is obtained by concatenation. , head and tail The second binary string is obtained by concatenation. .

[0125] Step 10224: Generate a second random integer within the bit length range of each first binary string and each second binary string, and use the second random integer as the mutation bit index of each parent pairing group.

[0126] In this step, the mutated bit index refers to the specific dimensional coordinates in the binary string where the value is about to be flipped.

[0127] In this embodiment of the application, a random number generator is used again based on each generated first binary string and second binary string. Within the interval... Generate an integer. This integer is then designated as the mutation index. Typically, to maintain algorithm stability, this step involves a very small mutation probability parameter to determine whether to actually generate the mutation index. If not triggered, subsequent flipping operations are skipped. Here, for the sake of describing the complete process, we assume mutation has occurred.

[0128] For example, according to Generate a second random integer 4 as the mutation bit index; according to Generate a second random integer 0 as the index of the mutated bit.

[0129] Step 10225: Reverse the binary value at the mutated bit index in each first binary string and each second binary string to obtain the child binary string.

[0130] In this embodiment, the bit corresponding to the mutation index in the first and second binary strings is located. The current value at that position is read. If the current value is 0, it is changed to 1; if the current value is 1, it is changed to 0. The binary sequence after the flip is the final generated child binary string, which carries some feature information from the parent and introduces new random perturbations, helping to escape local optima.

[0131] For example, for Flipping the 4th bit yields the child binary string. ;right Flip the 0th bit to obtain the child binary string. .

[0132] Step 10216: Update the initial population based on the binary string of the offspring until the updated population meets the preset iterative convergence condition, and use the binary string corresponding to the lowest fitness index in the updated population as the parameter mask.

[0133] In this step, the preset iterative convergence condition refers to the criterion for stopping the algorithm, which can be reaching the maximum number of iterations or the fitness index no longer decreasing significantly. The parameter mask refers to the optimal binary string output by the algorithm, which represents the optimal combination of core value attributes.

[0134] In this embodiment, individuals with lower fitness rankings in the initial population are replaced with the generated offspring binary strings, or a new generation population is directly constructed. The system checks whether the current iteration count has reached a preset upper limit, or whether the optimal fitness value has remained stable over several generations. If the conditions are not met, the loop continues until the conditions are met. If the conditions are met, the iteration stops, the current population is scanned, and the binary string with the smallest fitness index value is found. This binary string is then output as a parameter mask.

[0135] For example, after the algorithm converges iteratively, it outputs the binary string with the lowest fitness. As a parameter mask, it indicates that high mining height, electro-hydraulic control, and coal mining machine are the core value attributes of this type of material.

[0136] Step 103: Reconstruct the feature vector set spatially based on the parameter mask, and calculate the Euclidean distance between the reconstructed material to be procured vector in the reconstructed feature vector set and each reconstructed historical material to be procured vector, and generate a similarity sequence.

[0137] In this step, the reconstructed feature vector set refers to the data set composed of vectors after spatial reconstruction processing, which are then reduced in dimensionality. The similarity sequence is an ordered list of all reconstructed historical procurement material vectors sorted based on Euclidean distance, used to intuitively reflect the similarity between historical materials and materials to be procured at the core value level.

[0138] Step 10301: Determine the target location from the binary bits marked as reserved in the parameter mask.

[0139] In this step, the "retained state" refers to a state in the binary string where the value is 1, indicating that the technical indicator corresponding to that position belongs to the core value attribute. The "target position" refers to the set of index coordinates of all bits with a value of 1 in the parameter mask, indicating the data column in the feature vector set that needs to be retained.

[0140] In this embodiment of the application, the parameter mask is read. Iterate through each bit of the binary string. Check if the current bit is 1. If it is 1, record the index of the current bit. All recorded index values ​​are compiled into a list, which represents the target location. For example, if the parameter mask... Then the target location is determined as .

[0141] Step 10302: Remove the feature dimensions that are not located at the target position from the feature vector set to obtain the reconstructed feature vector set.

[0142] In this step, the feature dimension refers to the data columns in the feature vector set, with each column corresponding to a specific technical indicator.

[0143] In this embodiment of the application, the original feature vector set is loaded. Based on the determined list of target locations, the corresponding column vectors in the matrix are identified. These column vectors are extracted and recombined into a new matrix in their original order. Columns in the original matrix not selected by the target locations (i.e., dimensions with a mask of 0) are discarded. This new matrix is ​​the reconstructed feature vector set. Its number of rows is the same as the original matrix, but the number of columns is reduced, and it only includes core value attribute data. For example, for the original feature vector set... In the middle, the vector of materials to be procured Historical vector Its matrix form is:

[0144]

[0145] According to the target location After removing the second and third dimensions of data, the reconstructed feature vector set is obtained. Its matrix form is:

[0146]

[0147] Among them, the reconstructed feature vector set The first row corresponds to the reconstructed vector of materials to be procured. The second row corresponds to the reconstructed historical vector. The third row corresponds to the reconstructed historical vector. .

[0148] Step 10303: Calculate the Euclidean distance between the reconstructed material vector to be purchased and each reconstructed historical material vector in the reconstructed feature vector set.

[0149] In this step, Euclidean distance refers to the straight-line distance between two points in a multidimensional space, used to quantify the degree of difference between two materials in their core value attributes.

[0150] In this embodiment of the application, from the reconstructed feature vector set Extract the first row as the vector for reconstructing the materials to be procured. And subsequent rows serve as vectors for reconstructing historical materials. For each Calculate its relationship with formula (1) Distance:

[0151] (1)

[0152] in, To reconstruct the dimension of the vector, and The vectors at the th The value in the dimension. Repeat this calculation until the value is obtained. The distance value to all reconstructed historical material vectors.

[0153] For example, calculating the Euclidean distance: =0; It is 1.414.

[0154] Step 10304: Arrange the historical procurement material vectors corresponding to the reconstructed historical material vectors in ascending order of Euclidean distance to generate a similarity sequence.

[0155] In this embodiment, the calculated Euclidean distance is used to establish key-value pair associations with the corresponding historical procurement material vectors. Using the Euclidean distance values ​​as the sorting key, quicksort or mergesort algorithms are employed to sort these key-value pairs in ascending order. The smaller the distance, the earlier it appears in the list, indicating a greater similarity between the historical material and the material to be procured. After sorting, the sorted list of historical procurement material vectors is extracted, which constitutes the similarity sequence.

[0156] For example, the final similarity sequence is generated by sorting the results in ascending order of distance: The distance is 0. The distance is 1.414.

[0157] Step 104: Based on the similarity sequence, select the N historical procurement material vectors with the highest similarity as target samples, and use the weighted K-nearest neighbor algorithm to calculate the weighted average of the settlement unit price of the historical settlement records corresponding to the target samples to obtain the theoretical estimated price of the non-standard power materials to be procured.

[0158] In this step, the target sample refers to a limited number of historical material records selected from massive historical data that are closest to the materials to be procured in terms of core value attributes. The weighted k-nearest neighbor algorithm is an instance-based learning method that, in regression prediction, calculates predicted values ​​by assigning greater weight to closer neighbors. The theoretical estimated price refers to a reasonable reference price for the materials to be procured in the current market environment, derived based on the historical transaction data and similarity weights of the target samples.

[0159] In this embodiment, the top N historical procurement material vectors in the similarity sequence are selected. These N vectors and their corresponding original historical settlement records are determined as target samples. Next, the regression logic of the weighted K-nearest neighbor algorithm is invoked. For each target sample, its Euclidean distance to the material to be procured in the reconstruction space is obtained. Weights are calculated based on this distance, typically using an inverse distance weighting strategy, where smaller distances result in larger weights.

[0160] To prevent division by zero due to zero distance, a small smoothing factor can be added to the denominator. After calculating the weight of each target sample, the final settlement unit price recorded in the historical settlement data for each target sample is extracted. Finally, using the weighted average formula, the settlement unit prices of all target samples are multiplied by their corresponding weights and summed, then divided by the sum of all weights to obtain the theoretical estimated price of the non-standard power materials to be procured.

[0161] like Figure 2 As shown, Figure 2 The illustration shows a flowchart of a method for obtaining a theoretically estimated price according to an embodiment of this application.

[0162] Step 10401: Obtain the transaction timestamp of the historical settlement record corresponding to each target sample;

[0163] In this step, the transaction timestamp refers to the specific date and time information recorded in the historical settlement records for the completion of the transaction, the signing of the contract, or the financial settlement of the goods.

[0164] In this embodiment, N target samples are traversed. Based on the historical procurement material vector index of each target sample, the corresponding historical settlement record is located in the original historical settlement data. Time information is extracted from specific fields of this record, such as settlement date and contract signing date. The extracted time information is uniformly converted into a standard timestamp format and a correspondence is established with the target sample. For example, the target sample... Two months ago, It was 12 months ago.

[0165] Step 10402: The reciprocal of the sum of the Euclidean distance and the preset smoothing factor for each target sample is used as the spatial similarity weight for each target sample.

[0166] In this step, the preset smoothing factor is a very small positive number used to avoid division by zero errors when the Euclidean distance is zero.

[0167] In this embodiment, the Euclidean distance between each target sample and the reconstructed vector of materials to be procured is calculated. A smoothing factor is set. The Euclidean distance of each target sample is added to the smoothing factor and then the reciprocal is taken to obtain the spatial similarity weight corresponding to each sample. For example, if the smoothing factor is set to 0.1, then... Spatial weights are , The spatial weights are all .

[0168] Step 10403: Based on the time difference between each transaction timestamp and the current time, calculate the time decay weight of each target sample using the exponential decay function.

[0169] In this embodiment, the current system time is obtained. The difference between the transaction timestamp of each target sample and the current time is calculated. A time decay coefficient is set. The time decay weight is calculated using the exponential decay formula. As shown in formula (2):

[0170] (2)

[0171] in, The time decay coefficient, This is the time difference. is the base of the natural logarithm.

[0172] Step 10404: Multiply the spatial similarity weight and time decay weight of each target sample to obtain the target weight of each target sample.

[0173] In this embodiment of the application, for each target sample, the calculated spatial similarity weight is multiplied by the time decay weight to obtain the comprehensive target weight of the sample.

[0174] Step 10405: Based on the target weight, calculate the weighted average of the settlement unit price of the historical settlement records corresponding to each target sample to obtain the theoretical estimated price of the non-standard power materials to be purchased.

[0175] In this embodiment of the application, the historical settlement unit price corresponding to each target sample is extracted. The weighted average formula is used to calculate the theoretical price estimate. As shown in formula (3):

[0176] (3)

[0177] in, For the first The historical settlement unit price of each target sample For the first The target weights of each target sample This represents the number of target samples. For example, suppose... The unit price is 2 million. If the unit price is 1.8 million, then It amounted to 1.9876 million yuan.

[0178] Step 105: Calculate the difference ratio between the theoretical estimated price and the actual bid price, and generate price anomaly information when the difference ratio is greater than the preset ratio threshold, so as to realize intelligent monitoring of the price of power materials.

[0179] In this step, the preset ratio threshold refers to the maximum percentage limit for allowed price fluctuations set according to the company's procurement management regulations or historical statistical patterns. Price anomaly information refers to the alarm record triggered when the price deviation exceeds the safe range, which may include the abnormal material code, deviation range, reference estimate, and corresponding historical similar cases.

[0180] In this embodiment, the theoretical estimated price and the actual bid price are read. The absolute value of the difference between the two is calculated using a formula, and this difference is divided by the theoretical estimated price to obtain the difference ratio.

[0181] Next, the calculated difference ratio is compared with a preset threshold ratio. If the difference ratio is greater than the preset threshold, the current quote is deemed to have an abnormal risk. At this point, a price anomaly information message is generated, including information on the materials to be purchased, the actual quote, the theoretical estimate, and the difference ratio. This message is then pushed to the monitoring dashboard or sent to the procurement auditors via email for manual review or to initiate the price negotiation process. If the difference ratio is less than or equal to the threshold, the quote is deemed to be within a reasonable range, and the process ends normally or proceeds to the next stage.

[0182] For example, in a coal mine equipment procurement scenario, the theoretical estimated price of the non-standard coal mining machine to be purchased. Calculated as The quoted price was RMB 10,000, while the actual bid price R submitted by supplier C was RMB 2.4 million. First, the difference ratio was calculated to be 20.7%. The preset ratio threshold was set at 15%. It was determined that the bid price carried a risk of overpricing. Subsequently, price anomaly information was generated: The quoted price for the material [high-extraction electro-hydraulic controlled coal mining machine] was RMB 2.4 million, the theoretical estimated price was RMB 1.9876 million, the deviation was 20.7%, and verification was recommended.

[0183] Step 10501: Calculate the standard deviation of the settlement unit price of all target samples, and use the standard deviation of the preset multiple as the price fluctuation tolerance value.

[0184] In this step, the preset multiple refers to the confidence coefficient set based on statistical principles, usually with a value of 2 or 3, corresponding to a confidence level of approximately 95% or 99.7%, respectively.

[0185] In this embodiment, a set of historical settlement unit price data for N target samples is extracted. The standard deviation of this set is calculated using statistical formulas. Read the preset multiplier. Multiplying the standard deviation by a preset factor yields the price volatility tolerance value. This tolerance value reflects the reasonable fluctuation range of commodity prices at the current similarity level. If the price difference between target samples is large and the standard deviation is large, the tolerance value will increase accordingly, and vice versa, thus achieving dynamic adaptation of the monitoring threshold.

[0186] For example, if the unit prices of the three selected target samples are 2 million and 1.8 million respectively, calculate their standard deviations. The amount is 100,000 yuan; a preset multiplier is set. If the value is 1, then the price fluctuation tolerance value is... It is 100,000 yuan.

[0187] Step 10502: Construct a dynamic confidence interval based on theoretically estimated prices and price fluctuation tolerance values.

[0188] In this step, the dynamic confidence interval refers to the reasonable price range calculated based on theoretical valuation and historical volatility, including an upper limit and a lower limit.

[0189] In this embodiment, the lower limit of the confidence interval is obtained by subtracting the price fluctuation tolerance value from the theoretically estimated price. The upper limit of the confidence interval is obtained by adding the theoretically estimated price to the price fluctuation tolerance value. The closed interval formed by these two boundary values. This is called a dynamic confidence interval. The interval is no longer a fixed percentage, but rather tailored to the historical data distribution characteristics of each commodity. For example, given the theoretically estimated price... If the value is 10,000 yuan, then the lower limit of the confidence interval is... 10,000 yuan, maximum Ten thousand yuan.

[0190] Step 10503: If the actual bid price is less than the lower limit of the dynamic confidence interval or greater than the upper limit of the dynamic confidence interval, generate price anomaly information to achieve intelligent monitoring of power material prices.

[0191] In this embodiment of the application, the actual bid price submitted by the supplier is obtained. . judge Does it meet the conditions? or If any condition is met, i.e., the quoted price falls outside the confidence interval, it is considered an anomaly. In this case, an operation to generate anomaly information is triggered, recording the anomaly type (too low or too high), the deviation magnitude, and the confidence interval range, and outputting this information to relevant personnel for alert. If the quoted price is within the range, it is considered normal fluctuation. For example, the actual bid price... 10,000 yuan, meaning the quote is higher than the upper limit of the confidence interval. Therefore, an alert message is generated indicating that the price is abnormally high, prompting purchasing personnel that the quoted price exceeds the reasonable fluctuation range calculated based on historical data distribution.

[0192] This application's embodiments address the problem of insufficient valuation basis caused by differences in business segments and fragmented information by collecting comprehensive data on technical indicators and price information of materials from multiple fields such as coal mining, thermal power, and transportation. It also resolves the issue of chaotic price comparison dimensions caused by the complex parameter systems and mixed primary and secondary characteristics of non-standard materials. Furthermore, it avoids misjudgments of similarity due to differences in non-critical parameters. It overcomes the limitations of traditional precise matching models based on material codes, where the lack of samples due to missing or changed codes for non-standard materials makes valuation impossible. Finally, it solves the problems of low efficiency, strong subjectivity, and inconsistent standards associated with cross-category price comparisons based on manual experience, improving the accuracy, efficiency, and compliance of price monitoring in multi-business contexts.

[0193] Figure 4 This is a schematic diagram illustrating a specific implementation of an intelligent monitoring system for power material prices based on identical product identification, provided in this application. (Refer to...) Figure 4 The system may include:

[0194] Module 21 is used to acquire the technical specifications, historical settlement data and actual bid price of the non-standard power materials to be purchased. The historical settlement data includes the technical indicators and settlement price information of multiple historical settlement records.

[0195] The encoding module 22 is used to encode the technical indicators in the technical specification data and historical settlement data, construct a feature vector set, and use a genetic algorithm to iteratively filter the parameter mask of the technical indicators representing the core value attributes from the feature vector set. The feature vector set includes the vector of the material to be purchased and multiple historical material vectors.

[0196] The reconstruction module 23 is used to spatially reconstruct the feature vector set according to the parameter mask, and calculate the Euclidean distance between the reconstructed material to be purchased vector in the reconstructed feature vector set and each reconstructed historical material to be purchased vector, and generate a similarity sequence.

[0197] The calculation module 24 is used to select the N historical procurement material vectors with the highest similarity based on the similarity sequence as target samples, and use the weighted K-nearest neighbor algorithm to calculate the weighted average of the settlement unit price of the historical settlement records corresponding to the target samples to obtain the theoretical estimated price of the non-standard power materials to be procured.

[0198] The calculation module 24 is also used to calculate the difference ratio between the theoretical estimated price and the actual bid price, and to generate price anomaly information when the difference ratio is greater than a preset ratio threshold, so as to realize intelligent monitoring of the price of power materials.

[0199] This application provides an intelligent monitoring system for power material prices based on identical product identification, which is used to implement the aforementioned intelligent monitoring method for power material prices based on identical product identification. Therefore, the specific implementation of the intelligent monitoring system for power material prices based on identical product identification can be found in the embodiment section of the intelligent monitoring method for power material prices based on identical product identification mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0200] Figure 5 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0201] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the above-described intelligent monitoring method for power material prices based on same-item identification.

[0202] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0203] Specifically, the processor 510 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0204] Memory 520 may include mass storage for data or instructions. For example, and not limitingly, memory 520 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 520 may include removable or non-removable (or fixed) media. Where appropriate, memory 520 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 520 is non-volatile solid-state memory.

[0205] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to the first aspect of this disclosure.

[0206] The processor 510 reads and executes computer program instructions stored in the memory 520 to implement any of the intelligent monitoring methods for power material prices based on same-product identification in the above embodiments.

[0207] In one example, the electronic device may also include a communication interface 530 and a bus 540. Wherein, such as Figure 5 As shown, the processor 510, memory 520, and communication interface 530 are connected through bus 540 and complete communication with each other.

[0208] The communication interface 530 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0209] Bus 540 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 540 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0210] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described intelligent monitoring methods for power material prices based on identical product identification.

[0211] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0212] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the intelligent monitoring method for power material prices based on same-item identification.

[0213] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0214] The above provides a detailed description of the intelligent monitoring method and system for power material prices based on identical product identification provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for intelligent monitoring of electricity material prices based on identical product identification, characterized in that, include: Obtain the technical specifications, historical settlement data, and actual bid prices of the non-standard power materials to be procured. The historical settlement data includes the technical indicators and settlement price information of multiple historical settlement records. By encoding the technical specifications data and the technical indicators in the historical settlement data, a feature vector set is constructed, and a genetic algorithm is used to iteratively filter the parameter mask of the technical indicators representing the core value attributes from the feature vector set. The feature vector set includes a vector of materials to be procured and multiple vectors of historical procured materials. Based on the parameter mask, the feature vector set is spatially reconstructed, and the Euclidean distance between the reconstructed material to be procured vector in the reconstructed feature vector set and each reconstructed historical material to be procured vector is calculated to generate a similarity sequence. Based on the similarity sequence, the N historical procurement material vectors with the highest similarity are selected as target samples, and the weighted K-nearest neighbor algorithm is used to calculate the weighted average of the settlement unit price of the historical settlement records corresponding to the target samples to obtain the theoretical estimated price of the non-standard power materials to be procured. The difference ratio between the theoretical estimated price and the actual bid price is calculated, and price anomaly information is generated when the difference ratio is greater than a preset ratio threshold, so as to realize intelligent monitoring of the price of power materials.

2. The method according to claim 1, characterized in that, The weighted average settlement unit price of the historical settlement records corresponding to the target sample is calculated using the weighted K-nearest neighbor algorithm to obtain the theoretical estimated price of the non-standard power materials to be procured, including: Obtain the transaction timestamp of the historical settlement record corresponding to each target sample; The reciprocal of the sum of the Euclidean distance and the preset smoothing factor for each target sample is used as the spatial similarity weight for each target sample. Based on the time difference between each transaction timestamp and the current time, the time decay weight of each target sample is calculated using the exponential decay function; The spatial similarity weight and time decay weight of each target sample are multiplied to obtain the target weight of each target sample. Based on the target weight, the settlement unit price of the historical settlement records corresponding to each target sample is calculated by weighted average to obtain the theoretical estimated price of the non-standard power materials to be purchased.

3. The method according to claim 1, characterized in that, The method further includes: Calculate the standard deviation of the settlement unit price for all target samples, and use the standard deviation at a preset multiple as the price fluctuation tolerance value; Based on the theoretically estimated price and the price fluctuation tolerance value, a dynamic confidence interval is constructed; If the actual bid price is less than the lower limit of the dynamic confidence interval or greater than the upper limit of the dynamic confidence interval, price anomaly information is generated to achieve intelligent monitoring of power material prices.

4. The method according to claim 1, characterized in that, By encoding the technical specifications data and the technical indicators in the historical settlement data, a feature vector set is constructed, including: The text description field containing technical indicators in the technical specification data is extracted as the first text information, and the text description field containing technical indicators in each historical settlement record in the historical settlement data is extracted as the second text information. Using preset word segmentation rules, the first text information and each second text information are segmented into words to obtain a first keyword set and multiple second keyword sets; The first keyword set and all second keyword sets are deduplicated and merged to construct a feature index table; Based on the index position and occurrence status of each keyword in the first keyword set in the feature index table, construct a vector of materials to be procured; Based on the index position and occurrence status of each keyword in each second keyword set in the feature index table, construct multiple historical procurement material vectors; The vector of materials to be procured and all historical procurement vectors are combined to obtain a feature vector set.

5. The method according to claim 1, characterized in that, A genetic algorithm is used to iteratively select parameter masks representing technical indicators of core value attributes from the feature vector set, including: Using the number of feature dimensions of the feature vector set as the bit length of the binary string, an initial population consisting of multiple binary strings is randomly generated. Based on each binary string, the feature vector set is initially reconstructed, and based on the initially reconstructed feature vector set, a preset number of temporary historical samples that are closest to the Euclidean distance of the initially reconstructed vector of the material to be procured are selected. Calculate the dispersion of the settlement price information set of the temporary historical sample, and use the dispersion as the fitness index of the corresponding binary string, wherein the fitness index is used to measure whether the technical indicator reflects the core value attribute. The initial population is sorted in ascending order according to the fitness index, and the first preset number of binary strings in the sorting result are used as the parent binary strings. Perform crossover and mutation operations on the parent binary string to generate the child binary string; The initial population is updated based on the offspring binary string until the updated population meets the preset iterative convergence condition. The binary string corresponding to the lowest fitness index in the updated population is used as the parameter mask.

6. The method according to claim 5, characterized in that, Perform crossover and mutation operations on the parent binary string to generate the child binary string, including: The parent binary strings are randomly paired in pairs to obtain multiple parent pairing groups; A first random integer is generated within the bit length range of the parent binary string, and the first random integer is used as the intersection index of each parent pairing group; Swap the binary bits after the crosspoint index of the two parent binary strings in each parent pairing group to obtain the first binary string and the second binary string. Generate a second random integer within the bit length range of each first binary string and each second binary string, and use the second random integer as the mutation bit index of each parent pairing group; The binary values ​​at the mutation bit index in each first binary string and each second binary string are flipped to obtain the offspring binary string.

7. The method according to claim 1, characterized in that, Based on the parameter mask, the feature vector set is spatially reconstructed, and the Euclidean distance between the reconstructed material to be procured vector and each reconstructed historical material to be procured vector in the reconstructed feature vector set is calculated to generate a similarity sequence, including: The binary bits marked as reserved in the parameter mask are determined as the target position; The feature dimensions that are not located at the target position in the feature vector set are removed to obtain the reconstructed feature vector set; Calculate the Euclidean distance between the reconstructed material vector to be procured in the reconstructed feature vector set and each reconstructed historical material vector; The historical procurement material vectors corresponding to the reconstructed historical material vectors are arranged in ascending order according to the Euclidean distance to generate a similarity sequence.

8. A smart monitoring system for electricity material prices based on identical product identification, characterized in that, include: The acquisition module is used to acquire the technical specifications, historical settlement data and actual bid prices of the non-standard power materials to be procured. The historical settlement data includes the technical indicators and settlement price information of multiple historical settlement records. The encoding module is used to encode the technical indicators in the technical specification data and the historical settlement data to construct a feature vector set, and to use a genetic algorithm to iteratively filter the parameter mask of the technical indicators representing the core value attributes from the feature vector set. The feature vector set includes a vector of materials to be procured and multiple vectors of historical procured materials. The reconstruction module is used to spatially reconstruct the feature vector set according to the parameter mask, and calculate the Euclidean distance between the reconstructed material to be purchased vector in the reconstructed feature vector set and each reconstructed historical material to be purchased vector, and generate a similarity sequence. The calculation module is used to select the N historical procurement material vectors with the highest similarity as target samples based on the similarity sequence, and use the weighted K-nearest neighbor algorithm to calculate the weighted average of the settlement unit price of the historical settlement records corresponding to the target samples to obtain the theoretical estimated price of the non-standard power materials to be procured. The calculation module is also used to calculate the difference ratio between the theoretical estimated price and the actual bid price, and to generate price anomaly information when the difference ratio is greater than a preset ratio threshold, so as to realize intelligent monitoring of the price of power materials.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the intelligent monitoring method for power material prices based on same-item identification as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables a method for intelligent monitoring of power material prices based on identical product identification as described in any one of claims 1 to 7.

Images