Multi-dimensional lean-oriented power grid enterprise purchase contract intelligent management system

Abstract

The invention discloses a power grid enterprise purchase contract intelligent management system for multi-dimensional lean, and the system comprises the steps: achieving the full-process management and control of a purchase contract through multi-module cooperation, firstly obtaining the text data of a power grid purchase contract, calling a power grid contract multi-dimensional lean decision fusion model to carry out the correlation operation of the data, and carrying out the calculation of the data; an operation result is processed by a procurement cost lean optimization model to calculate cost related parameters, and entity relationship analysis is carried out on the cost parameters through a contract element entity relationship extraction algorithm; the analysis result is used for executing the storage, retrieval and updating operation of the purchase contract, monitoring the operation state of each operation link in real time, and dynamically adjusting the operation parameters such as the data transmission rate and the operation priority. The system integrates model algorithms such as multi-dimensional decision fusion, cost optimization and entity relation extraction, collaborative linkage of all links is achieved, the accuracy and efficiency of power grid enterprise purchase contract management are improved, and the multi-dimensional lean management requirement is met.

Description

Technical Field

[0001] This invention relates to the field of enterprise procurement contract management technology, and in particular to an intelligent management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean management. Background Technology

[0002] As the power grid industry continues to expand, the number of procurement contracts has surged, and the parameters involved have become increasingly complex, including key information such as contract amount, performance period, procurement quantity, and supplier level. Traditional manual management models are no longer sufficient to meet the demands for efficient processing. Currently, the procurement contract management process of power grid enterprises needs to simultaneously consider the coordination of multiple stages, such as contract element extraction, cost optimization, decision support, and contract storage and updates. However, data from each stage is often stored separately and processed independently, lacking a unified intelligent management platform to achieve data interoperability and module collaboration. This results in low contract management efficiency and fails to meet the actual needs of multi-dimensional lean management. There is an urgent need to build an intelligent management system that integrates multiple models and algorithms to achieve refined and intelligent control of the entire power grid procurement contract process.

[0003] Existing technologies for power grid enterprise procurement contract management have two significant drawbacks: First, existing systems often employ single models or algorithms to handle specific tasks, such as designing optimization models only for cost calculation or simply extracting contract elements. They fail to achieve a deep integration of multi-dimensional lean decision-making fusion models for power grid contracts, lean optimization models for procurement costs, and algorithms for extracting entity relationships of contract elements. This results in data from different stages not being effectively correlated and processed, making it difficult to form a complete decision support chain and affecting the overall accuracy of procurement contract management. Second, existing systems lack a collaborative control mechanism for various functional modules. Modules such as contract element collection, decision fusion, cost optimization, and intelligent management often operate independently. The data transmission rate and computational priority between modules cannot be dynamically adjusted according to the actual operating status, and there is no real-time monitoring and deviation correction of module operating status. This easily leads to problems such as data transmission delays and computational conflicts, making it impossible to guarantee the stable and efficient operation of the procurement contract management process. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of existing technologies, this invention provides an intelligent management system for power grid enterprise procurement contracts that is oriented towards multi-dimensional lean management.

[0005] The technical solution adopted in this invention is an intelligent management system for power grid enterprise procurement contracts oriented towards multi-dimensional lean management, comprising: a contract element acquisition module, a contract multi-dimensional decision fusion module, a cost lean optimization module, a decision support calculation module, an intelligent contract management module, and a system collaborative control module. The contract element acquisition module acquires contract text data through a power grid procurement contract text scanning interface and transmits the data to the contract multi-dimensional decision fusion module. The contract multi-dimensional decision fusion module calls the power grid contract multi-dimensional lean decision fusion model to perform correlation calculations on the element data, and the calculation results are transmitted to the cost lean optimization module. The cost lean optimization module loads the procurement cost lean optimization model to calculate cost parameters for the fused data, and the calculation results are sent to the decision support calculation module. The decision support calculation module runs a contract element entity relationship extraction algorithm to parse the entity relationships in the cost data, and the parsing results are transmitted to the intelligent contract management module. The intelligent contract management module performs storage, retrieval, and update operations on power grid enterprise procurement contracts based on the parsing results, and the operation instructions are synchronized to the system collaborative control module. The system collaborative control module monitors the operating status of different modules in real time and outputs module collaborative control signals to adjust the data transmission rate and calculation priority of different modules.

[0006] Furthermore, the expression for the power grid contract multidimensional lean decision fusion model in the contract multidimensional decision fusion module is as follows: ,in, For the result of multi-dimensional decision fusion of contracts, Let be the weighting coefficient between the i-th type of contract element and the j-th type of procurement parameter. Let be the correlation coefficient between the i-th type of contract element and the j-th type of procurement parameter. For the i-th type of contract element data, including contract amount, performance period, and warranty period, This is the j-th type of procurement parameter data, including procurement quantity, supplier level, and material unit price. For matrix dot product operation, Let be the decision confidence level for the i-th type of contract element and the j-th type of procurement parameter, and n be the total number of contract element categories. This represents the total number of procurement parameter categories.

[0007] Furthermore, the expression for the procurement cost lean optimization model in the cost lean optimization module is as follows: ,in, To optimize procurement costs, For the purchase quantity of material category k, Let k be the unit cost of the material. Let k be the discount rate for the k-th material. To improve the efficiency of supplying the first supplier, For the first l The duration of cooperation with each supplier, This is the cost adjustment coefficient. Let be the historical procurement cost deviation value for the rth time, p be the total number of material categories, q be the number of suppliers, and r be the number of historical procurements.

[0008] Furthermore, the expression for the contract element entity relationship extraction algorithm in the decision support computing module is as follows: ,in, For the extracted set of entity relationships of contractual elements, The s-th contract element entity includes the name of Party A, the name of Party B, and the contract number. The relationship type between entity u and entity v includes cooperation, payment, and performance relationships. To match the loss function for entity relationships, For entity feature mapping function, This is the relation feature weight matrix. It is the Sigmoid activation function. For relational bias terms, t represents the total number of contract element entities, and v represents the total number of entity relation types.

[0009] Furthermore, the smart contract management module satisfies the following during operation: ,in, A collection of effective operations for intelligent contract management. This is a function for storing the data of the a-th contract. The input is the contract text data, and the output is the storage address index. Let `a` be the function for retrieving the `a`th contract. The input is the search terms, and the output is the contract matching results. Let this be the update operation function for the a-th contract. The input is the data to be updated in the fields, and the output is the update confirmation signal. Let 'a' represent the data for the first power grid procurement contract, and 'b' represent the total number of system management contracts.

[0010] Furthermore, the cooperative control model expression of the system cooperative control module is as follows: ,in, For system coordinated control coefficients, Let c be the data transmission rate of the c-th module. The computation time of the c-th module, As a co-regulatory factor, Let be the deviation value of the operating status of the c-th module, which is the difference between the actual operating parameters and the standard operating parameters of the module. This represents the total number of system modules, which is 6.

[0011] Furthermore, the contract element acquisition module includes a text scanning unit, a data filtering unit, a format conversion unit, and a data verification unit. The text scanning unit scans the paper text of the power grid procurement contract page by page using a high-definition scanning device to generate high-definition image data. After grayscale processing of the image data, the text region contours are extracted. The data filtering unit performs noise filtering on the data within the extracted text region contours, removing light spots, shadows, and irrelevant background pixels from the image, while retaining valid text pixel information. The format conversion unit converts the filtered text pixel information into editable text format data, replacing and correcting any garbled characters that appear during the conversion process. The data verification unit performs integrity verification on the converted text data, comparing whether the contract identification fields in the text data are missing, including the contract number, procurement amount, and signing date, and outputs the verified contract text data.

[0012] Furthermore, the contract multidimensional decision fusion module includes a data receiving unit, a model loading unit, an association operation unit, and a result output unit. The data receiving unit receives contract text data transmitted by the contract element acquisition module, establishes a data cache queue, and sorts the data according to the data reception time. The model loading unit calls the power grid contract multidimensional lean decision fusion model from the system database, loads the weight coefficient matrix and association coefficient table required for model operation, and initializes the model parameters. The association operation unit inputs the sorted contract text data into the initialized model, performs matrix dot product operation and logarithmic operation, and generates intermediate results for multidimensional decision fusion. The result output unit performs data format standardization processing on the intermediate results, removes abnormal values, and transmits the standardized fusion results to the cost lean optimization module.

[0013] Furthermore, the cost lean optimization module includes a parameter import unit, a model running unit, a cost calculation unit, and a result sending unit. The parameter import unit receives the fusion results transmitted by the contract multi-dimensional decision fusion module, extracts the material quantity, unit cost, and supplier level parameters required for procurement cost calculation, and stores them in a temporary parameter database. The model running unit loads the procurement cost lean optimization model, reads the parameter data from the temporary parameter database, and sets the initial value of the cost adjustment coefficient. Based on the loaded model and parameter data, the cost calculation unit performs summation, division, and maximum value calculations to generate optimized procurement cost data and stores intermediate variables in real time during the calculation process. The result sending unit encapsulates the optimized procurement cost data according to a preset data transmission protocol and sends it to the decision support calculation module, while recording the data transmission time and the receiving module identifier.

[0014] A smart management system for power grid enterprise procurement contracts, oriented towards multidimensional lean optimization, comprises the following steps: First, the text scanning unit of the contract element acquisition module scans the power grid procurement contract text. Noise data is removed by the data filtering unit, the text format is converted by the format conversion unit, and the calibration fields are verified by the data verification unit to obtain valid contract text data. Second, the valid contract text data is transmitted to the contract multidimensional decision fusion module. After sorting by the data receiving unit, model initialization by the model loading unit, execution of calculations by the correlation calculation unit, and standardization by the result output unit, the contract multidimensional decision fusion result is obtained. Third, the fusion result is input into the cost lean optimization module. Parameters are extracted by the parameter import unit, the model is loaded by the model running unit, and cost... After the calculation unit performs calculations and the result sending unit encapsulates the data, the optimized procurement cost data is obtained. The fourth step involves transmitting the optimized procurement cost data to the decision support calculation module, where the contract element entity relationship extraction algorithm is run to parse the entity relationships in the cost data, generating entity relationship parsing results. The fifth step involves transmitting the parsing results to the intelligent contract management module, where contract storage operations are performed to save contract data, retrieval operations are performed to query historical contracts, and update operations are performed to modify contract fields, generating contract management operation records. The sixth step involves the system collaborative control module monitoring the operating status of different modules in real time, calculating the data transmission rate and computation time of different modules, adjusting the module collaborative control coefficients, and outputting control signals to maintain stable operation of different modules.

[0015] Beneficial Effects: This invention proposes an intelligent management system for power grid enterprise procurement contracts, oriented towards multi-dimensional lean management. This system deeply integrates a multi-dimensional lean decision-making fusion model for power grid contracts, a lean optimization model for procurement costs, and an algorithm for extracting entity relationships from contract elements. The multi-dimensional decision-making fusion module performs correlation calculations on the collected contract element data, the lean optimization module calculates cost parameters based on the fused data, and the decision support calculation module parses entity relationships, forming a complete decision support chain. This completely overcomes the shortcomings of existing technologies, such as single-model task processing and the inability to correlate data across different stages, thus improving the overall accuracy of procurement contract management. The system's collaborative control module monitors the operating status of each module in real time, dynamically adjusts the data transmission rate and computational priority, and outputs collaborative control signals to avoid data transmission delays and computational conflicts. Simultaneously, the intelligent contract management module achieves integrated operation of contract storage, retrieval, and updating, solving the problems of independent module operation and lack of collaborative control mechanisms in existing technologies. This significantly improves contract management efficiency, meets the multi-dimensional lean management needs of power grid enterprises, and realizes the transformation of procurement contract management from decentralized and manual to centralized and intelligent. Attached Figure Description

[0016] Figure 1 This is a diagram showing the system module composition of the present invention; Figure 2This is a flowchart of the system operation steps of the present invention. Detailed Implementation

[0017] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figure 1 As shown, a smart management system for procurement contracts of power grid enterprises with a focus on multidimensional lean management includes: a contract element acquisition module, a contract multidimensional decision fusion module, a cost lean optimization module, a decision support computing module, an intelligent contract management module, and a system collaborative control module; The contract element acquisition module obtains contract text data through the power grid procurement contract text scanning interface and transmits the data to the contract multi-dimensional decision fusion module. Specifically, the contract element acquisition module integrates a high-definition scanning device with a resolution of no less than 300 DPI to construct a power grid procurement contract text scanning interface. This interface supports automatic recognition and scanning of the two mainstream contract paper sizes, A4 and A3, with a scanning speed set at 2 pages per second to meet the needs of batch contract processing. During implementation, the paper contract is first scanned page by page using the scanning device to generate high-definition image data in RGB color mode. The image resolution is fixed at 300 DPI to ensure text clarity. Then, the image data undergoes grayscale processing, converting the RGB three-color channel values ​​into single-channel grayscale values ​​with weight ratios of 0.299, 0.587, and 0.114 to remove color interference. Next, an edge detection algorithm is used to extract the text region contours, with the contour extraction threshold set to 127 (based on the grayscale value range of 0-255 for 8-bit grayscale images) to ensure accurate segmentation of text and background areas. Finally, a median filtering algorithm is used to filter noise from the data within the text region, with the filtering window size set to 3×3 pixels to effectively remove light spots, shadows, and gray areas. Irrelevant background pixels with a deviation exceeding 20 are retained as valid text pixels. Then, the OCR recognition engine is invoked to convert the filtered text pixels into editable text format data encoded in UTF-8. For garbled characters (characters with a recognition confidence level below 0.8) appearing during the recognition process, they are replaced and corrected according to a common character library for contract text. Finally, a data integrity check is performed, covering key fields such as contract number (18-digit combination format), purchase amount (numerical format retaining 2 decimal places), and signing date (YYYY-MM-DD format). If any field is missing, a rescan instruction is triggered until all key fields of the complete contract text data are output. The time taken to collect a single contract is controlled within 10 seconds.

[0019] The contract multidimensional decision fusion module calls the power grid contract multidimensional lean decision fusion model to perform correlation calculations on the element data, and the calculation results are transmitted to the cost lean optimization module. Specifically, the contract multi-dimensional decision fusion module undertakes the data association and calculation function of contract elements. The hardware is equipped with a quad-core processor with a main frequency of no less than 2.8GHz and a memory configuration of no less than 16GB to ensure computing efficiency. At the software level, the basic parameter library required by the power grid contract multi-dimensional lean decision fusion model is pre-stored. During implementation, the module first receives contract text data transmitted from the contract element acquisition module via an RJ45 Ethernet interface, with a stable data transmission rate of 100Mbps. Simultaneously, a cache queue with a capacity of 100 contract data entries is established, and the data is sorted in ascending order according to the data reception timestamp (accurate to milliseconds) to avoid data processing order confusion. Then, the module retrieves the power grid contract multi-dimensional lean decision fusion model from the system's local database (using a MySQL relational database with a storage capacity of no less than 1TB). It loads the model's associated weight coefficient matrix (12×8 matrix dimensions, where 12 represents the number of contract element categories and 8 represents the number of procurement parameter categories; weight coefficients range from 0.1 to 0.9, determined through training with 3000 historical contract data entries) and the correlation coefficient table (correlation coefficients range from 0 to 1, calculated based on Pearson correlation coefficients). The initial learning rate is set to 0.01, and the number of iterations is set to 100 to ensure convergence. Next, the sorted contract text... Data is categorized and mapped according to contract element categories (including 12 categories such as contract amount, performance period, and warranty period) and procurement parameter categories (including 8 categories such as procurement quantity, supplier level, and material unit price), and input into the initialized model. When the model performs association operations, it first performs a matrix dot product operation on the contract element data and procurement parameter data of the corresponding categories, and then multiplies them by the corresponding weight coefficient and association coefficient. After that, it calculates the logarithm of base 2 (1 + decision confidence level) (the decision confidence level is calculated based on the supplier's historical performance records, with a value range of 0.6-1.0). The results of all categories are summed to obtain the intermediate result of multidimensional decision fusion. Finally, the intermediate result is processed to standardize the data format, normalize the numerical range to the 0-100 range, and remove outliers that exceed this range (the criteria for judging outliers is a deviation from the average value by more than 3 times the standard deviation). The standardized fusion result is transmitted to the cost lean optimization module through the same Ethernet interface. The fusion operation time for a single contract is controlled within 8 seconds.

[0020] The cost lean optimization module loads the procurement cost lean optimization model to calculate cost parameters from the fused data, and sends the calculation results to the decision support calculation module. Specifically, the lean cost optimization module focuses on calculating procurement cost parameters. Its procurement cost lean optimization model has been validated using 2000 sets of historical procurement data, with a calculation error rate of less than 5%. The module hardware is equipped with a dedicated GPU (with at least 6GB of video memory) to accelerate the calculation process. During implementation, the module first receives the fusion results transmitted from the contract multi-dimensional decision fusion module via a USB 3.0 interface, with transmission latency controlled within 500 milliseconds. Then, it initiates a parameter extraction program to extract parameters required for procurement cost calculation from the fusion results, including material quantity (unit: piece, range 10-10000 pieces), unit cost (unit: yuan, range 100-5000 yuan), and supplier level (divided into four levels: A, B, C, and D, corresponding to quantification values ​​of 1.0, 0.8, 0.6, and 0.4, respectively). After extraction, the parameters are stored in a 50GB capacity. A temporary database of parameters is created (stored on an SSD with a read / write speed of at least 500MB / s). Next, the lean optimization model for procurement costs is loaded, reading parameter data from the temporary database. Simultaneously, the initial cost adjustment coefficient is set to 0.05 (this value is based on a 10% fluctuation in procurement costs over the past year; adjustment coefficient = fluctuation range × 0.5). When the model performs cost calculations, it first calculates the product of the procurement quantity and unit cost of each type of material, then multiplies it by (1 - discount rate) (the discount rate ranges from 0 to 0.2, set according to the procurement quantity tiers; for procurement quantities of 1000 units or more, a discount rate is applied). The discount rate is 0.15 for 500-1000 units, 0.1 for 10-500 units, and 0.05 for 10-500 units. The numerator is obtained by summing the results for all categories. Then, the product of each supplier's supply efficiency (range 0.8-1.0, based on on-time delivery rate) and the cooperation period (in years, range 1-10 years) is calculated, and the denominator is obtained by summing the results for all suppliers. The numerator is divided by the denominator, and then the product of the cost adjustment coefficient and the maximum historical procurement cost deviation (historical procurement count is set to 50 times, deviation range from -5000 yuan to 5000 yuan) is added to obtain the final result. The system retrieves optimized procurement cost data; intermediate variables generated during the calculation process (such as intermediate costs of various materials, supplier product results, etc.) are stored in a temporary database in real time, with the storage record updated every 5 seconds; finally, the optimized procurement cost data is encapsulated in JSON format (data fields include cost value, calculation time, parameter source), and sent to the decision support calculation module via TCP / IP protocol, while recording the data sending time (accurate to the second) and the receiving module IP address (fixed to 192.168.1.100), and the calculation time for a single contract cost is controlled within 12 seconds.

[0021] The decision support computing module runs a contract element entity relationship extraction algorithm to parse the entity relationships in the cost data, and the parsing results are transmitted to the smart contract management module. Specifically, the core function of the decision support computing module is to parse the entity relationship of contract elements. The entity relationship extraction algorithm it runs has an entity recognition accuracy of no less than 92% and a relationship extraction accuracy of no less than 88% on the test set (including 1,500 power grid procurement contract texts). The module adopts a distributed computing architecture and supports parallel processing by 8 computing nodes. During implementation, the module receives optimized procurement cost data transmitted from the cost lean optimization module via a fiber optic interface. The data transmission bandwidth is no less than 1Gbps to ensure zero latency in batch data transmission. Then, the algorithm initialization program is started, loading the entity feature mapping function (trained based on the Word2Vec model, with a word vector dimension of 300, and a training corpus of 5000 power grid industry contract texts) and the relation feature weight matrix (matrix dimension 20×300, where 20 represents the number of entity relation types, 300 represents the word vector dimension, and the weight values ​​are optimized using the stochastic gradient descent algorithm). Simultaneously, the temperature parameter of the Sigmoid activation function is set to 1.0, and the relation bias term is initialized to 0. Next, entity annotation is performed on the cost data. The annotated contract element entities include the name of Party A (2-20 characters long, mostly the full name of the company), the name of Party B (format same as Party A's name), and the contract number (18-digit combination). Entity annotation adopts the BIO annotation system (B represents the start of entity). (I represents entities, O represents non-entities); then, entity relationship matching is performed, calculating the partial derivative of the entity feature mapping function output with respect to the relationship feature weight matrix, multiplying it by the relationship bias term processed by the Sigmoid activation function, to obtain the entity relationship matching score; the argmax function selects the relationship type with the highest matching score (including 20 categories such as cooperation relationship, payment relationship, performance relationship, etc.) to determine the relationship between entity pairs; all entities and their corresponding relationships are combined to form an entity relationship set, and redundant relationships in the set (such as duplicate payment relationship records) are deduplicated (the deduplication criterion is that the entity pair and the relationship type are completely consistent); finally, the parsed entity relationship set is formatted and stored in the triple format of "entity 1-relationship-entity 2", and a parsing report (including parsing accuracy, number of entities, and number of relationships) is generated. The entity relationship set and the parsing report are then transmitted to the smart contract management module, and the parsing time for a single contract relationship is controlled within 15 seconds.

[0022] The intelligent contract management module performs storage, retrieval, and update operations on power grid enterprise procurement contracts based on the parsing results, and the operation instructions are synchronized to the system collaborative control module; Specifically, the intelligent contract management module is responsible for the full lifecycle management of power grid enterprise procurement contracts. The database that the module connects to is an Oracle database with a storage capacity of no less than 5TB, supporting more than 100 concurrent accesses per second. It is also equipped with a data backup system that automatically performs a full backup at 2:00 AM every day, and the backup data is retained for 30 days. During implementation, the module first receives the entity relationship parsing results from the decision support computing module. It then verifies the completeness of the parsing results using a data verification program (they must include at least three types of entities: Party A's name, Party B's name, and contract number, along with their corresponding cooperative relationships). If verification fails, a re-parsing instruction is returned. After successful verification, the module performs a contract storage operation, indexing the contract text data, optimized procurement cost data, and entity relationship set by contract number, and storing them in the main contract table of the database (the table structure includes 25 fields such as contract number, Party A's name, Party B's name, procurement cost, and signing date). Sensitive data (such as procurement cost and payment method) is encrypted using the AES-256 encryption algorithm during storage, and the key is updated every 7 days. When a contract retrieval operation is required, the module receives user-input search keywords (supporting searches by contract number, Party A's name, Party B's name, etc.), builds an inverted index using the Lucene search engine, and achieves a retrieval response time of no more than 2 seconds. The returned results are sorted in descending order of matching degree. The system arranges and displays core contract information (contract number, Party A, Party B, and procurement cost). When performing a contract update operation, it receives updated field data submitted by the user (such as changes in performance period or adjustments to procurement quantity). First, it verifies the user's permissions (only system administrators and contract handlers have update permissions). After the permission verification is successful, it performs data format validation on the updated fields (e.g., date format must be YYYY-MM-DD, and values ​​must be positive). If the validation is successful, it performs a database UPDATE operation, generates an operation log after the update (including the updater, update time, and old and new values ​​of the updated fields), and stores the log in an audit table, which cannot be modified. During the module's operation, it monitors the database storage capacity in real time. When the remaining capacity is less than 10% of the total capacity, it triggers a capacity warning and automatically cleans up expired contract data from more than 3 years ago (a backup file is generated before cleaning). At the same time, it defragments active contract data to improve data read and write efficiency. The storage, retrieval, and update operations for a single contract are controlled within 5 seconds, 2 seconds, and 8 seconds, respectively.

[0023] The system collaborative control module monitors the operating status of different modules in real time and outputs collaborative control signals to adjust the data transmission rate and operation priority of different modules.

[0024] Specifically, the system collaborative control module serves as the scheduling core for the operation of each module. It adopts an industrial-grade PLC controller (with a response time of no less than 1ms) and is equipped with a 7-inch touch screen to display the module's operating status in real time. It also supports remote monitoring (accessible via a web interface, supporting Chrome and Firefox browsers).During implementation, the module establishes communication connections with five other modules via an RS485 bus. The communication baud rate is set to 115200bps, with 8 data bits, 1 stop bit, and no parity bit to ensure stable communication. Subsequently, a status monitoring program is started, collecting the operating parameters of each module every 2 seconds, including data transmission rate (in Mbps, with a collection accuracy of 0.1 Mbps), computation time (in seconds, with a collection accuracy of 0.1 seconds), CPU utilization (in %, with a collection range of 0-100%), and memory utilization (in %, with a collection range of 0-100%). The collected actual operating parameters are then compared with the preset standard operating parameters (data transmission rate standard). The standard value is 100Mbps. The standard value for computation time is set according to the modules, such as 10 seconds for the contract element acquisition module, with a standard value of CPU utilization ≤70% and a standard value of memory utilization ≤80%. A comparison is made to calculate the operational status deviation value (deviation value = actual value - standard value; positive deviation indicates exceeding the standard, negative deviation indicates falling below the standard). Based on the deviation value, the system collaborative control coefficient is calculated. First, the ratio of data transmission rate to computation time for each module is calculated (rate / time, in Mbps / second). The ratios of all modules are multiplied to obtain the base coefficient. Then, the collaborative adjustment factor is calculated (range 0.5-2.0, dynamically adjusted according to the overall system load; when the load rate is >80%). 2.0 (0.5 when load rate < 40%) is the product of the absolute value of the deviation value of each module's operating status and the natural exponent (exp(-product)). The base coefficient is multiplied by the exponent result to obtain the system coordination control coefficient (value range 0-10, coefficient > 8 represents excellent coordination status, 5-8 represents good, < 5 represents poor). When the control coefficient < 5, the module outputs a coordination control signal to adjust the operating parameters of each module: for modules with data transmission rates lower than the standard value, their network bandwidth allocation is increased (increased by 10Mbps each time, up to a maximum of 200Mbps); for modules with computation times longer than the standard value, their CPU priority is increased (priority is divided into 1-5 levels). (Default level 3, can be upgraded to level 4-5); For modules with excessive CPU or memory usage, non-critical tasks (such as data backup and log organization) are suspended to prioritize core operations; After the control signal is sent, the module's operating parameters are monitored every second until the control coefficient rises back to above 5; At the same time, the module displays the operating status, deviation value, control coefficient, and other data of each module on the touch screen in real time, and generates an operation report (including status statistics, anomaly records, and control operations) every 5 minutes. The report can be exported as a PDF and supports users to query historical reports (retaining 90 days of report data). The module's own failure rate is controlled within 0.1% to ensure long-term stable scheduling.

[0025] Preferably, the expression for the power grid contract multidimensional lean decision fusion model in the contract multidimensional decision fusion module is as follows: ,in, For the result of multi-dimensional decision fusion of contracts, Let be the weighting coefficient between the i-th type of contract element and the j-th type of procurement parameter. Let be the correlation coefficient between the i-th type of contract element and the j-th type of procurement parameter. For the i-th type of contract element data, including contract amount, performance period, and warranty period, This is the j-th type of procurement parameter data, including procurement quantity, supplier level, and material unit price. For matrix dot product operation, Let be the decision confidence level for the i-th type of contract element and the j-th type of procurement parameter, and n be the total number of contract element categories. This represents the total number of procurement parameter categories.

[0026] Specifically, the power grid contract multidimensional lean decision fusion model in the contract multidimensional decision fusion module uses weight coefficients and correlation coefficients determined through training on historical contract data. The weight coefficients range from 0.1 to 0.9, with specific values ​​determined by the relative importance of different contract elements and procurement parameters. For example, the weight coefficient for contract amount and material unit price is set at 0.9, and the weight coefficient for performance period and supplier level is set at 0.6. The correlation coefficient ranges from 0 to 1 and is calculated using the Pearson correlation coefficient. For instance, the correlation coefficient between procurement quantity and contract amount is calculated to be 0.85, and the correlation coefficient between warranty period and supplier level is 0.4. During implementation, the module first acquires contract element data and procurement parameter data. Contract elements include 12 categories such as contract amount, performance period, and warranty period, while procurement parameters include 8 categories such as procurement quantity, supplier level, and material unit price. Each category of data undergoes range validation; for example, the contract amount must be between 100,000 and 10 million yuan, and the procurement quantity must be between 10 and 10,000 units. Subsequently, correlation operations are performed on the data of the corresponding categories. First, data matching is performed to ensure that each type of contract element corresponds to the relevant procurement parameters. Then, the product of the two is calculated and multiplied by the corresponding weight coefficient and correlation coefficient. Next, the logarithm of (1 + decision confidence level) with base 2 is calculated. The decision confidence level is calculated based on the supplier's performance records over the past 3 years. The confidence level is 1.0 when the performance rate is 100%, 0.9 when the performance rate is 90% to 99%, and so on, with a minimum of 0.6. Finally, the calculation results of all categories are summed to obtain the multi-dimensional decision fusion result of the contract. This result must be controlled within the value range of 0 to 100. If it exceeds this range, a recalculation is triggered to ensure that the result is accurately used in subsequent cost optimization. The model calculation time for a single contract does not exceed 8 seconds, and the calculation error rate is less than 3%.

[0027] Preferably, the expression for the procurement cost lean optimization model in the cost lean optimization module is: ,in, To optimize procurement costs, For the purchase quantity of material category k, Let k be the unit cost of the material. Let k be the discount rate for the k-th material. To improve the efficiency of supplying the first supplier, For the first l The duration of cooperation with each supplier, This is the cost adjustment coefficient. Let be the historical procurement cost deviation value for the rth time, p be the total number of material categories, q be the number of suppliers, and r be the number of historical procurements.

[0028] Specifically, the lean cost optimization module includes a procurement cost optimization model. The model's parameters are set based on the actual procurement scenarios of power grid companies. The total number of material categories is set to 20, including commonly used power grid equipment such as transformers, cables, and insulators. The procurement quantity for each material category ranges from 10 to 10,000 units. The unit cost fluctuates between 100 and 5,000 yuan depending on the material type; for example, the unit cost of a transformer is 5,000 yuan, and the unit cost of a cable is 200 yuan. The discount rate is set in tiers based on the procurement quantity: 0.15 for purchases of 1,000 units or more, 0.1 for 500 to 1,000 units, and 0.05 for 10 to 500 units, ensuring compliance with the cost discount logic for bulk procurement. The system supports a maximum of 50 suppliers. Each supplier's delivery efficiency is calculated based on their on-time delivery rate over the past year. An efficiency value of 1.0 is assigned when the on-time delivery rate is 100%, 0.9 when it's 90%, and a minimum of 0.8. The cooperation period ranges from 1 to 10 years, with suppliers with 10 years of cooperation corresponding to a parameter of 10, and new suppliers to a parameter of 1. The initial cost adjustment coefficient is set at 0.05, based on a 10% fluctuation in the power grid company's procurement costs over the past year. For every 1% change in the fluctuation, the adjustment coefficient is adjusted by 0.005. The historical procurement count is set at 50 times. The cost deviation for each procurement is calculated as the difference between the actual procurement cost and the budgeted cost, with a range controlled between -5000 yuan and 5000 yuan. Deviations exceeding this range are considered abnormal data and are removed. During implementation, the module first extracts the above parameters and stores them in a temporary database. Then, it loads the model to perform calculations. First, it calculates the product of the purchase quantity and unit cost of various materials, multiplies it by (1 - discount rate), and sums the results to obtain the numerator. Next, it calculates the product of the supply efficiency and cooperation period of each supplier and sums the results to obtain the denominator. After dividing the numerator by the denominator, it adds the product of the cost adjustment coefficient and the maximum historical purchase cost deviation value to obtain the optimized purchase cost data. This data must be retained to two decimal places, the calculation error rate must be less than 5%, and the cost calculation time for a single contract must be controlled within 12 seconds.

[0029] Preferably, the expression for the contract element entity relationship extraction algorithm in the decision support computing module is: ,in, For the extracted set of entity relationships of contractual elements, The s-th contract element entity includes the name of Party A, the name of Party B, and the contract number. The relationship type between entity u and entity v includes cooperation, payment, and performance relationships. To match the loss function for entity relationships, For entity feature mapping function, This is the relation feature weight matrix. It is the Sigmoid activation function. For relational bias terms, t represents the total number of contract element entities, and v represents the total number of entity relation types.

[0030] Specifically, the contract element entity relationship extraction algorithm in the decision support computing module was tested and verified using a test set consisting of 1500 power grid procurement contract texts. Test results showed an entity recognition accuracy of no less than 92% and a relationship extraction accuracy of no less than 88%, ensuring the algorithm meets the accuracy requirements for practical applications. The contract element entities involved in the algorithm include 15 categories such as the name of Party A, the name of Party B, and the contract number. The names of Party A and Party B are limited to 2 to 20 characters in length and must conform to the full name format of the enterprise, such as "XX Province Power Company". The contract number is an 18-digit combination, with the first 6 digits representing the year and month, the middle 6 digits representing the region, and the last 6 digits representing the serial number. Entity relationship types are divided into 20 categories, such as cooperative relationships, payment relationships, and performance relationships. Each relationship type corresponds to specific judgment criteria; for example, payment relationships must include key information such as payment amount, payment time, and payment method. During algorithm implementation, the entity feature mapping function is first loaded. This function is trained based on the Word2Vec model using 5000 power grid industry contract texts as the training corpus. The word vector dimension is set to 300 dimensions to ensure accurate capture of text semantic features. Simultaneously, a relation feature weight matrix is ​​loaded. The matrix dimension is 20×300, and the weight values ​​are optimized using a stochastic gradient descent algorithm with a learning rate of 0.01 and 1000 iterations to ensure the rationality of the weight matrix. Next, entity annotation is performed on the cost data using the BIO annotation system, where "B" represents the start position of an entity, "I" represents an internal position of an entity, and "O" represents a non-entity position. Each character needs to be classified during annotation. Then, entity relation matching is performed. The partial derivative of the entity feature mapping function output with respect to the relation feature weight matrix is ​​calculated and multiplied by the relation bias term processed by the Sigmoid activation function (initially set to 0) to obtain the entity relation matching score. The relationship between entity pairs is determined by selecting the relation type with the highest score. Finally, the generated entity relationship set is deduplicated to remove redundant records where the entity pairs and relationship types are completely identical. After being organized in a fixed format, it is transmitted to the smart contract management module. The relationship parsing time for a single contract should not exceed 15 seconds, and the accuracy of the parsing results should be maintained above 88%.

[0031] Preferably, the smart contract management module satisfies the following conditions during operation: ,in, A collection of effective operations for intelligent contract management. This is a function for storing the data of the a-th contract. The input is the contract text data, and the output is the storage address index. Let `a` be the function for retrieving the `a`th contract. The input is the search terms, and the output is the contract matching results. Let this be the update operation function for the a-th contract. The input is the data to be updated in the fields, and the output is the update confirmation signal. Let 'a' represent the data for the first power grid procurement contract, and 'b' represent the total number of system management contracts.

[0032] Specifically, the operational logic of the intelligent contract management module is technically defined. The Oracle database connected to the module has a storage capacity of no less than 5TB and supports more than 100 concurrent accesses per second, ensuring it can meet the management needs of a large number of contracts from power grid companies. The contract operations performed by the module include three categories: storage, retrieval, and update. Each operation has clear parameter settings and execution procedures. In the contract storage operation, the contract data to be stored includes contract text data, optimized procurement cost data, and entity relationship sets. During storage, an index is created based on the contract number, which must conform to an 18-digit number format. The index is built using a B+ tree structure to improve data query efficiency. At the same time, the AES-256 encryption algorithm is used to encrypt sensitive data. The encryption key is automatically updated every 7 days. During the update process, the data encrypted with the old key must be re-encrypted to ensure data security. The storage time for a single contract is controlled within 5 seconds, and the storage success rate must reach 100%. Contract retrieval supports multi-field queries, including 10 categories of search keywords such as contract number, Party A's name, and Party B's name. During retrieval, an inverted index is built using the Lucene search engine, with an index update frequency of once per hour to ensure timely search results. The retrieval response time is no more than 2 seconds. Results are sorted in descending order of matching score, calculated based on keyword frequency and position, with weights set to 0.6 and 0.4 respectively. Core contract information is also displayed for quick access to key content. Contract update operations require user permission verification; only system administrators and contract handlers have access. After permission verification, the updated fields undergo data format validation. For example, date fields must conform to YYYY-MM-DD format, and numeric fields must be positive and within a reasonable range (e.g., purchase amount must be between 100,000 and 10 million RMB). After successful validation, a database UPDATE operation is executed, generating an unmodifiable operation log. The log includes the updater, update time, and the old and new values ​​of the updated fields. Updating a single contract takes no more than 8 seconds, and data consistency is ensured through a database transaction mechanism to prevent data tampering or loss.

[0033] Preferably, the collaborative control model expression of the system collaborative control module is: ,in, For system coordinated control coefficients, Let c be the data transmission rate of the c-th module. The computation time of the c-th module, As a co-regulatory factor, Let be the deviation value of the operating status of the c-th module, which is the difference between the actual operating parameters and the standard operating parameters of the module. This represents the total number of system modules, which is 6.

[0034] Specifically, the system's collaborative control module uses an industrial-grade PLC controller with a response time of no less than 1ms. It is equipped with a 7-inch touchscreen with a resolution of 800×480 pixels, supporting multi-touch for convenient real-time monitoring of module operation status. The remote monitoring function supports mainstream browsers such as Chrome and Firefox, with access port 8080 and HTTPS protocol for encrypted data transmission, ensuring the security of remote operations. The total number of system modules involved in model calculation is fixed at six: contract element acquisition, contract multi-dimensional decision fusion, cost lean optimization, decision support calculation, intelligent contract management, and system collaborative control. Each module has clearly defined standard values ​​for its operating parameters. The standard data transmission rate is 100Mbps, and the standard calculation time is set according to the module's function: 10 seconds for the contract element acquisition module, 8 seconds for the contract multi-dimensional decision fusion module, 12 seconds for the cost lean optimization module, 15 seconds for the decision support calculation module, and 5 seconds for storage operations, 2 seconds for retrieval, and 8 seconds for updates in the intelligent contract management module. The standard CPU utilization is ≤70%, and the standard memory utilization is ≤80%. During implementation, the module communicates with other modules via an RS485 bus at a baud rate of 115200bps, with 8 data bits, 1 stop bit, and no parity bit. Operating parameters of each module are collected every 2 seconds, and the deviation between the actual value and the standard value is calculated. A positive deviation indicates that the parameter exceeds the standard, while a negative deviation indicates that it is below the standard. The collaborative adjustment factor ranges from 0.5 to 2.0 and is dynamically adjusted according to the overall system load. The load rate is calculated by comprehensively considering CPU utilization, memory utilization, and data transmission volume. When the load rate is >80%, the adjustment factor is 2.0; when it is 60%-80%, it is 1.5; when it is 40%-60%, it is 1.0; and when it is <40%, it is 0.5. When calculating the system collaborative control coefficient, the basic coefficient is obtained by multiplying the ratio of the data transmission rate of each module to the computation time. Then, the reciprocal of the natural exponent of the product of the adjustment factor and the absolute value of the deviation of each module is calculated. Multiplying these two results in the final coefficient. A coefficient >8 is excellent, 5-8 is good, and <5 is poor. When the coefficient is less than 5, the module outputs control signals to adjust parameters, such as increasing bandwidth by 10Mbps each time (maximum 200Mbps), increasing CPU priority (levels 1-5), pausing non-critical tasks, monitoring parameter changes every second until the coefficient rises back to above 5, and generating a PDF format operation report every 5 minutes, retaining 90 days of data, and keeping the module's failure rate below 0.1%.

[0035] Preferably, the contract element acquisition module includes a text scanning unit, a data filtering unit, a format conversion unit, and a data verification unit. The text scanning unit scans the paper text of the power grid procurement contract page by page using a high-definition scanning device to generate high-definition image data. After grayscale processing of the image data, the text region contours are extracted. The data filtering unit performs noise filtering on the data within the extracted text region contours, removing light spots, shadows, and irrelevant background pixels from the image, while retaining valid text pixel information. The format conversion unit converts the filtered text pixel information into editable text format data, replacing and correcting any garbled characters that appear during the conversion process. The data verification unit performs integrity verification on the converted text data, comparing whether the contract identification fields in the text data are missing, including the contract number, procurement amount, and signing date, and outputs the verified contract text data.

[0036] Specifically, the contract element acquisition module comprises four units. The text scanning unit is equipped with a high-definition scanner with a resolution of 300 DPI, supporting automatic recognition of A4 and A3 paper. The scanning speed is set to 2 pages per second. During implementation, the paper contract is first scanned page by page to generate RGB mode images, and then grayscale processing is performed. The three color channels are converted into single-channel grayscale values ​​according to the weight ratios of 0.299, 0.587, and 0.114. Subsequently, the text area contour is extracted, and the contour extraction threshold is set to 127 (based on the grayscale value range of 0-255 of an 8-bit grayscale image). The data filtering unit uses a 3×3 pixel median filter window to filter noise in the text area data, removing light spots, shadows, and background pixels with grayscale value deviations exceeding 20, while retaining valid text pixels. The format conversion unit calls the OCR recognition engine to convert the filtered text pixels into UTF-8 encoded editable text. For garbled characters with a recognition confidence of less than 0.8, they are replaced and corrected according to the common character library of power grid contracts. The data verification unit performs integrity checks on key fields such as contract number (18-digit format), purchase amount (with 2 decimal places), and signing date (YYYY-MM-DD format). If any field is missing, a rescan instruction is triggered until the complete contract text data with key fields is output. The total processing time for a single contract by the four units is controlled within 10 seconds, and the data verification pass rate must reach over 99% to ensure that the data received by subsequent modules is accurate and effective.

[0037] Preferably, the contract multidimensional decision fusion module includes a data receiving unit, a model loading unit, an association operation unit, and a result output unit. The data receiving unit receives contract text data transmitted by the contract element acquisition module, establishes a data cache queue, and sorts the data according to the data reception time. The model loading unit calls the power grid contract multidimensional lean decision fusion model from the system database, loads the weight coefficient matrix and association coefficient table required for model operation, and initializes the model parameters. The association operation unit inputs the sorted contract text data into the initialized model, performs matrix dot product operation and logarithmic operation, and generates intermediate results for multidimensional decision fusion. The result output unit performs data format standardization processing on the intermediate results, removes abnormal values, and transmits the standardized fusion results to the cost lean optimization module.

[0038] Specifically, the contract multidimensional decision fusion module comprises four units. The data receiving unit receives data transmitted from the contract element acquisition module via an RJ45 Ethernet interface, with a stable transmission rate of 100Mbps. During implementation, a cache queue with a capacity of 100 contract data entries is established, and the data is sorted in ascending order by the data reception timestamp (accurate to milliseconds) to avoid processing order confusion. The model loading unit retrieves the power grid contract multidimensional decision fusion model from a 1TB local MySQL database, loading a 12×8 dimension weight coefficient matrix (weight values ​​0.1-0.9) and a correlation coefficient table (correlation degree 0-1). Simultaneously, the initial learning rate is set to 0.01, and the number of iterations is set to 100, completing parameter initialization. The correlation operation unit maps the sorted contract text data into 12 categories of contract elements and 8 categories of procurement parameters, inputs it into the initialization model, and sequentially performs matrix dot product operations, multiplication operations with weight coefficients and correlation coefficients, and logarithmic operations with base 2 (1 + decision confidence level 0.6-1.0). The results of all categories are then summed to generate intermediate fusion results. The output unit normalizes intermediate results to a value range of 0-100, removes outliers that deviate from the average by more than three standard deviations, processes them according to a standard format, and transmits them to the cost lean optimization module via an Ethernet interface. The calculation time for a single contract association does not exceed 6 seconds, and the accuracy of the output results must be consistent with the model calculation error rate, which must be less than 3%, to ensure that the cost optimization process receives high-quality fused data.

[0039] Preferably, the cost lean optimization module includes a parameter import unit, a model running unit, a cost calculation unit, and a result sending unit. The parameter import unit receives the fusion results transmitted by the contract multi-dimensional decision fusion module, extracts the material quantity, unit cost, and supplier level parameters required for procurement cost calculation, and stores them in a temporary parameter database. The model running unit loads the procurement cost lean optimization model, reads the parameter data from the temporary parameter database, and sets the initial value of the cost adjustment coefficient. The cost calculation unit performs summation, division, and maximum value calculations based on the loaded model and parameter data to generate optimized procurement cost data and stores intermediate variables in real time during the calculation process. The result sending unit encapsulates the optimized procurement cost data according to a preset data transmission protocol and sends it to the decision support calculation module, while recording the data sending time and the receiving module identifier.

[0040] Specifically, the cost lean optimization module comprises four units. The parameter import unit receives data from the contract multi-dimensional decision fusion module via a USB 3.0 interface, with the interface transmission latency controlled within 500 milliseconds. During implementation, it extracts cost calculation parameters such as purchase quantity (10-10000 units), unit cost (100-5000 yuan), and supplier level (AD level corresponding to 1.0, 0.8, 0.6, and 0.4), and stores them in a 50GB SSD temporary parameter database (read / write speed ≥500MB / s). During storage, an integrity check is performed every 10 parameter data entries to prevent parameter loss. The model running unit loads the procurement cost lean optimization model, reads the parameters from the temporary database, sets the initial value of the cost adjustment coefficient to 0.05 (based on a 10% cost fluctuation range), and loads 50 historical procurement cost deviation values ​​(from -5000 yuan to 5000 yuan). The cost calculation unit, based on the loaded model and parameters, first calculates the sum of various material purchase quantities × unit cost × (1 - discount rate) (discount rate 0-0.2, set according to purchase quantity tiers) as the numerator. Then, it calculates the sum of each supplier's supply efficiency (0.8-1.0) × cooperation period (1-10 years) as the denominator. After performing the numerator-denominator division operation, it adds the cost adjustment coefficient × the maximum historical deviation value to obtain the optimized purchase cost data. During the calculation process, intermediate variables are stored every 5 seconds. The result sending unit encapsulates the optimized cost data in JSON format (including cost value, calculation time, and parameter source fields) and sends it to the decision support calculation module (receiving end IP 192.168.1.100) via TCP / IP protocol, while recording the sending time and module identifier. The cost calculation time for a single contract is controlled within 12 seconds, and the result sending success rate must reach 100%, providing accurate cost parameter support for subsequent relationship analysis.

[0041] The multi-dimensional lean decision-making fusion model for power grid contracts of this invention is a core model for associating contract elements with procurement parameters and generating decision-making basis. Specifically, it integrates 12 types of contract elements (including contract amount, performance period, warranty period, etc.) and 8 types of procurement parameters (including procurement quantity, supplier level, material unit price, etc.) and achieves deep data fusion through multi-dimensional calculation. The implementation is as follows: First, a pre-set 12×8 dimension weight coefficient matrix (weight values ​​0.1-0.9, set according to the importance of the relationship between elements and parameters) and a correlation coefficient table (correlation degree 0-1, calculated based on Pearson correlation coefficient) are loaded from the system's MySQL database (capacity not less than 1TB). The initial learning rate of the model is set to 0.01 and the number of iterations is set to 100 to complete the initialization. Then, text data transmitted from the contract element collection module is received, mapped by category, and input into the model. Matrix dot product operation, multiplication operation with weights and correlation coefficients, and logarithmic operation with base 2 (1 + decision confidence level) (confidence level 0.6-1.0, determined based on supplier performance records) are performed sequentially. The accumulated results generate fused data, which is finally normalized to the 0-100 range and outliers are removed. This model breaks down the data barriers between contract elements and procurement parameters, generating coherent decision support data; it provides an accurate multi-dimensional data foundation for subsequent cost optimization, avoids the limitations of single-data decision-making, and improves the scientific and comprehensive nature of power grid procurement contract decisions.

[0042] The lean optimization model for procurement costs of this invention is a key model for calculating the optimal cost of power grid procurement contracts and achieving cost control. Specifically, it is a technical model that combines material procurement data, supplier data, and historical cost data to derive the optimized procurement cost through multi-parameter calculations. The implementation method is as follows: First, the parameter import unit extracts parameters such as purchase quantity (10-10000 units), unit cost (100-5000 yuan), and supplier level (AD level corresponds to a quantitative value of 1.0-0.4) from the multi-dimensional decision fusion results of the contract, and stores them in a 50GB SSD temporary database (read and write speed ≥500MB / s); then, the model is loaded and the initial value of the cost adjustment coefficient is set to 0.05 (based on a 10% cost fluctuation range over 1 year), and 50 historical purchase cost deviation values ​​(-5000 yuan to 5000 yuan) are imported at the same time; then, the calculation is performed, first calculating the sum of purchase quantity of various materials × unit cost × (1-discount rate) (discount rate 0-0.2, set according to purchase quantity tiers) as the numerator, and the sum of the supply efficiency of each supplier (0.8-1.0) × cooperation period (1-10 years) as the denominator, calculating the numerator / denominator, and adding the product of the adjustment coefficient × the maximum historical deviation to obtain the optimized cost. This model accurately calculates the optimal procurement cost, reducing unnecessary expenses; it helps power grid companies achieve lean management of procurement costs, reduce cost waste, improve the efficiency of procurement fund utilization, and ensure the economic efficiency of procurement.

[0043] The contract element entity relationship extraction algorithm of this invention is a core algorithm used to analyze the association between contract elements and clarify the relationship logic. Specifically, it is a technical algorithm for identifying contract entities and determining the relationships between entities from procurement cost data to generate a structured entity relationship set. Its implementation is as follows: First, a Word2Vec entity feature mapping function (300-dimensional word vector) trained on 5000 power grid contract texts and a 20×300-dimensional relationship feature weight matrix (optimized by stochastic gradient descent, learning rate 0.01, 1000 iterations) are loaded. The Sigmoid activation function temperature parameter is set to 1.0, and the initial value of the relationship bias term is 0. Then, the cost data is labeled with 15 types of entities (including the name of the client, the name of the contractor, the contract number, etc.) using the BIO annotation system. Next, the partial derivatives of the entity features with respect to the weight matrix are calculated, multiplied by the activated bias term to obtain the matching score, and the 20 relationship types with the highest scores (including cooperation, payment, performance relationships, etc.) are selected. Finally, duplicate and redundant relationships are removed to generate a set of "entity-relationship-entity" triples. This algorithm transforms unstructured data into structured entity relationship data, clarifying the connections between elements; it provides clear relational logic support for intelligent contract management, avoids confusion in contract element relationships, improves the structuring and accuracy of contract data processing, and ensures the orderly management of contracts.

[0044] The power grid procurement decision support platform of this invention is a comprehensive platform that integrates multiple modules and models / algorithms to provide full-process decision support for power grid procurement contract management. Specifically, it is a technology platform that integrates data acquisition, decision fusion, cost optimization, relationship parsing, contract management, and collaborative control. Its implementation is as follows: A core architecture of six modules is used. The contract element acquisition module acquires contract text data; the contract multi-dimensional decision fusion module runs the corresponding model to generate fusion results; the cost lean optimization module runs the cost model to calculate optimized costs; the decision support calculation module parses entity relationships through extraction algorithms; the intelligent contract management module performs contract storage (AES-256 encryption), retrieval (Lucene engine), and updates (permission verification + log recording); and the system collaborative control module (industrial-grade PLC controller) collects the operating parameters of each module (transmission rate, computation time, etc.) every 2 seconds and calculates the collaborative control coefficient (0-10). When the coefficient is <5, parameters such as bandwidth and CPU priority are adjusted. This platform provides full-process technical support for power grid procurement contract management, enabling collaborative operation of all aspects; it breaks the traditional decentralized management model, builds an integrated intelligent management system, improves the accuracy, efficiency and collaboration of power grid procurement contract management, meets multi-dimensional lean management needs, and promotes the digital transformation of power grid procurement contract management.

[0045] like Figure 2As shown, an intelligent management system for power grid enterprise procurement contracts, oriented towards multi-dimensional lean optimization, is described. The system operates as follows: First, the text scanning unit of the contract element acquisition module scans the power grid procurement contract text. Noise data is removed by the data filtering unit, the text format is converted by the format conversion unit, and the calibration fields are verified by the data verification unit to obtain valid contract text data. Second, the valid contract text data is transmitted to the contract multi-dimensional decision fusion module. After sorting by the data receiving unit, model initialization by the model loading unit, execution of calculations by the correlation calculation unit, and standardization by the result output unit, the contract multi-dimensional decision fusion result is obtained. Third, the fusion result is input into the cost lean optimization module. Parameters are extracted by the parameter import unit, the model is loaded by the model running unit, and the cost is calculated by the result output unit. After the calculation unit performs calculations and the result sending unit encapsulates the data, the optimized procurement cost data is obtained. The fourth step involves transmitting the optimized procurement cost data to the decision support calculation module, where the contract element entity relationship extraction algorithm is run to parse the entity relationships in the cost data, generating entity relationship parsing results. The fifth step involves transmitting the parsing results to the intelligent contract management module, where contract storage operations are performed to save contract data, retrieval operations are performed to query historical contracts, and update operations are performed to modify contract fields, generating contract management operation records. The sixth step involves the system collaborative control module monitoring the operating status of different modules in real time, calculating the data transmission rate and computation time of different modules, adjusting the module collaborative control coefficients, and outputting control signals to maintain stable operation of different modules.

[0046] A smart management system for power grid enterprise procurement contracts, designed for multidimensional lean management, organically combines a multidimensional lean decision-making fusion model for power grid contracts, a lean optimization model for procurement costs, and an algorithm for extracting entity relationships from contract elements through a multidimensional contract decision fusion module, a cost lean optimization module, and a decision support calculation module. The system first acquires complete contract data through a contract element acquisition module, then performs correlation calculations on the element data through the multidimensional decision fusion module, calculates optimized costs based on the fusion results, and the decision support calculation module parses entity relationships, forming a complete chain from data acquisition to decision output. This completely overcomes the shortcomings of existing technologies, such as the inability to correlate data at different stages and the lack of coherent decision support, significantly improving the accuracy of procurement contract management and ensuring that every management decision is based on comprehensive and relevant contract and procurement parameters.

[0047] This system establishes a comprehensive modular collaborative control mechanism, addressing the issues of independent module operation and lack of coordination in existing technologies. The system's collaborative control module monitors the operational status of all modules in real time, including contract element collection, multi-dimensional decision fusion, cost optimization, decision support, and intelligent management. It dynamically adjusts the data transmission rate and computational priority of each module, avoiding data transmission delays and computational conflicts by outputting collaborative control signals. Simultaneously, the intelligent contract management module achieves integrated operation of contract storage, retrieval, and updates, seamlessly connecting with other modules to form an efficient management closed loop. This design completely changes the current situation where modules operate independently, significantly improving contract management efficiency and ensuring the stable and efficient operation of power grid company procurement contracts from data collection to final management, meeting the practical needs of multi-dimensional lean management.

[0048] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean manufacturing, characterized in that, include: Contract element acquisition module, contract multi-dimensional decision fusion module, cost lean optimization module, decision support computing module, intelligent contract management module, and system collaborative control module; The contract element acquisition module obtains contract text data through the power grid procurement contract text scanning interface and transmits the data to the contract multi-dimensional decision fusion module. The contract multidimensional decision fusion module calls the power grid contract multidimensional lean decision fusion model to perform correlation calculations on the element data, and the calculation results are transmitted to the cost lean optimization module. The cost lean optimization module loads the procurement cost lean optimization model to calculate cost parameters from the fused data, and sends the calculation results to the decision support calculation module; the decision support calculation module runs the contract element entity relationship extraction algorithm to parse the entity relationships in the cost data, and the parsing results are transmitted to the intelligent contract management module. The intelligent contract management module performs storage, retrieval, and update operations on power grid enterprise procurement contracts based on the parsing results, and the operation instructions are synchronized to the system collaborative control module. The system collaborative control module monitors the operating status of different modules in real time and outputs module collaborative control signals to adjust the data transmission rate and operation priority of different modules.

2. The intelligent management system for power grid enterprise procurement contracts oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The expression for the power grid contract multidimensional lean decision fusion model in the contract multidimensional decision fusion module is as follows: ,in, For the result of multi-dimensional decision fusion of contracts, Let be the weighting coefficient between the i-th type of contract element and the j-th type of procurement parameter. Let be the correlation coefficient between the i-th type of contract element and the j-th type of procurement parameter. For the i-th type of contract element data, including contract amount, performance period, and warranty period, This is the j-th type of procurement parameter data, including procurement quantity, supplier level, and material unit price. For matrix dot product operation, Let be the decision confidence level for the i-th type of contract element and the j-th type of procurement parameter, and n be the total number of contract element categories. This represents the total number of procurement parameter categories.

3. The intelligent management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The expression for the procurement cost lean optimization model in the cost lean optimization module is as follows: ,in, To optimize procurement costs, For the purchase quantity of material category k, Let k be the unit cost of the material. Let k be the discount rate for the material category k. To improve the efficiency of supplying the first supplier, For the first l The duration of cooperation with each supplier, This is the cost adjustment coefficient. Let be the historical procurement cost deviation value for the rth time, p be the total number of material categories, q be the number of suppliers, and r be the number of historical procurements.

4. The intelligent management system for power grid enterprise procurement contracts oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The algorithm expression for extracting entity relationships of contract elements in the decision support computing module is as follows: ,in, For the extracted set of entity relationships of contractual elements, The s-th contract element entity includes the name of Party A, the name of Party B, and the contract number. The relationship type between entity u and entity v includes cooperation, payment, and performance relationships. Loss function for matching entity relationships, For entity feature mapping function, This is the relation feature weight matrix. It is the Sigmoid activation function. For relational bias terms, t represents the total number of contract element entities, and v represents the total number of entity relation types.

5. The intelligent management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The smart contract management module meets the following requirements during operation: ,in, A collection of effective operations for smart contract management. This is a function for storing the data of the a-th contract. The input is the contract text data, and the output is the storage address index. Let `a` be the function for retrieving the `a`th contract. The input is the search terms, and the output is the contract matching results. This is the update operation function for the a-th contract. The input is the data to be updated in the fields, and the output is the update confirmation signal. Let a be the data for the first power grid procurement contract, and b be the total number of system management contracts.

6. The intelligent management system for power grid enterprise procurement contracts oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The collaborative control model expression of the system collaborative control module is as follows: ,in, For system coordinated control coefficients, Let c be the data transmission rate of the c-th module. The computation time of the c-th module, As a co-regulatory factor, Let be the deviation value of the operating status of the c-th module, which is the difference between the actual operating parameters and the standard operating parameters of the module. This represents the total number of system modules, which is 6.

7. The intelligent management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The contract element acquisition module includes a text scanning unit, a data filtering unit, a format conversion unit, and a data verification unit. The text scanning unit scans the paper text of the power grid procurement contract page by page using a high-definition scanning device, generating high-definition image data. After grayscale processing of the image data, the text region contours are extracted. The data filtering unit filters noise within the extracted text region contours, removing light spots, shadows, and irrelevant background pixels from the image, retaining valid text pixel information. The format conversion unit converts the filtered text pixel information into editable text format data, replacing and correcting any garbled characters that appear during the conversion process. The data verification unit performs integrity verification on the converted text data, comparing whether contract identification fields in the text data are missing, including the contract number, procurement amount, and signing date, and outputs the verified contract text data.

8. The intelligent management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The contract multidimensional decision fusion module includes a data receiving unit, a model loading unit, an association operation unit, and a result output unit. The data receiving unit receives contract text data transmitted by the contract element acquisition module, establishes a data cache queue, and sorts the data according to the data reception time. The model loading unit calls the power grid contract multidimensional lean decision fusion model from the system database, loads the weight coefficient matrix and association coefficient table required for model operation, and initializes the model parameters. The association operation unit inputs the sorted contract text data into the initialized model, performs matrix dot product operation and logarithmic operation, and generates intermediate results for multidimensional decision fusion. The result output unit performs data format standardization processing on the intermediate results, removes outlier values, and transmits the standardized fusion results to the cost lean optimization module.

9. The intelligent management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean manufacturing, as described in claim 1, is characterized in that... The lean cost optimization module includes a parameter import unit, a model running unit, a cost calculation unit, and a result sending unit. The parameter import unit receives the fusion results transmitted by the contract multi-dimensional decision fusion module, extracts the material quantity, unit cost, and supplier level parameters required for procurement cost calculation, and stores them in a temporary parameter database. The model running unit loads the lean cost optimization model, reads the parameter data from the temporary parameter database, and sets the initial value of the cost adjustment coefficient. Based on the loaded model and parameter data, the cost calculation unit performs summation, division, and maximum value calculations to generate optimized procurement cost data and stores intermediate variables in real time during the calculation process. The result sending unit encapsulates the optimized procurement cost data according to a preset data transmission protocol and sends it to the decision support calculation module, while recording the data transmission time and the receiving module identifier.

10. A smart management system for procurement contracts of power grid enterprises oriented towards multi-dimensional lean manufacturing, as described in any one of claims 1-9, characterized in that, The system operation includes the following steps: First, the text scanning unit of the contract element acquisition module scans the power grid procurement contract text. Noise is removed by the data filtering unit, the text format is converted by the format conversion unit, and the calibration fields are verified by the data verification unit to obtain valid contract text data. Second, the valid contract text data is transmitted to the contract multi-dimensional decision fusion module. After sorting by the data receiving unit, model initialization by the model loading unit, calculation by the correlation operation unit, and standardization by the result output unit, the contract multi-dimensional decision fusion result is obtained. Third, the fusion result is input into the cost lean optimization module. After parameter extraction by the parameter import unit, model loading by the model running unit, calculation by the cost calculation unit, and result output, the result is output. After the unit encapsulates the data, optimized procurement cost data is obtained. The fourth step involves transmitting the optimized procurement cost data to the decision support computing module, where a contract element entity relationship extraction algorithm is run to parse the entity relationships in the cost data, generating entity relationship parsing results. The fifth step involves transmitting the parsing results to the intelligent contract management module, performing contract storage operations to save contract data, performing retrieval operations to query historical contracts, performing update operations to modify contract fields, and generating contract management operation records. The sixth step involves the system collaborative control module monitoring the operating status of different modules in real time, calculating the data transmission rate and computation time of different modules, adjusting the module collaborative control coefficients, and outputting control signals to maintain stable operation of different modules.

Images