Multi-source procurement resource integration and optimization method for special mining area in power grid

Abstract

The invention relates to the technical field of electric power material supply chain management, and discloses a multi-source purchase resource integration and optimization method for a power grid internal mining area, and the method comprises the steps: obtaining multi-source data, and carrying out the semantic mapping and normalization processing; extracting features to generate a structured project demand portrait, and analyzing technical specifications to establish a standard clause constraint mapping library; generating an alternative candidate set according to the demand portrait and the constraint library; calculating difference characteristics, outputting a risk assessment result, and analyzing multi-project dependence to establish a cross-project coupling relation object; constructing a mathematical model based on the candidate set, the risk result and the coupling relationship, and solving to obtain a resource integration optimization scheme; and the scheme is disassembled into instructions to be issued, and model parameters are calibrated based on feedback data. Through the digital constraint and global optimization model, the problems of resource conflict and compliance verification in a multi-project parallel scene are solved, and automatic integration and global configuration optimization of purchase resources are realized.

Description

Technical Field

[0001] This invention relates to the field of power supply chain management technology, specifically a method for integrating and optimizing multi-source procurement resources for power grid internal procurement zones. Background Technology

[0002] In the operational system of modern power grid enterprises, the internal procurement zone plays a crucial role in supporting the material supply for multi-level and multi-regional engineering projects. With the expansion of power grid construction and the iteration of technical standards, the data involved in material procurement is widely distributed across multiple heterogeneous systems such as material master data, equipment ledgers, engineering design, procurement management, and warehousing and logistics, exhibiting characteristics of fragmentation and semantic separation.

[0003] Existing procurement resource integration methods typically use material codes or category catalogs as aggregation dimensions, constructing optimization objectives through explicit indicators such as price and delivery time, and generating scheduling suggestions based on hard constraints such as inventory levels. These methods primarily treat materials as interchangeable units based on codes, which can improve efficiency to some extent in scenarios with a high degree of standardization. However, in practical engineering applications, the applicability of materials is constrained by multiple factors, including equipment interface compatibility, operating environment (such as high altitude or heavy pollution), technical standard versions, and acceptance criteria. These constraints often exist in unstructured technical specifications or design instructions and are difficult for systems based solely on codes to identify.

[0004] Because existing technologies lack the digital processing capabilities to address the aforementioned unstructured technical constraints, they cannot accurately characterize the substitution boundaries of materials under different operating conditions. Furthermore, existing methods often overlook the implicit coupling relationships that exist when multiple projects are running concurrently, such as competition among multiple projects for the same inventory batch or the same supplier's capacity window. This limitation leads to the system easily merging physically incompatible materials during the resource integration phase, or generating optimization solutions that fail to consider capacity and technical risks, ultimately resulting in materials failing to pass acceptance upon arrival, inventory backlogs, and delays on the critical path of the project, making it difficult to achieve effective allocation of global resources and risk control. Summary of the Invention

[0005] To address the problems in the existing power grid material procurement process, such as low efficiency in procurement resource integration, difficulty in verifying the compliance of substitution, and uneven global resource allocation due to semantic heterogeneity of multi-source data, difficulty in digitizing technical specification constraints, lack of quantitative assessment of substitution risks, and resource coupling conflicts between multiple projects, this invention provides a multi-source procurement resource integration optimization method for power grid internal procurement zones.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for integrating and optimizing multi-source procurement resources for power grid internal procurement zones includes the following steps:

[0008] Acquire heterogeneous data from multiple sources, perform semantic mapping and normalization, and output standardized data with a unified semantic format;

[0009] A structured project requirement profile is generated based on the material and environmental characteristics extracted from the standardized data, and a standard clause constraint mapping library is established based on the technical specification parsing of the standard documents in the standardized data.

[0010] Based on the structured project requirement profile, initial materials are retrieved, and the initial materials are verified using the standard clause constraint mapping library to generate a set of alternative candidates.

[0011] The difference features between the alternative candidate set and the structured project demand profile are calculated to output the alternative risk assessment results, and cross-project coupling relationship objects are established by analyzing inventory, capacity and logical dependencies in multi-project parallel scenarios.

[0012] Based on the set of alternative candidates, the results of the alternative risk assessment, and the cross-project coupling relationship objects, a mathematical model is constructed to solve for the resource integration optimization scheme.

[0013] The resource integration and optimization scheme is broken down into instruction issuance, and the parameters of the mathematical model are calibrated based on the collected execution feedback data.

[0014] Furthermore, the specific steps for generating a structured project requirement profile are as follows: Initialize a project requirement profile object containing six dimensions: material basis, equipment context, environment and operating conditions, standards and compliance, acceptance criteria, and project constraints; Fill the material basis dimension of the project requirement profile object with the material ID from the standardized data, and extract parent equipment information and physical interface features from the standardized data to fill the equipment context dimension; Use GIS association and text extraction based on site coordinates in the standardized data to fill the environment and operating conditions dimension of the project requirement profile object; Write standard compliance, acceptance criteria, and project constraint information to the project requirement profile object to generate a structured project requirement profile.

[0015] Furthermore, the specific steps for establishing the standard clause constraint mapping library are as follows: Perform digital preprocessing, including OCR recognition and table parsing, and key entity extraction on the standard documents in the standardized data, outputting entity labels covering constraint objects, triggering conditions, and constraint thresholds; map the entity labels to system variable codes using a standardized attribute dictionary, and then assemble structured rules; determine the constraint strength of the structured rules, marking rules containing no exception mandatory keywords as hard constraints and rules containing advisory keywords as flexible constraints; collect the rules marked with the hard constraints and the flexible constraints to establish the standard clause constraint mapping library.

[0016] Further, the specific steps for generating the alternative candidate set are as follows: performing a search based on the key material attributes in the structured project requirement profile to form an initial candidate list; validating the initial candidate list using the hard constraints in the standard clause constraint mapping library, retaining the material set that satisfies the hard constraint triggering conditions and verification logic; validating the material set that satisfies the hard constraint triggering conditions and verification logic using the flexible constraints in the standard clause constraint mapping library, generating a verification action list for materials that violate the rules, forming a candidate set that only satisfies the flexible constraints, and marking materials that do not violate the rules as directly usable materials; integrating the directly usable materials with the candidate set that only satisfies the flexible constraints to form a complete alternative candidate set.

[0017] Further, the specific steps for outputting the alternative risk assessment result are as follows: calculate the key parameter differences between the candidate materials in the alternative candidate set and the original demand corresponding to the structured project demand profile, and construct a difference feature vector including parameter deviation rate and matching status; determine whether there is a historical substitution record between the candidate materials and the original demand; if not, calculate the distance of the difference feature vector based on the preset expert weight vector to output a basic risk score; if so, input the difference feature vector into a pre-trained machine learning model to output the probability of substitution acceptance failure; convert the basic risk score or the probability of substitution acceptance failure into a monetized risk cost, and generate a generalized cost as the input for solving the resource integration optimization scheme.

[0018] Furthermore, the specific steps for establishing cross-project coupling relationship objects are as follows: traverse the alternative candidate set of all pending project requirements, identify the same inventory material item commonly referenced by multiple project requirements, and establish an inventory resource coupling object; identify multiple non-inventory requirements pointing to the same supplier's capacity within a preset time window, and establish a supplier capacity coupling object; define strongly correlated technical components in the standardized data as coupling groups and set a guarantee coefficient, and establish a logical substitution coupling object; integrate the inventory resource coupling object, the supplier capacity coupling object, and the logical substitution coupling object to form a cross-project coupling relationship object.

[0019] Furthermore, the specific steps for constructing the mathematical model and solving for the resource integration optimization scheme are as follows: Define binary decision variables to represent material selection and verification task execution respectively, and use the domain formula of decision variables to limit the range of values ​​for the binary decision variables; construct a generalized cost minimization objective function based on the binary decision variables and the generalized cost; call the branch and bound algorithm to solve for the optimal solution of the generalized cost minimization objective function, and output a resource integration optimization scheme containing selected materials, paths, and verification lists.

[0020] Furthermore, the mathematical model also sets the following constraints: a single-source hard constraint satisfaction formula is set to ensure that each demand is satisfied by only one candidate material; a shared inventory resource upper limit formula and a supplier capacity constraint formula are set to limit the total resource allocation; a cross-project coupling group guarantee formula is set to ensure that the consistency ratio of technical routes within the coupling group meets the guarantee coefficient; and a flexible substitution and verification task linkage formula is set to force that the corresponding verification task must be triggered when selecting a substitute that only satisfies the flexible constraint.

[0021] Furthermore, the specific steps for decomposing the resource integration and optimization scheme into instruction issuance are as follows: based on the binary decision variables in the resource integration and optimization scheme, the scheme is decoupled into a logistics scheduling instruction stream and a quality verification instruction stream; the logistics scheduling instruction stream is encapsulated and pushed to the warehousing terminal, and the quality verification instruction stream is encapsulated and routed to the quality inspection terminal; the terminal's response is monitored, and if an abnormal feedback is received, the resource integration and optimization scheme is recalculated and updated.

[0022] Furthermore, the specific steps for calibrating the model parameters are as follows: collecting measured data and aggregating it into a standard alternative case record; analyzing the flexible alternative cases that failed verification in the standard alternative case record, and upgrading the attributes of the technical parameters that caused the failure in the standard clause constraint mapping library to hard constraints; based on the verification results in the standard alternative case record, adjusting the expert weight vector used when calculating the basic risk score using a risk weight adaptive update formula.

[0023] This invention provides a method for integrating and optimizing multi-source procurement resources for power grid internal procurement zones. It has the following beneficial effects:

[0024] 1. This invention constructs a structured project requirement profile encompassing six dimensions, including material basis, working environment, and standard compliance, and analyzes technical specifications to establish a standard clause constraint mapping library that distinguishes between hard and flexible rules. This achieves digital docking between procurement requirements and technical standards. This method transforms discrete business data and unstructured technical text into computer-readable logical rules, which can automatically identify and verify special constraints under complex working conditions. It solves the problem of poor compliance caused by the inability of traditional manual matching methods to fully cover the details of technical specifications.

[0025] 2. This invention introduces a dual-track risk assessment mechanism, which calculates the basic risk score or the probability of substitution failure by calculating the difference feature vector, and quantifies the risk into a generalized cost in monetary form and incorporates it into the optimization model. With the linkage constraint of flexible substitution and verification tasks, this method can weigh physical costs and technical risks in the mathematical model. For material generation pre-verification tasks with parameter deviations but meeting flexible constraints, it can effectively expand the supply range of alternative resources while ensuring quality control.

[0026] 3. This invention solves the resource conflict problem in multi-project parallel scenarios by establishing cross-project coupling relationship objects covering inventory resources, supplier capacity and logical replacement components, and transforming these coupling relationships into global constraints in a mixed integer programming model. This method takes into account the inventory competition, capacity bottleneck restrictions and technical route consistency requirements of complete sets of equipment among multiple projects, avoids the overall configuration imbalance caused by local resource grabbing, and realizes the optimized allocation of global resources in the power grid internal procurement zone. Attached Figure Description

[0027] Figure 1 This invention provides a structural block diagram of a multi-source procurement resource integration and optimization system for power grid internal procurement zones.

[0028] Figure 2 The overall flowchart of the multi-source procurement resource integration and optimization method for power grid internal procurement zones provided by this invention;

[0029] Figure 3 This is a logical diagram of the resource integration optimization solution engine in this invention. Detailed Implementation

[0030] The technical solutions in the embodiments 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, and 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.

[0031] See attached document Figure 1 To implement the above method, the system architecture constructed in this embodiment mainly includes a data access module, a knowledge computing module, an optimization decision-making module, and an execution interaction module.

[0032] The data access module communicates with the enterprise's material master data, engineering project management, procurement management, warehousing and logistics, quality inspection, and standard document library, among other business systems. This module is responsible for extracting, cleaning, transforming, and loading multi-source heterogeneous data. Through pre-set unit conversion rules and a thesaurus, it converts scattered procurement, engineering, quality, and standard data into a unified semantic format.

[0033] The knowledge computing module includes a project requirement profile generation unit, a standard clause parsing unit, and an alternative risk assessment unit. The project requirement profile generation unit aggregates discrete project data into structured requirement objects; the standard clause parsing unit uses natural language processing technology to transform textual technical specifications into computer-readable logical constraint rules; and the alternative risk assessment unit uses rule models or machine learning models to quantitatively score the feasibility of material substitution.

[0034] The optimization decision-making module integrates a resource integration optimization solution engine, which receives a set of alternative candidates, alternative risk assessment results, and cross-project coupling relationship objects from the knowledge computing module. Based on a mixed-integer programming algorithm, this engine balances cost, risk, and resource efficiency while meeting hard technical constraints. The output includes an optimization solution comprising selected candidate materials, route suggestions, supplier batches, and a list of verification actions.

[0035] The execution interaction module connects the optimization decision results with the business execution system. This module breaks down the optimization plan into logistics scheduling instructions and quality verification task instructions and pushes them to the corresponding terminals. At the same time, it collects execution feedback data and sends it back to the knowledge computing module to drive knowledge base updates and model parameter calibration.

[0036] See attached document Figure 2 ,based on Figure 1 The system architecture shown in the invention provides a multi-source procurement resource integration and optimization method for power grid internal procurement zones, comprising the following steps:

[0037] S1, Perform multi-source heterogeneous data access and consistency processing: The system obtains original procurement, engineering and standard data from business terminals, and outputs standardized data with a unified semantic format through semantic mapping and unit normalization processing.

[0038] S2, Generate a structured project requirement profile: Based on standardized data with a unified semantic format, extract basic resource parameters, equipment context and environmental conditions to generate a structured project requirement profile.

[0039] S3, Construct a standard clause constraint mapping library: parse entity objects and logical thresholds in technical specification texts, and establish a standard clause constraint mapping library containing hard and flexible constraint rules;

[0040] S4, Generate and filter alternative candidate set: Retrieve initial materials based on the structured project requirement profile, and verify them using the standard clause constraint mapping library, outputting an alternative candidate set that has been logically filtered and includes a list of verification actions;

[0041] S5, Perform a dual-track alternative risk assessment: Calculate the differences between each option in the alternative candidate set and the structured project requirement profile, and output quantitative alternative risk assessment results;

[0042] S6, Identify implicit coupling relationships across projects: Analyze inventory competition, capacity constraints and logical dependencies among multiple projects, and establish objects with cross-project coupling relationships;

[0043] S7, Perform resource integration optimization solution: Use the alternative candidate set, alternative risk assessment results and cross-project coupling relationship objects as input parameters to build a mathematical model and calculate the globally optimal resource integration optimization solution;

[0044] S8, Task-based execution: Decompose the resource integration and optimization plan into logistics scheduling instructions and quality verification task instructions, send them to the execution end and collect execution feedback data;

[0045] S9, Result Structured Feedback and Closed-Loop Evolution: Calibrate risk model parameters and adjust constraint rule strength based on the feedback data of the returned execution.

[0046] The technical implementation details of each of the above steps will be described in detail below with reference to specific embodiments.

[0047] Step S1: Perform multi-source heterogeneous data access and consistency processing.

[0048] S101, the data access module connects to systems such as material master data, engineering project management, procurement management, quality inspection, and standard document libraries via ODBC (Open Database Connectivity) or API (Application Programming Interface). The accessed data covers: procurement-side data including material codes and technical parameters; engineering-side data including BOM (Bill of Materials), WBS (Work Breakdown Structure), and GIS (Geographic Information System) information; quality-side data including inspection reports and defect logs; and various technical standard documents.

[0049] S102, the data access module performs normalization processing using a pre-built standard unit library. It iterates through numerical fields, identifies unit symbols, and uses a conversion factor table to uniformly convert physical parameters into SI or industry standard units (such as mm, kV) to ensure consistent numerical comparison benchmarks.

[0050] S103, the data access module performs semantic alignment using a synonym mapping table. It segments and extracts features from text fields, mapping non-standard descriptions to unified feature labels (e.g., mapping "high-altitude prototype" to Env.Altitude_High); for descriptive words not present in the mapping table, they are marked as fields to be verified.

[0051] Step S2, generate a structured project demand portrait (PDP):

[0052] S201, the Project Requirements Profile Generation Unit in the Knowledge Computing Module initializes the PDP object, which includes six data dimensions: material basis, equipment context, environment and operating conditions, standards and compliance, acceptance criteria and project constraints.

[0053] S202, fill in the basic dimensions of the material (such as rated voltage and current) based on the material ID; analyze the equipment topology relationship in the engineering data, extract the parent equipment information and physical interface characteristics (such as terminal model and mounting hole distance) and fill them into the equipment context dimension.

[0054] S203 uses GIS to obtain meteorological and geographical data (such as altitude and temperature) from the site and populates the environmental dimension; it uses regular expressions to extract special working condition requirements from the design specifications (such as extracting salt spray level through the pattern "corrosion resistance level") and stores them in key-value pairs.

[0055] S204: Based on the project type, associate technical standards and write them into the standards and compliance dimension; statistically analyze the historical high-frequency defects of similar materials and write them into the acceptance criteria dimension; extract the latest delivery time and critical path status from the schedule plan and write them into the project constraint dimension, and finally generate a structured PDP object for subsequent processing.

[0056] Step S3, construct the standard clause constraint mapping library, specifically including:

[0057] S301, the standard clause parsing unit in the knowledge computing module, performs digital preprocessing on standard documents in PDF or image format. It uses an OCR component (Optical Character Recognition) to restore the text content and uses a table parsing algorithm to convert table data into a sequence of key-value pairs.

[0058] S302 performs key entity extraction based on the BERT-BiLSTM-CRF deep learning model (a combination of deep learning models for named entity recognition). Preprocessed text is input into the model, which outputs character-level entity labels. Entity labels include: constraint objects, triggering conditions, logical operators, constraint thresholds, and units of measurement.

[0059] S303, Construct constraint rule triples and logical transformations. Use a standardized attribute dictionary to map extracted entity names to system variable codes, assembling structured rules. For clauses containing conditional restrictions, set specific triggering condition logic (e.g., Env.Altitude > 1000) and constraint verification logic; for general clauses without conditional restrictions, set the triggering condition logic to Global.

[0060] S304 determines the constraint strength and generates a standard clause constraint mapping library. Keyword matching is performed based on modal verbs: clauses containing "must," "strictly prohibited," or "should" without exceptions are marked as Hard (hard constraint); clauses containing "should" or "recommended" are marked as Soft (soft constraint). Finally, the structured rule records and their original document sources are stored in a graph database.

[0061] Step S4: Generate the alternative candidate set and perform two-stage filtering.

[0062] S401, the knowledge computing module performs an initial candidate search. Based on the material group code, rated voltage, and rated capacity key attributes in the project requirement profile (PDP), it searches the ERP system material master data (ERP, Enterprise Resource Planning system) for inventory materials with the same material group and attribute deviations within a preset range (such as ±10%) to form an initial candidate list.

[0063] S402, the standard clause parsing unit in the knowledge computing module performs the first-level filtering based on hard constraints. It iterates through the initial candidate list and verifies it using rules marked "Hard" from the standard clause constraint mapping library. Materials that fail to meet the hard constraint triggering conditions or verification logic are removed, and the remaining set of materials is denoted as [list of materials]. .

[0064] S403, the standard clause parsing unit performs a second-level filtering and verification task based on flexible constraints to generate [something]. The materials are validated using flexible constraint (Soft) rules: if no violation occurs, they are marked as directly usable; if a violation occurs, a pre-set "deviation verification action mapping table" is queried to generate a corresponding list of verification actions (such as field adaptability testing). These materials and their verification tasks constitute a candidate set that only satisfies the flexible constraints. Ultimately, the knowledge computing module integrates directly available resources and... This constitutes a complete set of alternative candidates. .

[0065] Step S5: The alternative risk assessment unit in the knowledge computing module performs a dual-track alternative risk assessment.

[0066] S501, Alternative Risk Assessment Unit for Calculating Candidate Materials Compared with the original demand Key parameter differences, constructing difference feature vectors This vector contains the relative deviation rate of numerical parameters and the matching status of enumerated parameters. For numerical parameters, their relative deviation rate is calculated; for enumerated parameters (such as installation method and insulation medium), they are converted into numerical values ​​through binary encoding (0 for matching and 1 for non-matching). The final vector is... It consists of all the above numerical elements and is used for subsequent distance and modulus calculations.

[0067] S502, perform dual-track assessment calculation. Determine if a historical substitute record exists: If not (cold start), enter the expert rule track: based on the pre-set expert weight vector, use the weighted Euclidean distance algorithm to calculate the distance between the difference feature vector and the zero vector, and output the basic risk score. If a substitute record exists (mature scenario), enter the machine learning track: [The text abruptly ends here, likely due to an incomplete sentence or missing information.] The XGBoost classification model is pre-trained with the project environment characteristics as input, and the output is the probability of failure of alternative acceptance.

[0068] S503, generating generalized cost The risk score or failure probability output by S502 is converted into a monetized risk cost (calculated as: failure probability multiplied by the sum of delay loss and re-mining cost). This risk cost is then added to the physical supply cost of materials to obtain the... The project requirements are selected first. The generalized cost of each candidate material This serves as the input for the optimization decision-making module.

[0069] See attached document Figure 3 The optimization decision-making module receives three types of structured input data from the knowledge computing module: a set of alternative candidates, alternative risk assessment results, and cross-project coupling relationship objects. Based on the above data, the optimization decision-making module constructs a mixed-integer programming model and outputs an optimization solution that includes selected candidate materials, route suggestions, supplier batches, and a list of verification actions.

[0070] Step S6 involves the identification and modeling of implicit coupling relationships across projects. In a multi-project parallel scenario, the knowledge computing module is responsible for identifying the dependencies between different project requirements arising from shared resources or technological connections, and generating cross-project coupling relationship objects. This includes the following sub-steps:

[0071] S601, the knowledge computing module establishes a coupled object for inventory resources. The knowledge computing module iterates through the alternative candidate set of all pending project requirements, identifying the same inventory item referenced by multiple project requirements. When a certain inventory item... Simultaneously appearing in demand and demand When a candidate item is selected, the knowledge computing module establishes an inventory competition relationship object and records the current actual total inventory of that item. This serves as a rigid upper limit for subsequent resource allocation.

[0072] S602, the knowledge computing module establishes supplier capacity coupling objects. For non-inventory procured materials, the knowledge computing module identifies the same supplier based on the supplier mapping relationship provided by the data access module. Remaining available capacity within a specific time window When multiple projects require similar materials produced by the same supplier, the knowledge computing module generates a capacity coupling object to prevent delivery delays caused by concentrated order placement.

[0073] S603, the knowledge computing module establishes logical replacement coupling objects. For complete sets of equipment or strongly related technical components, the knowledge computing module defines them as coupling groups. When substituting materials within a coupling group, the technical approach must remain consistent. For example, if a domestic alternative is chosen for the main equipment, its supporting components must also be compatible domestic models. The knowledge computing module sets the coupling group's protection coefficient. This is used to constrain the consistency ratio of resource allocation within a group.

[0074] Step S7 involves the operation mechanism of the resource integration optimization solution engine. Based on the above input data, the optimization decision module constructs a mixed-integer programming mathematical model. This process includes three stages: variable definition, objective function construction, and constraint setting.

[0075] S701, the optimization decision module defines the decision variables of the model. The optimization decision module sets binary decision variables. Indicates the first Should the project requirements be selected as the first one? One candidate material, and a binary decision variable Indicates whether to execute the first step. Verification task.

[0076] The optimization decision-making module uses the domain definition formula for decision variables to limit the range of variables. The domain definition formula for decision variables is as follows:

[0077] ;

[0078] In the formula, Indicates selection. This indicates that the selection is not made; the same applies to [other selections]. .

[0079] S702, the optimization decision module constructs the optimization objective function. The system aims to find the resource allocation scheme with the lowest generalized cost, which includes the comprehensive supply cost of materials (including physical cost and risk cost) and the execution cost of the verification task. The optimization decision module uses the generalized cost minimization objective function for calculation, which is:

[0080] ;

[0081] In the formula, This represents the operation of finding the minimum value; This represents the summation operation; This indicates the total quantity required for the project; Indicates the number index of project requirements ( ); Indicates the first A set of alternative candidates for each requirement; Indicates the index of the candidate materials; Indicates the first The project requirements are selected first. The generalized cost of each candidate material (calculated from the preceding steps); Indicates the total number of verification tasks; Indicates the number index of the verification task ( ); Indicates execution of the first The fixed cost of each verification task; Indicates whether to execute the first step. Verification task.

[0082] S703, the optimization decision-making module sets a single-source hard constraint. To ensure that each demand ultimately falls on only a specific resource, the optimization decision-making module uses a single-source hard constraint satisfaction formula for calculation. The single-source hard constraint satisfaction formula is:

[0083] ;

[0084] In the formula, Indicates the first A complete set of alternative candidates for each requirement (including a subset of candidates that meet the hard constraints). Compared with candidate subsets that only satisfy flexible constraints ); This means that the condition holds true for any element. This constraint ensures that every requirement... Must be and can only be one candidate material The requirements are met.

[0085] S704, the optimization decision-making module sets resource ceiling constraints. To prevent over-issuance of materials and over-ordering, the optimization decision-making module sets constraints for both inventory and production capacity. For shared inventory, the optimization decision-making module uses the shared inventory resource ceiling formula for calculation. The shared inventory resource ceiling formula is as follows:

[0086] ;

[0087] In the formula, Indicates the first Individual demand for materials The number of requests per transaction; Indicates supplies The current actual total inventory; This represents the collection of all inventory materials. Regarding supplier capacity, the optimization decision module uses a supplier capacity constraint formula for calculation. The supplier capacity constraint formula is as follows:

[0088] ;

[0089] In the formula, Indicates the supplier's number index; This indicates that the needs can be met. The supplier set; Indicate demand Assigned to supplier Capacity utilization at that time; Indicates supplier The remaining available capacity; This represents the set of all suppliers.

[0090] S705, the optimization decision-making module sets cross-project coupling group guarantee constraints. To ensure the consistency of technical approaches within the coupling group, the optimization decision-making module uses the cross-project coupling group guarantee formula for calculation. The cross-project coupling group guarantee formula is as follows:

[0091] ;

[0092] In the formula, Represents the set of all coupled groups; Indicates the number index of the coupled group; Indicates the first The set of project requirements contained within each coupled group; This indicates the total demand within the group; This refers to the set of target critical materials that need to be prioritized or kept consistent within the group (corresponding to the coupling group critical materials defined in S603). Represents the protection coefficient of the coupling group ( ).

[0093] S706, the optimization decision module sets constraints for the linkage between flexible substitution and verification tasks.

[0094] When the system selects an alternative that only satisfies the flexibility constraint (i.e., has a technical deviation), the corresponding verification task must be forcibly triggered. The optimization decision module uses a formula linking flexible substitution and verification tasks for calculation. The formula for linking flexible substitution and verification tasks is as follows:

[0095] ;

[0096] In the formula, Indicates the first A subset of candidates that only meet the flexible constraints among the requirements; This represents the set of verification tasks related to this flexible alternative; This represents the linkage coefficient, which is usually set to 1, indicating mandatory execution.

[0097] S707, the optimization decision module performs the solution and output. The optimization decision module calls the built-in branch-and-bound algorithm to solve the above mixed-integer programming model, obtaining the solution that minimizes the generalized cost objective function. and The optimal solution is then determined. Based on this optimal solution, the optimization decision module generates the final optimization plan, which specifically includes: the final material code selected for each project requirement, logistics route suggestions for cross-regional allocation, supplier batch allocation plan for purchased parts, and a list of mandatory verification actions for flexible alternatives. This plan is then passed to the execution interaction module for task distribution.

[0098] Step S8, execute task-based distribution:

[0099] S801, the execution interaction module receives the optimization plan and, based on the decision variables... (Resource allocation) and (Verification Task) The solution is decoupled into a logistics scheduling instruction flow and a quality verification instruction flow. The logistics instruction includes the material code, storage location number, destination, and quantity; the quality inspection instruction includes the material ID, inspection parameters, and threshold.

[0100] S802 executes the interaction module's protocol conversion and distribution: encapsulating logistics instructions into XML messages and pushing them to WMS or PDA terminals (PDA, Personal Digital Assistant); converting quality inspection instructions into JSON task sheets and routing them to LIMS terminals (LIMS, Laboratory Information Management System). For "flexible alternative" materials, a "high-risk" label is added to the instructions, mandating the return of original testing data.

[0101] S803, the execution interaction module listens for receipts through the message queue. If it receives a "frozen inventory" or "physical shortage" anomaly from the warehouse terminal, it sends a recalculation request to the optimization decision module, marks the corresponding material as unavailable, and triggers a re-optimization solution after removal.

[0102] Step S9, structured backpropagation and closed-loop evolution of results:

[0103] S901, the execution interaction module collects the measured data from the quality inspection terminal, aggregates it into a standard alternative case record containing the original requirement parameters, alternative parameters, parameter deviation vector, verification conclusion and cost, and sends it back to the knowledge computing module.

[0104] S902, the standard clause parsing unit in the knowledge computing module analyzes flexible alternative cases for "verification failure". If the failure is caused by a deviation of a specific technical parameter, the attribute of that parameter in the standard clause constraint mapping library is upgraded from Soft (flexible constraint) to Hard (hard constraint) to automatically filter out similar risks in subsequent calculations.

[0105] S903, the alternative risk assessment unit in the knowledge computing module adjusts the expert rule weights based on the validation results using an adaptive risk weight update formula:

[0106] ;

[0107] In the formula, Indicates the number index of the technical parameters; Indicates the iteration round; Indicates the updated number Risk weights for each technical parameter; Indicates the number before the update The current risk weight of each technical parameter; This represents the preset learning rate coefficient; This represents the feedback signal value when the verification result is "failure". The value is 1, indicating that the verification result is "successful". The value can be 0 or a negative value; Indicates the first case in the current case The deviation values ​​of each parameter; This represents the differential feature vector of the current case; Represents the absolute value of the parameter deviation; Represents the differential feature vector The modulus (norm).

[0108] System Deployment and Architecture:

[0109] This system adopts a microservice architecture and is deployed on a private cloud or server cluster.

[0110] Hardware configuration: Server nodes are configured with multi-core processors (e.g., 64 cores or more) to support NLP parallel computing (NLP, Natural Language Processing); large-capacity memory (e.g., 256GB or more) is configured to carry graph database indexes; solid-state drive arrays are used to ensure master data read and write.

[0111] Software Architecture: The data access module connects to systems such as ERP, PMS, and LIMS (Project Management System) via an API gateway; the knowledge computing module uses a graph database (such as Neo4j) to store and parse data; the optimization decision-making module integrates a mathematical programming solver (such as Gurobi or CPLEX); and the execution interaction module uses a distributed message queue (such as Kafka) to achieve asynchronous processing of instructions and feedback. The system supports incremental updates and full system refactoring.

[0112] To verify the technical effects of the present invention, the following embodiments and comparative examples are provided.

[0113] Example 1:

[0114] Twelve parallel distribution network automation upgrade projects were selected from a provincial power grid company, involving 87 types of materials (such as ring main units, DTU terminals, etc.), with a total demand of approximately 3,200 units / sets.

[0115] Data processing: The system integrates procurement, engineering, quality, and standards data (referencing standards such as Q / GDW1374.2). The knowledge computing module generates 87 requirement profiles and loads 412 standard mapping records (287 hard constraints).

[0116] Candidate generation: 1423 initial candidates, 986 were retained after two-level filtering (723 passed hard constraints, 263 were feasible conditions).

[0117] Coupling identification: Three sets of cross-project coupling relationships were identified:

[0118] Coupling Group A (Protocol Constraints): Five projects compete for a limited DTU framework protocol quota (DTU, Data Terminal Unit).

[0119] Coupling Group B (Capacity Constraint): 4 projects are concentrated at Supplier X, exceeding its monthly capacity of 400 units;

[0120] Coupling Group C (Environmental Limitations): IP54 protection rating is mandatory for 3 high-altitude projects.

[0121] Solution results: The optimization decision module completed the solution in 1.8 hours. The solution satisfies all hard constraints, with a 100% resource availability rate for the critical path and a 38% conditionally feasible solution for non-critical projects. The execution interaction module issued 152 verification tasks.

[0122] Comparative Example 1:

[0123] The traditional matching method centered on material codes ignores project environment and dynamic standard constraints. Under the same input, this leads to:

[0124] Technical incompatibility: Five projects were rejected due to incompatible DTU communication protocols (one party used the 101 protocol, and the other used the 104 protocol); two high-altitude projects were misconfigured with IP44 equipment, resulting in condensation failures.

[0125] Resource conflict: Failure to identify capacity bottlenecks led to delays in three projects due to supplier shortages.

[0126] Validation deficiency: The lack of pre-validation tasks means that quality issues are only exposed during the acceptance phase.

[0127] Comparison of effects:

[0128] The key indicators are compared in the table below:

[0129] Evaluation Dimensions Example 1 (Invention) Comparative Example 1 (Traditional Method) Alternative solution first pass rate 98.7% 62.3% Critical path project resource availability rate 100% 76.5% Acceptance return and exchange rate 1.2% 18.9% Number of cross-project resource conflicts 0 7 Average optimization solution time 1.8 hours 0.3 hours Number of days of delay due to replacement failure 0 days Average 9.4 days / project Pre-verification coverage 100% 0%

[0130] Data shows that this invention improves the feasibility of the solution and the resource availability rate through refined profiling and coupled modeling.

[0131] Example 2:

[0132] This system was applied to a maintenance project at an ultra-high voltage converter station (involving highly constrained materials). The knowledge computing module, adhering to strict standards (such as IEC60633), eliminated 92% of non-compliant candidates during the initial screening stage. Addressing the issue of overlapping maintenance windows, the system identified "valve cooling pump inventory" as a critical shared resource. By setting a minimum guaranteed quantity constraint, priority supply was ensured for main channel maintenance, ultimately achieving zero-risk, on-time delivery.

Claims

1. A method for integrating and optimizing multi-source procurement resources for power grid internal procurement zones, characterized in that, Includes the following steps: Acquire heterogeneous data from multiple sources, perform semantic mapping and normalization processing, and output standardized data with a unified semantic format; A structured project requirement profile is generated based on the material and environmental characteristics extracted from the standardized data, and a standard clause constraint mapping library is established based on the technical specification parsing of the standard documents in the standardized data. Based on the structured project requirement profile, initial materials are retrieved, and the initial materials are verified using the standard clause constraint mapping library to generate a set of alternative candidates. The difference characteristics between the alternative candidate set and the structured project demand profile are calculated to output the alternative risk assessment results, and cross-project coupling relationship objects are established by analyzing inventory, capacity and logical dependency relationships in multi-project parallel scenarios. Based on the set of alternative candidates, the results of the alternative risk assessment, and the cross-project coupling relationship objects, a mathematical model is constructed to solve for the resource integration optimization scheme. The resource integration and optimization scheme is broken down into instruction issuance, and the parameters of the mathematical model are calibrated based on the collected execution feedback data.

2. The method for multi-source procurement resource integration and optimization for power grid internal procurement zones according to claim 1, characterized in that, The generated structured project requirements profile includes: Initialize a project requirement profile object that includes six dimensions: material basis, equipment context, environment and operating conditions, standards and compliance, acceptance criteria and project constraints; Based on the material ID in the standardized data, fill the material basic dimension in the project requirement profile object, and extract the parent equipment information and physical interface features from the standardized data to fill the equipment context dimension. The environment and working condition dimensions of the project requirement profile object are populated using GIS association and text extraction methods based on the site coordinates in the standardized data. Write standard compliance, acceptance criteria, and project constraint information into the project requirement profile object to generate a structured project requirement profile.

3. The method for multi-source procurement resource integration and optimization for power grid internal procurement zones according to claim 1, characterized in that, The establishment of the standard clause constraint mapping library includes: The standardized documents in the standardized data undergo digital preprocessing, including OCR recognition and table parsing, as well as key entity extraction, and the output entity labels cover the constraint objects, triggering conditions, and constraint thresholds. The entity labels are mapped to system variable codes using a standardized attribute dictionary, and then structured rules are assembled. Determine the constraint strength of the structured rules, mark rules containing no exception mandatory keywords as hard constraints, and mark rules containing advisory keywords as flexible constraints; The rules that mark the hard constraints and the flexible constraints are collected to establish a standard clause constraint mapping library.

4. The method for multi-source procurement resource integration and optimization for power grid internal procurement zones according to claim 3, characterized in that, The generation of the alternative candidate set includes: Based on the key material attributes in the structured project requirement profile, a search is performed to form an initial candidate list; The initial candidate list is validated using the hard constraints in the standard clause constraint mapping library, and the set of materials that meet the hard constraint triggering conditions and validation logic is retained. The set of materials for verifying the triggering conditions and verification logic of the hard constraints is verified using the flexible constraints in the standard clause constraint mapping library. A list of verification actions is generated for materials that violate the rules, forming a candidate set that only satisfies the flexible constraints, and materials that do not violate the rules are marked as directly usable materials. The directly available resources are integrated with the candidate set that only meets the flexible constraints to form a complete alternative candidate set.

5. The method for multi-source procurement resource integration and optimization for power grid internal procurement zones according to claim 1, characterized in that, The output substitution risk assessment results include: Calculate the key parameter differences between the candidate materials in the alternative candidate set and the original requirements corresponding to the structured project requirement profile, and construct a difference feature vector including parameter deviation rate and matching status; Determine whether there is a historical substitution record between the candidate material and the original demand. If not, calculate the distance of the difference feature vector based on the preset expert weight vector to output the basic risk score. If there is, input the difference feature vector into the pre-trained machine learning model to output the substitution acceptance failure probability. The basic risk score or the probability of failure of alternative acceptance is converted into a monetized risk cost, which is then used to generate a generalized cost as input for solving the resource integration optimization scheme.

6. The method for multi-source procurement resource integration and optimization for power grid internal procurement zones according to claim 1, characterized in that, The objects used to establish cross-project coupling relationships include: Iterate through the set of alternative candidates for all pending project requirements, identify the same inventory item that is referenced by multiple project requirements, and establish an inventory resource coupling object. Identify multiple non-inventory demands that point to the same supplier's capacity within a preset time window, and establish supplier capacity coupling objects; Define the strongly correlated technical components in the standardized data as coupling groups and set a guarantee coefficient to establish logical replacement coupling objects; The inventory resource coupling object, the supplier capacity coupling object, and the logical substitution coupling object are integrated to form a cross-project coupling relationship object.

7. The method for multi-source procurement resource integration and optimization for power grid internal procurement zones according to claim 5, characterized in that, The process of constructing a mathematical model and solving it to derive a resource integration optimization scheme includes: Define binary decision variables to represent the tasks of material selection and verification, and use the domain formula of decision variables to limit the range of values ​​of the binary decision variables; Based on the binary decision variables and the generalized cost, a generalized cost minimization objective function is constructed. The branch and bound algorithm is invoked to find the optimal solution to the generalized cost minimization objective function, and the output is a resource integration optimization scheme that includes selected materials, routes, and a verification list.

8. The method for multi-source procurement resource integration and optimization for power grid internal procurement zones according to claim 7, characterized in that, The construction of the mathematical model also includes setting the following constraints: A single-source hard constraint satisfaction formula is set to ensure that each requirement is satisfied by only one candidate material; Set formulas for the upper limit of shared inventory resources and supplier capacity constraints to limit the total amount of resources allocated. Define a cross-project coupling group guarantee formula to ensure that the consistency ratio of technical routes within the coupling group meets the guarantee coefficient; A formula linking flexible substitution and verification tasks is set up, which mandates that the corresponding verification task must be triggered when a substitute that only satisfies the flexible constraints is selected.

9. The multi-source procurement resource integration and optimization method for power grid internal procurement zones according to claim 7, characterized in that, The disassembly into instruction issuance includes: Based on the binary decision variables in the resource integration and optimization scheme, the scheme is decoupled into a logistics scheduling instruction flow and a quality verification instruction flow; The logistics scheduling instruction stream is encapsulated and pushed to the warehousing terminal, and the quality verification instruction stream is encapsulated and routed to the quality inspection terminal. Listen for terminal feedback; if abnormal feedback is received, trigger a recalculation and update of the resource integration and optimization scheme.

10. A method for integrating and optimizing multi-source procurement resources for power grid internal procurement zones according to claim 5, characterized in that, The calibration model parameters include: Collect and aggregate measured data to create standard alternative case records; Analyze the flexible alternative cases that failed verification in the standard alternative case record, and upgrade the attributes of the technical parameters that caused the failure in the standard clause constraint mapping library to hard constraints; Based on the verification results in the standard alternative case records, the expert weight vector used in calculating the basic risk score is adjusted using a risk weight adaptive update formula.

Images