Intelligent power supply chain procurement method and system based on digital employee model and skill package arrangement

Abstract

This application proposes an intelligent procurement method and system for the power supply chain based on a digital employee model and skill package orchestration. The method includes: collecting multi-source heterogeneous data from the power supply chain; fusing maintenance record data and equipment sensor monitoring data to generate a dual-view feature vector; performing interpretable clustering analysis, dynamically merging similar clusters based on the similarity of cluster points, and generating supplier or equipment classification results containing key decision-making features; selecting target skill packages from a standardized skill package resource pool, constructing an execution process using an orchestration engine, and dynamically scheduling the digital employee model to automatically execute procurement tasks; and using privacy computing technology for data interaction, automatically executing acceptance and settlement, and updating supplier lifecycle credit data. This method improves the accuracy and traceability of procurement decisions, enables flexible configuration and efficient collaboration of the procurement process, significantly reduces deployment costs, and improves the overall response efficiency of the supply chain.

Description

Technical Field

[0001] This application relates to the field of power supply chain operation technology, and in particular to a smart procurement method and system for power supply chain based on digital employee models and skills package orchestration. Background Technology

[0002] Currently, the power supply chain, as a core hub of the energy industry, is widely used for centralized management and control of materials. With the development of digitalization, related technologies, through the collaborative operation of AI large-scale models, RPA process robots, and blockchain, have constructed a full-chain intelligent system covering bidding, evaluation, and settlement. Specifically, this system encompasses key links from unstructured data parsing to multi-dimensional decision analysis, significantly improving the level of business automation.

[0003] However, intelligent procurement solutions based on related technologies often rely on a single black-box model for decision-making, lacking a dual-view fusion mechanism that integrates maintenance and sensor data. This can lead to fluctuations in the accuracy of complex equipment assessments and difficulties in tracing responsibility. Furthermore, the lack of multi-source data standards and rigid skill sets make it difficult to adapt to the diverse needs of SMEs, resulting in severe information silos and low collaboration efficiency. This ultimately hinders the overall agile response and secure controllability of the supply chain. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in the related art.

[0005] Therefore, the first objective of this application is to propose an intelligent procurement method for the power supply chain based on a digital employee model and skills package orchestration.

[0006] The second objective of this application is to propose an intelligent procurement system for the power supply chain based on a digital employee model and skills package orchestration.

[0007] The third objective of this application is to provide a computer-readable storage medium.

[0008] To achieve the above objectives, the first aspect of this application is to propose an intelligent procurement method for the power supply chain based on a digital employee model and skills package orchestration, comprising the following steps:

[0009] Collect multi-source heterogeneous data from the power supply chain, and through intelligent cleaning and unified coding mapping processing, fuse maintenance record data and equipment sensor monitoring data from the multi-source heterogeneous data to generate a dual-view feature vector; Based on the dual-view feature vector, interpretable clustering analysis is performed, and clusters of the same type are dynamically merged according to the similarity of the cluster particles to generate supplier or equipment classification results containing key decision features. Based on the procurement business needs, target skill packages are selected from the standardized skill package resource pool, and the execution process is constructed using the orchestration engine. The procurement task is automatically executed by dynamically scheduling the digital employee model based on a greedy strategy. In the process of cross-enterprise collaboration, data interaction is carried out through privacy computing technology, and acceptance settlement is automatically executed using formally verified smart contracts. Supplier full lifecycle credit data is updated based on the performance evaluation model.

[0010] To achieve the above objectives, a second aspect of this application also proposes an intelligent procurement system for the power supply chain based on a digital employee model and skills package orchestration, comprising the following modules: The acquisition module is used to collect multi-source heterogeneous data from the power supply chain. Through intelligent cleaning and unified coding mapping processing, the maintenance record data and equipment sensor monitoring data in the multi-source heterogeneous data are fused to generate a dual-view feature vector. The generation module is used to perform interpretable clustering analysis based on the dual-view feature vectors, dynamically merge similar clusters according to the similarity of cluster particles, and generate supplier or equipment classification results containing key decision features. The execution module is used to select target skill packages from the standardized skill package resource pool according to the procurement business needs, build the execution process using the orchestration engine, and automatically execute the procurement tasks based on the intelligent employee model with a greedy strategy. The interaction module is used to exchange data through privacy computing technology during cross-enterprise collaboration, automatically execute acceptance and settlement using formally verified smart contracts, and update the supplier's full lifecycle credit data based on the performance evaluation model.

[0011] To achieve the above objectives, a third aspect of this application also proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the intelligent procurement method for the power supply chain based on the digital employee model and skills package orchestration as described in any of the first aspects above.

[0012] To achieve the above objectives, the fourth aspect of this application also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent procurement method for the power supply chain based on the digital employee model and skills package orchestration described in any of the first aspects above.

[0013] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects: This application effectively solves the problems of AI decision-making black boxes and insufficient adaptability to complex equipment evaluation by using dual-view data fusion and interpretable clustering algorithms, significantly improving the accuracy of decision traceability and evaluation accuracy. Based on standardized skill package orchestration and lightweight adaptation mechanisms, it breaks down data silos across multiple systems and reduces deployment costs, achieving efficient collaboration and secure control throughout the entire power supply chain. This application improves the accuracy and traceability of procurement decisions, realizes flexible configuration and efficient collaboration of the procurement process, significantly reduces deployment costs, and improves the overall response efficiency of the supply chain.

[0014] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0015] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating an intelligent procurement method for the power supply chain based on a digital employee model and skills package orchestration, as proposed in this application embodiment; Figure 2 This is a schematic diagram of the structure of an intelligent procurement system for the power supply chain based on a digital employee model and skills package orchestration, as proposed in an embodiment of this application. Detailed Implementation

[0016] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0017] The following description, with reference to the accompanying drawings, illustrates an intelligent procurement method and system for the power supply chain based on a digital employee model and skills package orchestration, as proposed in an embodiment of this application.

[0018] Example 1 Figure 1 This is a flowchart illustrating an intelligent procurement method for the power supply chain based on a digital employee model and skills package orchestration, as proposed in an embodiment of this application. Figure 1 As shown, the method includes the following steps: Step S101: Collect multi-source heterogeneous data from the power supply chain, and through intelligent cleaning and unified coding mapping processing, fuse maintenance record data and equipment sensor monitoring data from the multi-source heterogeneous data to generate a dual-view feature vector.

[0019] Specifically, this step aims to address the technical challenges in the power supply chain, such as inconsistent standards for multi-source data, a high proportion of unstructured information, and incomplete feature representation due to the fragmentation of maintenance and operation data. Its core lies in building a universal data fusion and processing mechanism. First, it collects multi-source heterogeneous data covering business records and physical monitoring through an adaptive interface. Then, it uses intelligent cleaning algorithms to remove redundancy and fill in missing items. Finally, based on a unified coding system, it maps data from different sources to a standard semantic space, eliminating information silos caused by coding differences.

[0020] Based on this, this step adopts a dual-view feature fusion strategy, which integrates maintenance record data reflecting the historical status of the equipment with the equipment sensor monitoring data reflecting the real-time operating conditions in a weighted manner, generating a dual-view feature vector that can comprehensively characterize the features of the object, providing a high-confidence data foundation for subsequent decision-making.

[0021] As a specific implementation method, a multi-source data adapter can be configured to access ERP and IoT systems. A redundancy detection algorithm based on cosine similarity is used to merge duplicate supplier information, and dynamic weights are calculated based on information entropy. Through formula maintenance data feature vector With sensor data feature vector Fusion, where the weights are determined by This ensures that the data structuring rate is increased to over 98%.

[0022] Therefore, by performing the above data processing and fusion steps, the semantic gap between multi-source heterogeneous data can be significantly eliminated, the quality and consistency of input data can be greatly improved, and the generated dual-view feature vector can be ensured to have both historical maintenance depth and real-time monitoring accuracy. This effectively overcomes the decision bias caused by a single data source and lays a solid and interpretable data foundation for subsequent clustering analysis and intelligent scheduling.

[0023] Step S102: Perform interpretable clustering analysis based on dual-view feature vectors, dynamically merge similar clusters according to the similarity of cluster points, and generate supplier or equipment classification results containing key decision features.

[0024] Specifically, interpretable clustering analysis based on dual-view feature vectors aims to solve the black box problem of decision-making caused by the complexity of data dimensions in the power supply chain. Its core lies in building a processing mechanism that can integrate the features of multi-source heterogeneous data and output classification results with clear business meaning.

[0025] In practice, this step first receives a dual-view feature vector formed by fusing maintenance records and equipment sensor monitoring data. Then, a clustering algorithm is used to initially classify suppliers or equipment in the feature space. Subsequently, an interpretable component is introduced to extract the key decision features of each cluster, ensuring that the classification criteria can be traced back to specific business indicators such as equipment failure frequency and production capacity achievement rate.

[0026] Based on this, the system dynamically identifies and merges highly similar clusters by calculating the similarity between particles in different clusters, thereby adaptively optimizing the classification granularity and avoiding over-segmentation or under-segmentation caused by deviations in initial parameter settings, ultimately generating classification results that are both accurate and interpretable.

[0027] As a specific implementation method, an improved K-means algorithm can be used, through the distance formula... Iterative optimization of cluster partitioning, based on the cluster particle similarity formula. Set a threshold, and automatically perform cluster merging operation when the similarity meets the conditions. At the same time, use the decision tree visualization component to output the corresponding key features to enhance the transparency of decision-making.

[0028] Therefore, this step, through dual-view fusion and dynamic cluster merging mechanisms, significantly improves the accuracy and consistency of power material classification and supplier evaluation, effectively overcoming the problem of insufficient adaptability of traditional AI models in complex scenarios. Simultaneously, interpretability analysis provides clear logical support for the clustering results, greatly reducing the cost of manual review and providing a reliable data foundation for subsequent procurement decisions.

[0029] Step S103: Select the target skill package from the standardized skill package resource pool according to the procurement business needs, build the execution process using the orchestration engine, and automatically execute the procurement task based on the intelligent employee model with a greedy strategy.

[0030] Specifically, this step aims to address the issues of poor technology compatibility and high deployment costs in the power supply chain procurement scenario. Its core lies in building an automated execution mechanism based on modular skill encapsulation and dynamic strategy scheduling.

[0031] Specifically, this method pre-encapsulates atomic business capabilities across the entire procurement chain into several standardized skill packages and aggregates them into a resource pool. These skill packages cover functional units of different granularities, from tender document preparation and intelligent bid evaluation to contract settlement. During the execution phase, the system first parses the current procurement business requirements, matches and selects the corresponding target skill packages from the resource pool, and then uses an orchestration engine to combine these discrete skill packages according to the business logic sequence to construct an executable procurement process model.

[0032] Based on this, a greedy strategy is introduced as the core of the dynamic scheduling algorithm. With the goal of minimizing the overall process execution time, the algorithm comprehensively considers the independent execution time of each skill package and the switching overhead between skill packages. It automatically calculates and determines the optimal execution sequence of the digital employee model, thereby driving the digital employee model to automatically complete the procurement task.

[0033] As a specific implementation method, capabilities such as AI bidding and RPA contract processing can be broken down into basic or advanced skill packages. The process can then be configured through a visual orchestration engine and based on the objective function. Perform scheduling, among which Indicates the first The execution time of each skill pack Indicates skill pack arrive The switching time.

[0034] Therefore, by employing a combination of standardized skill package orchestration and greedy strategy-based dynamic scheduling, this step significantly improves the flexibility and efficiency of the procurement process. On one hand, the modular skill package packaging allows the system to be combined on demand according to the actual needs of enterprises of different sizes, greatly reducing the technical deployment threshold and cost. On the other hand, the dynamic scheduling mechanism based on the greedy strategy effectively optimizes the task execution order, reduces the time loss caused by skill switching, and ensures the efficiency and consistency of complex procurement tasks under the execution of the digital employee model.

[0035] Step S104: During cross-enterprise collaboration, data interaction is conducted through privacy computing technology, and acceptance settlement is automatically executed using formally verified smart contracts. Supplier full lifecycle credit data is updated based on the performance evaluation model.

[0036] Specifically, in the process of cross-enterprise collaboration, a trusted data interaction environment is built to achieve secure information sharing among multiple participating entities. Privacy computing technology is used to complete necessary data interaction and joint computing while ensuring that the original data does not leave the domain, thereby resolving the contradiction between protecting trade secrets and business collaboration.

[0037] Based on this, formally verified smart contracts are invoked to automatically trigger and execute the material acceptance and fund settlement processes. Mathematical logic proofs ensure that the contract code is free of logical conflicts and permission overflow vulnerabilities, achieving automated and tamper-proof transaction execution. Simultaneously, a pre-defined performance evaluation model is used to quantitatively assess the supplier's performance during the collaboration process, dynamically updating their full lifecycle credit data to form a closed-loop credit management system.

[0038] As a specific implementation method, a privacy protection scheme combining homomorphic encryption and zero-knowledge proofs can be adopted, ensuring that only the calculation results are exposed during data sharing without disclosing sensitive information such as supplier quotations. This scheme is also compatible with various domestic consortium blockchain architectures to enhance high-frequency transaction processing capabilities. Formal verification algorithms are embedded in the smart contract auditing process to automatically detect code vulnerabilities. In the credit update process, a formula is used... The system comprehensively calculates capacity compliance rate, on-time delivery rate, credit score, and digital adaptation, and updates supplier credit status in real time.

[0039] Therefore, this step, through the combination of privacy-preserving computation and formal verification technologies, significantly reduces the risk of information leakage and security vulnerabilities in cross-enterprise data interaction, ensuring data sovereignty and transaction credibility in the supply chain collaboration process. Simultaneously, dynamically updating credit data based on a multi-dimensional performance evaluation model effectively improves the real-time perception and risk warning accuracy of supplier performance capabilities, promoting a virtuous cycle and efficient collaboration within the supply chain ecosystem.

[0040] Example 2 Based on the above embodiments, this embodiment provides a detailed description of the specific implementation of collecting multi-source heterogeneous data from the power supply chain in step S101, and fusing maintenance record data and equipment sensor monitoring data to generate dual-view feature vectors through intelligent cleaning and unified coding mapping.

[0041] In this embodiment, maintenance record data and equipment sensor monitoring data are fused to generate a dual-view feature vector through intelligent cleaning and unified coding mapping. This includes: accessing mainstream power industry system data through a multi-source data adapter, and converting and semantically extracting unstructured tender documents and contract execution formats to obtain a standardized original dataset; performing cosine similarity-based redundancy detection on supplier registration information in the standardized original dataset, and automatically merging duplicate data when the similarity calculation result exceeds a preset threshold to obtain deduplicated cleaned data; performing unified coding mapping on the deduplicated cleaned data based on the power material classification standard, and weightedly fusing the mapped maintenance record data feature vector with the equipment sensor monitoring data feature vector to generate a dual-view feature vector.

[0042] Specifically, firstly, the system accesses data from twelve mainstream power industry systems, including ERP, WMS, and bidding platforms, through a multi-source data adapter to obtain raw data streams containing unstructured tender documents and contracts as input sources. The processing actions include performing format conversion and semantic extraction based on natural language processing on unstructured documents, transforming them into machine-readable structured fields. The output is a standardized raw dataset, which unifies the data interface format and provides a foundation for subsequent processing.

[0043] Next, using supplier registration information from the standardized raw dataset as input, the system extracts supplier registration codes, qualification information, and production capacity data to construct the first and second vectors respectively, and then applies the formula... Calculate the cosine similarity between two vectors, where, a i Let be the first vector. b i This is the second vector. The processing action involves comparing the calculated similarity value with a preset similarity threshold of 0.85. When the value is greater than or equal to the threshold, the corresponding supplier information is determined to be duplicate data, and an automatic merging operation is performed to remove redundant records. The output result is the cleaned data after deduplication, which effectively solves the problem of decision-making misjudgment caused by duplicate information.

[0044] Then, using the cleaned data after deduplication as input, the system loads power material classification standards, such as GB / T14684-2021, as the baseline coding system. A semantic mapping algorithm is used to identify differences in material coding within the data. The processing involves automatically mapping the identified differences to standard codes within the baseline coding system, eliminating the risk of duplicate procurement caused by inconsistent coding. Furthermore, feature vectors from maintenance records and equipment sensor monitoring data are separated. The final processing involves weighted fusion of these two feature vectors to generate a dual-view feature vector as output. This vector simultaneously integrates equipment maintenance history and real-time operating status, providing high-dimensional feature input for subsequent interpretable clustering analysis.

[0045] Therefore, this implementation significantly improves the data structuring rate through multi-source adaptation and semantic extraction, greatly reduces the duplication rate of supplier information by using redundancy detection based on cosine similarity, and eliminates material coding differences by combining unified coding mapping. The resulting dual-view feature vector effectively integrates multi-source heterogeneous data, laying a solid data foundation for improving the accuracy and interpretability of procurement decisions.

[0046] Example 3 Based on the above embodiments, this embodiment provides a detailed description of the specific implementation method of performing interpretable clustering analysis based on dual-view feature vectors in step S102, dynamically merging similar clusters according to the similarity of cluster point particles, and generating supplier or equipment classification results containing key decision features.

[0047] In this embodiment, interpretable clustering analysis is performed based on dual-view feature vectors, and clusters of the same type are dynamically merged according to the similarity of the cluster particles. This includes: calculating the information entropy values ​​of maintenance data and sensor data based on information entropy, and using the formula... Calculate dynamic weights using the formula Generate a unified feature space, where,F M For maintenance data feature vectors, F S For sensor data feature vectors, For dynamic weights, H Information entropy; Improved K-means two-view clustering is performed in a unified feature space, using the distance formula. Iterative optimization of cluster partitioning yields an initial set of clusters; the similarity of particles between any two clusters in the initial set is calculated, and when the similarity satisfies the formula... It automatically merges the corresponding clusters and their respective datasets to generate supplier or equipment classification results containing key decision features.

[0048] Specifically, firstly, the system input source is the preprocessed feature vector of maintenance record data. Feature vector of monitoring data from equipment sensors The processing steps include calculating the information entropy values ​​for the two types of data respectively. and Using the formula Calculate the dynamic weights that reflect the reliability of the data. .

[0049] Then according to the formula The two types of features are weighted and fused. The output of this step is a dual-view fused feature vector in a unified feature space. This eliminates the bias of relying on a single data source.

[0050] Next, using this unified feature vector As input, improved K-means two-view clustering is performed in a unified feature space. The processing action is based on the distance formula. Iterative calculation of sample points and the centers of each cluster The Euclidean distance is used to continuously adjust the cluster partitioning until convergence. The output of this step is the initial set of clusters.

[0051] Furthermore, taking any two clusters from the initial cluster set... and The centroid eigenvectors are used as input to calculate the similarity of their mass points. The processing action is to apply the formula... Cosine similarity is calculated, and the result is checked against a threshold of 0.9. If the threshold is met, the corresponding clusters and their respective datasets are automatically merged. The output of this step is the optimized supplier or device classification result.

[0052] Finally, to enhance decision interpretability, the system inputs the optimized classification results into the decision tree visualization component. The processing step involves extracting and outputting key decision features for each cluster, specifically indicators such as equipment failure frequency and production capacity achievement rate. The final output of this step is an interpretable classification report with over 90% traceability accuracy, clearly defining the classification criteria.

[0053] Therefore, this implementation method effectively solves the evaluation bias problem caused by a single data source by dynamically weighting and fusing multi-source heterogeneous data through information entropy. At the same time, the adaptive cluster merging mechanism based on the similarity of mass points significantly improves the consistency of classifying similar equipment or suppliers. Combined with the key features of the decision tree visualization output, it achieves a high degree of transparency and traceability in the procurement decision-making process.

[0054] Example 4 Based on the above embodiments, this embodiment describes in detail the specific implementation method of step S103 above, which involves selecting target skill packages from the standardized skill package resource pool according to procurement business needs, constructing an execution process using an orchestration engine, and automatically executing procurement tasks by dynamically scheduling the intelligent employee model based on a greedy strategy.

[0055] In one embodiment of this application, a target skill package is selected from a standardized skill package resource pool based on procurement business needs. An orchestration engine is used to construct an execution process, and a digital employee model is dynamically scheduled to automatically execute procurement tasks based on a greedy strategy. This includes: breaking down AI bidding, RPA contract processing, and blockchain traceability capabilities into standardized skill packages and storing them in a skill package resource pool; configuring the procurement business process through a visual orchestration engine; and constructing an objective function based on a greedy algorithm. ,in, For the first k Execution time of each skill package To optimize the process efficiency, the execution order of skill packages is determined based on the switching time from skill package i to j. The digital employee model is dynamically scheduled to call the corresponding target skill package according to the determined execution order, automatically execute bidding, bid evaluation, contract signing and settlement tasks, and provide real-time feedback on the execution progress.

[0056] Specifically, this embodiment first uses pre-packaged AI bidding, RPA contract processing, and blockchain traceability capabilities as input sources. These capabilities have been broken down into more than 20 standardized skill packages, including basic and advanced versions, and stored in the skill package resource pool. The processing action is to receive the procurement business process instructions configured by the user through a visual orchestration engine, and automatically match the skill permissions of the digital employee model according to the enterprise size.

[0057] Specifically, only the basic bidding skills package and simple settlement skills package are used for small and medium-sized enterprises, while the full-link collaboration skills package is used for large enterprises. The output of this step is a subset of skills packages adapted to the specific needs of the enterprise and an initial business process topology, thereby achieving a lightweight application with a 70% reduction in deployment costs.

[0058] Next, using the aforementioned subset of skill packages and their estimated execution parameters as input, the processing action is based on a greedy algorithm to construct the objective function. The algorithm iteratively calculates and finds the skill pack sequence that minimizes the total process time T; the output of this step is a list of optimal skill pack execution order.

[0059] Finally, using the optimal execution order list as input, the processing action is to dynamically schedule the intelligent employee model to call the corresponding target skill packages in sequence, automatically execute the bidding, evaluation, contract signing and settlement tasks in turn, and collect execution status data in real time at each node; the output of this step is the complete execution of the procurement business results and real-time feedback of execution progress information, ensuring that the output data of the previous task is seamlessly transferred to the next task as input, forming a logical closed loop.

[0060] Therefore, this implementation method, by combining skill package decomposition with dynamic permission matching, significantly reduces the threshold and cost of digital deployment for SMEs. At the same time, by using a greedy algorithm to optimize the execution sequence, it effectively reduces the time spent switching skill packages and improves the overall execution efficiency and response speed of power supply chain procurement tasks.

[0061] Example 5 Based on the above embodiments, this embodiment describes in detail the specific implementation method of using privacy computing technology to perform data interaction in the cross-enterprise collaboration process in step S104, using formally verified smart contracts to automatically execute acceptance settlement, and updating the supplier's full life cycle credit data based on the performance evaluation model.

[0062] In one embodiment of this application, data interaction is performed using privacy-preserving computation technology during cross-enterprise collaboration. Formalized smart contracts are used to automatically execute acceptance and settlement, and supplier lifecycle credit data is updated based on a performance evaluation model. This includes: employing a combination of homomorphic encryption and zero-knowledge proofs for data interaction during cross-enterprise collaboration to ensure that computation results are exposed without revealing the original supplier quotes and base prices; embedding formal verification algorithms into the smart contract code to automatically detect logical conflicts and permission overflow vulnerabilities; and automatically executing order acceptance and payment operations using verified smart contracts; based on the formula... Calculate supplier performance score, where, The capacity compliance rate is verified by on-chain production data. To improve on-time performance, For credit scores recorded on the blockchain, To digitally adapt the interface integration capabilities, and update the supplier's full lifecycle credit data based on the supplier performance score.

[0063] Specifically, this embodiment first involves secure data interaction during cross-enterprise collaboration. The input source consists of sensitive business data to be shared by the participating enterprise nodes, specifically supplier quotation information and minimum purchase price information. The processing employs a combination of homomorphic encryption and zero-knowledge proof to encrypt the original data and generate verifiable proof credentials, enabling the data to be calculated and verified in an encrypted state. The output is interactive data that only exposes the calculation and comparison results without completely revealing the original supplier quotation and minimum purchase price information, thereby ensuring privacy and security during data interaction.

[0064] Next, based on the aforementioned secure interactive data, the automatic execution mechanism of the smart contract is initiated. The input sources are order status data verified through privacy computation and pre-built smart contract code. The processing involves embedding a formal verification algorithm into the smart contract code before contract deployment to automatically detect logical conflicts and permission overflow vulnerabilities. Contract execution logic is triggered only after successful verification, automatically comparing order acceptance conditions and executing payment instructions. The output is a transaction record confirming automatic acceptance and fund settlement, which is simultaneously stored on the blockchain.

[0065] Subsequently, the supplier's full lifecycle credit data is updated based on the settlement results. Input sources include the capacity compliance rate obtained from on-chain production data verification. On-time performance rate Credit scores recorded on the blockchain and digital adaptability Among them, the digital adaptation for SMEs Data is acquired through a lightweight data acquisition plugin, supporting direct import of Excel data or the use of OCR technology to recognize unstructured documents. It can be integrated into the supply chain collaboration portal without modifying existing systems. The processing action involves substituting the aforementioned four indicators into the formula. Perform a weighted calculation. The output is the final supplier performance score. The system updates the supplier's full lifecycle credit database in real time based on this score, serving as the basis for subsequent procurement decisions.

[0066] Therefore, this implementation method effectively addresses the privacy leakage risks in cross-enterprise data sharing through a combination of homomorphic encryption and zero-knowledge proofs, while formal verification ensures the security and reliability of smart contract execution. By combining a lightweight data acquisition plugin with a multi-dimensional performance evaluation formula, it significantly improves the data access rate for SMEs and the accuracy of supplier evaluation, achieving an efficient and secure closed loop for supply chain collaboration.

[0067] Example 6 This embodiment will provide a detailed description of the complete implementation of the intelligent procurement method for the power supply chain based on the digital employee model and skill package orchestration, focusing on the specific implementation details of the entire chain from data access, feature fusion, clustering decision-making, skill orchestration to collaborative settlement.

[0068] In this embodiment, the method first constructs a five-layer integrated intelligent procurement automation architecture, consisting of a data adaptation layer, a skill package resource pool, an orchestration engine layer, a digital employee execution layer, and an application interaction layer, from bottom to top. The data adaptation layer is responsible for connecting to 12 mainstream business systems in the power industry, including Enterprise Resource Planning (ERP) systems, Warehouse Management Systems (WMS), and various electronic bidding platforms, solving the information silo problem through multi-source data adapters. The skill package resource pool pre-configures over 20 standardized skill packages, covering core capabilities such as AI bidding evaluation, RPA contract processing, and blockchain traceability, and is divided into basic and advanced versions to suit the needs of enterprises of different sizes. The orchestration engine layer provides a visual process configuration interface based on the BPMN 2.0 standard, supporting drag-and-drop operations. The digital employee execution layer automatically calls the corresponding skill packages to complete tasks according to orchestration instructions, while the application interaction layer provides users with progress feedback and decision traceability views.

[0069] Specifically, in the data access and preprocessing stage, the system collects multi-source heterogeneous data from the power supply chain through a multi-source data adapter. For unstructured tender documents and contract texts, the system performs format conversion and semantic extraction operations to transform them into structured data, increasing the overall data structuring rate to over 98%. Subsequently, the system performs redundancy detection on the supplier registration information in the standardized raw dataset. Specifically, it extracts the supplier's registration code, qualification information, and capacity data to construct the first vector. With the second vector , where vector elements and These represent numerical features such as registration code, qualifications, and production capacity. The system calculates the similarity between two vectors based on the cosine similarity formula:

[0070] In this embodiment, the similarity threshold is set to 0.85. When the calculated similarity value is greater than or equal to this threshold, the system determines that the two records are duplicate data and performs an automatic merging operation to obtain cleaned data after deduplication, effectively reducing the duplication rate of supplier information from 15% to below 1%.

[0071] Furthermore, to eliminate the risk of duplicate procurement due to inconsistent material coding, the system uses the power material classification standard GB / T 14684-2021 as the baseline coding system. A semantic mapping algorithm identifies differences in material coding within the cleaned, deduplicated data, automatically mapping the identified differences to the standard codes in the baseline coding system. This process generates mapped feature vectors for maintenance record data and equipment sensor monitoring data, laying the foundation for subsequent dual-view fusion.

[0072] During the dual-view feature fusion and clustering analysis phase, the system synchronously accesses maintenance data (including fault records, repair cycles, etc.) and sensor data (including equipment operating parameters, energy consumption, etc.). First, the information entropy values ​​of the maintenance data and sensor data are calculated based on information entropy, denoted as follows: and Calculate dynamic weights using the entropy weight method. The calculation formula is as follows:

[0073] in, This reflects the contribution of data reliability to the fusion result. Subsequently, a unified feature space is generated based on the dual-view feature fusion formula. :

[0074] In the formula, For maintenance data feature vectors, The data consists of sensor data feature vectors. Within the generated unified feature space, the system executes an improved K-means two-view clustering algorithm. After initializing the cluster centers, the cluster partitioning is iteratively optimized using a distance formula, defined as:

[0075] in, For sample points, For the first The system calculates the centers of each cluster. After obtaining the initial set of clusters, the system calculates the centers of any two clusters. and The particle similarity is calculated using the following formula:

[0076] When the similarity satisfies At that time, the system automatically merges the corresponding clusters and their respective datasets. This adaptive cluster merging strategy improves decision-making consistency for similar equipment or suppliers. After clustering is completed, the system outputs key decision features corresponding to the clustering results through the decision tree visualization component, such as equipment failure frequency and production capacity achievement rate, so that the decision traceability accuracy reaches more than 90%, effectively solving the "black box" problem of AI models.

[0077] In the skill package orchestration and dynamic scheduling phase, users configure procurement business processes through a visual orchestration engine. Typical processes include bidding, bid evaluation, contract signing, and settlement. The system automatically matches skill permissions of the digital employee model based on the enterprise's size: for SMEs, only the "Basic Bidding" and "Simple Settlement" skill packages are enabled, achieving lightweight application and reducing deployment costs by 70%; for large enterprises, the full-link collaborative skill package is enabled. After determining the target skill package, the system constructs an objective function based on a greedy algorithm to determine the optimal execution order. The objective function is defined as:

[0078] in, For the first The execution time of each skill pack For skill pack arrive The switching time is limited. The system aims for optimal process efficiency by dynamically scheduling the digital employee model to call the corresponding target skill packages, automatically executing various procurement tasks, and providing real-time feedback on the execution progress to the application interaction layer.

[0079] In terms of cross-enterprise collaboration and security management, this embodiment adopts consortium blockchain technology, compatible with domestic blockchain platforms such as Chang'an Chain and Huadian Chain, supporting the on-chain recording of data such as equipment traceability and performance records, and increasing the transaction processing capacity (TPS) to over 2000 to meet the needs of high-frequency transactions. During data interaction, a combination of "homomorphic encryption + zero-knowledge proof" is used to ensure that only the calculation results are exposed without disclosing original sensitive information such as supplier quotations and base prices, reducing the risk of sensitive information leakage by 95%. For smart contracts, the system embeds a formal verification algorithm to automatically detect logical conflicts and permission overflow vulnerabilities in the code, with a vulnerability identification rate of 98%. Verified smart contracts will automatically execute order acceptance and payment operations, significantly shortening the settlement cycle.

[0080] Finally, in the supplier lifecycle management module, the system updates supplier credit data in real time based on the performance evaluation model. Specifically, it acquires Excel import data or OCR recognition data from SMEs through a lightweight data acquisition plugin, allowing them to connect to the supply chain collaboration portal without modifying their existing systems, thereby obtaining digital adaptation data. Simultaneously, the system obtains the production capacity compliance rate verified by production data from the blockchain. On-time performance rate And credit scores recorded on the blockchain Substitute the above four indicators into the supplier performance scoring formula for weighted calculation:

[0081] Calculated supplier performance score This will be directly used to update suppliers' full lifecycle credit data, serving as an important basis for subsequent procurement decisions. The assessment model improves the accuracy of supplier performance evaluation to 90% and reduces the project delay rate from 15% to 3%.

[0082] In summary, this embodiment achieves intelligent and automated procurement across the entire power supply chain through the deep integration of a digital employee model and skill package orchestration. By employing dual-view feature fusion and interpretable clustering analysis, it significantly improves the accuracy of complex equipment evaluation and decision-making transparency. Flexible orchestration and greedy scheduling of standardized skill packages lower the technical application threshold and deployment costs for SMEs. Privacy-preserving computation and formally verified smart contracts ensure data security and transaction credibility in cross-enterprise collaboration. The overall solution improves procurement efficiency by over 60% and reduces procurement costs by over 25%, effectively addressing technical issues in existing technologies such as poor data quality, difficult system integration, numerous security risks, and cost-benefit imbalances.

[0083] To achieve the above embodiments, this application also proposes an intelligent procurement system for the power supply chain based on a digital employee model and skills package orchestration. Figure 2 This is a schematic diagram of the structure of an intelligent procurement system for the power supply chain based on a digital employee model and skill package orchestration, as proposed in an embodiment of this application. Figure 2 As shown, the system includes: The acquisition module 100 is used to acquire multi-source heterogeneous data from the power supply chain. Through intelligent cleaning and unified coding mapping processing, the maintenance record data and equipment sensor monitoring data in the multi-source heterogeneous data are fused to generate a dual-view feature vector. The generation module 200 is used to perform interpretable clustering analysis based on the dual-view feature vector, dynamically merge similar clusters according to the similarity of cluster particles, and generate supplier or equipment classification results containing key decision features. The execution module 300 is used to select target skill packages from the standardized skill package resource pool according to the procurement business needs, build the execution process using the orchestration engine, and automatically execute the procurement task based on the dynamic scheduling of the digital employee model using a greedy strategy. Interaction module 400 is used to exchange data through privacy computing technology during cross-enterprise collaboration, automatically execute acceptance settlement using formally verified smart contracts, and update supplier full lifecycle credit data based on performance evaluation models.

[0084] It should be noted that the foregoing explanation of the embodiment of the intelligent procurement method for the power supply chain based on the digital employee model and skill package orchestration also applies to the system of this embodiment, and will not be repeated here.

[0085] In summary, the intelligent procurement system for the power supply chain based on the digital employee model and skills package orchestration in this application embodiment improves the accuracy and traceability of procurement decisions, realizes flexible configuration and efficient collaboration of the procurement process, significantly reduces deployment costs and improves the overall response efficiency of the supply chain.

[0086] To implement the above embodiments, this application also proposes an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the intelligent procurement method for the power supply chain based on the digital employee model and skills package orchestration as described in any of the first aspect embodiments above.

[0087] To implement the above embodiments, this application also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the intelligent procurement method for the power supply chain based on the digital employee model and skill package orchestration as described in any one of the first aspect embodiments above.

[0088] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0089] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0090] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0091] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0092] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0093] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0094] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0095] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A smart procurement method for the power supply chain based on a digital employee model and skills package orchestration, characterized in that, Includes the following steps: Collect multi-source heterogeneous data from the power supply chain, and through intelligent cleaning and unified coding mapping processing, fuse maintenance record data and equipment sensor monitoring data from the multi-source heterogeneous data to generate a dual-view feature vector; Based on the dual-view feature vector, interpretable clustering analysis is performed, and clusters of the same type are dynamically merged according to the similarity of the cluster particles to generate supplier or equipment classification results containing key decision features. Based on the procurement business needs, target skill packages are selected from the standardized skill package resource pool, and the execution process is constructed using the orchestration engine. The procurement task is automatically executed by dynamically scheduling the digital employee model based on a greedy strategy. In the process of cross-enterprise collaboration, data interaction is carried out through privacy computing technology, and acceptance settlement is automatically executed using formally verified smart contracts. Supplier full lifecycle credit data is updated based on the performance evaluation model.

2. The method according to claim 1, characterized in that, The process of intelligent cleaning and unified coding mapping integrates maintenance record data with equipment sensor monitoring data to generate a dual-view feature vector, including: By accessing mainstream power industry system data through multi-source data adapters, and converting and semantically extracting unstructured tender documents and contract execution formats, a standardized raw dataset is obtained. Redundancy detection based on cosine similarity is performed on the supplier registration information in the standardized original dataset. When the similarity calculation result exceeds a preset threshold, duplicate data is automatically merged to obtain cleaned data after deduplication. Based on the power material classification standard, the deduplicated cleaning data is subjected to unified encoding mapping, and the feature vector of the mapped maintenance record data is weighted and fused with the feature vector of the equipment sensor monitoring data to generate the dual-view feature vector.

3. The method according to claim 2, characterized in that, The process involves performing cosine similarity-based redundancy detection on the supplier registration information in the standardized original dataset, and automatically merging duplicate data when the similarity calculation result exceeds a preset threshold, including: Extract supplier registration codes, qualification information, and production capacity data to construct a first vector and a second vector, and calculate the similarity between the first vector and the second vector using the following formula: in, a i Let be the first vector. b i The second vector; The similarity value is compared with a preset similarity threshold. When the similarity value is greater than or equal to the similarity threshold, it is determined to be duplicate data and an automatic merging operation is performed.

4. The method according to claim 1, characterized in that, The step of performing interpretable clustering analysis based on the dual-view feature vectors, and dynamically merging similar clusters according to the similarity of cluster particles, includes: The information entropy values ​​of maintenance data and sensor data are calculated based on information entropy, and then the formula is used. Calculate dynamic weights using the formula Generate a unified feature space, where, F M For maintenance data feature vectors, F S For sensor data feature vectors, For dynamic weights, H Information entropy; Improved K-means two-view clustering is performed in the unified feature space, using the distance formula. Iterative optimization of cluster partitioning yields an initial set of clusters. Calculate the particle similarity between any two clusters in the initial cluster set. When the similarity satisfies the formula... It automatically merges the corresponding clusters and their respective datasets to generate supplier or equipment classification results containing key decision features.

5. The method according to claim 1, characterized in that, The process of selecting target skill packages from a standardized skill package resource pool based on procurement business needs, constructing an execution flow using an orchestration engine, and automatically executing procurement tasks based on a greedy strategy and dynamic scheduling of a digital employee model includes: The capabilities of AI bidding, RPA contract processing, and blockchain traceability are broken down into standardized skill packages and stored in a skill package resource pool. The procurement business process is then configured through a visual orchestration engine. Constructing the objective function based on a greedy algorithm ,in, For the first k Execution time of each skill package To determine the execution order of skill packs based on the switching time from skill pack i to j, with the goal of optimizing process efficiency; Based on the determined execution order, the digital employee model dynamically schedules the corresponding target skill packages to automatically execute bidding, evaluation, contract signing and settlement tasks, and provides real-time feedback on the execution progress.

6. The method according to claim 1, characterized in that, The process of data interaction through privacy computing technology in cross-enterprise collaboration, automatic execution of acceptance and settlement using formally verified smart contracts, and updating of supplier lifecycle credit data based on a performance evaluation model include: In cross-enterprise collaboration, a combination of homomorphic encryption and zero-knowledge proof is used for data interaction to ensure that the calculation results are exposed without revealing the original information of supplier quotations and bottom prices. Formal verification algorithms are embedded in smart contract code to automatically detect logical conflicts and privilege overflow vulnerabilities, and the verified smart contracts are used to automatically execute order acceptance and payment operations. According to the formula Calculate supplier performance score, where, The capacity compliance rate is verified by on-chain production data. To improve on-time performance, For credit scores recorded on the blockchain, To digitally adapt the interface integration capabilities, and update the supplier's full lifecycle credit data based on the supplier performance score.

7. The method according to claim 6, characterized in that, The calculation of the supplier performance score includes: Acquire Excel import data or OCR recognition data from SMEs through lightweight data acquisition plugins, and connect to the supply chain collaboration portal to obtain digital adaptation data. The on-chain production data verification results in the capacity achievement rate, on-time delivery rate, on-chain credit score, and digital adaptation data are substituted into the formula. A weighted calculation is performed to obtain the supplier performance score.

8. The method according to claim 2, characterized in that, The process of performing unified encoding mapping on the deduplicated cleaned data based on the power material classification standard includes: The relevant power material classification standards are loaded as the benchmark coding system, and the differences in material codes in the deduplicated clean data are identified through a semantic mapping algorithm. The identified differential codes are automatically mapped to the standard codes in the benchmark coding system, eliminating the risk of duplicate procurement caused by inconsistent codes, and obtaining the mapped feature vectors of maintenance record data and equipment sensor monitoring data.

9. A smart procurement system for the power supply chain based on a digital employee model and skills package orchestration, characterized in that, Includes the following modules: The acquisition module is used to collect multi-source heterogeneous data from the power supply chain. Through intelligent cleaning and unified coding mapping processing, the maintenance record data and equipment sensor monitoring data in the multi-source heterogeneous data are fused to generate a dual-view feature vector. The generation module is used to perform interpretable clustering analysis based on the dual-view feature vectors, dynamically merge similar clusters according to the similarity of cluster particles, and generate supplier or equipment classification results containing key decision features. The execution module is used to select target skill packages from the standardized skill package resource pool according to the procurement business needs, build the execution process using the orchestration engine, and automatically execute the procurement tasks based on the intelligent employee model with a greedy strategy. The interaction module is used to exchange data through privacy computing technology during cross-enterprise collaboration, automatically execute acceptance and settlement using formally verified smart contracts, and update the supplier's full lifecycle credit data based on the performance evaluation model.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent procurement method for the power supply chain based on the digital employee model and skill package orchestration as described in any one of claims 1-8.

Images