Business requirement disassembling method and system integrating fine tuning small model and knowledge retrieval

Abstract

The invention discloses a business requirement disassembling method and system integrating fine tuning of a small model and knowledge retrieval. The method comprises the following steps: receiving a business demand text expressed in a natural language; performing business intention recognition on the business demand text through a fine tuning small model to obtain a preliminary business intention; knowledge retrieval is started, related data are extracted from an internal knowledge base of the power enterprise to optimize the preliminary business intention, the power business intention is obtained, and the knowledge retrieval comprises at least one of vector embedding retrieval, rule matching and retrieval enhancement generation; and reasoning based on the power business intention to generate a structured business instruction chain. According to the technical scheme provided by the invention, the business demand is cooperatively processed through fine adjustment of the small model and knowledge retrieval, the intention recognition precision and the accuracy of the business instruction chain are improved, intelligent and structured disassembly of the business demand is realized, and the business automation degree and response efficiency of a power enterprise are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of business task identification and automatic decomposition of electric power enterprises, and in particular to a business demand decomposition method and system integrating fine-tuning of a small model and knowledge retrieval. Background Art

[0002] With the continued advancement of enterprise digital transformation, demand for business automation and intelligent management is increasing across various industries. This is especially true in specialized fields like electric power, where daily operations involve a large amount of demand information expressed in natural language, such as equipment maintenance, troubleshooting, and task dispatching. Traditionally, the analysis and breakdown of business needs has relied heavily on manual judgment and accumulated experience, resulting in low analysis efficiency and significant limitations on personnel expertise. Furthermore, since electric power companies typically have standardized operating instructions and extensive business knowledge bases, effectively leveraging this knowledge to assist in business needs analysis has become key to driving business automation.

[0003] In recent years, the development of artificial intelligence (AI), particularly the rise of fine-tuned language models (such as small-scale pre-trained language models) and knowledge retrieval technologies (such as vector embedding retrieval, rule matching, and retrieval enhancement generation), has provided new solutions to these problems. While existing technologies have gradually applied natural language processing and knowledge retrieval technologies to different fields, there are still limitations and gaps in the synergistic integration of fine-tuned small models with knowledge retrieval mechanisms, especially in the automated decomposition of highly standardized business requirements within the power industry. In particular, there is a lack of efficient, accurate, and domain-specific technical solutions for effectively leveraging existing internal knowledge resources to accurately identify business intent and invoke instructions. Properly addressing these issues has become a pressing issue for the industry. Summary of the Invention

[0004] The present invention provides a business demand decomposition method and system that integrates fine-tuning of small models and knowledge retrieval, so as to improve the accuracy and efficiency of business intent recognition, realize automatic and precise decomposition of business demands, and enhance the management efficiency of power enterprises.

[0005] According to a first aspect of the present invention, a method for decomposing business requirements by integrating fine-tuning a small model and knowledge retrieval is provided. The method comprises: Receive business requirement text expressed in natural language; By fine-tuning the small model, the business intent of the business requirement text is identified to obtain preliminary business intent; Initiate knowledge retrieval to extract relevant information from the internal knowledge base of the power enterprise to optimize the preliminary business intent and obtain the power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching, and retrieval enhancement generation; Reasoning is performed based on the power business intention to generate a structured business instruction chain.

[0006] In one embodiment, it further includes: Fine-tuning the pre-trained natural language model on corpus in the field of power equipment operation and maintenance management to obtain the fine-tuned small model; The preliminary business intent in the business requirement text is identified by the fine-tuned small model.

[0007] In one embodiment, the knowledge retrieval includes vector embedding retrieval, including: Convert the business requirement text into a vector representation; Based on the vector representation, a similarity search is performed in the vector space of the internal knowledge base of the electric power enterprise to extract information related to the business requirement text.

[0008] In one embodiment, the knowledge retrieval process includes rule matching retrieval, including: Match key content in the business requirement text through predefined rules or templates; Based on the matching results, the corresponding knowledge entries are retrieved from the internal knowledge base of the power enterprise.

[0009] In one embodiment, the knowledge retrieval process includes retrieval enhancement generation, including: Performing a preliminary analysis of the business requirement text using the fine-tuned small model; Retrieving relevant supplementary information from the internal knowledge base of the power enterprise based on the preliminary analysis results; The retrieved supplementary information is provided to the fine-tuned mini-model to enhance the understanding of the business requirements and assist in generating the structured business instruction chain.

[0010] In one embodiment, it further includes: Providing the generated structured business instruction chain in the form of an instruction identifier sequence; According to the instruction identification sequence, a pre-stored specific instruction is called from an instruction registration center.

[0011] According to a second aspect of the present invention, a business requirement decomposition system integrating fine-tuning of a small model and knowledge retrieval is provided, comprising: A receiving module, used for receiving a business requirement text expressed in a natural language; An acquisition module is used to identify the business intent of the business requirement text by fine-tuning a small model to obtain preliminary business intent; A retrieval module is used to initiate knowledge retrieval, extract relevant information from the internal knowledge base of the power enterprise to optimize the preliminary business intent, and obtain the power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching, and retrieval enhancement generation; A generation module is used to perform reasoning based on the power business intention and generate a structured business instruction chain.

[0012] In one embodiment, the receiving module, the acquiring module, the retrieving module and the generating module are controlled to execute any one of the above-mentioned business requirement decomposition methods integrating fine-tuning of small models and knowledge retrieval.

[0013] According to a third aspect of the present invention, there is provided an electronic device, the electronic device comprising: a communication interface, a processor, and a memory; Wherein, the memory is used to store program instructions, and when the program instructions are executed by the processor that is communicatively connected to the memory through the communication interface, any of the above-mentioned business demand decomposition methods integrating fine-tuning of small models and knowledge retrieval is implemented.

[0014] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a computer (for example, a processor in a computer), any of the above-mentioned business requirement decomposition methods integrating fine-tuning of a small model and knowledge retrieval is implemented.

[0015] In summary, the present invention provides a method and system for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval, the method comprising: receiving a business requirement text expressed in natural language; identifying the business intent of the business requirement text by fine-tuning a small model to obtain a preliminary business intent; initiating knowledge retrieval, extracting relevant information from the internal knowledge base of the electric power enterprise to optimize the preliminary business intent, and obtaining the electric power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching, and retrieval enhancement generation; reasoning based on the electric power business intent to generate a structured business instruction chain. The technical solution of the present application collaboratively processes business requirements through fine-tuning a small model and knowledge retrieval, improves the accuracy of intent recognition and the accuracy of the business instruction chain, realizes intelligent and structured decomposition of business requirements, and improves the business automation level and response efficiency of the electric power enterprise.

[0016] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0017] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 A flowchart of a business requirement decomposition method that integrates fine-tuning of a small model and knowledge retrieval provided by an embodiment of the present invention; Figure 2 A flowchart of another method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval provided by an embodiment of the present invention; Figure 3 A flowchart of another method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval provided by an embodiment of the present invention; Figure 4 A flowchart of another method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval provided by an embodiment of the present invention; Figure 5 A flowchart of another method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval provided by an embodiment of the present invention; Figure 6 A flowchart of another method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval provided by an embodiment of the present invention; Figure 7 A structural diagram of a business requirement decomposition system integrating fine-tuning of small models and knowledge retrieval provided by an embodiment of the present invention; Figure 8 A structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present application by illustrating the examples of the present application.

[0021] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, the elements defined by the phrase "comprising..." do not exclude the presence of other identical elements in the process, method, article, or device comprising the elements.

[0022] like Figure 1 As shown, the present invention provides a method for disassembling business requirements by integrating fine-tuning a small model and knowledge retrieval, and the method comprises: In step S11, a business requirement text expressed in natural language is received; In step S12, the business intent of the business requirement text is identified by fine-tuning the small model to obtain preliminary business intent; In step S13, knowledge retrieval is initiated to extract relevant information from the internal knowledge base of the power enterprise to optimize the preliminary business intent and obtain the power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching, and retrieval enhancement generation; In step S14, reasoning is performed based on the power business intention to generate a structured business instruction chain.

[0023] In one embodiment, the system automatically converts business requirements expressed in natural language into executable business instruction chains. The system receives the user's submitted business requirement text in natural language and uses a small, domain-tuned model to identify the business intent of the text, extracting the initial business intent of the text. Based on this, it further initiates knowledge retrieval, extracting relevant information from the power company's internal knowledge base related to the business requirement text. This supplements and optimizes the initial business intent to generate a more accurate power business intent. Based on the optimized power business intent, it performs logical reasoning to generate a structured business instruction chain.

[0024] At the initial stage of business demand processing, a business demand text expressed in natural language is received, such as a work request submitted by an operations and maintenance personnel via a mobile device. Natural language text often contains rich context and diverse expressions. A fine-tuned small AI model is introduced to identify business intent from the user's natural language input. This fine-tuned small model is trained for specific business scenarios in the power industry and can identify the business demand categories and key elements implicit in the text, thereby generating a preliminary business intent representation. For example, when a user inputs "Transformer No. 5 is faulty, please arrange for repair as soon as possible," the fine-tuned small model converts the non-standard name into a standard name. For example, if the standard name for transformer No. 5 in the power company's internal knowledge base is "309 110kv Longshan Substation 5#," the model can convert the urgent repair request for transformer No. 5 into an urgent repair request for the "309 110kv Longshan Substation 5#" transformer. Key information such as the device identification and urgency level is extracted as preliminary business intent, transforming the unstructured natural language request into a structured preliminary business intent.

[0025] To further enhance the professional accuracy of business intent identification, a knowledge retrieval step is introduced after obtaining the preliminary business intent to optimize and refine it. The knowledge retrieval module extracts background information and regulatory information related to the current need from the power company's internal knowledge base. This internal knowledge base encompasses a wealth of specialized knowledge in the power industry, such as equipment archives, historical troubleshooting records, operation and maintenance procedures, dispatching rules, and business process specifications. Knowledge retrieval can employ strategies such as vector embedding retrieval, rule matching, or retrieval-enhanced generation. Semantic matching is used to locate relevant documents and cases from the knowledge base. Specific information, such as the urgency of the need, is parsed based on pre-set rules. When necessary, the generated model supplements the context with the retrieval content. Through this retrieval process, the preliminary business intent is enriched into a more comprehensive and accurate power business intent. For example, in the case of an equipment emergency repair request, after identifying the emergency repair request, the technical manual, recent inspection reports, and relevant safety regulations for the equipment are retrieved to clarify the key steps and safety precautions required for the repair, thus enriching and refining the preliminary intent. This effectively integrates user needs with enterprise knowledge, making the understanding of business intent both professional and contextual.

[0026] After accurately obtaining the power business intent, the inference and decision-making phase begins, automatically generating a structured business instruction chain based on the business intent. The inference module uses the optimized business intent as input and, combined with power industry process knowledge and logical rules, generates an ordered sequence of instructions to meet the business requirement. This structured business instruction chain contains a series of specific instructions required to complete the business requirement. Each instruction clearly defines the action content, execution subject, and sequence. For example, for the aforementioned emergency transformer maintenance request, an instruction chain can be automatically generated according to power operation and maintenance specifications. First, the operation and maintenance dispatch center is notified of the fault at transformer No. 5, requiring urgent attention. A maintenance team is then dispatched, with designated personnel, required tools, and an estimated arrival time. Safety measures such as on-site power outages are then implemented, and maintenance operations are carried out (such as replacing damaged parts and testing power restoration). Finally, a report and system recovery instructions are generated after the maintenance is completed. This business instruction chain is organized in a logical business sequence, enabling all relevant departments and systems to work collaboratively according to established processes. The automatic generation of business instruction chains significantly improves response speed, ensuring that even complex requests are handled quickly and accurately.

[0027] This automated solution, which transforms natural language requests into business execution, refines user needs in the business intent recognition phase, enhances understanding by incorporating enterprise knowledge in the knowledge retrieval phase, and transforms the resulting understanding into action plans in the reasoning generation phase. Early intent recognition results provide guidance for subsequent retrieval, and the output of knowledge retrieval forms the basis for the final reasoning decision. For example, in an intelligent inspection and dispatching scenario, a dispatcher can use natural language to request, "Please schedule a routine inspection for Line 3 next week." After identifying this request as an inspection and dispatching intent, the system retrieves the operating status and relevant safety regulations for Line 3 from the knowledge base and specifies the inspection task details. The reasoning module then generates an instruction chain, including steps such as creating an inspection task order, assigning inspection personnel, pre-scheduling a power outage, and executing the inspection at the designated time. This system can flexibly leverage knowledge and generate corresponding execution plans based on different business needs, making it widely applicable in common business scenarios such as power equipment emergency repair, intelligent dispatching, and inspection and dispatching.

[0028] By combining natural language processing, industry knowledge retrieval and automatic reasoning, it realizes the intelligent understanding and processing of power business needs, and can directly generate structured and executable business instruction chains from the user's unstructured demand description. It improves the automation and accuracy of the internal business process response of power companies. On the one hand, it reduces the cost of manual analysis and communication, and shortens the time from demand proposal to implementation; on the other hand, it makes full use of the company's accumulated professional knowledge and rules to ensure that the generated instructions meet industry standards and safety requirements. From emergency response to equipment failures to daily operation and maintenance scheduling, it can be efficiently applied. The technical solution in this embodiment fine-tunes the small model and knowledge retrieval to coordinate the processing of business needs, improves the accuracy of intent recognition and the accuracy of the business instruction chain, realizes the intelligent and structured decomposition of business needs, and improves the business automation level and response efficiency of power companies.

[0029] In one embodiment, Figure 2 As shown, the following steps S21-S22 are also included: In step S21, the pre-trained natural language model is fine-tuned on the corpus in the field of power equipment operation and maintenance management to obtain the fine-tuned small model; In step S22, the preliminary business intent in the business requirement text is identified by using the fine-tuned small model.

[0030] In one embodiment, the innovative application of artificial intelligence in the automated processing of business requirements in the field of power equipment operation and maintenance management is demonstrated. This is to enhance the professional understanding ability of the natural language model through domain fine-tuning, thereby achieving efficient recognition of the preliminary intent of the business requirement text. It is proposed to fine-tune the pre-trained natural language processing model on the corpus in the field of power equipment operation and maintenance management. Although the pre-trained model has general language understanding capabilities, it often has difficulty accurately grasping the key points when faced with the terminology, expression habits and business processes unique to the power industry. By fine-tuning on a large amount of professional corpus such as power equipment operation and maintenance related documents, fault reports, operating specifications and dispatch instructions, the model can learn and solidify the expression logic and knowledge system unique to this field. The fine-tuned small model can not only more accurately understand key information such as equipment names, fault phenomena, operation and maintenance requests, but also improve sensitivity to industry standards and business context.

[0031] After obtaining the fine-tuned small model, the fine-tuned small model is used to perform preliminary business intent recognition on the business requirement text. Specifically, when the operation and maintenance personnel submit business requirements in natural language through a mobile terminal or management platform, such as "The main transformer has an abnormal temperature rise, please arrange for technical personnel to handle it on site as soon as possible", the fine-tuned small model can identify the key intentions and core elements in the text, such as equipment category, fault type, urgency, etc., and preliminarily classify the request into categories such as equipment abnormality handling, emergency repair tasks or routine maintenance. This automated recognition optimized based on industry knowledge not only greatly improves the response speed of business flow, but also effectively reduces the risk of human misjudgment and information omission, and further provides data and logical basis for the generation of subsequent business instruction chains and intelligent applications such as automatic dispatching and process collaboration. It realizes the customization and intelligent upgrade of artificial intelligence in the operation and maintenance management of power equipment. The fine-tuned small model has transformed a large amount of business demand analysis work that originally relied on human understanding and experience judgment into an efficient and standardized intelligent process.

[0032] In one embodiment, Figure 3 As shown, the following steps S31-S32 are also included: In step S31, the business requirement text is converted into a vector representation; In step S32, a similarity search is performed in the vector space of the internal knowledge base of the electric power enterprise based on the vector representation to extract information related to the business requirement text.

[0033] In one embodiment, knowledge retrieval includes vector embedding retrieval, the core of which is to use modern natural language processing technology to achieve efficient and intelligent retrieval of the knowledge base of the power enterprise. The business demand text expressed in natural language is encoded in a vectorized manner, so as to perform high-dimensional similarity search in the vector space of the knowledge base and realize deep semantic matching between demand and knowledge. Specifically, the business demand text is input into the embedding model (such as a language model based on deep learning), and the model maps the text into a high-dimensional vector representation. This high-dimensional vector not only carries the literal information of the text, but also captures the association of the semantic level, enabling the system to go beyond the surface matching of traditional keyword retrieval and deeply understand the real needs in business scenarios and professional contexts.

[0034] After vectorization is completed, this high-dimensional vector is used as the retrieval key to conduct efficient searches in the knowledge base vector space constructed within the power company. The knowledge base of the power company contains a large amount of structured or unstructured data, such as operating specifications, equipment information, historical operation and maintenance records, fault cases, etc. These materials are also vectorized and stored in the database. The retrieval system can quickly locate the knowledge fragments that are most relevant to the current business needs by calculating the distance or similarity between the demand vector and the vectors of each entry in the knowledge base. For example, when the business needs involve the abnormal handling of transformer temperature rise, not only can direct records containing temperature rise or transformers be retrieved, but also related historical cases, operating specifications and experience in handling similar problems can be found, thereby improving the relevance and accuracy of knowledge retrieval.

[0035] Deeply integrating vector embedded search into the power industry's knowledge service system enables the automatic and efficient association of natural language requirements with professional knowledge and data. Compared to traditional rule-based or keyword-based search, vector search offers significant advantages, including strong semantic understanding, high scalability, and rapid retrieval speed. Its logical implications are not only reflected in its technical implementation, but also in its ability to promote intelligent upgrades to business processes and improve knowledge utilization efficiency. In practical applications within power companies, this technology can effectively support diverse scenarios such as equipment fault analysis, emergency repair dispatch, and intelligent inspections, enabling real-time and accurate responses to complex and massive amounts of business information.

[0036] In one embodiment, Figure 4 As shown, the following steps S41-S42 are also included: In step S41, key content in the business requirement text is matched using predefined rules or templates; In step S42, corresponding knowledge items are retrieved from the internal knowledge base of the electric power enterprise based on the matching result.

[0037] In one embodiment, the knowledge retrieval process includes rule-matching retrieval, which reflects the organic combination of domain knowledge and automated text analysis. Through the rules, templates, and specifications accumulated by power companies in long-term business management, a mechanism is constructed that can automatically match and identify key information in business demand texts. Specifically, a series of business rules or matching templates are pre-set in the system. These rules can cover equipment categories, common fault types, standard operating instructions, emergency response procedures, etc. For example, for high-frequency keywords or semantic patterns such as "inspection", "emergency repair", and "abnormal temperature rise", corresponding matching rules are preset for automatic identification in actual business texts.

[0038] After receiving a user-submitted natural language business requirement text, the system parses and scans the text according to established rules or templates, identifying content that matches the rule conditions. This system not only accurately locates key information within the text, such as device names, event types, and operational requirements, but also automatically categorizes the requirement into the corresponding business topic or process within the enterprise knowledge base. In this way, rule-matching retrieval effectively maps natural language expressions with standardized knowledge systems, significantly reducing the need for human intervention and improving the accuracy and efficiency of text parsing and knowledge association.

[0039] Furthermore, based on the results of rule or template matching, knowledge items that are highly relevant to business needs can be directly retrieved from the internal knowledge base of the power enterprise. These knowledge items may include historical processing records, operating instructions, emergency response specifications, equipment instructions, and other content. By automatically mapping the demand text with the knowledge items, not only is the precise connection from demand to knowledge achieved, but it also provides a solid data foundation for subsequent business process automation, task dispatching, and instruction generation. This method is suitable for business scenarios with high standardization and clear processes, especially routine inspections, fault repairs, equipment inspections, and other links that are common in power companies. It can significantly improve the company's intelligent processing capabilities and response efficiency for complex business needs.

[0040] In one embodiment, Figure 5 As shown, the following steps S51-S53 are also included: In step S51, the business requirement text is preliminarily parsed by the fine-tuned small model; In step S52, based on the preliminary analysis results, the relevant supplementary information is retrieved from the internal knowledge base of the power enterprise; In step S53, the retrieved supplementary information is provided to the fine-tuning mini-model to enhance the understanding of the business requirements and assist in generating the structured business instruction chain.

[0041] In one embodiment, the knowledge retrieval process includes enhanced retrieval generation, highlighting the deep synergy between AI models and knowledge bases. This dynamic interaction between the model and knowledge retrieval enables a high-level understanding of business requirements and precise task breakdown. Based on a small model fine-tuned in the power sector, a preliminary analysis of user-submitted business requirement text is performed. This enables the fine-tuned model to discern the main intent and core elements of the text, such as identifying key equipment, anomalies, and operational objectives.

[0042] After obtaining preliminary analysis results, the results are used as a guide to search the power company's internal knowledge base for supplementary information closely related to business needs. The power company's internal knowledge base contains a wealth of professional materials, such as equipment operation manuals, standard operating procedures, historical processing cases, etc. By combining model analysis with knowledge base retrieval, it is possible to specifically locate the knowledge fragments that best supplement the initial understanding and achieve a precise connection between knowledge and intent. For example, when an "abnormal temperature rise" is identified in a certain device, the historical processing steps, safety precautions, and relevant operation and maintenance standards for the device's temperature rise problem can be retrieved, thereby expanding the model's knowledge coverage of the current scenario.

[0043] Furthermore, the retrieved supplementary knowledge is re-entered into the fine-tuned mini-model and fed into it alongside the original business requirement text, helping the model deepen its understanding of the business requirement and directly contributing to the generation of a structured business instruction chain. Retrieval-enhanced generation organically integrates enterprise knowledge and model reasoning capabilities through a knowledge-driven feedback loop, ensuring that the generated instruction chain not only meets user needs but also fully reflects enterprise standards and best practices. It not only provides answers based on memory but also integrates actual enterprise knowledge and scenario-based needs to output more accurate, professional, and executable structured instructions.

[0044] In one embodiment, Figure 6 As shown, the following steps S61-S62 are also included: In step S61, the generated structured business instruction chain is provided in the form of an instruction identification sequence; In step S62, a pre-stored specific instruction is called from an instruction registration center according to the instruction identification sequence.

[0045] In one embodiment, the generated structured business instruction chain is provided in the form of an instruction identification sequence; based on the instruction identification sequence, the pre-stored specific instructions are called from the instruction registration center. Through intelligent analysis and reasoning, a structured business instruction chain is generated, and complex business requirements are automatically decomposed into orderly operation steps, and represented by a set of unique instruction identifiers. Each instruction identifier corresponds to a specific business operation preset and standardized by the enterprise. It not only ensures the high degree of structuring and clarity of the business process decomposition results, but also lays the foundation for subsequent automated and standardized execution. Taking power companies as an example, whether it is equipment emergency repair, inspection or daily operation and maintenance scheduling, the structured instruction chain can resolve the ambiguity and fuzziness in free text into a traceable and manageable set of standard instructions.

[0046] The integration of instruction identification sequences with the instruction registration center embodies a highly modular and resource-reusable system design. The instruction registration center, serving as the enterprise's unified instruction management and distribution platform, pre-stores all standardized business instructions (such as operation templates, safety guidelines, and process flows). Business instruction chains utilize identification sequences to access existing instruction resources in the registration center, avoiding duplication and redundant definitions. This also centralizes, unifies, and facilitates traceability of instruction content updates and maintenance. This not only improves system scalability and maintenance efficiency, but also ensures that all business operations adhere to the latest enterprise standards and best practices, reducing the potential for human error. For example, in a specific business instruction chain, steps such as "power outage operation," "fault confirmation," and "site safety isolation" are accurately matched to corresponding standard instructions in the instruction registration center through identification sequences. The system automatically retrieves detailed operational details and precautions, significantly enhancing the automation and intelligence of business processes. The seamless integration of structured instruction chains and the instruction registration center achieves a complete, closed-loop process from intelligent business requirement understanding to automated, standardized execution. It not only ensures the traceability and consistency of business decomposition results, but also greatly improves the company's response speed, execution accuracy and system security when dealing with large-scale and complex business needs.

[0047] In one embodiment, Figure 7 This is a block diagram of a business requirement decomposition system that integrates fine-tuning of small models and knowledge retrieval according to an exemplary embodiment. Figure 7 As shown, the business requirement decomposition system integrating fine-tuning of small models and knowledge retrieval includes a receiving module 71, an acquisition module 72, a retrieval module 73 and a generation module 74.

[0048] The receiving module 71 is used to receive a business requirement text expressed in a natural language; The acquisition module 72 is used to identify the business intent of the business requirement text by fine-tuning the small model to obtain the preliminary business intent; The retrieval module 73 is used to initiate knowledge retrieval, extract relevant information from the internal knowledge base of the power enterprise to optimize the preliminary business intent and obtain the power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching and retrieval enhancement generation; The generation module 74 is configured to perform reasoning based on the power business intention and generate a structured business instruction chain.

[0049] The receiving module 71, the acquiring module 72, the retrieving module 73 and the generating module 74 included in the business requirement decomposition system block diagram of the integrated fine-tuning small model and knowledge retrieval are controlled to execute the business requirement decomposition method of the integrated fine-tuning small model and knowledge retrieval described in any of the above embodiments.

[0050] like Figure 8 As shown, the present invention provides an electronic device 800, which includes: a communication interface, a processor 801, and a memory 802; The memory 802 is used to store program instructions, which, when executed by the processor 801 that is communicatively connected to the memory 802 through the communication interface, receives a business requirement text expressed in natural language; identifies the business intent of the business requirement text by fine-tuning a small model to obtain a preliminary business intent; initiates knowledge retrieval to extract relevant information from the internal knowledge base of the power enterprise to optimize the preliminary business intent and obtain the power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching and retrieval enhancement generation; and performs reasoning based on the power business intent to generate a structured business instruction chain.

[0051] The present invention provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, a business requirement text expressed in natural language is received; business intent is identified on the business requirement text by fine-tuning a small model to obtain a preliminary business intent; knowledge retrieval is initiated to extract relevant information from an internal knowledge base of an electric power enterprise to optimize the preliminary business intent and obtain the electric power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching, and retrieval enhancement generation; reasoning is performed based on the electric power business intent to generate a structured business instruction chain.

[0052] It should be understood that the specific features, operations and details described herein above with respect to the method of the present invention may also be similarly applied to the system of the present invention, or vice versa. In addition, each step of the method of the present invention described above may be performed by the corresponding components or units of the system of the present invention.

[0053] It should be understood that the various modules / units of the system of the present invention can be implemented in whole or in part by software, hardware, firmware, or a combination thereof. Each module / unit can be embedded in the processor of a computer device in the form of hardware or firmware or independent of the processor, or can be stored in the memory of a computer device in the form of software for the processor to call to perform the operations of each module / unit. Each module / unit can be implemented as an independent component or module, or two or more modules / units can be implemented as a single component or module.

[0054] In one embodiment, a computer device is provided, comprising a memory and a processor. The memory stores computer instructions executable by the processor, which, when executed by the processor, instruct the processor to perform the steps of the method according to an embodiment of the present invention. The computer device can be broadly defined as a server, a terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device can include a processor, memory, a network interface, a communication interface, etc. connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device can include a non-volatile storage medium and an internal memory. An operating system, a computer program, etc. can be stored in or on the non-volatile storage medium. The internal memory can provide an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface and communication interface of the computer device can be used to connect to and communicate with external devices via a network. When the computer program is executed by the processor, the steps of the method according to the present invention are performed.

[0055] The present invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the method of an embodiment of the present invention to be performed. In one embodiment, the computer program is distributed on a plurality of computer devices or processors coupled to a network so that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, can be performed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations can be performed by one or more computer devices or processors, and one or more other method steps / operations can be performed by one or more other computer devices or processors. One or more computer devices or processors can perform a single method step / operation, or perform two or more method steps / operations.

[0056] Those skilled in the art will appreciate that the method steps of the present invention can be performed by instructing related hardware, such as a computer device or processor, through a computer program. The computer program can be stored in a non-transitory computer-readable storage medium, and when the computer program is executed, the steps of the present invention are performed. Depending on the circumstances, any reference herein to memory, storage, database, or other media may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (GGPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state disk, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0057] The various technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification as long as such combination does not conflict.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A business requirement decomposition method integrating fine-tuning of small models and knowledge retrieval, characterized by: include: Receive business requirement text expressed in natural language; By fine-tuning the small model, the business intent of the business requirement text is identified to obtain preliminary business intent; Initiate knowledge retrieval to extract relevant information from the internal knowledge base of the power enterprise to optimize the preliminary business intent and obtain the power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching, and retrieval enhancement generation; Reasoning is performed based on the power business intention to generate a structured business instruction chain.

2. The method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval according to claim 1, characterized in that: Also includes: Fine-tuning the pre-trained natural language model on corpus in the field of power equipment operation and maintenance management to obtain the fine-tuned small model; The preliminary business intent in the business requirement text is identified by the fine-tuned small model.

3. The method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval according to claim 1, characterized in that: The knowledge retrieval includes vector embedding retrieval, including: Convert the business requirement text into a vector representation; Based on the vector representation, a similarity search is performed in the vector space of the internal knowledge base of the electric power enterprise to extract information related to the business requirement text.

4. The method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval according to claim 1, characterized in that: The knowledge retrieval process includes rule matching retrieval, including: Match key content in the business requirement text through predefined rules or templates; Based on the matching results, the corresponding knowledge entries are retrieved from the internal knowledge base of the power enterprise.

5. The method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval according to claim 1, characterized in that: The knowledge retrieval process includes retrieval enhancement generation, including: Performing a preliminary analysis of the business requirement text using the fine-tuned small model; Retrieving relevant supplementary information from the internal knowledge base of the power enterprise based on the preliminary analysis results; The retrieved supplementary information is provided to the fine-tuned mini-model to enhance the understanding of the business requirements and assist in generating the structured business instruction chain.

6. The method for decomposing business requirements by integrating fine-tuning a small model with knowledge retrieval according to claim 1, characterized in that: Also includes: Providing the generated structured business instruction chain in the form of an instruction identifier sequence; According to the instruction identification sequence, a pre-stored specific instruction is called from an instruction registration center.

7. A business requirement disassembly system integrating fine-tuning small models and knowledge retrieval, characterized by: include: A receiving module, used for receiving a business requirement text expressed in a natural language; An acquisition module is used to identify the business intent of the business requirement text by fine-tuning a small model to obtain preliminary business intent; A retrieval module is used to initiate knowledge retrieval, extract relevant information from the internal knowledge base of the power enterprise to optimize the preliminary business intent, and obtain the power business intent, wherein the knowledge retrieval includes at least one of vector embedding retrieval, rule matching, and retrieval enhancement generation; A generation module is used to perform reasoning based on the power business intention and generate a structured business instruction chain.

8. The business requirement decomposition system integrating fine-tuning of small models and knowledge retrieval according to claim 7 is characterized by: The receiving module, the acquiring module, the retrieval module and the generating module are controlled to execute the business requirement decomposition method of integrating fine-tuning small models and knowledge retrieval as described in any one of claims 1 to 6.

9. An electronic device, characterized in that: include: Communication interface, processor, memory; Wherein, the memory is used to store program instructions, and when the program instructions are executed by the processor that is communicatively connected to the memory through the communication interface, the electronic device implements the business requirement decomposition method of integrating fine-tuning small models and knowledge retrieval as described in any one of claims 1 to 6.

10. A computer-readable storage medium having program instructions stored thereon, characterized in that: When the program instructions are executed by a computer, the computer implements the business requirement decomposition method of integrating fine-tuning small models and knowledge retrieval as described in any one of claims 1 to 6.

Images