Energy-saving processing method and device of network unit, computer equipment, readable storage medium and program product

Abstract

The invention relates to an energy-saving processing method and device for a network unit, equipment, a storage medium and a program product, and relates to the technical field of artificial intelligence and mobile communication. The end-to-end processing method for energy-saving control of the network unit can be provided, and the intelligent degree of energy-saving control of the network unit is improved. The method comprises the following steps: acquiring description information for energy-saving control of a network unit; if the big model identifies that the corresponding intention category belongs to the energy-saving control operation, obtaining the operation category through the big model; determining a primary network unit according to the operation category and control demand information corresponding to description information acquired by the large model, and generalizing the primary network unit to obtain a target network unit set; according to the operation category, obtaining an energy-saving control processing scheme by utilizing the large model according to the unit state information corresponding to the target network unit set and the available execution tool set; and obtaining processing summary information for energy-saving control of the network unit by utilizing the large model according to the description information and an execution result of the energy-saving control processing scheme.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence and mobile communication technology, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for energy-saving processing of a network unit. Background Art

[0002] In the field of mobile communications, various network units are involved, such as base stations, IT (Information Technology) equipment, and IDC (Internet Data Center) equipment in computer rooms. The energy saving of these network units has cost reduction and green benefits. They need to be able to recover quickly in emergencies without affecting network quality and user perception. Therefore, they need to be quickly located and safely awakened in scenarios such as emergencies.

[0003] Traditional energy-saving control of network units relies on manual monitoring and analysis, which has the problem of low intelligence level of energy-saving control of network units and is difficult to meet the energy-saving control needs of network units in scenarios such as emergencies. Summary of the invention

[0004] Based on this, it is necessary to provide a network unit energy-saving processing method, device, computer equipment, computer-readable storage medium and computer program product to address the above technical problems.

[0005] In a first aspect, the present application provides a method for energy saving processing of a network unit, comprising:

[0006] Get the description information of energy saving control for network unit;

[0007] If the large model identifies that the intention category corresponding to the description information belongs to an energy-saving control operation, obtaining the operation category corresponding to the description information through the large model;

[0008] Determine a preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalize the preliminary selected network unit to obtain a target network unit set;

[0009] According to the operation category, using the large model, according to the unit state information corresponding to the target network unit set and the available execution tool set, an energy-saving control processing scheme for execution of the target network unit set is obtained;

[0010] The large model is used to obtain processing summary information for network unit energy-saving control according to the description information and the execution result of the energy-saving control processing solution.

[0011] In a second aspect, the present application further provides an energy-saving processing device for a network unit, comprising:

[0012] An information acquisition module, used to acquire description information for energy-saving control of a network unit;

[0013] a category identification module, configured to obtain the operation category corresponding to the description information through the large model if the large model identifies that the intention category corresponding to the description information belongs to an energy-saving control operation;

[0014] A unit determination module, used to determine the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalize the preliminary selected network unit to obtain a target network unit set;

[0015] An energy-saving control module, configured to obtain an energy-saving control processing scheme for execution of the target network unit set according to the operation category and using the large model according to the unit state information corresponding to the target network unit set and a set of available execution tools;

[0016] The processing summary module is used to obtain processing summary information for network unit energy-saving control based on the large model according to the description information and the execution result of the energy-saving control processing solution.

[0017] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0018] Obtain descriptive information for energy-saving control of network units; if the large model recognizes that the intention category corresponding to the descriptive information belongs to an energy-saving control operation, obtain the operation category corresponding to the descriptive information through the large model; determine the preliminary network units according to the operation category and the control requirement information corresponding to the descriptive information obtained by the large model, and generalize the preliminary network units to obtain a target network unit set; according to the operation category, use the large model to obtain an energy-saving control processing scheme for execution of the target network unit set based on the unit status information corresponding to the target network unit set and a set of available execution tools; use the large model to obtain processing summary information for energy-saving control of network units based on the descriptive information and the execution results of the energy-saving control processing scheme.

[0019] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0020] Obtain descriptive information for energy-saving control of network units; if the large model recognizes that the intention category corresponding to the descriptive information belongs to an energy-saving control operation, obtain the operation category corresponding to the descriptive information through the large model; determine the preliminary network units according to the operation category and the control requirement information corresponding to the descriptive information obtained by the large model, and generalize the preliminary network units to obtain a target network unit set; according to the operation category, use the large model to obtain an energy-saving control processing scheme for execution of the target network unit set based on the unit status information corresponding to the target network unit set and a set of available execution tools; use the large model to obtain processing summary information for energy-saving control of network units based on the descriptive information and the execution results of the energy-saving control processing scheme.

[0021] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0022] Obtain descriptive information for energy-saving control of network units; if the large model recognizes that the intention category corresponding to the descriptive information belongs to an energy-saving control operation, obtain the operation category corresponding to the descriptive information through the large model; determine the preliminary network units according to the operation category and the control requirement information corresponding to the descriptive information obtained by the large model, and generalize the preliminary network units to obtain a target network unit set; according to the operation category, use the large model to obtain an energy-saving control processing scheme for execution of the target network unit set based on the unit status information corresponding to the target network unit set and a set of available execution tools; use the large model to obtain processing summary information for energy-saving control of network units based on the descriptive information and the execution results of the energy-saving control processing scheme.

[0023] The energy-saving processing method, device, computer equipment, computer-readable storage medium and computer program product of the above-mentioned network unit obtain descriptive information for energy-saving control of the network unit; if the big model recognizes that the intention category corresponding to the descriptive information belongs to the energy-saving control operation, the operation category corresponding to the descriptive information is obtained through the big model; the preliminary network unit is determined according to the operation category and the control requirement information corresponding to the descriptive information obtained by the big model, and the preliminary network unit is generalized to obtain the target network unit set; according to the operation category, the big model is used to obtain the energy-saving control processing plan for the target network unit set according to the unit status information corresponding to the target network unit set and the set of available execution tools; the big model is used to obtain the processing summary information for the energy-saving control of the network unit according to the execution results of the descriptive information and the energy-saving control processing plan. This solution can provide an end-to-end processing method for energy-saving control of network units, from the acquisition of descriptive information to the accurate identification of intent categories and operation categories, to the effective positioning of network units and the formation and execution of their control processing solutions, and finally to the processing summary. This avoids the defects of traditional technologies that rely on manual monitoring and analysis, improves the intelligence level of energy-saving control of network units, and can meet the energy-saving control needs of network units in scenarios such as emergency with efficient end-to-end information extraction and transmission and full-chain automated operation means. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 is an application environment diagram of an energy-saving processing method for a network unit in one embodiment;

[0026] Figure 2 is a flow chart of a method for energy saving processing of a network unit in one embodiment;

[0027] Figure 3 A schematic diagram of a paradigm of first category prompt information in one embodiment;

[0028] Figure 4 An example schematic diagram of an intent category recognition interaction in one embodiment;

[0029] Figure 5 A schematic diagram of a paradigm for extracting prompt information in one embodiment;

[0030] Figure 6It is a schematic diagram of an example of controlling the interaction of extracting demand information in one embodiment;

[0031] Figure 7 is a schematic diagram of a knowledge graph in one embodiment;

[0032] Figure 8 A schematic diagram of a paradigm for searching prompt information in one embodiment;

[0033] Fig. 9 A schematic diagram of an example of an interaction for searching for a noun associated with an address in one embodiment;

[0034] Fig.10 A schematic diagram of an associated network unit search in one embodiment;

[0035] Fig.11 A schematic diagram of a paradigm for processing prompt information in one embodiment;

[0036] Fig.12 A schematic diagram of an example of an interaction for generating a processing scheme in one embodiment;

[0037] Fig.13 A schematic diagram of a paradigm of the first summary prompt information in one embodiment;

[0038] Fig.14 and Fig.15 A schematic diagram of an example of processing summary information generation interaction in one embodiment;

[0039] Fig.16 It is a schematic diagram of the overall framework of a method for energy saving processing of a network unit in one embodiment;

[0040] Fig.17 A schematic diagram of the overall framework of a method for obtaining a target network unit set in one embodiment;

[0041] Fig.18 A schematic diagram of an implementation service of an energy-saving processing method of a network unit in one embodiment;

[0042] Fig.19 An example schematic diagram of intent recognition classification in one embodiment;

[0043] Fig. 20 is a structural block diagram of an energy-saving processing device of a network unit in one embodiment;

[0044] Fig.21 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0046] Introduction to terms related to this application:

[0047] Cell wake-up: refers to the operation process of an energy-saving cell exiting the energy-saving state.

[0048] Word vector: Words with similar meanings have similar word vectors. Word vector technology can transform words into dense vectors, making similar words close in distance in the vector space, making it easy to know the close relationship between words.

[0049] Knowledge graph: By building a structured semantic knowledge base, entities, relationships, and attributes can be described in symbolic form. Entities in the data and their relationships can be mapped to the knowledge graph. The basic structure can include nodes (entities) and edges (relationships). Each node can represent an entity, and the nodes are connected to each other through edges (relationships) to form a network of knowledge structures.

[0050] Prompt Engineering: refers to a technique that improves the performance of generative AI models by designing and optimizing prompts. Large models can be guided to generate desired outputs by using specific text or instructions.

[0051] Thinking Chain: refers to a technology that can be used in the field of artificial intelligence to improve the reasoning ability of large language models. It can break down complex logical problems into a series of logically related thinking steps to form a complete thinking process, thereby guiding the large model to generate more accurate and logical answers. The application scenarios are mainly concentrated on problems that require precise reasoning. By encouraging large language models to show their reasoning process, Thinking Chain can significantly improve the performance and interpretability of large language models on complex reasoning tasks.

[0052] Emergency: refers to the emergency actions taken in response to an emergency in a communication base station and the cells within its service area. Base station cell emergency refers to a series of pre-preparation and response measures taken within the cell served by a communication base station to deal with possible emergencies.

[0053] In the field of mobile communications, network units such as base stations, IT equipment, and IDC equipment in computer rooms often require efficient end-to-end information extraction and transmission and full-chain automated operation methods for energy-saving control in scenarios such as emergency. Traditional network unit energy-saving control relies on manual monitoring and analysis, and its intelligent level of energy-saving control of network units is low, making it difficult to meet the energy-saving control needs in scenarios such as emergency.

[0054] In this regard, taking the emergency response scenario of base station energy saving as an example, in the traditional energy-saving control scheme, in terms of information extraction and transmission, manual monitoring and analysis require a large investment, the operation and maintenance of the equipment is specialized and detailed, and there are many domain and layered links. The "hand-made workshop-style" customized verification is not efficient enough. Moreover, due to the lack of standardization and variety of nouns in the location information of the cell, the actual geographic information is also complex and numerous, and it is difficult to identify the cells in the area through a simple fuzzy search, which relies heavily on front-line maintenance experience. In addition, in terms of the use and learning guidance of automated emergency tools, the learning cost of multiple systems is high, and complex operations are prone to subjective errors and omissions. For example, there is a lack of efficient and flexible connection between intent recognition and emergency tools. Therefore, there is room for improvement in its emergency intelligence level.

[0055] In this regard, the energy-saving processing method of the network unit of the present application can provide an end-to-end processing method for energy-saving control of the network unit, from the acquisition of descriptive information to the accurate identification of intent categories and operation categories, to the effective positioning of the network unit and the formation and execution of its control processing scheme, and finally to the processing summary. This avoids the defects of traditional technology that relies on manual monitoring and analysis, improves the intelligence level of energy-saving control of the network unit, and can meet the energy-saving control needs of the network unit in scenarios such as emergency with efficient end-to-end information extraction and transmission and full-chain automated operation means.

[0056] It can be understood that the energy-saving processing method of the network unit of the present application can be applied to energy-saving scenarios of various network units. As an example, it can be applied to base station emergency, IT equipment scheduling, computer room IDC air conditioning energy saving and other scenarios.

[0057] The energy-saving processing method of the network unit provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown in the figure, the terminal can communicate with the server through the network. The data storage system can store the data that the server needs to process. The data storage system can be integrated on the server, or placed on the cloud or other network servers. The energy-saving processing method of the network unit of the present application can be as follows: Figure 1The terminal shown may be executed, or it may be executed by a server, or it may be executed by a terminal and a server in cooperation. Among them, the terminal may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, projection devices, etc. Portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted devices may be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. The server may be an independent physical server, or a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0058] In an exemplary embodiment, Figure 2 As shown, a method for energy saving processing of a network unit is provided, and the method can be performed as follows Figure 1 The terminal shown in the figure is executed, and the method may include the following steps:

[0059] Step S201: Acquire description information for energy-saving control of a network unit.

[0060] Among them, the descriptive information for the energy-saving control of the network unit refers to the information used to describe the energy-saving control related to the network unit. Among them, the form of the descriptive information can be text, voice, video, picture, etc. The network element unit can be a physical network unit, such as a base station, etc., or it can be a virtual / logical network unit, such as a base station cell, etc. Among them, the energy-saving control can include energy-saving shutdown, energy-saving wake-up, etc. in various scenarios. As an embodiment, the descriptive information for the energy-saving control of the network unit can be an emergency description text for the energy-saving control of the base station cell, and the text can be a text used to describe the emergency wake-up aspects of the energy-saving control of the base station cell. As an example, the emergency description text can be: "What are the means of cell emergency?", or it can be: "An emergency incident occurred in the x new city in A area, and all 3.5GNR energy-saving cells of all 5G networks in the area and within a radius of 1 km around were awakened" and so on.

[0061] In some embodiments, the description information for the energy-saving control of the network unit may be a user question, such as the user inputting the description information of the energy-saving control of the network unit when asking a question, or may be a third-party information source, such as a third-party emergency information source. That is, the user may provide the description information for the energy-saving control of the network unit to the terminal by inputting the question information. The terminal may also obtain the description information for the energy-saving control of the network unit sent by the third-party information source by accessing the third-party information source. This may help to realize the full process automation of the energy-saving control processing of the network unit.

[0062] Step S202: If the large model identifies that the intention category corresponding to the description information belongs to an energy-saving control operation, the operation category corresponding to the description information is obtained through the large model.

[0063] Among them, the intent category corresponding to the description information can be identified through a large model (also called a large language model). The intent category is used to represent the intent corresponding to the description information. Different intent categories can be processed through different thinking chains, and each thinking chain can correspond to an intent category. As an embodiment, the intent category corresponding to the description information may include energy-saving consultation and energy-saving control operations, etc. Among them, thinking chain 1 can correspond to energy-saving consultation, thinking chain 2 can correspond to energy-saving control operations, and so on. As an example, in an emergency scenario of energy saving in a base station cell, energy-saving consultation can be emergency consultation, and energy-saving control operations can be emergency operations.

[0064] In this step, if the large model identifies that the intention category corresponding to the description information belongs to an energy-saving control operation, the terminal can process it through the thought chain 2 corresponding to the energy-saving control operation. In this thought chain 2, the terminal can further identify the operation category corresponding to the description information through the large model. Among them, the operation category can be used to indicate which energy-saving control operation or operations need to be performed. The operation category can include various operation categories such as energy-saving wake-up and energy-saving shutdown. Different operation categories can construct corresponding processing procedures for processing. As an example, in the emergency scenario of energy saving in the base station cell, energy-saving wake-up can be emergency wake-up.

[0065] Step S203, determining the preliminary selected network units according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalizing the preliminary selected network units to obtain a target network unit set.

[0066] In this step, the terminal may further obtain control requirement information corresponding to the description information through the large model. The control requirement information may be extracted from the description information through the large model. The control requirement information refers to relevant information used to represent its energy-saving control requirements. As an example, the energy-saving control requirements may include unit location information (network unit location information) that requires energy-saving control, unit type information (network unit type information), and the like. After obtaining the control requirement information, the terminal may determine the preliminary network unit according to the operation category and the control requirement information and generalize the preliminary network unit to obtain a target network unit set. The preliminary network unit may be one or more network units determined as a basis according to the control requirement information. As an example, a base station may be determined as the preliminary network unit according to the control requirement information, a base station cell may be determined as the preliminary network unit, and several base station cells may be determined as the preliminary network units, and the like. The generalization process means that the preliminary selected network unit can be generalized according to the operation category based on the preliminary selected network unit's unit identification information (network unit identification information), unit location information, and unit performance information (network unit performance information). For example, other network units with similar names, similar locations, and / or similar performances can be used as target network units together with the preliminary selected network unit, thereby forming a target network unit set. The information or information based on which the preliminary selected network unit is generalized can be determined according to the operation category, and different generalization methods can be adopted according to different operation categories. Therefore, this step can obtain a range of network unit sets that are suitable for or more secure in energy-saving control of network units.

[0067] Step S204, according to the operation category, using the large model according to the unit status information corresponding to the target network unit set and the available execution tool set, obtain an energy-saving control processing solution for the target network unit set to execute.

[0068] In this step, after determining the target network unit set, the unit state information (network unit state information) of each target network unit in the target network unit set can be obtained to obtain the unit state information corresponding to the target network unit set. The unit state information of the target network unit can be used to indicate the state of the target network unit, which can include the current energy-saving state, the whitelist state, etc. The current energy-saving state can be used to indicate whether the target network unit is currently energy-saving, etc., and the whitelist state can be used to indicate whether the target network unit is on the whitelist or the blacklist, etc.

[0069] In this step, the available execution tool set refers to a tool set that can be used for the target network unit to execute, which can be a set of control instructions. As an example, the available execution tool set can include: "Exit Energy Saving", "Set Whitelist", "Do Not Execute Operation" and other available execution tools.

[0070] In this step, the terminal can obtain an energy-saving control processing scheme for the target network unit set according to the unit state information corresponding to the target network unit set and the set of available execution tools using the large model according to the operation category. Among them, the energy-saving control processing scheme may include an energy-saving control processing method for each target network unit in the target network unit set. The energy-saving control processing scheme can be obtained by combining the large model with the operation category, and different operation categories (energy-saving wake-up, energy-saving shutdown, etc.) can output different energy-saving control processing schemes. In the energy-saving control processing scheme, the energy-saving control processing method for the target network unit can be represented according to the available execution tools that can be taken for it under the operation category. As an example, in the emergency scenario of base station cell energy saving, the available execution tool of "exit energy saving" can be used for the target cell 001-1 (base station ID-cell ID) in the target cell set, and the available execution tools of "setting whitelist" and "exit energy saving" can be used for the target cell 001-4 in the target cell set, and so on.

[0071] In some embodiments, after obtaining the energy-saving control processing scheme for execution of the target network unit set, the terminal can instruct each target network unit in the target network unit set to execute the corresponding available execution tool according to the energy-saving control processing method of each target network unit in the energy-saving control processing scheme, and then the terminal can obtain the execution result of each target network unit, thereby the terminal can obtain the execution result of the energy-saving control processing scheme according to the execution result of each target network unit.

[0072] Step S205, using the large model to obtain processing summary information for the energy-saving control of the network unit according to the description information and the execution result of the energy-saving control processing solution.

[0073] In this step, after obtaining the execution result of the energy-saving control processing scheme, the terminal can use the large model to obtain the processing summary information for the energy-saving control of the network unit according to the description information and the execution result of the energy-saving control processing scheme. Among them, the processing summary information can be the summary information of this energy-saving control processing, which can include the analysis of the energy-saving control processing process and the summary information made based on the analysis, and the summary information can be used to feedback whether this energy-saving control processing can meet the user's requirements for energy-saving control processing. As an embodiment, the terminal can hand over the relevant information in this energy-saving control processing to the large model for processing summary information generation. The relevant text can include the above-mentioned description information and the execution result of the energy-saving control processing scheme, and can also include intent categories, operation categories, preliminary network units, target network unit sets, unit status information, available execution tool sets, etc., so that the large model can analyze and summarize the energy-saving control processing process according to these information to obtain accurate and comprehensive processing summary information.

[0074] The energy-saving processing method of the network unit of this embodiment obtains description information for energy-saving control of the network unit; if the large model identifies that the intention category corresponding to the description information belongs to the energy-saving control operation, the operation category corresponding to the description information is obtained through the large model; the preliminary network unit is determined according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and the preliminary network unit is generalized to obtain the target network unit set; according to the operation category, the large model is used to obtain the energy-saving control processing scheme for the target network unit set according to the unit state information corresponding to the target network unit set and the available execution tool set; the large model is used to obtain the processing summary information for the energy-saving control of the network unit according to the execution result of the description information and the energy-saving control processing scheme. The scheme can provide an end-to-end processing method for energy-saving control of the network unit from the acquisition of description information to the accurate identification of intention category and operation category, to the effective positioning of the network unit and the formation and execution of its control processing scheme, and finally to the processing summary, which avoids the defect of relying on manual monitoring and analysis in traditional technology, improves the intelligence level of energy-saving control of the network unit, and can meet the energy-saving control requirements of efficient end-to-end information extraction and transmission and full-chain automated operation means of the network unit in scenarios such as emergency.

[0075] In an exemplary embodiment, after obtaining description information for energy-saving control of a network unit in step S201, the method of the present application may further include the following steps:

[0076] According to the description information and the first classification prompt information, the large model is called to identify the intent category corresponding to the description information from multiple intent categories. The first classification prompt information includes multiple intent categories.

[0077] In this embodiment, the terminal may call the large model according to the description information and the first classification prompt information, where the first classification prompt information includes multiple intent categories, and the large model identifies the intent category corresponding to the description information from the multiple intent categories.

[0078] As an example, in an emergency scenario of energy saving in a base station cell, such as Figure 3 The paradigm of the first classification prompt information is shown in FIG. Multiple intent categories can be preset in the first classification prompt information, and the multiple intent categories can include emergency consultation and emergency operation, etc. In some embodiments, the big model can also identify the intent category through multiple rounds of dialogue, and the history records of multiple rounds of dialogue can be provided in the first classification prompt information.

[0079] As an example, in an emergency scenario of energy saving in a base station cell, such as Figure 4 As shown, the description information is a description text, which can be "An emergency event occurred in x new town in A place, and all 3.5GNR energy-saving cells of 5G networks in the area and within a radius of 1 km around it were awakened." The description information and the first classification prompt information containing multiple intent categories can be provided to the large model, and the large model can also be instructed in the first classification prompt information to extract keywords related to the classification from the description information. Therefore, the large model can output the corresponding intent category as "emergency operation" according to the description information and the first classification prompt information, and can also output keywords "x new town in A place", "emergency event", "5G network", "3.5GNR", "energy-saving cell", "wake up", etc.

[0080] The solution of this embodiment can accurately and effectively guide the large model to identify the intent category corresponding to the description information, so as to perform subsequent processing on it through the corresponding thinking chain.

[0081] In an exemplary embodiment, the operation category corresponding to the description information obtained by the large model in step S202 may include:

[0082] According to the description information, the intention category corresponding to the description information and the second classification prompt information, the large model is called to identify the operation category corresponding to the description information from multiple operation categories. The second classification prompt information includes multiple operation categories.

[0083] In this embodiment, the terminal can call the big model according to the description information, the intent category corresponding to the description information, and the second classification prompt information. The second classification prompt information includes multiple operation categories, wherein the multiple operation categories may include various operation categories such as energy-saving wake-up and energy-saving shutdown. The big model can identify the operation category corresponding to the description information from the multiple operation categories.

[0084] As an example, in an emergency scenario of base station cell energy saving, the paradigm of the second classification prompt information can refer to the paradigm of the first classification prompt information mentioned above, and the multiple intention categories in the paradigm of the first classification prompt information can be replaced by the multiple operation categories, and the multiple operation categories can include emergency wake-up, etc.

[0085] In this embodiment, the terminal can concatenate the description information and the intent category corresponding to the description information from multiple sources and provide them to the large model. The large model identifies the operation category corresponding to the description information from multiple operation categories based on the multi-source concatenated description information and its intent category and the second classification prompt information.

[0086] The solution of this embodiment can accurately and effectively guide the large model to identify the operation category corresponding to the description information, so as to process it through subsequent corresponding processing methods.

[0087] In an exemplary embodiment, before determining the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model in step S203, and generalizing the preliminary selected network unit to obtain the target network unit set, the method of the present application may further include the following steps:

[0088] According to the description information and the extraction prompt information, the large model is called to obtain the control demand information corresponding to the control demand extraction target according to the description information. The extraction prompt information includes the control demand extraction target; the control demand extraction target is determined according to the operation category.

[0089] In this embodiment, the terminal can call the large model according to the description information and the refining prompt information, and the refining prompt information includes a control demand refining target, and the control demand refining target is used to indicate that the large model needs to refine the corresponding control demand information according to the description information. The control demand refining target can be determined according to the operation category, and different operation categories can correspond to different control demand refining targets. As an example, the operation category can include energy-saving shutdown, energy-saving wake-up, etc., and the control demand refining target can include a shutdown demand refining target, a wake-up demand refining target, etc., so that the corresponding shutdown demand information, wake-up demand information, etc. can be refined.

[0090] In some embodiments, the wake-up demand extraction target may include location, etc., and may also include unit type and / or generalization degree based on the location. Among them, the location can be used to instruct the large model to extract the wake-up demand information related to the location representing the network unit according to the description information. The unit type can be used to indicate the type of network unit targeted by the network unit energy-saving control, such as "5G", "3.5GNR", etc.; the generalization degree can be used to indicate the generalization degree of the subsequent generalization processing of the preliminary network unit, such as the generalization degree "1 km", etc. This can improve the pertinence and effectiveness of the network unit energy-saving control.

[0091] Based on this, the large model can obtain the control requirement information corresponding to the control requirement extraction target according to the description information and the extraction prompt information.

[0092] As an example, Figure 5 The paradigm of refining prompt information is shown in FIG. 1 , and the control requirement refining target can be set in the refined prompt information.

[0093] As an example, in an emergency scenario of energy saving in a base station cell, such as Figure 6 As shown, the description information is a description text, which may be "an emergency occurs in x new town in area A, and all 3.5GNR energy-saving cells of all 5G networks in the area and within a radius of 1 km are awakened." The control demand refining target is a wake-up demand refining target, which may include a location, which may be "a first-level division, a second-level division, emergency address information", etc. Thus, the large model can output the refined control demand information as wake-up demand information according to the description information and the refined prompt information, and the wake-up demand information may include: first-level division: A1; second-level division: A2; emergency address information: x new town, etc.

[0094] It should be noted that the large model can extract the control demand information corresponding to the control demand extraction target from the description information based on the description information and the extraction prompt information. The large model can further process the extracted control demand information through naming association completion and content standardization to obtain perfect control demand information to improve the comprehensibility of the backward model. For example, the large model can identify "A place" from the description information. The "A place" can be a text representing the above-mentioned secondary division but lacking the secondary division unit. The large model can further form the "primary division: A1; secondary division: A2" in the control demand information through automatic completion.

[0095] The solution of this embodiment can accurately and effectively guide the large model to identify the control requirement information corresponding to the description information.

[0096] In an exemplary embodiment, determining the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model in step S203, and generalizing the preliminary selected network unit to obtain the target network unit set may include:

[0097] If the operation category belongs to energy-saving wake-up, the preliminary network unit is determined according to the wake-up requirement information corresponding to the description information obtained by the large model, and the preliminary network unit is generalized based on the unit position information of the preliminary network unit to obtain the target network unit set.

[0098] In this embodiment, if the operation category is further identified as energy-saving wakeup through the large model, the terminal can further obtain the wakeup requirement information corresponding to the description information through the large model. The wakeup requirement information can be extracted from the description information through the large model. The wakeup requirement information refers to the relevant information used to indicate its wakeup requirement. As an example, the wakeup requirement information may include the unit location information, unit type information, etc. of the wakeup requirement.

[0099] After obtaining the wake-up requirement information, the terminal can determine the preliminary network unit according to the wake-up requirement information, and generalize the preliminary network unit based on the unit location information of the preliminary network unit, so as to obtain a target network unit set. As an example, in an emergency scenario of energy saving of a base station cell, the preliminary network unit may be a base station cell. In this embodiment, the generalization process may be to generalize the network unit to be selected based on the unit location information of the preliminary network unit. For example, other base station cells with similar names, similar locations, etc. and the preliminary base station cells may be used as a target network unit set. Therefore, the solution of this embodiment can obtain a more guaranteed target network unit set range for energy-saving wake-up.

[0100] In an exemplary embodiment, determining the preliminary selected network unit according to the wake-up requirement information corresponding to the description information obtained by the large model in the above embodiment may include the following steps:

[0101] According to the network unit query condition corresponding to the wake-up requirement information, a preliminary network unit is queried and obtained.

[0102] In this embodiment, the wake-up requirement information can be conditionally escaped to obtain the network unit query condition, and the preliminary network unit can be obtained by querying the network unit query condition and the industrial parameter table. Among them, the SQL (Structured Query Language) condition can be combined through the where condition mapping. For example, the wake-up requirement information can include the first-level division A1 and the second-level division A2, and the unit type is 5G. The SQL condition after conditional escape can be select * from the industrial parameter table where the first-level division ID = 123 and the second-level division ID = 321 and network = 5G. After executing the network unit query condition, the preliminary network unit can be queried.

[0103] The solution of this embodiment can convert the wake-up requirement information into corresponding network unit query conditions and query to obtain the preliminary selected network unit.

[0104] In an exemplary embodiment, the generalization process of the preliminarily selected network units based on the unit location information of the preliminarily selected network units in the above embodiment to obtain the target network unit set may include:

[0105] Obtain the generalized location information corresponding to the unit location information of the preliminary selected network unit; query and obtain the generalized network unit according to the network unit query condition corresponding to the generalized location information; obtain the target network unit set according to the preliminary selected network unit and the generalized network unit.

[0106] In this embodiment, the terminal may first obtain the generalized location information corresponding to the unit location information of the pre-selected service area. The generalized location information refers to the relevant information that can be used to indicate the location of the generalized network unit to be queried. The generalized network unit refers to the network unit obtained by generalizing the pre-selected network unit. The generalized location information may be location information associated with the unit location information of the pre-selected network unit, may be specific address information, may be information indicating the search range, etc., which is not specifically limited here.

[0107] After obtaining the generalized location information, the terminal can further query the generalized network unit according to the network unit query condition corresponding to the generalized location information. In this regard, the terminal can query the generalized network unit in the above-mentioned way of querying the initial network unit, and can obtain the network unit query condition by conditional translation of the generalized location information, and obtain the generalized network unit by querying the network unit query condition and the engineering parameter table. Among them, the number of the generalized network units can be multiple.

[0108] Thus, the terminal can use both the primary network unit and the generalized network unit as target network units to obtain a target network unit set.

[0109] The solution of this embodiment can obtain the generalized network unit of the preliminary selected network unit by acquiring the generalized location information corresponding to the unit location information of the preliminary selected network unit, and form a target network unit set.

[0110] In an exemplary embodiment, the step of obtaining the generalized location information corresponding to the unit location information of the preliminarily selected network unit in the above embodiment may include:

[0111] According to the unit location information of the preliminary network unit, the associated address subject words with the target association relationship with the unit location information of the preliminary network unit are obtained in the knowledge graph; the knowledge graph is constructed based on each address subject word and the association relationship between each address subject word; the address subject word and the association relationship are determined based on the unit location information and the target subject of each network unit; and / or, according to the unit location information of the preliminary network unit and the search prompt information, the large model is called to search and obtain the associated address nouns associated with the unit location information of the preliminary network unit; the search prompt information is used to guide the large model to obtain the associated address nouns based on the regional search; and / or, according to the unit location information of the preliminary network unit and the regional search target information, the associated network unit search information is obtained; the regional search target information includes regional scope information and / or associated network unit feature information. Finally, generalized location information can be obtained based on one or more of the associated address subject words, associated address nouns and associated network unit search information.

[0112] In this embodiment, the generalized location information corresponding to the unit location information of the primary selected network unit can be obtained by combining multiple generalization methods, and the generalized location information can be obtained based on one or more of the associated address keywords, associated address nouns and associated network unit search information.

[0113] For the associated address subject words: the terminal can obtain the associated address subject words with the target association relationship with the unit location information of the pre-selected network unit in the knowledge graph based on the unit location information of the pre-selected network unit. The knowledge graph is constructed based on each address subject word and the association relationship between each address subject word. The address subject words and the association relationship are determined based on the unit location information of each network unit and the target subject. The address subject word refers to the noun with address as the subject in the knowledge graph. The specific target subject can be set by the user, such as Figure 7 As shown, a knowledge graph can be constructed with "living facilities" as the target theme. The address keywords in the knowledge graph can be, for example, x new town, y2 bus stop, etc. The association between the address keywords in the knowledge graph can include attributes such as transportation and hospital. The association relationship can also be measured by the vectors between words. The vectors between words can be used to determine whether the two meet a certain association relationship, etc.

[0114] As an example, in an emergency scenario of energy saving in a base station cell, such as Figure 7As shown, the cell location information (unit location information) such as the base station name, cell name, divisions at all levels, and address information of each cell (all cells) can be spliced, and then according to the implementation order of cleaning, attribute extraction, word segmentation, clustering, dimensionality reduction, and association, according to the target theme such as "life support", the address subject words and the relationship vectors therebetween can be obtained, and then the knowledge graph can be constructed with the address subject words as points and the association relationships as edges. Based on this, after obtaining the cell location information (unit location information) of the preliminary selected cell (preliminary selected network unit), the base station name, cell name, and cell address in the cell location information can be combined, and the corresponding address noun set can be obtained by word segmentation. In the knowledge graph, the address subject words within a distance of, for example, 1 degree from the address nouns in the address noun set are selected as the above-mentioned associated address subject words with a target association relationship with the cell location information of the preliminary selected cell. In this way, the preliminary selected network units such as the preliminary selected cell can be generalized with the help of the knowledge graph of the address subject words.

[0115] For the associated address noun: the terminal can call the large model according to the unit location information of the preliminary network unit and the search prompt information, and the search prompt information is used to guide the large model to obtain the associated address noun based on the regional search. The regional search can be based on the area where the preliminary network unit is located. The associated address noun can be an address noun associated with the unit location information of the preliminary network unit in the area, such as Figure 8 The paradigm of the search prompt information is shown in . Thus, the large model can search for the associated address nouns associated with the unit location information of the preliminary selected network unit in the area where the preliminary selected network unit is located according to the unit location information of the preliminary selected network unit and the search prompt information.

[0116] As an example, in an emergency scenario of energy saving in a base station cell, such as Fig. 9 As shown, the cell location information (unit location information) of the preliminary cell (preliminary network unit) can be "x New Town". The large model can perform a regional search for "x New Town" according to the search prompt information, and can obtain the associated address nouns such as surrounding roads and surrounding bus stations associated with it. The large model can also complete and standardize the naming relationship of the associated address nouns obtained by its search as described above. In this way, the preliminary network units such as the preliminary cell can be generalized with the help of the large model.

[0117] For the associated network unit search information, the terminal can obtain the associated network unit search information based on the unit location information of the pre-selected network unit and the regional search target information. Among them, the regional search target information refers to the information used to indicate the area and its search target, which may include regional range information and / or associated network unit characteristic information. Among them, the regional range information can be used to indicate the search range based on the area where the pre-selected network unit is located, such as within 1 kilometer, etc.; the associated network unit characteristic information can be used to indicate the characteristic information of the associated network unit to be searched, such as characteristic information of unit types such as "5G", "3.5GNR", etc.

[0118] As an example, in an emergency scenario of energy saving in a base station cell, such as Fig.10 As shown, the terminal can determine the associated cell search information (associated network unit search information) with the latitude and longitude of the primary selected cell as the center and the search range as the radius according to the latitude and longitude in the cell location information (unit location information) of the primary selected cell (primary selected network unit) and the search range in the regional search target information. The associated cell search information can be used to search the associated cells (associated network units) of the primary selected cell to obtain the generalized cells (generalized network units) of the primary selected cell. In this way, the primary selected cells and other primary selected network units can be generalized with the help of the latitude and longitude and search range information of the primary selected cells and other primary selected network units.

[0119] After obtaining one or more of the associated address keywords, associated address nouns and associated network unit search information, one or more of the associated address keywords, associated address nouns and associated network unit search information can be used as generalized location information corresponding to the unit location information of the primary network unit, thereby obtaining the generalized network unit according to the network unit query condition corresponding to the generalized location information, thereby achieving generalized retrieval of the network unit through multiple generalization paradigms on the basis of accurate retrieval of the initial network unit, thereby providing full guarantee for energy-saving controls such as energy-saving wake-up of the network unit.

[0120] In an exemplary embodiment, further, the area range information in the above-mentioned area search target information is derived from the wake-up requirement information corresponding to the degree of generalization; and / or the associated network unit feature information in the area search target information is derived from the wake-up requirement information corresponding to the unit type; wherein the control requirement refinement target under the energy-saving wake-up operation category includes the degree of generalization and / or the unit type.

[0121] Among them, as mentioned above, the generalization degree can be used as a wake-up demand refining target, and the wake-up demand information corresponding to the generalization degree can be used to indicate the generalization degree of the subsequent generalization processing of the primary network unit. For example, the wake-up demand information "1 km" corresponding to the generalization degree can be used as the source of regional range information. As mentioned above, the unit type can also be used as a wake-up demand refining target, and the wake-up demand information corresponding to the unit type can be used to indicate the unit type targeted by the network unit energy-saving control. For example, the wake-up demand information "5G" and "3.5GNR" corresponding to the unit type can be used as the source of the associated network unit characteristic information. Therefore, in the search for associated network units, the user does not need to provide additional information, and the regional range information and associated network unit characteristic information in the regional search target information can be obtained based on the wake-up demand information corresponding to the description information, thereby improving the efficiency and intelligence of the energy-saving control processing of the network unit.

[0122] In an exemplary embodiment, determining the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model in step S203, and generalizing the preliminary selected network unit to obtain the target network unit set may include:

[0123] If the operation category belongs to energy-saving shutdown, the preliminary network unit is determined according to the shutdown requirement information corresponding to the description information obtained by the large model, and the preliminary network unit is generalized based on the unit performance information of the preliminary network unit to obtain the target network unit set.

[0124] In this embodiment, if the operation category is further identified as energy-saving shutdown through the large model, the terminal can further obtain shutdown requirement information corresponding to the description information through the large model. The shutdown requirement information can be extracted from the description information through the large model. The shutdown requirement information refers to relevant information used to indicate its shutdown requirement. As an example, the shutdown requirement information may include unit performance information, unit location information, unit type information, etc. of the shutdown requirement. As an example, the unit performance information may include signal strength, number of users, signal-to-noise ratio, etc.

[0125] After obtaining the shutdown requirement information, the terminal can determine the pre-selected network unit according to the shutdown requirement information, and generalize the pre-selected network unit based on the unit performance information of the pre-selected network unit, so as to obtain a target network unit set. For example, other base station cells (generalized network units) with similar unit performance information within a certain area range of the pre-selected base station cell (the area range can be preset) and the pre-selected base station cell can be used as a target network unit set. Among them, for the pre-selected network unit, it can also be determined by referring to the relevant means of energy-saving wake-up mentioned above. For example, the shutdown requirement information can be conditionally escaped to obtain the network unit query condition, and the pre-selected network unit is obtained by querying the network unit query condition and the working parameter table, wherein the shutdown requirement information can also include the first-level division A1 and the second-level division A2, the unit type is 5G, etc. Among them, for the determination of the above-mentioned generalized network unit, it can also be obtained by querying the corresponding network unit query conditions such as the identification information (base station identification-cell identification, etc.) of other base station cells with similar unit performance information. Therefore, the solution of this embodiment can obtain the target network unit set range adapted for energy-saving shutdown.

[0126] In an exemplary embodiment, before obtaining the energy-saving control processing scheme for the target network unit set to be executed by the target network unit set according to the operation category and the unit state information and the available execution tool set corresponding to the target network unit set by using the large model in step S204, the method of the present application may further include the following steps:

[0127] According to the target network unit set, the unit state model is called to obtain the unit state information corresponding to the target network unit set.

[0128] In this embodiment, the terminal can call the unit state model to obtain the unit state information of each target network unit according to each target network unit in the target network unit set, so as to obtain the unit state information corresponding to the target network unit set. As an example, the unit state information may include the current energy-saving state, the whitelist state, etc., and each unit state information can be obtained through the corresponding unit state model. For example, the small model of "current energy-saving state" can be called to obtain the current energy-saving state, and the small model of "whitelist state" can be called to obtain the whitelist state, etc. Therefore, the solution of this embodiment can perform energy-saving control processing in a collaborative manner between a large model and a small model, thereby improving its processing efficiency and accuracy.

[0129] In some embodiments, step S204, according to the operation category, using the large model according to the unit status information corresponding to the target network unit set and the available execution tool set, obtains the energy-saving control processing scheme for the target network unit set to execute, which may include:

[0130] According to the unit status information and processing prompt information corresponding to the target network unit set, the large model is called to obtain the energy-saving control processing scheme for the target network unit set to execute, wherein the processing prompt information includes the available execution tool set and the processing logic corresponding to the operation category.

[0131] In this embodiment, the terminal can call the large model according to the unit state information and processing prompt information corresponding to the target network unit set, and the processing prompt information includes a set of available execution tools and processing logic corresponding to the operation category. Among them, the processing logic may include information such as the use of the available execution tools under the operation category for various unit state information, and the operation category may include energy-saving shutdown, energy-saving wake-up, etc. Therefore, the large model can obtain an energy-saving control processing scheme for the target network unit set to execute according to the unit state information corresponding to the target network unit set. Among them, according to different operation categories, the corresponding energy-saving control processing scheme can be obtained, for example, for energy-saving shutdown, a shutdown processing scheme can be obtained, for energy-saving wake-up, a wake-up processing scheme can be obtained, and so on.

[0132] As an example, in an emergency scenario of energy saving of a base station cell, the terminal can provide the cell status information corresponding to the target cell set (target network unit set) to the large model in the form of cell list information (which may include base station name, cell name, base station ID, cell ID, longitude and latitude, address information, energy saving start time, energy saving end time, current energy saving state, shutdown type, latest execution time, whitelist state, etc.) containing the cell status information (unit status information) corresponding to the target cell set (target network unit set). The large model can obtain an emergency processing plan (energy saving control processing plan) for execution of the target cell set based on the cell status information corresponding to the target cell set in the cell list information and other information, combined with the processing prompt information of emergency wake-up. Among them, for emergency scenarios of energy saving of base station cells, such as Fig.11 Shows the paradigm for processing prompt information.

[0133] As an example, in an emergency scenario of energy saving in a base station cell, such as Fig.12 As shown, the terminal can provide the cell list information and the corresponding processing prompt information to the big model. The cell list information may include the base station name, cell name, base station identifier (ID), cell identifier (ID), latitude and longitude, address information, energy saving start time, energy saving end time, current energy saving status, shutdown type, latest execution time, whitelist status, etc. The big model selects and executes the available execution tools for each cell in the cell list information, and can obtain and execute the emergency processing plan for the target cell set, and can also collect the execution results, which may include shutdown type, success, failure, waiting, timeout and execution log, etc.

[0134] The solution of this embodiment can accurately and effectively guide the large model to obtain and execute the energy-saving control processing solution and obtain the execution result of the energy-saving control processing solution according to the unit status information corresponding to the target network unit set.

[0135] In an exemplary embodiment, obtaining the processing summary information for the energy-saving control of the network unit by using the large model according to the description information and the execution result of the energy-saving control processing scheme in step S205 may include:

[0136] According to the description information, the execution results of the energy-saving control processing solution and the first summary prompt information, the large model is called to obtain the processing summary information; the first summary prompt information is used to indicate whether the large model analysis execution results meet the description information.

[0137] In this embodiment, the terminal can call the big model based on the execution results of the description information and the energy-saving control processing plan and the first summary prompt information. The first summary prompt information can be used to indicate the satisfaction of the big model analysis execution results with the description information, such as whether the execution results can meet the requirements reflected by the description information, the degree to which the execution results meet the requirements, etc.

[0138] As an example, in an emergency scenario of energy saving in a base station cell, such as Fig.13 The paradigm of the first summary prompt information is shown in . Based on this, the large model can analyze and summarize according to the description information and the execution results of the wake-up processing solution to obtain the processing summary information.

[0139] As an example, in an emergency scenario of energy saving in a base station cell, such as Fig.14 and Fig.15 As shown, the terminal can provide the description text (description information), the execution results of the emergency handling plan and other information as well as the first summary prompt information to the large model. The large model can analyze and summarize based on the information provided, and can give corresponding result summaries or suggestions. It can also display its analysis process of the execution results to obtain processing summary information for users to refer to the execution status and results of this emergency wake-up, for example.

[0140] In an exemplary embodiment, after obtaining the description information for the energy-saving control of the network unit in step S201, the method of the present application may further include the following steps:

[0141] If the large model recognizes that the intent category corresponding to the description information belongs to energy-saving consultation, then based on the description information and the energy-saving knowledge base, the large model is called to obtain the energy-saving consultation answer corresponding to the description information.

[0142] Among them, the intent category corresponding to the description information can be identified through the large model. The intent category is used to represent the intent corresponding to the description information. Different intent categories can be processed through different thinking chains, and each thinking chain can correspond to an intent category. The intent categories corresponding to the description information may include energy-saving consultation and energy-saving control operations, etc. Among them, thinking chain 1 may correspond to energy-saving consultation, thinking chain 2 may correspond to energy-saving control operations, and so on. As an example, in an emergency scenario of energy saving in a base station cell, energy-saving consultation may be an emergency consultation, and energy-saving control operations may be emergency operations.

[0143] In this embodiment, if the large model identifies that the intent category corresponding to the description information belongs to energy-saving consultation, the terminal can process it through the thinking chain 1 corresponding to the energy-saving consultation. In this thinking chain 1, the terminal can further call the large model to obtain the energy-saving consultation answer corresponding to the description information based on the description information and the energy-saving knowledge base. Among them, the energy-saving knowledge base can be knowledge related to the energy-saving control of the network unit, which can be provided to the large model for use to answer the description information. In the emergency scenario of energy saving in the base station cell, the energy-saving knowledge base can be an emergency knowledge base, which can store emergency knowledge related to base station energy saving. Specifically, in thinking chain 1, the terminal can use RAG (Retrieval-Augmented Generation) and the energy-saving knowledge base to realize retrieval-augmented answer description information, and obtain the energy-saving consultation answer corresponding to the description information. In this way, the description information belonging to energy-saving consultation can be quickly and accurately answered with the help of the large model and the energy-saving knowledge base.

[0144] In an exemplary embodiment, further, the method of the present application may also include the following steps:

[0145] According to the description information, the energy-saving consultation answer and the second summary prompt information, the big model is called to obtain the consultation summary information for the network unit energy-saving control; the second summary prompt information is used to instruct the big model to analyze the satisfaction of the energy-saving consultation answer with the description information.

[0146] In this embodiment, for the descriptive information belonging to energy-saving consultation, the terminal can call the big model according to the descriptive information, the answer to the energy-saving consultation and the second summary prompt information. The second summary prompt information can be used to instruct the big model to analyze the satisfaction of the answer to the energy-saving consultation with the descriptive information, such as whether the answer to the energy-saving consultation can answer the relevant questions in the descriptive information, how many questions in the descriptive information can be answered, etc. Based on this, whether in Thinking Link 1 or Thinking Link 2, the big model can be used to analyze and summarize the descriptive information and its corresponding response results, promoting the full chain automation of energy-saving control processing of the network unit.

[0147] In an application embodiment, taking the emergency scenario of energy saving of a base station cell as an example, the energy saving processing method of the network unit of the present application is described, such as Fig.16 As shown, this embodiment provides an energy-saving processing method for a network unit.

[0148] The method of this application embodiment can provide intent insights, tool recommendations, and automatic closed-loop capabilities for emergency response by introducing the intelligent features of semantic analysis, retrieval generation, reasoning generalization, and text summarization of large models in the key links of emergency support. It can help to quickly identify and judge operations in the regional scope, realize end-to-end drive from emergency intent text to live network operations, and adaptively select tools to arrange execution. It is an application example of embodied intelligence. The method of this application embodiment can effectively improve the flexibility and intelligence level of emergency response, alleviate the difficulties of maintenance personnel's tedious hard queries and the difficulty of some community names being hit by intent, and connect factors such as user intent, community collection, and behavioral operations.

[0149] In the overall architecture, the method of this application embodiment can construct an emergency thinking chain and 5 types of prompt projects, apply multiple rounds of large models for complex reasoning and planning, and cooperate with the emergency knowledge base, tool library, community engineering parameter library, and energy-saving status small model to achieve data screening generalization, tool execution, and text summary. In the process, key thinking steps and input and output can be fed back to guide users to make decisions. Among them, the steps can be decomposed into 7 steps and divided into 4 stages.

[0150] In Phase 1: Steps 1 to 3 can perform emergency intent classification and achieve business diversion through classification prompt engineering.

[0151] In Phase 2: Step 4, emergency area (target cell set) identification can be performed. Among them, the two-stage enhanced Text2SQL (text to SQL) method of precise screening and generalized retrieval is used to improve the accuracy and recall rate of area identification. The primary selection can use the named body to extract key label information to obtain accurate primary cells (sets). Then, the primary cells (sets) are generalized through address approximate blood relationship keywords, knowledge graphs, and geometric longitude and latitude methods to obtain a more secure target cell set.

[0152] In stage 3: Steps 5 to 6 are status analysis and orchestration execution. The energy-saving status and whitelist status can be obtained through a small model according to the target cell set. The tool adaptive selection is completed by building a large model of the orchestration tool prompt words.

[0153] In Phase 4: Step 7, a comprehensive analysis and summary of the end-to-end process of user intentions and implementation results is conducted. Emergency consultations can be answered directly in conjunction with the knowledge base.

[0154] like Fig.16As shown, the method of this application embodiment can enhance the comprehensive balance of emergency queries and operations in terms of accuracy and recall by introducing large-model semantic analysis, retrieval generation, generalization, and summarization capabilities, adopting technologies such as named entity recognition and word vector blood relationship, and effectively guiding large and small collaboration and tool use through prompt engineering.

[0155] Step 1: Emergency intent recognition and classification. User questions or third-party source text W1 is used as input. In some practical cases, the whole process can be automated by connecting to third-party emergency information sources. The large model can be called by building a "classification" prompt engineering paradigm A and returning the classification result S1. Its paradigm and examples are as follows: Figure 3 and 4 In some practical cases, the two chains of thinking can be divided into step 2.

[0156] Step 2 (Thinking Chain 1): Emergency Consultation. After the user asks W1 and identifies S1 “Emergency Consultation”, the RAG and emergency knowledge base can be used to implement retrieval-enhanced answer R2.1, and then jump to step 7.

[0157] Step 2 (Thinking Chain 2): Emergency operation. According to W1, get S1 and jump to step 3.

[0158] Step 3: Identification and classification of operation categories (operation intentions). Based on the classification result S1 and the source text W1, multiple sources are concatenated into W2 to obtain the operation classification S2. This "classification" method can also come from the prompt engineering paradigm A, and the preset category content can be updated.

[0159] Step 4: Identify the emergency area (target cell set). Fig.17 As shown, Fig.16 The expansion method of "target cell identification" in the service inputs the question text and outputs the regional cell list. The method can be divided into (1) named body recognition and (2) address noun knowledge graph / word vector lineage / geometric longitude and latitude. The target cell set is determined by precise screening and generalized retrieval of each level of Text2SQL architecture. The semantic recognition and generalization capabilities of the large model can be used multiple times and provide users with secondary confirmation. Among them, as mentioned above, in the energy-saving shutdown scenario, the shutdown requirement information can also be converted into a network unit query condition through conditional escape, and the preliminary network unit can be obtained by querying the network unit query condition and the work parameter table. Then, the target network unit set can be obtained according to the corresponding network unit query condition query based on the identification information of other base station cells with similar unit performance information. That is, in the energy-saving shutdown scenario, the target cell set can also be determined by precise screening and generalized retrieval of each level of Text2SQL architecture.

[0160] (1) First precision screening.

[0161] A "regional naming body" can be constructed to prompt the engineering paradigm B. The paradigm is as follows Figure 5 As shown, the example can be Figure 6 As shown. The words of location information, unit type, and degree of generalization can be extracted from the text W1. The backward model can be improved in comprehensibility through naming association completion and content standardization (for example, if the user input only provides "x city" without the division unit, the first-level division A1 and the second-level division A2 can be automatically completed through the large model to initially screen the preliminary selected cells. Among them, the method can be to combine SQL through where condition mapping. For example, if the location contains the first-level division A1, the second-level division A2, and the unit type is 5G, the SQL condition after condition escape is select*from the industrial parameter table where the first-level division ID=123and the second-level division ID=321andnetwork=5G. After executing the cell query condition, the preliminary selected cells can be queried. Thereafter, the address information such as the base station name, cell name and cell address of the preliminary selected cell can be combined and its key address noun set F1 can be generated after word segmentation. The unit type and degree of generalization information are optional. If missing, the full amount or default value can be used.

[0162] (2) Secondary generalization search: The following three methods can be used to generalize the data respectively, and the results are combined to obtain the generalized target cell set.

[0163] 1) Pre-construct a knowledge graph of address keywords. The method can be to splice the base station names, community names, various levels of divisions and address information of all cells, and then follow the implementation sequence of cleaning, attribute extraction, word segmentation, clustering, dimensionality reduction, and association. According to the theme of "life facilities", the relationship vectors between words are obtained, and the knowledge graph is constructed with the theme nouns as points and the relationships as edges, such as Figure 7 As shown, the subject words within 1 degree of distance from the F1 word set are selected for secondary retrieval.

[0164] 2) The word vector method can be to construct a "regional similarity word" prompt engineering paradigm C, such as Figure 8 As shown, for example Fig. 9 As shown. The large model can be called and the resulting nouns can be parsed for secondary retrieval, and naming relationship completion and standardization similar to step 4(1) can be applied.

[0165] 3) The latitude and longitude method can be to use the latitude and longitude of the initial selected cell as the center of the circle, and use the distance extracted from the text information as the radius to implement regional expansion, and obtain the generalized cell under the region, such as Fig.10 shown.

[0166] Step 5: Status analysis. According to the query result cell (divisions at all levels, base station ID, cell ID, time), call small models such as "current energy-saving status" and "white list status" to output cell list information (for example, energy-saving services may not be limited to: base station name, cell name, base station ID, cell ID, longitude and latitude, address, energy-saving start time, energy-saving end time, current energy-saving status, shutdown type, latest execution time, white list, etc.).

[0167] Step 6: Orchestrate execution. Based on the list L and tags output in step 5 and the set of available execution tools T, we construct an "orchestration tool" to prompt the engineering paradigm D, and select the optimal tools and execution for each cell. The paradigm is as follows: Fig.11 As shown, for example Fig.12 . The execution results (such as shutdown type, success, failure, waiting, timeout, execution log, etc.) can be collected.

[0168] Step 7: Text summary. The text information obtained in the above steps (including R2.1, S1, W2, Result, etc.) can be integrated, and the "text summary" can be used to prompt the engineering paradigm E, give the corresponding result summary or suggestions, and output the emergency treatment summary information R2.2. Its paradigm is as follows Fig.13 As shown, for example Fig.14 and Fig.15 shown.

[0169] As a specific example of emergency wake-up, the scenario and process of implementing the service of the method of this application embodiment are as follows: Fig.18 As shown. Specifically:

[0170] 1) Source text: "An emergency occurred in the new town x in area A, and all 3.5G NR energy-saving cells of the 5G network in the area and within a radius of 1 km were awakened."

[0171] 2) Based on intent recognition, such as Fig.19 get:

[0172] Thinking chain 2: emergency operation; operation category: emergency awakening.

[0173] 3) Emergency area (target cell set) identification:

[0174] (1) The first precise screening, through the nomenclature screening, obtained:

[0175] First-level zoning: A1; Second-level zoning: A2; Emergency address information: x New Town;

[0176] Equipment: 5G 3.5GNR; Cell: 001-1; Generalization: 1000m distance;

[0177] (2) Secondary generalization retrieval, we get:

[0178] Approximate address keywords, such as xxx Road, xxx Road, etc. Secondary search areas can include 0005 / 6 / 7, etc.

[0179] Latitude and longitude: With cell 001-1 as the center, obtain other 5G 3.5G NR cells within a radius of 1000m. Secondary search can output cells such as 002 / 3 / 4.

[0180] Address knowledge graph: Select address keywords within 1 degree of fine-screening words, such as y4 primary school, y7 subway station, etc. Secondary search of communities can include 0038 / 39, etc.

[0181] 4) Status analysis and orchestration execution: The above cell sets can be combined to call the small model to obtain the corresponding whitelist status and energy-saving status, and the corresponding tool recommendation "tool to be executed" can be obtained through the "tool selection" capability. The following table 1 shows some implementation examples:

[0182] Table 1

[0183]

[0184] 5) Text summary:

[0185] The large model can generate summary text as follows:

[0186] Executive summary: In the emergency incident of x new town in A area, a total of 6 3.5GNR energy-saving cells of 5G network were tried to be awakened. 5 cells were successfully awakened, accounting for 83.33% of the total attempted awakening cells. One cell failed to be awakened successfully, and the failure reason was "timeout".

[0187] Conclusion: The execution results basically meet the user's requirements, because most of the cells were successfully awakened, but one cell failed to be awakened successfully due to "timeout". It is necessary to further investigate the specific reasons why the cell failed to be awakened, and take corresponding measures to ensure that all cells can respond in time in emergency situations.

[0188] Based on the embodiments of the present application, the present application realizes energy-saving control intention recognition and classification selection through a large language model, and diverts to different thinking chains. In one type of thinking chain case, after the operation intention is recognized, the energy-saving control area recognition and performance status analysis can be realized through conversational data acquisition, command body recognition, enhanced Text2Sql, and large and small collaboration. During this period, the generalized retrieval and precise screening of the cell can be realized through the word vector kinship, knowledge graph text matching, and geometric longitude and latitude. Then, the tool and orchestration execution can be adaptively selected through the semantic analysis capability of the large model, such as providing intelligent orchestration to wake up the energy-saving sector. Finally, the implementation results and questions and answers can be summarized. This method can be oriented to scenarios such as emergency protection of network unit energy-saving control, combined with the key technologies of AI large models, so that the processing of network unit energy-saving control has the characteristics of efficient, flexible and generalized intent recognition classification, end-to-end, and regional recognition, giving play to the advantages of intelligence, and is expected to realize unmanned emergency response, release manual 24-hour standby, and improve the level of network protection.

[0189] Specifically, the method of the present application has the following characteristics as a whole:

[0190] 1) Strong end-to-end practicality: It has end-to-end solutions in multiple stages, including intelligent identification of users' emergency intentions, accurate identification of network units, adaptive selection of processing tools, and post-event result summary.

[0191] 2) High level of intelligence: Combining the semantic analysis, retrieval generation, reasoning generalization, and text summarization features of large models, the intelligence level of energy-saving operations is improved, and data interpretation, policy insights, intelligent orchestration, and automatic closed-loop capabilities are provided. It can accommodate more user energy-saving control needs and intelligently connect factors such as user intentions, cell collections, and behavioral operations.

[0192] 3) Identification guarantee: Improve the recognition accuracy and recall rate of network units and achieve an effective balance between the two. Use a two-level query structure + three generalization capabilities to provide decision support that maximizes the guarantee effect.

[0193] 4) Efficient and unmanned emergency operation: It can solve the problem of long multi-departmental collaboration process in energy-saving control scenarios such as emergency, alleviate the complicated hard query operation, and achieve what you want is what you get. In addition, by introducing third-party (emergency) information sources, it can help achieve full unmanned automation.

[0194] 5) Flexibility: Tool categories, thought chain categories, and operation categories can be quickly adjusted or expanded to enhance the flexibility of energy-saving control operations and have adaptability to multiple scenarios.

[0195] 6) Security: It can ensure the continuity and security of communications between network units such as base stations.

[0196] 7) Summary: By collecting objective online energy-saving control processing implementation data and combining it with the energy-saving knowledge base, it supports post-data analysis and review.

[0197] In addition, the method of the present application is practical:

[0198] 1) It can be applied to, for example, refined management of base station cell locations and rapid identification of regional cell sets based on semantics. 2) It can be applied to scenarios such as smart energy saving and predictive maintenance of base station cells, and can be extended to pan-emergency scenarios to improve the comprehensive security capabilities of cloud networks. 3) It is conducive to unmanned operation, maximized security, and improved energy-saving and safety operation and maintenance efficiency.

[0199] In the method of the present application, the 4-stage 7-step method based on intention recognition and thinking chain classification includes intention recognition classification, operation recognition classification, energy-saving consultation and classification drive of energy-saving control operations, regional recognition, state analysis, arrangement execution, and text summary, which effectively improves the semantic understanding of the description information of the user's energy-saving control, and realizes intelligent selection tool arrangement and end-to-end execution, with the advantages of building efficient, flexible, easy to expand, and intelligent interaction. In regional recognition, it can include the use of a two-level architecture enhanced Text2SQL recognition, which is the first accurate screening and the second generalized retrieval, which can enhance the comprehensive balance of energy-saving control (energy-saving shutdown, energy-saving wake-up, etc.) queries in terms of accuracy and recall. In addition, other processes of regional recognition can also include named body label recognition based on a large model, approximate subject word vector generalization, knowledge graph generalization, and geometric longitude and latitude combination generalization methods, which increase the recognition of recognition and the interpretability of the call through the collaboration of large and small models. This method can meet the requirements of emergency and high-security scenarios under the difficult problems of limited recognition of location parameters and diverse specifications, and efficiently screen cell sets. The method of this application also applies various prompt engineering paradigms, which are reusable and help guide the large model to think, analyze, and make decisions in the direction of user intentions.

[0200] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0201] Based on the same inventive concept, the embodiment of the present application also provides an energy-saving processing device for a network unit for implementing the energy-saving processing method for the network unit involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the energy-saving processing device embodiments of one or more network units provided below can refer to the limitations of the energy-saving processing method for the network unit above, and will not be repeated here.

[0202] In an exemplary embodiment, Fig. 20 As shown, a network unit energy-saving processing device is provided, and the device 2000 may include:

[0203] The information acquisition module 2001 is used to acquire the description information for the energy saving control of the network unit;

[0204] A category identification module 2002 is used to obtain the operation category corresponding to the description information through the large model if the large model identifies that the intention category corresponding to the description information belongs to the energy-saving control operation;

[0205] The unit determination module 2003 is used to determine the preliminary network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalize the preliminary network unit to obtain a target network unit set;

[0206] The energy-saving control module 2004 is used to obtain an energy-saving control processing scheme for execution of the target network unit set according to the operation category and using the large model according to the unit state information corresponding to the target network unit set and the available execution tool set;

[0207] The processing summary module 2005 is used to obtain processing summary information for network unit energy-saving control based on the large model according to the description information and the execution result of the energy-saving control processing solution.

[0208] In an exemplary embodiment, the category identification module 2002 is also used to call the large model to identify the intent category corresponding to the description information from multiple intent categories based on the description information and the first classification prompt information; the first classification prompt information contains the multiple intent categories.

[0209] In an exemplary embodiment, the category identification module 2002 is used to call the large model to identify the operation category corresponding to the description information from multiple operation categories based on the description information, the intention category corresponding to the description information, and the second classification prompt information; the second classification prompt information contains the multiple operation categories.

[0210] In an exemplary embodiment, the unit determination module 2003 is also used to call the large model to obtain control requirement information corresponding to the control requirement refining target according to the description information and the refined prompt information; the refined prompt information includes the control requirement refining target; the control requirement refining target is determined according to the operation category.

[0211] In an exemplary embodiment, the unit determination module 2003 is used to determine the preliminary network unit according to the wake-up requirement information corresponding to the description information obtained by the large model if the operation category belongs to energy-saving wake-up, and generalize the preliminary network unit based on the unit location information of the preliminary network unit to obtain a target network unit set.

[0212] In an exemplary embodiment, the unit determination module 2003 is used to query and obtain the preliminary selected network unit according to the network unit query condition corresponding to the wake-up requirement information.

[0213] In an exemplary embodiment, the unit determination module 2003 is used to obtain the generalized location information corresponding to the unit location information of the preliminary selected network unit; query and obtain the generalized network unit according to the network unit query condition corresponding to the generalized location information; and obtain the target network unit set according to the preliminary selected network unit and the generalized network unit.

[0214] In an exemplary embodiment, the unit determination module 2003 is used to obtain, in the knowledge graph, associated address keywords having a target association relationship with the unit location information of the preliminary network unit according to the unit location information of the preliminary network unit; the knowledge graph is constructed based on each address keyword and the association relationship between each address keyword; the address keyword and the association relationship are determined based on the unit location information and target topic of each network unit; and / or, based on the unit location information and search prompt information of the preliminary network unit, call the large model to search for associated address nouns associated with the unit location information of the preliminary network unit; the search prompt information is used to guide the large model to obtain the associated address nouns based on regional search; and / or, based on the unit location information and regional search target information of the preliminary network unit, obtain the associated network unit search information; the regional search target information includes regional range information and / or associated network unit feature information; the generalized location information is obtained according to one or more of the associated address keywords, the associated address nouns and the associated network unit search information.

[0215] In an exemplary embodiment, the area range information in the area search target information is derived from the wake-up requirement information corresponding to the degree of generalization; and / or the associated network unit characteristic information in the area search target information is derived from the wake-up requirement information corresponding to the unit type; wherein the control requirement refinement target under the energy-saving wake-up operation category includes the degree of generalization and / or the unit type.

[0216] In an exemplary embodiment, the unit determination module 2003 is used to determine the preliminary network unit according to the shutdown requirement information corresponding to the description information obtained by the large model if the operation category belongs to energy-saving shutdown, and generalize the preliminary network unit based on the unit performance information of the preliminary network unit to obtain a target network unit set.

[0217] In an exemplary embodiment, the energy-saving control module 2004 is further configured to call a unit state model according to the target network unit set to obtain unit state information corresponding to the target network unit set.

[0218] In an exemplary embodiment, the processing summary module 2005 is used to call the large model to obtain the processing summary information based on the description information, the execution results of the energy-saving control processing scheme and the first summary prompt information; the first summary prompt information is used to instruct the large model to analyze the satisfaction of the execution results with the description information.

[0219] In an exemplary embodiment, the category identification module 2002 is also used to call the big model to obtain the energy-saving consultation answer corresponding to the description information based on the description information and the energy-saving knowledge base if the big model identifies that the intention category corresponding to the description information belongs to energy-saving consultation.

[0220] In an exemplary embodiment, the processing summary module 2005 is also used to call the big model to obtain the consulting summary information for the energy-saving control of the network unit based on the descriptive information, the energy-saving consulting answer and the second summary prompt information; the second summary prompt information is used to instruct the big model to analyze the satisfaction of the energy-saving consulting answer with the descriptive information.

[0221] Each module in the energy-saving processing device of the above network unit can be implemented in whole or in part by software, hardware and a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module above.

[0222] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.21 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (Near Field Communication, NFC) or other technologies. When the computer program is executed by the processor, a method for energy-saving processing of a network unit is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0223] Those skilled in the art will understand that Fig.21 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0224] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0225] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0226] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0227] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0228] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0229] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0230] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for energy saving of a network unit, characterized in that: The method comprises: Obtaining description information for energy-saving control of network units; If the large model identifies that the intention category corresponding to the description information belongs to an energy-saving control operation, obtaining the operation category corresponding to the description information through the large model; Determine a preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalize the preliminary selected network unit to obtain a target network unit set; According to the operation category, using the large model, according to the unit state information corresponding to the target network unit set and the available execution tool set, an energy-saving control processing scheme for execution of the target network unit set is obtained; The large model is used to obtain processing summary information for network unit energy-saving control according to the description information and the execution result of the energy-saving control processing solution.

2. The method according to claim 1, characterized in that After acquiring the description information for the energy-saving control of the network unit, the method further includes: According to the description information and the first classification prompt information, the large model is called to identify the intent category corresponding to the description information from multiple intent categories; the first classification prompt information includes the multiple intent categories.

3. The method according to claim 1, characterized in that The obtaining, through the large model, an operation category corresponding to the description information includes: According to the description information, the intention category corresponding to the description information and the second classification prompt information, the large model is called to identify the operation category corresponding to the description information from multiple operation categories; the second classification prompt information includes the multiple operation categories.

4. The method according to claim 1, characterized in that: Before determining the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalizing the preliminary selected network unit to obtain a target network unit set, the method further includes: According to the description information and the refined prompt information, the large model is called to obtain the control requirement information corresponding to the control requirement refined target according to the description information; the refined prompt information includes the control requirement refined target; the control requirement refined target is determined according to the operation category.

5. The method according to claim 1, characterized in that The determining of the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalizing the preliminary selected network unit to obtain a target network unit set includes: If the operation category belongs to energy-saving wake-up, the preliminary network unit is determined according to the wake-up requirement information corresponding to the description information obtained by the large model, and the preliminary network unit is generalized based on the unit position information of the preliminary network unit to obtain the target network unit set.

6. The method according to claim 5, characterized in that The determining of the preliminary selected network unit according to the wake-up requirement information corresponding to the description information obtained by the large model includes: The preliminary selected network unit is obtained by querying according to the network unit query condition corresponding to the wake-up requirement information.

7. The method according to claim 5, characterized in that The generalizing the preliminarily selected network units based on the unit position information of the preliminarily selected network units to obtain a target network unit set includes: Obtaining generalized position information corresponding to the unit position information of the initially selected network unit; According to the network unit query condition corresponding to the generalized location information, query and obtain the generalized network unit; The target network unit set is obtained according to the initially selected network unit and the generalized network unit.

8. The method according to claim 7, characterized in that The obtaining of generalized location information corresponding to the unit location information of the preliminarily selected network unit includes: According to the unit location information of the preliminarily selected network unit, an associated address subject word having a target association relationship with the unit location information of the preliminarily selected network unit is obtained in the knowledge graph; the knowledge graph is constructed based on each address subject word and the association relationship between each address subject word; the address subject word and the association relationship are determined based on the unit location information of each network unit and the target subject; and / or, based on the unit location information and search prompt information of the preliminarily selected network unit, calling the large model to search for associated address nouns associated with the unit location information of the preliminarily selected network unit; the search prompt information is used to guide the large model to obtain the associated address nouns based on regional search; and / or, obtaining associated network unit search information according to the unit location information and the regional search target information of the preliminarily selected network unit; the regional search target information includes regional range information and / or associated network unit feature information; The generalized location information is obtained according to one or more of the associated address subject words, the associated address nouns and the associated network unit search information.

9. The method according to claim 8, characterized in that The area range information in the area search target information is derived from the wake-up requirement information corresponding to the degree of generalization; and / or the associated network unit feature information in the area search target information is derived from the wake-up requirement information corresponding to the unit type; wherein the control requirement refinement target under the energy-saving wake-up operation category includes the degree of generalization and / or the unit type.

10. The method according to claim 1, characterized in that The determining of the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalizing the preliminary selected network unit to obtain a target network unit set includes: If the operation category belongs to energy-saving shutdown, the preliminary network unit is determined according to the shutdown requirement information corresponding to the description information obtained by the large model, and the preliminary network unit is generalized based on the unit performance information of the preliminary network unit to obtain the target network unit set.

11. The method according to claim 1, characterized in that: Before obtaining, according to the operation category, an energy-saving control processing scheme for execution of the target network unit set by using the large model according to the unit state information corresponding to the target network unit set and a set of available execution tools, the method further includes: According to the target network unit set, a unit state model is called to obtain unit state information corresponding to the target network unit set.

12. The method according to claim 1, characterized in that The using of the large model to obtain processing summary information for network unit energy-saving control according to the description information and the execution result of the energy-saving control processing scheme includes: According to the description information, the execution result of the energy-saving control processing scheme and the first summary prompt information, the large model is called to obtain the processing summary information; the first summary prompt information is used to instruct the large model to analyze the satisfaction of the execution result with the description information.

13. The method according to any one of claims 1 to 12, characterized in that After acquiring the description information for the energy-saving control of the network unit, the method further includes: If the large model recognizes that the intent category corresponding to the description information belongs to energy-saving consultation, then based on the description information and the energy-saving knowledge base, the large model is called to obtain the energy-saving consultation answer corresponding to the description information.

14. The method according to claim 13, characterized in that The method further comprises: According to the description information, the energy-saving consultation answer and the second summary prompt information, the big model is called to obtain the consultation summary information for the energy-saving control of the network unit; the second summary prompt information is used to instruct the big model to analyze the satisfaction of the energy-saving consultation answer with the description information.

15. An energy-saving processing device for a network unit, characterized in that: The device comprises: An information acquisition module, used to acquire description information for energy-saving control of a network unit; a category identification module, configured to obtain the operation category corresponding to the description information through the large model if the large model identifies that the intention category corresponding to the description information belongs to an energy-saving control operation; A unit determination module, used to determine the preliminary selected network unit according to the operation category and the control requirement information corresponding to the description information obtained by the large model, and generalize the preliminary selected network unit to obtain a target network unit set; An energy-saving control module, configured to obtain an energy-saving control processing scheme for execution of the target network unit set according to the operation category and using the large model according to the unit state information corresponding to the target network unit set and a set of available execution tools; The processing summary module is used to obtain processing summary information for network unit energy-saving control based on the large model according to the description information and the execution result of the energy-saving control processing solution.

16. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 14 are implemented.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 14 are implemented.

18. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 14 are implemented.

Images