Energy management methods, devices, equipment, media and program products

Abstract

The present invention provides an energy management method, device, equipment, medium and program product, which are applied to the field of energy management technology. The method includes: obtaining the operation instruction of energy equipment and the energy management type corresponding to the operation instruction; preprocessing the operation instruction to obtain target text information, the target text information includes the operation information in the operation instruction, the keywords in the operation information and the context dependency in the operation information; according to the energy management type, the target text information is processed through the retrieval enhancement generation and reward rules in the pre-trained large language model to generate the control instructions corresponding to the operation instruction; and the control instructions are sent to the energy equipment. The efficiency of processing the operation instruction of energy equipment using the embodiment of the present invention is high, and the accuracy of the control instructions and the operation instruction is high.

Description

Technical Field

[0001] The present invention relates to the field of energy management technology, and in particular to an energy management method, device, equipment, medium and program product. Background Art

[0002] The Standard Operating Procedure (SOP) in the Energy Management System (EMS) is somewhat too complicated, causing confusion to operators. At the same time, due to insufficient internal personnel training in the enterprise, operators may not be able to perform operations correctly without adequate training. In addition, with the development of technology and the improvement of processes, the regular update of the operating instructions may cause operators to misunderstand and ignore the latest operating requirements. In addition, the understanding of the operating instructions is inconsistent, and different people may have different understandings of the same operating instructions, resulting in confusion during the execution process. Therefore, the current process of implementing standard operating instructions in energy management systems is full of uncertainty, with low execution efficiency and low accuracy. Summary of the invention

[0003] The purpose of the present invention is to provide an energy management method, device, equipment, medium and program product, which are used to solve the problem of low efficiency and low accuracy in executing standard operating instructions in energy management systems in the prior art.

[0004] In order to solve the above technical problems, an embodiment of the present invention provides an energy management method, including:

[0005] Obtaining an operating instruction of the energy equipment and an energy management type corresponding to the operating instruction;

[0006] Preprocessing the operation instruction to obtain target text information, wherein the target text information includes operation information in the operation instruction, keywords in the operation information, and context dependencies in the operation information;

[0007] According to the energy management type, the target text information is processed by using the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction;

[0008] The control instruction is sent to the energy device.

[0009] Optionally, the target text information is processed according to the energy management type by using the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction, including:

[0010] Based on the energy management type, the target text information is retrieved and enhanced by a pre-trained large language model to obtain code information corresponding to the work instruction;

[0011] The code information is scored according to the reward rule, and when the score of the code information is greater than a first threshold, a control instruction is generated according to the code information.

[0012] Optionally, the reward rules include one or more of the following:

[0013] When the accuracy rate between the code information and the operation instruction is greater than a second threshold, increasing the score of the code information;

[0014] When the generation time of the code information is less than a third threshold, increasing the score of the code information;

[0015] In the case where the code information includes an energy-saving strategy, increasing the score of the code information; wherein the energy-saving strategy is to adjust the energy consumption of the energy device according to the energy consumption status of the energy device;

[0016] In the case where the code information includes a non-safety policy, the score of the code information is reduced; wherein the non-safety policy is used to indicate a policy that causes a safety hazard to the energy device.

[0017] Optionally, based on the energy management type, the target text information is retrieved and enhanced by a pre-trained large language model to obtain code information corresponding to the work instruction, including:

[0018] Extracting questions from the target text information to obtain questions to be queried corresponding to the target text information;

[0019] Based on the energy management type, the query question is searched in multiple knowledge vector libraries corresponding to the energy management type by using a retrieval enhancement model in a pre-trained large language model, and the retrieval information corresponding to each of the knowledge vector libraries and the matching degree between the retrieval information and the query question are obtained;

[0020] The code information corresponding to the work instruction is generated according to the question to be queried and the search information whose matching degree is greater than a fourth threshold.

[0021] Optionally, the preprocessing of the work instruction to obtain target text information includes:

[0022] Removing irrelevant information and extracting structured information from the operation instruction in sequence to obtain operation information in the operation instruction;

[0023] Performing keyword recognition on the operation information to obtain keywords in the operation information;

[0024] Performing sentence segmentation and dependency analysis on the operation information to obtain context dependency in the operation information;

[0025] The target text information is constructed according to the operation information in the work instruction, the keywords in the operation information and the context dependency relationship in the operation information.

[0026] Optionally, the removing irrelevant information and extracting structured information from the operation instruction in sequence to obtain the operation information in the operation instruction includes:

[0027] Performing format conversion and denoising processing on the work instruction by using regular expressions or natural language processing technology to obtain first text information;

[0028] Structural information is extracted from the first text information to obtain operation information in the work instruction book.

[0029] Optionally, performing keyword recognition on the operation information to obtain keywords in the operation information includes:

[0030] The operation information is subjected to keyword recognition through named entity recognition technology NER and / or a dictionary corresponding to the energy device to obtain keywords in the operation information.

[0031] Optionally, performing sentence segmentation and dependency analysis on the operation information to obtain context dependency in the operation information includes:

[0032] Segment the operation information using a word segmentation tool in natural language processing technology to divide the operation information into multiple words and / or phrases;

[0033] Dependency analysis is performed on the words and / or phrases to obtain context dependencies in the operation information.

[0034] An embodiment of the present invention also provides an energy management device, including:

[0035] A first acquisition module is used to acquire an operation instruction of an energy device and an energy management type corresponding to the operation instruction;

[0036] A first processing module is used to pre-process the operation instruction to obtain target text information, wherein the target text information includes operation information in the operation instruction, keywords in the operation information, and context dependencies in the operation information;

[0037] A second processing module is used to process the target text information according to the energy management type through the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction;

[0038] The first sending module is used to send the control instruction to the energy device.

[0039] An embodiment of the present invention further provides a network device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program implements the energy management method as described in any one of the above items when executed by the processor.

[0040] An embodiment of the present invention further provides a readable storage medium, comprising: a program is stored on the readable storage medium, and when the program is executed by a processor, the steps of the energy management method as described in any one of the above items are implemented.

[0041] An embodiment of the present invention further provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps of any of the above energy management methods.

[0042] At least one of the above technical solutions of the present invention has the following beneficial effects:

[0043] In the above scheme, the operation instructions corresponding to the energy equipment and the energy management type corresponding to the operation instructions are obtained; the operation instructions are processed, first, the operation instructions are pre-processed, the operation information in the operation instructions, the keywords in the operation information and the context dependency in the operation information are extracted to form the target text information, and then, based on the different specific control methods corresponding to the operation instructions of different energy management types, the target text information is processed through the retrieval enhancement generation and reward rules of the large language model according to the energy management type of the operation instructions, and the control instructions corresponding to the operation instructions are generated; the corresponding control instructions are sent to the energy equipment to control the energy equipment to execute the contents in the operation instructions. The embodiment of the present invention can automatically process the operation instructions of the energy equipment, generate the corresponding control instructions, and the efficiency of processing the operation instructions is high, and the accuracy of the control instructions and the operation instructions is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the structure of an energy management system according to an embodiment of the present invention;

[0045] Figure 2 A schematic diagram of a flow chart of an energy management method according to an embodiment of the present invention;

[0046] Figure 3 Schematic diagram of the structure of an energy management device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] like Figure 1 As shown, the present invention provides an energy management system, which includes: an operation node 120, a large language model node 110 and an Internet of Things node 130; wherein the operation node 120 is used to manage relevant asset information of energy equipment, including receiving and managing standard operating instructions SOP; the large language model node 110 is used to process the standard operating instructions SOP and generate corresponding code information; the Internet of Things node 130 is used to convert the code information into corresponding control instructions, send corresponding instructions to the energy equipment, and receive the execution results fed back by the energy equipment.

[0049] Optionally, the operation node 120 also includes one or more of the following functions: providing a visual dashboard to display the real-time energy consumption of each energy area, energy equipment or energy system; automatically identifying abnormal energy consumption patterns and issuing timely alarms via email, text messages or system notifications; built-in data analysis engine to support trend analysis, peak load analysis, energy consumption baseline setting and comparison, etc.; allowing staff to set energy-saving and emission reduction targets, and track progress in real time to evaluate the achievement of targets.

[0050] Optionally, the IoT node 130 also includes the following functions: obtaining real-time and historical energy data and status monitoring of energy equipment managed by the operation node 120 .

[0051] like Figure 2 As shown, an embodiment of the present invention provides an energy management method, which is applied to the energy management system provided by the present invention, including:

[0052] Step S201, obtaining an operation instruction of an energy device and an energy management type corresponding to the operation instruction;

[0053] In step S201, the large language model node 110 obtains the Standard Operating Procedure (SOP) and the energy management type of the SOP stored in the operation node 120, wherein the energy management type includes but is not limited to power management, water management, and gas management.

[0054] Specifically, the work instructions are automatically classified through a text classification algorithm (such as a support vector machine or a deep learning model) to obtain the energy management type of the work instructions.

[0055] Generally, the operation instructions stored in the operation node 120 are uploaded by the staff, but the present invention is not limited to this.

[0056] Step S202, preprocessing the operation instruction to obtain target text information, wherein the target text information includes operation information in the operation instruction, keywords in the operation information, and context dependencies in the operation information;

[0057] In step S202, the large language model node 110 preprocesses the operation instruction and extracts the target text information, where the target text information includes but is not limited to the operation information in the operation instruction, the keywords in the operation information and the context dependency in the operation information; wherein the operation information includes but is not limited to the operation steps, parameter settings and conditional judgments, etc.

[0058] Step S203, according to the energy management type, the target text information is processed by using the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction;

[0059] In step S203, the pre-trained large language model is trained based on a historical data set of energy equipment, where the historical data set includes historical operation instructions and control instructions corresponding to the historical operation instructions;

[0060] The large language model node 110 uses a pre-trained large language model to process the target text information. During the processing, the Retrieval Augmented Generation (RAG) function and reward rules in the large language model are used to generate control instructions corresponding to the work instructions. In addition, during the processing using the large language model, the energy management type must also be considered to make the generated control instructions more practical and more accurate.

[0061] Step S204: sending the control instruction to the energy device.

[0062] In step S204, the IoT node 130 obtains a control instruction and sends the control instruction to the energy device.

[0063] In an embodiment of the present invention, the operation instructions of energy equipment are preprocessed, and the operation information in the operation instructions, the keywords in the operation information, and the context dependency in the operation information are extracted to form the target text information. Then, based on the different specific control methods corresponding to the operation instructions of different energy management types, the target text information is processed through the retrieval enhancement generation and reward rules of the large language model according to the energy management type of the operation instructions to generate control instructions corresponding to the operation instructions; the corresponding control instructions are sent to the energy equipment to control the energy equipment to execute the contents in the operation instructions. The embodiment of the present invention can automatically process the operation instructions of energy equipment, generate corresponding control instructions, and process the operation instructions with high efficiency and high accuracy of the control instructions and the operation instructions.

[0064] Optionally, the target text information is processed according to the energy management type by using the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction, including:

[0065] Based on the energy management type, the target text information is retrieved and enhanced by a pre-trained large language model to obtain code information corresponding to the work instruction;

[0066] The code information is scored according to the reward rule, and when the score of the code information is greater than a first threshold, a control instruction is generated according to the code information.

[0067] In an embodiment of the present invention, the context understanding ability of a large language model is utilized to parse the logical relationship between the operation steps corresponding to the operation information in the target text information. First, the target text information is converted into corresponding code information. Furthermore, in order to make the generated code information more accurate and safer, according to the energy management type corresponding to the target text information, the target text information is first RAGed in a database of the corresponding energy management type, and then the corresponding code information is generated. Then, based on the reward rules, the code information is scored. When the score of the code information is greater than a first threshold, the code information is converted into a control instruction.

[0068] Optionally, the reward rules include one or more of the following:

[0069] When the accuracy rate between the code information and the operation instruction is greater than a second threshold, increasing the score of the code information;

[0070] When the generation time of the code information is less than a third threshold, increasing the score of the code information;

[0071] In the case where the code information includes an energy-saving strategy, increasing the score of the code information; wherein the energy-saving strategy is to adjust the energy consumption of the energy device according to the energy consumption status of the energy device;

[0072] In the case where the code information includes a non-safety policy, the score of the code information is reduced; wherein the non-safety policy is used to indicate a policy that causes a safety hazard to the energy device.

[0073] In the embodiment of the present invention, when the accuracy rate between the generated code information and the operation instruction is high, it means that the operation information in the operation instruction is correctly converted into the code information, and a positive reward is given to the code information, that is, the score of the code information is increased;

[0074] In the case where the time required to generate the code information is relatively short, it indicates that the efficiency of converting the work instruction into the code information is relatively high, and a positive reward is given to the code information, that is, the score of the code information is increased;

[0075] In the case where the code information includes an energy-saving strategy, the energy consumption of the energy equipment can be automatically adjusted in different time periods to achieve the purpose of energy saving, and a positive reward is given to the code information, that is, the score of the code information is increased;

[0076] In the case where the code information includes a non-safety strategy, for example, the operation indicated in the code information may cause safety hazards such as overheating and short circuit, a negative reward is given to the code information, that is, the score of the code information is reduced.

[0077] It should be noted that in order to ensure positive rewards, an automated testing program can be used to evaluate the generated code information when pre-training the large language model.

[0078] Optionally, based on the energy management type, the target text information is retrieved and enhanced by a pre-trained large language model to obtain code information corresponding to the work instruction, including:

[0079] Extracting questions from the target text information to obtain questions to be queried corresponding to the target text information;

[0080] Based on the energy management type, the query question is searched in multiple knowledge vector libraries corresponding to the energy management type by using a retrieval enhancement model in a pre-trained large language model, and the retrieval information corresponding to each of the knowledge vector libraries and the matching degree between the retrieval information and the query question are obtained;

[0081] According to the question to be queried and the search information whose matching degree is greater than a fourth threshold, code information corresponding to the work instruction is generated.

[0082] In the embodiment of the present invention, the question to be queried in the target text information is firstly extracted, and then RAG is performed on the question to be queried to generate retrieval information of the question to be queried.

[0083] Finally, code information is generated based on the retrieval information. The RAG function in the large language model is implemented by the retrieval enhancement model in the large language model. The retrieval enhancement model includes multiple knowledge vector libraries. The specific process is as follows:

[0084] Step 1: Based on the keywords and context dependencies of the operation information in the target text information, question extraction is performed on the operation information to obtain the query questions corresponding to the target text information;

[0085] The second step is to search the query questions in multiple knowledge vector libraries corresponding to the energy management type through the retrieval enhancement model. Specifically, the query questions are searched according to the prompt words of the query questions to obtain the retrieval information fed back by the multiple knowledge vector libraries and the matching degree between the retrieval information and the query questions.

[0086] The third step is to sort the search information from high to low according to the matching degree, and determine the search information with a matching degree greater than a fourth threshold as the target search information;

[0087] The fourth step is to use the retrieval enhancement model to generate text enhancement for the query question and target retrieval information, and output the answer text;

[0088] The fifth step is to generate the code information corresponding to the work instruction based on the answer text.

[0089] Among them, in the process of generating code information in the fifth step, because under general conditions, the operation information includes complex conditional judgments and loop structures, that is, some operation steps need to be executed under specific conditions, and the generated answer text is similar to the operation information, and also includes complex conditional judgments and loop structures. Therefore, when generating code information, you can first generate an intermediate representation (such as pseudocode) through a large language model, and then generate the final required code information based on the intermediate representation. Among them, in the intermediate representation, you can use a data structure (such as JSON or XML) to represent the various operation steps of the target text information.

[0090] For example, the operation information includes operation steps, parameter settings and conditional judgments, where the operation steps are: start the device; the parameter settings are: voltage value 220V, current value 10A; the conditional judgment is: start the device when the temperature is <50℃. Before generating the final code information, define a JSON object as an intermediate representation to store the above operation information:

[0091] ```json

[0092] {

[0093] "operation":"Start device",

[0094] "parameters":[

[0095] {"name":"voltage","value":"220V"},

[0096] {"name":"Current","value":"10A"}

[0097] ],

[0098] "conditions":[

[0099] {"type":"if","condition":"Temperature<50℃","then":"Start the device"} ]

[0101] }.

[0102] Optionally, the preprocessing of the work instruction to obtain target text information includes:

[0103] Removing irrelevant information and extracting structured information from the operation instruction in sequence to obtain operation information in the operation instruction;

[0104] Performing keyword recognition on the operation information to obtain keywords in the operation information;

[0105] Performing sentence segmentation and dependency analysis on the operation information to obtain context dependency in the operation information;

[0106] The target text information is constructed according to the operation information in the work instruction, the keywords in the operation information and the context dependency relationship in the operation information.

[0107] In the embodiment of the present invention, irrelevant information includes but is not limited to headers, footers, and directories; keywords include common keywords and professional keywords, common keywords include but are not limited to device names, parameter values, and operation steps, and professional keywords are professional terms in the field of energy management;

[0108] Specifically, the operation of preprocessing the work instruction in step S202 is described as follows:

[0109] In the first step, irrelevant information is removed from the work instructions, redundant formatting marks are removed, and the plain text content is retained. Then, structured information is extracted from the retained plain text content to obtain the operation information in the work instructions. The operation information includes but is not limited to operation steps, parameter settings, and conditional judgments;

[0110] The second step is to perform keyword recognition on the operation information, obtain the keywords in the operation information and mark the keywords in the operation information, so as to facilitate better understanding and processing in subsequent large language model processing;

[0111] The third step is to first perform sentence segmentation on the operation information to divide the text information into words or phrases, and then perform dependency analysis based on the segmented words or phrases to extract the subject-predicate-object relationship and other important information to obtain the context dependency relationship in the operation information.

[0112] The operation information in the work instruction, the keywords in the operation information and the context dependency in the operation information together constitute the target text information to be processed.

[0113] Optionally, the removing irrelevant information and extracting structured information from the operation instruction in sequence to obtain the operation information in the operation instruction includes:

[0114] Performing format conversion and denoising processing on the work instruction by using regular expressions or natural language processing technology to obtain first text information;

[0115] Structural information is extracted from the first text information to obtain operation information in the work instruction book.

[0116] In an embodiment of the present invention, the work instruction is formatted and denoised by regular expressions or natural language processing technology (such as word segmentation or named entity recognition (NER) technology), irrelevant information is removed, and redundant formatting tags are removed to obtain first text information; then structured information is extracted from the first text information to obtain operation information in the work instruction.

[0117] Optionally, performing keyword recognition on the operation information to obtain keywords in the operation information includes:

[0118] The operation information is subjected to keyword recognition through named entity recognition technology NER and / or a dictionary corresponding to the energy device to obtain keywords in the operation information.

[0119] In the embodiment of the present invention, the keywords include common keywords and professional keywords. The common keywords include but are not limited to device names, parameter values, and operation steps, and the professional keywords are professional terms in the field of energy management.

[0120] For commonly used keywords, the named entity recognition technology NER is used to identify the commonly used keywords in the operation information, and the identified keywords are marked out; for terms in professional fields, the named entity recognition technology NER may not be able to recognize them, so the professional keywords in the operation information are identified through the pre-configured dictionary corresponding to the energy equipment, and the identified keywords are marked out.

[0121] Optionally, performing sentence segmentation and dependency analysis on the operation information to obtain context dependency in the operation information includes:

[0122] Segment the operation information using a word segmentation tool in natural language processing technology to divide the operation information into multiple words and / or phrases;

[0123] Dependency analysis is performed on the words and / or phrases to obtain context dependencies in the operation information.

[0124] In an embodiment of the present invention, a word segmentation tool in natural language processing technology performs sentence segmentation on the operation information, dividing the operation information into multiple words or phrases; then, dependency analysis is performed based on the segmented words or phrases to extract the subject-predicate-object relationship and other important information, thereby obtaining the context dependency relationship in the operation information.

[0125] like Figure 3 As shown, an embodiment of the present invention also provides an energy management device, including:

[0126] A first acquisition module 301 is used to acquire an operation instruction of an energy device and an energy management type corresponding to the operation instruction;

[0127] A first processing module 302 is used to pre-process the operation instruction to obtain target text information, where the target text information includes operation information in the operation instruction, keywords in the operation information, and context dependencies in the operation information;

[0128] The second processing module 303 is used to process the target text information according to the energy management type through the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction;

[0129] The first sending module 304 is used to send the control instruction to the energy device.

[0130] Optionally, the second processing module 303 includes:

[0131] A first processing unit is used to retrieve, enhance and generate the target text information through a pre-trained large language model based on the energy management type, and obtain code information corresponding to the work instruction;

[0132] The first scoring unit is used to score the code information according to the reward rule, and generate a control instruction according to the code information when the score of the code information is greater than a first threshold.

[0133] Optionally, the reward rules in the first scoring unit include one or more of the following:

[0134] A second scoring unit, configured to increase the score of the code information if the accuracy rate between the code information and the work instruction is greater than a second threshold;

[0135] a third scoring unit, configured to increase the score of the code information if the generation time of the code information is less than a third threshold;

[0136] a fourth scoring unit, configured to increase the score of the code information when the code information includes an energy-saving strategy; wherein the energy-saving strategy is to adjust the energy consumption of the energy device according to the energy consumption status of the energy device;

[0137] A fifth scoring unit is used to reduce the score of the code information when the code information includes a non-safety strategy; wherein the non-safety strategy is used to indicate a strategy that causes a safety hazard to the energy device.

[0138] Optionally, the first processing unit includes:

[0139] A first extraction unit, configured to extract questions from the target text information to obtain questions to be queried corresponding to the target text information;

[0140] A first retrieval unit is configured to retrieve the question to be queried in a plurality of knowledge vector libraries corresponding to the energy management type respectively based on the energy management type through a retrieval enhancement model in a pre-trained large language model, and obtain retrieval information corresponding to each of the knowledge vector libraries and a matching degree between the retrieval information and the question to be queried;

[0141] The first generating unit is used to generate code information corresponding to the work instruction according to the question to be queried and the search information with a matching degree greater than a fourth threshold.

[0142] Optionally, the first processing module 302 includes:

[0143] A second extraction unit is used to remove irrelevant information and extract structured information from the operation instruction in sequence to obtain operation information in the operation instruction;

[0144] A first recognition unit, configured to perform keyword recognition on the operation information to obtain keywords in the operation information;

[0145] A first analysis unit, configured to perform sentence segmentation and dependency analysis on the operation information to obtain a context dependency relationship in the operation information;

[0146] The first forming unit is used to form the target text information according to the operating information in the work instruction, the keywords in the operating information and the dependency relationship of the context in the operating information.

[0147] Optionally, the second extraction unit includes:

[0148] A second processing unit is used to perform format conversion and noise removal on the work instruction by using regular expressions or natural language processing technology to obtain first text information;

[0149] The third extraction unit is used to extract structured information from the first text information to obtain the operation information in the work instruction book.

[0150] Optionally, the first identification unit includes:

[0151] The second recognition unit is used to perform keyword recognition on the operation information through named entity recognition technology NER and / or a dictionary corresponding to the energy device to obtain keywords in the operation information.

[0152] Optionally, the first analysis unit includes:

[0153] A first segmentation unit, configured to perform sentence segmentation on the operation information by using a word segmentation tool in natural language processing technology, and divide the operation information into a plurality of words and / or phrases;

[0154] The second analysis unit is used to perform dependency analysis on the words and / or phrases to obtain context dependency relationships in the operation information.

[0155] It should be noted that the embodiment of the device is a device corresponding to the embodiment of the above method, and all implementation methods in the embodiment of the above method are applicable to the embodiment of the device and can achieve the same technical effect.

[0156] An embodiment of the present invention also provides a network device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, the program implements the energy management method as described in any one of the above items and can achieve the same technical effect. To avoid repetition, it will not be described here.

[0157] The embodiment of the present invention further provides a readable storage medium, including: a program is stored on the readable storage medium, and when the program is executed by a processor, the steps of the energy management method described in any one of the above items are implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0158] An embodiment of the present invention further provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the steps of the energy management method as described in any of the above items are implemented and the same technical effect can be achieved. To avoid repetition, they are not described here.

[0159] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0160] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An energy management method, characterized in that: include: Obtaining an operating instruction of the energy equipment and an energy management type corresponding to the operating instruction; Preprocessing the operation instruction to obtain target text information, wherein the target text information includes operation information in the operation instruction, keywords in the operation information, and context dependencies in the operation information; According to the energy management type, the target text information is processed by using the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction; The control instruction is sent to the energy device.

2. The energy management method according to claim 1, characterized in that: The target text information is processed according to the energy management type through the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction, including: Based on the energy management type, the target text information is retrieved and enhanced by a pre-trained large language model to obtain code information corresponding to the work instruction; The code information is scored according to the reward rule, and when the score of the code information is greater than a first threshold, a control instruction is generated according to the code information.

3. The energy management method according to claim 2, characterized in that: The reward rules include one or more of the following: When the accuracy rate between the code information and the operation instruction is greater than a second threshold, increasing the score of the code information; When the generation time of the code information is less than a third threshold, increasing the score of the code information; In the case where the code information includes an energy-saving strategy, increasing the score of the code information; wherein the energy-saving strategy is to adjust the energy consumption of the energy device according to the energy consumption status of the energy device; In the case where the code information includes a non-safety policy, the score of the code information is reduced; wherein the non-safety policy is used to indicate a policy that causes a safety hazard to the energy device.

4. The energy management method according to claim 2, characterized in that: The method of retrieving, enhancing and generating the target text information based on the energy management type in a pre-trained large language model to obtain code information corresponding to the work instruction includes: Extracting questions from the target text information to obtain questions to be queried corresponding to the target text information; Based on the energy management type, the query question is searched in multiple knowledge vector libraries corresponding to the energy management type by using a retrieval enhancement model in a pre-trained large language model, and the retrieval information corresponding to each of the knowledge vector libraries and the matching degree between the retrieval information and the query question are obtained; According to the question to be queried and the search information whose matching degree is greater than a fourth threshold, code information corresponding to the work instruction is generated.

5. The energy management method according to claim 1, characterized in that: The preprocessing of the operation instruction to obtain target text information includes: Removing irrelevant information and extracting structured information from the operation instruction in sequence to obtain operation information in the operation instruction; Performing keyword recognition on the operation information to obtain keywords in the operation information; Performing sentence segmentation and dependency analysis on the operation information to obtain context dependency in the operation information; The target text information is constructed according to the operation information in the work instruction, the keywords in the operation information and the context dependency relationship in the operation information.

6. The energy management method according to claim 5, characterized in that: The step of removing irrelevant information and extracting structured information from the operation instruction in sequence to obtain operation information in the operation instruction includes: Performing format conversion and denoising processing on the work instruction by using regular expressions or natural language processing technology to obtain first text information; Structural information is extracted from the first text information to obtain operation information in the work instruction book.

7. The energy management method according to claim 5, characterized in that: Performing keyword recognition on the operation information to obtain keywords in the operation information includes: The operation information is subjected to keyword recognition through named entity recognition technology NER and / or a dictionary corresponding to the energy device to obtain keywords in the operation information.

8. The energy management method according to claim 5, characterized in that: Sentence segmentation and dependency analysis are performed on the operation information to obtain context dependency relationships in the operation information, including: Segment the operation information using a word segmentation tool in natural language processing technology to divide the operation information into multiple words and / or phrases; Dependency analysis is performed on the words and / or phrases to obtain contextual dependencies in the operation information.

9. An energy management device, characterized in that: include: A first acquisition module is used to acquire an operation instruction of an energy device and an energy management type corresponding to the operation instruction; A first processing module is used to pre-process the operation instruction to obtain target text information, wherein the target text information includes operation information in the operation instruction, keywords in the operation information, and context dependencies in the operation information; A second processing module is used to process the target text information according to the energy management type through the retrieval enhancement generation and reward rules in the pre-trained large language model to generate control instructions corresponding to the work instruction; The first sending module is used to send the control instruction to the energy device.

10. A network device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program implements the energy management method according to any one of claims 1 to 8 when executed by the processor.

11. A readable storage medium, characterized in that: include: The readable storage medium stores a program, and when the program is executed by a processor, the steps of the energy management method according to any one of claims 1 to 8 are implemented.

12. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the energy management method according to any one of claims 1 to 8.

Images