Greenhouse gas emissions management method

Abstract

To provide a method of enabling efficient greenhouse gas emissions management, including computation of greenhouse gas emissions by a business operator.SOLUTION: A greenhouse gas emissions management method according to an embodiment of the present invention comprises receiving image data of invoice information from a business operator terminal, extracting information on an energy consumption based on a large-scale language model using the image data as input data, and deriving greenhouse gas emissions based on the information on the energy consumption.SELECTED DRAWING: Figure 8

Description

Technical Field

[0001] The present invention relates to a method for managing greenhouse gas emissions.

Background Art

[0002] Regarding the greenhouse gas emissions of businesses associated with the use of fuels, electricity, etc., a reporting system targeting SCOPE1 emissions (direct emissions of the company itself) and SCOPE2 emissions (indirect emissions of the company itself) has become widespread, and the calculation and reduction efforts of emissions in SCOPE1 and SCOPE2 have been progressing.

[0003] In Non-Patent Document 1, in order to further reduce the greenhouse gas emissions emitted by businesses, as emissions other than SCOPE1 and SCOPE2, there are suggestions regarding the calculation of SCOPE3 emissions, that is, the emissions of the supply chain (the entire series of flows such as raw material procurement, manufacturing, logistics, sales, waste disposal, etc.) of other related businesses, etc.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, although the technology disclosed in Non-Patent Document 1 discloses a method for calculating greenhouse gas emissions related to SCOPE3, etc., for each business operator, especially companies and local governments, etc., a huge amount of data collection and input is required for calculating emissions, calculating the emissions, and managing the calculation results, which is extremely time-consuming and labor-intensive. In particular, in the field of GHG emissions management, the SCOPE of emissions to be calculated is expanding, the data on which emissions calculations are based varies widely for each SCOPE, and the data management methods also differ for each business operator. Therefore, the introduction of advanced technologies for improving business efficiency has been delayed.

[0006] Therefore, an object of the present invention is to provide a method for efficiently realizing emissions management by reducing the man-hours of work by utilizing advanced technologies in the field of GHG emissions management such as the calculation of greenhouse gas emissions by business operators.

Means for Solving the Problems

[0007] A method for managing greenhouse gas emissions according to an embodiment of the present invention, comprising receiving image data of invoice information from a business operator terminal, using the image data as input data, and extracting information on the amount of energy used based on a large language model, and calculating the greenhouse gas emissions based on the information on the amount of use.

Effects of the Invention

[0008] According to the present invention, it is possible to provide a method for efficiently realizing management such as the calculation of greenhouse gas emissions by business operators.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Mode for Carrying Out the Invention

[0010] The contents of the embodiments of the present invention will be listed and described. A greenhouse gas emissions management system (hereinafter simply referred to as "system") according to an embodiment of the present invention has the following configuration. [Item 1] A method for managing greenhouse gas emissions, receiving image data of invoice information from a business operator terminal, using the image data as input data, extracting information on energy usage based on a large language model, A method for calculating greenhouse gas emissions based on the information on usage. [Item 2] The invoice information includes information on energy usage, and the management method according to Item 1. [Item 3] The management method according to item 1, wherein extracting information regarding the usage amount of the energy includes searching for a vector value that approximates a vector value obtained by converting character information included in the input data. [Item 4] The management method according to item 1, wherein extracting information regarding the usage amount of the energy includes searching for a vector value that approximates a vector value obtained by converting character information included in the input data, and predicting a trading partner included in the invoice information and / or a usage fee of the energy. [Item 5] The management method according to item 1, which instructs the large language model to extract information regarding the usage amount of the energy based on past invoice information.

[0011] <First Embodiment> Hereinafter, a system according to an embodiment of the present invention will be described with reference to the drawings.

[0012] FIG. 1 is a diagram for explaining a greenhouse gas emission management system according to the first embodiment of the present invention.

[0013] As shown in FIG. 1, in the emission management system 1 in the present embodiment, a manager terminal 100 and a plurality of operator terminals 200A and 200B are interconnected via a communication network NW.

[0014] For example, the management terminal 100 receives basic information regarding the operator and input information (for example, image data of invoice information) for calculating the greenhouse gas (for example, CO2) emission amount from the operator terminals 200A and 200B.

[0015] In addition, the management terminal 100 analyzes the received image data of the invoice information by machine learning, extracts necessary items of the invoice information included in the image data, and calculates the greenhouse gas emission amount. Further, the management terminal 100 analyzes the calculated change (for example, increase or decrease) in the greenhouse gas emission amount over time by machine learning and predicts the cause of the change.

[0016] Furthermore, the management terminal 100 has a wallet and is connected to the public blockchain network NW. Based on the information on greenhouse gas emissions for each of the above-specified periods, the management terminal 100 generates a single hash value using SHA256 or other hash functions and records it as transaction information in the blockchain network. On the blockchain network, a block is generated based on the transaction information, the hash value recorded in the previous block, and the nonce value mined by the node, and is recorded following the previous block, forming a blockchain. Here, the hash generation and / or the recording of the transaction information in the blockchain can also be performed not by the management terminal 100 but via another terminal. In this case, the management terminal 100 transmits the greenhouse gas emissions calculated by the matching process to the other terminal. Furthermore, the management terminal 100 can record the information on greenhouse gas emissions in the blockchain network as a smart contract. By using the smart contract, based on the information on the emissions, a contract regarding the emissions trading with other operators can be automatically generated, approved, and executed without the intervention of a third party. Also, with the smart contract, each operator can refer to the transaction information without going through the management terminal, enhancing the service convenience and reducing the operation cost. Furthermore, the management terminal 100 is connected to a large language model (LLM, e.g., GPT) 300 that receives a predetermined question from another terminal via the network through an API (Applicaton Progamming Interface), analyzes the question, and generates an answer.

[0017] Here, as described above, since public blockchain approves transactions not by specific administrators but by an unspecified number of nodes or miners, compared with private blockchain, it can ensure higher non-tampering property and fault tolerance of data. Therefore, since the safety of transactions is ensured, it is preferable to use public blockchain as the destination for recording power transactions in this embodiment. Representative public blockchains include Bitcoin, Ethereum, etc. For example, Ethereum has higher non-tampering property and reliability among public blockchains.

[0018] In addition, the management terminal 100 can associate information regarding greenhouse gas emissions with an identifier or the like and record it as a non-fungible token (Non-Fungible Token (hereinafter referred to as "NFT")) in the blockchain network. The NFT is, for example, a token issued according to the "ERC721" standard of Etherium, which is a platform of the blockchain network, and is a unit of data recorded in the blockchain network and has a non-fungible nature. The NFT is recorded on the blockchain together with a smart contract and can be traced, so it can prove transaction information including details and history such as the information of the operator managing the greenhouse gas emissions.

[0019] FIG. 2 is a functional block diagram of the management terminal constituting the emission management system.

[0020] The communication unit 110 is a communication interface for communicating with an external terminal via the network NW, and communication is performed according to a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol).

[0021] The storage unit 120 stores programs for executing various control processes and each function in the control unit 130, input data, etc., and is composed of a RAM (Random Access Memory), a ROM (Read Only Memory), etc. Further, the storage unit 120 has a business operator data storage unit 121 that stores various data related to business operators, and an AI model storage unit 122 that stores learning data and a learning model learned by AI (Artificial Intelligence) from the learning data. Note that a database (not shown) storing various data may be constructed outside the storage unit 120 or the management terminal 100.

[0022] The control unit 130 controls the overall operation of the management terminal 100 by executing the programs stored in the storage unit 120, and is composed of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), etc. As functions of the control unit 130, there are an information reception unit 131 that receives information from external terminals such as the business operator terminal 200, an image analysis unit 132 that analyzes image data such as invoice information received from the business operator terminal and calculates the greenhouse gas emission amount, a cause analysis unit 133 that analyzes the cause of the time-series change of the greenhouse gas emission amount calculated based on the information included in the extracted invoice information by analyzing the image data, a transaction processing unit 134 that generates a hash value by summarizing information on the greenhouse gas emission amount for a predetermined period and records it as transaction information in the blockchain network, and a report generation unit 135 that generates and transmits report data for outputting the greenhouse gas emission amount and the result of the cause analysis of the change in the emission amount to the business operator every predetermined period.

[0023] Further, although not shown, the control unit 130 has an image generation unit and generates screen information to be displayed via a user interface of an external terminal such as the operator terminal 200. For example, using the images and text data stored in the storage unit 120 as materials, various images and texts are arranged in a predetermined area of the user interface based on a predetermined layout rule, thereby generating the information to be displayed on the user interface. The processing related to the image generation unit can also be executed by a GPU (Graphics Processing Unit).

[0024] Further, the management terminal 100 further has a (not shown) wallet necessary for recording transaction information in the blockchain network. Note that this wallet can also be provided outside the management terminal 100.

[0025] FIG. 3 is a functional block diagram of an operator terminal constituting the emission management system.

[0026] The operator terminal 200 includes a communication unit 210, a display operation unit 220, a storage unit 230, and a control unit 240.

[0027] The communication unit 210 is a communication interface for communicating with the management terminal 100 via the network NW, and communication is performed according to a communication protocol such as TCP / IP.

[0028] The display operation unit 220 is a user interface used for an operator to input instructions and display texts, images, etc. according to input data from the control unit 240. When the operator terminal 200 is configured by a personal computer, it is composed of a display, a keyboard, and a mouse. When the operator terminal 200 is configured by a smartphone or a tablet terminal, it is composed of a touch panel or the like. This display operation unit 220 is activated by a control program stored in the storage unit 230 and executed by the operator terminal 200 which is a computer (electronic computer).

[0029] The storage unit 230 stores programs for executing various control processes and each function in the control unit 240, input data, etc., and is composed of a RAM, a ROM, etc. Further, the storage unit 230 temporarily stores the communication content with the management terminal 100.

[0030] The control unit 240 controls the overall operation of the business operator terminal 200 by executing the programs stored in the storage unit 230, and is composed of a CPU, a GPU, etc.

[0031] FIG. 4 is a diagram for explaining the details of the business operator data according to the first embodiment of the present invention.

[0032] The business operator data 1000 shown in FIG. 4 stores various data related to the business operator acquired from the business operator via the business operator terminal 200. In FIG. 4, for the sake of convenience of explanation, an example of one business operator (the business operator identified by the business operator ID "10001") is shown, but information of a plurality of business operators can be stored. As various data related to the business operator, for example, basic information of the business operator (for example, corporate name of the business operator, user name, address, business type, contact information, email address, business office name, related company name, business operator name related in the supply chain, etc.), input information (for example, image data of invoice information, etc.), analysis information (for example, information related to invoices extracted from image data, greenhouse gas emissions, prediction of the cause of changes in greenhouse gas emissions, etc.), customer information (for example, customer ID, blockchain address, etc.), and offset report information (for example, TXID, NFTID, etc.) can be included.

[0033] FIG. 8 is a flowchart diagram showing an example of the calculation process of greenhouse gas emissions according to the first embodiment of the present invention.

[0034] First, as the process of step S101, the information acquisition unit 131 of the control unit 130 of the management terminal 100 acquires, from the business operator terminal 200 via the network NW, image data including invoice information collected on the business operator side. The business operator uploads invoices, receipts, vouchers, etc. (collectively referred to as "invoices" in this embodiment) in file formats such as PDF, Excel, JPG, etc. (collectively referred to as "image data" in this embodiment) to the management terminal 100 via the business operator terminal 200. The image data acquired by the information acquisition unit 131 is stored as input information in the business operator data storage unit 121 of the storage unit 120.

[0035] Subsequently, as the process of step S102, the image analysis unit 132 of the control unit 130 of the management terminal 100 analyzes the image data acquired in the previous step by machine learning. Here, when performing image analysis, a method called so-called OCR is used. The image analysis unit 132 of the control unit 130 of the management terminal 100 uses a learning model generated by previously learning image data of a plurality of various styles of invoices stored in the AI model storage unit 122 of the storage unit 120 to recognize text from the image data and extract items included in the invoice information as structured string data. Here, when performing image analysis, it is also possible to use an image analysis engine (such as an OCR engine) provided by a business operator other than the management terminal 100 and linked by an API.

[0036] Regarding image analysis, for example, as shown in FIG. 5, it is performed by recognizing and extracting text from image data containing claim information. As shown in FIG. 5, examples of claim information include various items included in the claim form. For example, items such as the breakdown name of the electricity bill, the amount (yen) for each breakdown, the contract power (kW), the electricity consumption (kWh) for each breakdown, the total amount (yen), the date (year and month), etc. are included. In this example, the breakdown of the electricity bill is illustrated, but it may be a claim form for the usage fees of other energies including gas and fuel in addition to electricity, or alternatively, for example, a receipt for transportation expenses of business trips, a receipt for commuting expenses of employees, a claim form associated with transactions with freight forwarders, or a breakdown of a claim form associated with transactions with waste operators. The image analysis unit 132 can extract the amount information, the following activity amount information, etc. included in the claim information as text by analyzing the image data of these claim information. The extracted claim information is stored as analysis information in the business operator data storage unit 121 of the storage unit 120. In this way, through image analysis by machine learning, without the business operator inputting claim information manually, etc., a vast amount of necessary information for calculating greenhouse gas emissions can be obtained as image data, and through highly accurate image recognition, the information necessary for calculating greenhouse gas emissions can be accurately extracted, thus realizing the efficiency improvement and high accuracy of calculating greenhouse gas emissions. Here, as an image analysis method, there is a method of understanding the format of the claim form based on angle correction included in the image data, approximation of patterns, feature points (distance specification from characters such as size ratio starting from a logo, circular notation, kiloliter, etc.), etc., and extracting the items and amounts included in the claim form as structured data. Furthermore, by sending an instruction to the LLM 300 from the management terminal 100 to calculate the energy consumption using the claim information as input data, the LLM 300 can also be made to calculate the energy consumption. More specifically, the LLM 300 can calculate the energy consumption by searching for a vector value that approximates the vector value obtained by converting the character information (text information) included in the input data and predicting information regarding the business partner and / or the energy consumption included in the claim information (energy usage fees such as gasoline costs, electricity bills, etc.).At this time, the instruction text (prompt text) sent to the LLM 300 can include text such as "output what seems to be the fee (in the claim)" and "answer what seems to be the business partner (in the claim)". Also, for example, when the LLM 300 returns an answer with multiple fee information candidates, refer to the past claim information of the business operator stored as business operator data 1000 in the storage unit 120 of the management terminal 100 as a data set, and use the information regarding the energy consumption amount (such as the fee described in the first item of the claim being the fee related to the energy consumption amount) included in the claim information registered as the master. Set an instruction text (prompt text) to predict and extract the energy consumption amount, send it from the management terminal 100 to the LLM 300, and it is also possible to obtain the regenerated answer. Additionally, the management terminal 100 can also generate an instruction text in advance to refer to the past claim information of the business operator and send it to the LLM 300.

[0037] Next, in step S103, the image analysis unit 132 of the control unit 130 calculates the greenhouse gas emissions based on the claim information extracted from the image data or based on the information regarding the energy consumption obtained from the LLM300. Here, the greenhouse gas emissions are classified into SCOPE1, SCOPE2, and SCOPE3. SCOPE1 refers to the direct emissions of greenhouse gases by the operator itself (for example, emissions associated with fuel combustion and industrial processes), SCOPE2 refers to the indirect emissions associated with the use of electricity, heat, gas, etc. supplied from other companies to the operator, and further, SCOPE3 is the calculation standard for the emissions of the entire supply chain of an organization issued by the GHG Protocol, referring to the emissions of the operator's supply chain (the entire series of flows such as raw material procurement, manufacturing, logistics, sales, and waste disposal). SCOPE3 is further classified into 15 categories: (1) purchased products / services, (2) capital goods, (3) fuel and energy-related activities not included in SCOPE1 and SCOPE2, (4) transportation, distribution (upstream), (5) waste from the business, (6) business trips, (7) commuting of employees, (8) leased assets (upstream), (9) transportation, distribution (downstream), (10) processing of sold products, (11) use of sold products, (12) disposal of sold products, (13) leased assets (downstream), (14) franchises, (15) investments). Here, the greenhouse gases include carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), hydrofluorocarbons (HFCs), perfluorocarbons (PFCs), sulfur hexafluoride (SF6), and nitrogen trifluoride (NF3). In this embodiment, CO2 will be described as an example.

[0038] Also, the amount of greenhouse gas emissions is defined by taking the operator's electricity consumption, cargo transportation volume, waste treatment volume, and various transaction amounts as activity volumes, and multiplying the activity volumes by the emission factors per unit of activity, namely, the CO2 emissions per 1 kWh of electricity used, the CO2 emissions per 1 ton of cargo transported, and the CO2 emissions per 1 ton of waste incinerated. The amount of greenhouse gas emissions is calculated separately for the above SCOPE1, SCOPE2, and SCOPE3 (for SCOPE3, by category of the 15 categories), and the total emissions are calculated as the supply chain emissions.

[0039] In this embodiment, the image analysis unit 132 extracts relevant claim information for each of SCOPE1, SCOPE2, and SCOPE3 (and for SCOPE3, further categorized by category), and among the claim information, calculates the emissions based on the above calculation method, for example, based on the electricity consumption in kWh. The calculated emissions are stored as analysis information in the operator data storage unit 121 of the storage unit 120.

[0040] Subsequently, as the process of step S104, the report generation unit 135 of the control unit 130 generates a visualized report showing the breakdown of emissions over time for each SCOPE (and for SCOPE3, further categorized by category) based on the information regarding the calculated emissions.

[0041] FIG. 9 is a flowchart showing an example of the cause prediction process for changes in greenhouse gas emissions according to the first embodiment of the present invention.

[0042] First, as the process of step S201, the cause analysis unit 133 of the control unit 130 of the management terminal 100 refers to the information regarding the greenhouse gas emissions of the operator calculated in step S103 of FIG. 8. Here, regarding the greenhouse gas emissions, the emissions for each SCOPE (and for SCOPE3, further categorized by category) are referred to. Also, the cause analysis unit 133 can confirm the change (increase or decrease) in emissions by referring to the past emissions data of the same operator for the greenhouse gas emissions. As described above, the emissions are stored as analysis information in the operator data storage unit 121 of the storage unit 120.

[0043] Subsequently, as the process of step S202, based on the information regarding the emissions amount referred to above, the cause analysis unit 133 analyzes and predicts the cause of the change in the emissions amount by machine learning. Here, in the cause analysis, the cause analysis unit 133 of the control unit 130 of the management terminal 100 uses the information on the emissions amount referred to above, the factors that affect the change (increase or decrease) in the emissions amount, and the learning model generated by learning the data regarding the factors that affect the change (increase or decrease) in the emissions amount, which is stored in the AI model storage unit 122 of the storage unit 120 in advance, to predict the cause of the change in the emissions amount for each SCOPE (for SCOPE3, further categorized by category).

[0044] Here, as factors that affect the change (increase or decrease) in the emissions amount, for example, factors such as weather, temperature, product demand and / or factory operation, store or factory business or operation hours, changes in equipment or facilities, measures by software, energy-saving actions, fuel conversion, energy menu changes, changes in business trips or commuting volume, and the power generation amount of self-generation can be mentioned. Each of these factors is a factor that affects the emissions amount of any SCOPE. For example, the factor of weather affects precipitation, wind volume, sunshine duration, and temperature. Precipitation affects the small hydropower generation amount, wind volume affects the wind power generation amount, sunshine duration affects the solar power generation amount, temperature affects air conditioning, and furthermore, the power generation amount affects the self-generation amount, and the self-generation amount affects the CO2 emissions amount by electricity, thereby affecting the change in the emissions amount of SCOPE2. On the other hand, air conditioning affects the gas usage amount, and the gas usage amount affects the CO2 emissions amount by gas combustion, thereby affecting the change in the emissions amount of SCOPE1. Also, energy-saving activities, factory operation due to product demand, and business hours affect the electricity usage amount and affect SCOPE2. Also, EMS, replacement of refrigeration equipment, introduction of energy-saving equipment, and automobile usage amount also affect the electricity usage amount and affect SCOPE2. In addition, automobile usage amount, fuel consumption, boiler usage amount, and boiler efficiency affect the fuel usage amount and affect the CO2 emissions amount by fuel, thereby affecting SCOPE1.

[0045] In addition, the factor of product sales volume affects Categories 1, 9, 10, 11, and 12 of SCOPE3, equipment investment affects Category 2, the renewable energy ratio and the amount of purchased energy affect Category 3, the number of transportation trips and transportation route changes affect Categories 4 and 9, the product loss rate affects Category 5, the number of business trips and the number of employees going to the office affect Category 6, the number of commuting employees and the number of employees going to the office affect Category 7, power consumption affects Category 8, reduction of processing due to product improvement affects Category 10, improvement to energy-saving products affects Category 11, increase in the recycling rate affects Category 12, the office power of tenants affects Category 13, the emissions of franchises affect Category 14, and the emissions of investment destinations affect Category 15, respectively.

[0046] In this way, by using machine learning to learn which factors affect which SCOPE or category, and by obtaining information on emissions and information on each factor from the business operator, it is possible to predict the causes of changes in emissions. Here, by predicting the causes of emissions using machine learning, it is possible to efficiently and accurately predict the factors that affect the changes in greenhouse gas emissions for each business operator and for each SCOPE.

[0047] Subsequently, as the process of step S203, the report generation unit 135 of the control unit 130 generates a visualized report on the causes of changes in emissions for each SCOPE (and further for each category in the case of SCOPE3) based on the information on the predicted causes of changes in the analyzed emissions.

[0048] FIG. 10 is a flowchart showing an example of the transaction process of greenhouse gas emissions according to the first embodiment of the present invention.

[0049] First, as the process of step S301, the transaction processing unit 134 of the control unit 130 of the management terminal 100 refers to the business operator data stored in the business operator data storage unit 121 of the storage unit 120. Here, the business operator data to be referred to includes the analysis information of the business operator (greenhouse gas emissions for each SCOPE), etc.

[0050] Next, as the processing in step S302, the transaction processing unit 134 generates a hash value based on the operator data referred to in step S301. That is, the transaction processing unit 134 generates a hash value for one line using a hash function for the greenhouse gas emissions during a predetermined period, and records the hash value as transaction information in the public blockchain. On the blockchain network, a block is generated based on the transaction information, the hash value recorded in the immediately preceding block, and the nonce value mined by the node, and is recorded following the immediately preceding block to form a blockchain. Here, in this example, in order to reduce the cost related to blockchain recording, it is assumed that the recording is made in layer 2 (for example, a side chain) different from the main blockchain (so-called layer 1).

[0051] In addition, the transaction processing unit 134 can assign and manage an NFT ID in association with the blockchain recording of the greenhouse gas emissions of the operator. More specifically, as shown in FIG. 4, in the operator data 1000, the customer ID of the operator is assigned as customer information, and the blockchain address to be referred to is stored, and the NFT ID and TX ID can be assigned as offset report information.

[0052] As shown in FIG. 6, on the blockchain network, blockchain addresses are associated with each NFT ID. In the management terminal 100, NFT IDs and customer IDs are managed. For example, regarding the information on the greenhouse gas emissions of the operator corresponding to the customer ID "2", by referring to the blockchain address for each NFT ID, such as NFT IDs "13" and "14", the details of the emissions information can be read out as shown in FIG. 7. FIG. 7 shows the information regarding the offset report associated with the NFT ID "14". A TXID is assigned in association with the offset report, and the CO2 emissions by SCOPE, the target year and month, and the issuance date of the report are included in the offset report. In addition to the CO2 emissions in the target year and month in this example, the CO2 emissions in the most recent years, the reduced CO2 emissions, and the offset CO2 emissions can also be NFTized. In this way, by managing CO2 emissions as NFTs, the operator can conduct transactions of NFTized certificates in a state where non-tampering and transaction reliability are ensured, and can also prove the emissions to a third party.

[0053] The above-described embodiments are merely examples for facilitating the understanding of the present invention and are not intended to limit the interpretation of the present invention. The present invention can be changed and improved without departing from its gist, and it goes without saying that equivalents of the present invention are included therein.

Explanation of Reference Numerals

[0054] 100 Management terminal 200 Operator terminal

Claims

1. A method for managing greenhouse gas emissions, comprising: receiving image data of invoice information from an operator terminal; using the image data as input data to extract information on energy usage based on a large language model; calculating greenhouse gas emissions based on the information on energy usage.

2. The management method according to claim 1, wherein the invoice information includes information on energy usage.

3. The management method according to claim 1, wherein extracting information on energy usage includes searching for a vector value that approximates a vector value obtained by converting character information included in the input data.

4. The management method according to claim 1, wherein extracting information on energy usage includes searching for a vector value that approximates a vector value obtained by converting character information included in the input data, and predicting a counterparty and / or energy usage fee included in the invoice information.

5. The management method according to claim 1, further comprising instructing the large language model to extract information on energy usage based on past invoice information.

Images