Environmental protection management method and system for food industry based on large model

Abstract

The invention belongs to the technical field of environmental protection, and provides an environmental protection management method and system for the food industry based on a large model. The method comprises the following steps: acquiring various real-time data of each sensor, and filling a preset data template with the real-time data to obtain a real-time data report; performing cue word extraction from the real-time data report based on a large language model to obtain a first cue word set, and obtaining a plurality of first knowledge fragments from the dynamically updated knowledge graph based on the first cue word set; performing cue word extraction from the first knowledge fragment based on a large language model to obtain a second cue word set, and combining the cue word sets to form an optimal cue word set; the large language model takes a preset evaluation item as an index, and generates a first initial evaluation text based on the optimal prompt word set; and segmenting the first initial evaluation text according to a preset evaluation report template, and filling the first initial evaluation text to a corresponding position of the evaluation report template to obtain a real-time evaluation report. The method is high in efficiency and reliable in evaluation result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental protection, in particular to an environmental protection management method and system for the food industry based on a large model. BACKGROUND

[0002] In the food industry, monitoring and analyzing the discharge of pollutants including various solid wastes and wastewater is not only a requirement of policies and regulations, but also a demand of environmental protection, and is closely related to food safety and human health.

[0003] The solid waste includes defective products, spoiled food, waste oil, waste organic solvent, etc. The wastewater contains harmful substances such as animal and vegetable oil, ammonia nitrogen, pesticide residues, heavy metals, etc. In the prior art, the monitoring and analysis of solid waste and wastewater often require the environmental protection department of the enterprise or professional institutions to manually sample and analyze with the assistance of related instruments.

[0004] However, the above-mentioned method has the following defects: first, the manual sampling method is inefficient and cannot effectively monitor continuously. Second, the limited knowledge reserve, retrieval ability of the analyst, and the dynamic update of the relevant regulations, the information gap caused by the iterative update of waste storage technology and wastewater treatment technology, etc. also lead to the inaccuracy of the analysis results. SUMMARY

[0005] The present application aims to provide an environmental protection management method and system for the food industry based on a large model to solve the technical problems of low efficiency and low accuracy of analysis results in the prior art based on manual sampling and analysis of pollutants.

[0006] To achieve the above-mentioned purpose, the present application proposes the following technical solutions: In a first aspect, the present application provides an environmental protection management method for the food industry based on a large model, comprising: acquiring various types of real-time data of each sensor according to a preset period and filling into a preset data template to obtain a real-time data report; extracting prompt words from the real-time data report based on a large language model to obtain a first prompt word set, and acquiring a plurality of first knowledge fragments from the knowledge graph that is automatically updated based on the first prompt word set; extracting prompt words from each first knowledge fragment based on a large language model to obtain a second prompt word set, and merging the first prompt word set and the second prompt word set to form an optimal prompt word set; The large language model indexes preset evaluation items and generates first initial evaluation text based on the optimal prompt set; wherein, the evaluation items include one-to-one corresponding data type sub-items, pollution level sub-items and optimization scheme sub-items, the large language model introduces a self-attention mechanism to dynamically allocate weights to each data type sub-item according to the latest environmental protection regulations; wherein, the large language model is periodically iteratively updated at a preset frequency, and the latest pollution-related data of each industry is used to construct data samples during initial training and each iteration update; The first initial evaluation text is segmented according to the preset evaluation report template, and is filled into the corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report.

[0007] Further, after the large language model indexes preset evaluation items and generates first initial evaluation text based on the optimal prompt set; comprising: input the first initial evaluation text into the dynamically updated knowledge graph to obtain a plurality of second knowledge segments; The large language model compares the first initial evaluation text and the second knowledge segment, and when the part of the content in the first initial evaluation text and the corresponding content in the second knowledge segment are opposite in semantics, the part of the content in the first initial evaluation text is replaced by the corresponding content in the second knowledge segment to obtain an updated evaluation text.

[0008] Further, after the first initial evaluation text is segmented according to the preset evaluation report template, and filled into the corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report; comprising: synchronize the real-time evaluation report to the government affairs system; When there is an emission anomaly, issue a penalty instruction to the enterprise corresponding to the real-time evaluation report, and adjust the acquisition period of each type of real-time data according to the rectification period to obtain real-time rectified emission data.

[0009] Further, the method comprises: acquire the number of penalty instructions received by any enterprise within a first preset period of time; dynamically adjust the acquisition period of each type of real-time data of the corresponding enterprise based on the number of instructions; wherein, the acquisition period and the number of instructions are inversely proportional.

[0010] Further, comprising: acquire all real-time evaluation reports of any enterprise within a second preset period of time; analyzing and processing the real-time evaluation report based on a large language model to obtain a second initial report text; segmenting the second initial evaluation text according to a preset summary report template, and filling the second initial evaluation text into corresponding positions of the summary report template to obtain an evaluation summary report corresponding to all real-time evaluation reports.

[0011] In a second aspect, the technical scheme provides an environmental protection management system for the food industry based on a large model, comprising: A first obtaining module is configured to obtain various types of real-time data of various sensors according to a preset period, and fill the real-time data into a preset data template to obtain a real-time data report. A second obtaining module is configured to extract first prompt words from the real-time data report based on a large language model to obtain a first prompt word set, and obtain a plurality of first knowledge segments from an automatically updated knowledge graph based on the first prompt word set. A third obtaining module is configured to extract second prompt words from each first knowledge segment based on a large language model to obtain a second prompt word set, and merge the first prompt word set and the second prompt word set to form an optimal prompt word set. A text generation module is configured to index a large language model with a preset evaluation item, and generate a first initial evaluation text based on the optimal prompt word set; wherein the evaluation item comprises a one-to-one corresponding data type sub-item, a pollution level sub-item and an optimization scheme sub-item, the large language model introduces a self-attention mechanism to dynamically allocate weights to each data type sub-item according to the latest environmental protection regulations; wherein the large language model is periodically iteratively updated at a preset frequency, and at the initial training and each iteration update, the latest pollution related data of each industry is used to construct data samples. A report generation module is configured to segment the first initial evaluation text according to a preset evaluation report template, and fill the first initial evaluation text into corresponding positions of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report.

[0012] Further comprising: A fourth obtaining module is configured to input the first initial evaluation text into a dynamically updated knowledge graph to obtain a plurality of second knowledge segments. A text correction module is configured to compare the first initial evaluation text and the second knowledge segments based on a large language model, and when part of the content in the first initial evaluation text is opposite in semantics to the corresponding content in the second knowledge segment, replace the part of the content in the first initial evaluation text with the corresponding content in the second knowledge segment to obtain an updated evaluation text.

[0013] Further comprising: A synchronous updating module is configured to synchronously update the real-time evaluation report to a government affairs system. The periodic updating module is configured to issue a penalty instruction to an enterprise corresponding to the real-time evaluation report when there is an emission abnormality, and adjust the acquisition period of various types of real-time data according to a rectification period to acquire and analyze the rectified emission data in real time.

[0014] In a third aspect, the technical solution provides an electronic device, comprising at least one processor, the processor is coupled with a memory, the memory stores a computer program, and the computer program is configured to be executed by the processor to execute the method.

[0015] In a fourth aspect, the technical solution provides a computer readable storage medium, characterized in that a computer program is stored thereon, and the computer program is executed by a computer to implement the method.

[0016] Advantages: From the above technical solution, the technical solution of the present application provides an environmental management method for the food industry based on a large model to solve the technical defects of low efficiency and low accuracy of analysis results in the prior art when pollutants are sampled and analyzed manually.

[0017] The method comprises: firstly, obtaining various types of real-time data of each sensor at a preset period, and filling into a preset data template to obtain a real-time data report. At this time, the real-time data reporting based on the sensor improves the efficiency and sustainability of data acquisition. At the same time, the real-time data is sorted according to the data template to standardize the data, which also improves the accuracy of subsequent processing based on the large language model. Continue, the large language model extracts the prompt words from the real-time data report to obtain a first prompt word set, and obtains a plurality of first knowledge fragments based on the first prompt word set in the automatically updated knowledge graph. The large language model extracts the prompt words from each first knowledge fragment to obtain a second prompt word set, and merges the first prompt word set and the second prompt word set to form an optimal prompt word set. At this time, since the dynamic updated knowledge graph is introduced in the prompt word acquisition, the comprehensiveness and accuracy of each prompt word in the optimal prompt word set are ensured. Then, the large language model takes the preset evaluation item as the index, and generates a first initial evaluation text based on the optimal prompt word set. Specifically, the evaluation item includes a one-to-one correspondence of data type subitem, pollution level subitem and optimization scheme subitem. The large language model is periodically iteratively updated at a preset frequency, and the latest pollution related data of each industry is used to construct data samples during initial training and each iteration update. The self-attention mechanism is also introduced into the large language model to dynamically allocate weights to each data type subitem according to the latest environmental protection regulations, so as to improve the accuracy of the text output of the large language model. Finally, the first initial evaluation text is segmented according to the preset evaluation report template, and filled into the corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report. At this time, the real-time evaluation report has higher accuracy compared with the artificial analysis report.

[0018] It should be understood that all combinations of the foregoing concepts and additional concepts described in greater detail below (provided such concepts are not mutually inconsistent) are contemplated as being part of the inventive subject matter.

[0019] The foregoing and other aspects, embodiments and features of the present teachings can be better understood from the following detailed description taken in conjunction with the accompanying drawings. Additional features of the present teachings will be described or will become apparent in the course of the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical, or nearly identical, component that is illustrated in various figures is represented with a like numeral. For purposes of clarity, not every component is called out in every drawing. Embodiments of various aspects of the present teachings will now be described, by way of example only, with reference to the drawings in which: 1. A method for generating a real-time evaluation report based on real-time data of a plurality of sensors, comprising: obtaining various types of real-time data of each sensor at a preset period, and filling into a preset data template to obtain a real-time data report; extracting prompt words from the real-time data report based on a large language model to obtain a first prompt word set, and obtaining a plurality of first knowledge fragments based on the first prompt word set in an automatically updated knowledge graph; extracting prompt words from each first knowledge fragment based on the large language model to obtain a second prompt word set, and merging the first prompt word set and the second prompt word set to form an optimal prompt word set; taking a preset evaluation item as an index based on the large language model, and generating a first initial evaluation text based on the optimal prompt word set; segmenting the first initial evaluation text according to a preset evaluation report template, and filling into a corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report; wherein the evaluation item comprises a one-to-one correspondence of data type subitem, pollution level subitem and optimization scheme subitem; the large language model is periodically iteratively updated at a preset frequency, and the latest pollution related data of each industry is used to construct data samples during initial training and each iteration update; the self-attention mechanism is introduced into the large language model to dynamically allocate weights to each data type subitem according to the latest environmental protection regulations, so as to improve the accuracy of the text output of the large language model. Figure 1 A flowchart of the environmental protection management method for the food industry based on a large model described in the present embodiment; Figure 2 A flowchart of the first initial evaluation text update; Figure 3 A flowchart of the real-time evaluation report synchronization to the government affairs system; Figure 4 A flowchart of the real-time data acquisition period adjustment; Figure 5 A flowchart of the evaluation summary report acquisition; Figure 6 A structural block diagram of the environmental protection management system for the food industry based on a large model described in the present embodiment; Figure 7 A structural block diagram of the electronic device described in the present embodiment. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be described below in detail with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the described embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the scope of the present application. Unless otherwise defined, the technical terms or scientific terms used herein should have their usual meanings understood by those of ordinary skill in the art to which the present application belongs.

[0022] The terms "first", "second", and similar terms used in the present application specification and claims do not represent any order, number, or importance, but are only used to distinguish different components. Similarly, unless the context clearly indicates otherwise, the singular form "one", "an", or "the" and the like do not represent a quantity limitation, but represent the existence of at least one. The terms "include" or "contain" and the like mean that the elements or objects appearing before "include" or "contain" cover the features, whole, steps, operations, elements, and / or components listed after "include" or "contain", and do not exclude the existence or addition of one or more other features, whole, steps, operations, elements, components, and / or sets thereof. "Up", "down", "left", "right", and the like are only used to represent relative positional relationships, and when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0023] In the food industry, monitoring and analyzing the emission of pollutants is not only a requirement of policies and regulations, but also a demand of environmental protection, which is closely related to food safety and human health. However, the existing technology is mostly based on manual methods for the above processes. Further, there are the following defects: the efficiency of manual sampling is low, and continuous monitoring is not possible. The limited knowledge reserve and retrieval ability of the analyst, the dynamic update of the relevant regulations, the information gap generated by the iterative update of waste storage technology and wastewater treatment technology, and the accuracy of the analysis results cannot be guaranteed. Based on this, the present embodiment aims to provide an environmental management method for the food industry based on a large model to solve the above technical defects.

[0024] The environmental management method for the food industry based on a large model described in the present embodiment will be described in detail below with reference to the accompanying drawings.

[0025] In combination with Figure 1 As shown in the figure, the method comprises the following steps: Step S202, obtain various types of real-time data of each sensor according to a preset period, and fill them into a preset data template to obtain a real-time data report.

[0026] In the present embodiment, the real-time data includes COD emission concentration, BOD emission concentration, SS emission concentration, animal and vegetable oil emission concentration, ammonia emission concentration, nitrogen emission concentration, etc.

[0027] Due to the differences in sensor data types and accuracy of each enterprise, in order to improve the accuracy of the subsequent large language model in extracting prompt words, the real-time data is filled into the preset data template for data preprocessing. At the same time, this method can also further detect whether the data is missing to feedback to the enterprise for timely confirmation and processing of various abnormal situations including sensor abnormalities and data link abnormalities. It can also avoid data falsification that may occur during manual collection.

[0028] Step S204, based on the large language model, extract prompt words from the real-time data report to obtain a first set of prompt words, and obtain a plurality of first knowledge fragments based on the first set of prompt words in the automatically updated knowledge graph.

[0029] Step S206, based on the large language model, extract prompt words from each first knowledge fragment to obtain a second set of prompt words, and merge the first set of prompt words and the second set of prompt words to form an optimal set of prompt words.

[0030] Based on the steps S204-S206, since the dynamically updated knowledge graph is introduced when obtaining the prompt words, the comprehensiveness and accuracy of each prompt word in the optimal set of prompt words are ensured.

[0031] Step S208, the large language model indexes the preset evaluation items, and generates a first initial evaluation text based on the optimal prompt set.

[0032] In this embodiment, the evaluation items include one-to-one corresponding data type sub-items, pollution level sub-items, and optimization scheme sub-items. The data type sub-items correspond to the types of the real-time data.

[0033] From the perspective of the prediction processing process, in order to improve the accuracy of the output of the large language model, in the model training process, data samples are constructed based on the latest pollution-related data of various industries; at the same time, periodic iterative updates are performed at a preset frequency. Secondly, in the prediction process, a self-attention mechanism is introduced to dynamically allocate weights to each data type sub-item according to the latest environmental protection regulations to capture the semantic association between different abstraction levels.

[0034] As a preferred implementation, in combination with Figure 2 As shown in FIG. 8, after step S208, the following steps are further included: Step S20902, inputting the first initial evaluation text into the dynamically updated knowledge graph to obtain a plurality of second knowledge segments.

[0035] Step S20904, comparing the first initial evaluation text and the second knowledge segments based on the large language model, and when part of the content in the first initial evaluation text is opposite in semantics to the corresponding content in the second knowledge segment, replacing the part of the content in the first initial evaluation text with the corresponding content in the second knowledge segment to obtain an updated evaluation text.

[0036] At this time, based on steps S20902-S20904, the first initial evaluation text output by the large language model can be factually checked and corrected based on the knowledge graph, thereby further ensuring the reliability of the subsequent output real-time evaluation report.

[0037] Step S210, segmenting the first initial evaluation text according to a preset evaluation report template, and filling the corresponding positions of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report.

[0038] As a preferred implementation, considering the regulatory needs of the government environmental protection department on the pollution situation, in combination with Figure 3 As shown in FIG. 8, after step S210, the following steps are further included: Step S21202, synchronously updating the real-time evaluation report to the government affairs system.

[0039] Step S21204, when it is judged that there is an emission abnormality, a punishment instruction is issued to the enterprise corresponding to the real-time evaluation report, and the acquisition period of various types of real-time data is adjusted according to the rectification period to acquire and analyze the rectified emission data in real time.

[0040] At this time, based on steps S21202-S21204, the government department can directly issue corresponding instructions based on the real-time evaluation report, and the instruction issuance is more reliable, thereby improving the compliance and objectivity of the instruction itself. At the same time, the efficiency of government processing is improved. Furthermore, the data acquisition period is also adjusted in real time according to the rectification period, thereby improving the timely tracking of subsequent rectification.

[0041] Further preferably, based on steps S21202-S21204, in combination with Figure 4 As shown in the figure, the data acquisition period of step S202 is also adjusted as follows: Step S20202, acquiring the instruction number of the punishment instruction received by any enterprise within the first preset period of time.

[0042] Step S20204, dynamically adjusting the acquisition period of various types of real-time data of the corresponding enterprise based on the instruction number.

[0043] In this step, the acquisition period is inversely proportional to the instruction number. At this time, based on steps S20202-S20204, the rationality of the data acquisition frequency can be improved, the consumption of computing resources is avoided, and the rationality of resource allocation is improved.

[0044] And considering the summary demand within a certain period of time in actual monitoring in order to control the whole situation and make subsequent pollution or work plans, in combination with Figure 5 As shown in the figure, the following steps are further included after step S210: Step S21222, acquiring all real-time evaluation reports of any enterprise within the second preset period of time.

[0045] Step S21224, analyzing and processing the real-time evaluation report based on a large language model to acquire a second initial report text.

[0046] Step S21226, segmenting the second initial evaluation text according to a preset summary report template, and filling it into the corresponding position of the summary report template to acquire an evaluation summary report corresponding to all real-time evaluation reports.

[0047] At this time, based on steps S21222-S21224, the evaluation summary report can be directly based on the large language model, which not only has higher efficiency, but also has higher accuracy. Thus, reliable basis is provided for subsequent work planning.

[0048] As can be seen from the above, in the environmental protection management method, online monitoring, large language model and dynamic knowledge graph are introduced at the same time, which not only improves the data monitoring efficiency, but also eliminates the information difference between the dynamically updated knowledge data and the artificial knowledge reserve, effectively improving the reliability of the final obtained analysis report. At the same time, the analysis report is associated with the government affairs system to improve the efficiency and compliance of the enterprise pollution supervision.

[0049] The above program can run in the processor, or can also be stored in the memory (or called computer readable storage medium), the computer readable medium includes permanent and non-permanent, movable and non-movable medium can realize information storage by any method or technology. Information can be computer readable instructions, data structure, program module or other data. Examples of computer storage medium include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape magnetic disk storage or other magnetic storage device or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. According to the definition in this paper, computer readable medium does not include temporary computer readable medium, such as modulated data signal and carrier wave.

[0050] These computer programs can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate computer implemented processing, so that the instructions executed on the computer or other programmable device provide a process for implementing the functions specified in the flow Figure 1 One flow or multiple flows and / or blocks Figure 1 The steps of the functions specified in one block or multiple blocks correspond to different steps, which can be realized by different modules.

[0051] The embodiment also provides an environmental protection management system for food industry based on large model, which combines the advantages of online monitoring, large language model and dynamic knowledge graph. Figure 6 As shown in the figure, it includes the following functional modules: The first acquisition module is used for acquiring various types of real-time data of each sensor according to a preset period, and filling the data template to obtain a real-time data report.

[0052] The second acquisition module is configured to perform prompt word extraction from the real-time data report based on the large language model to obtain a first prompt word set, and acquire a plurality of first knowledge segments from the automatically updated knowledge graph based on the first prompt word set.

[0053] The third acquisition module performs prompt word extraction from each first knowledge segment based on the large language model to obtain a second prompt word set, and merges the first prompt word set and the second prompt word set to form an optimal prompt word set.

[0054] The text generation module is configured to index the large language model based on a preset evaluation item, and generate a first initial evaluation text based on the optimal prompt word set; wherein the evaluation item includes a one-to-one corresponding data type sub-item, a pollution level sub-item, and an optimization scheme sub-item, the large language model introduces a self-attention mechanism to dynamically allocate weights to each data type sub-item according to the latest environmental protection regulations; wherein the large language model is periodically iteratively updated at a preset frequency, and the latest pollution-related data of each industry is used to construct data samples during initial training and each iteration update.

[0055] The report generation module is configured to segment the first initial evaluation text according to a preset evaluation report template, and fill in the corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report.

[0056] Since the system is built based on the method, the above description is not repeated here.

[0057] For example, it also includes: The fourth acquisition module is configured to input the first initial evaluation text into the dynamically updated knowledge graph to obtain a plurality of second knowledge segments.

[0058] The text correction module is configured to compare the first initial evaluation text and the second knowledge segment based on the large language model, and when part of the content in the first initial evaluation text is opposite in semantics to the corresponding content in the second knowledge segment, replace the part of the content in the first initial evaluation text with the corresponding content in the second knowledge segment to obtain an updated evaluation text.

[0059] For example, it also includes: The synchronous update module is configured to synchronously update the real-time evaluation report to the government affairs system.

[0060] The periodic update module is configured to issue a penalty instruction to an enterprise corresponding to the real-time evaluation report when there is an emission anomaly, and adjust the acquisition period of each type of real-time data according to the rectification period to obtain real-time rectified emission data.

[0061] Meanwhile, an electronic device is also provided, which is combined with the method Figure 7 As shown in the method, the electronic device comprises at least one processor, which is coupled with a memory, and the memory stores a computer program configured to be run by the processor to execute the method.

[0062] Further, the embodiment also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a computer to implement the method.

[0063] Since the system, the electronic device and the storage medium are all based on the method and are used to implement the method, they also have the advantages of high efficiency and high reliability of output report in actual application.

[0064] Although the present application has been disclosed with the preferred embodiments as above, it is not intended to limit the present application. Those skilled in the art can make various modifications and improvements without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application shall be subject to the scope defined by the claims.

Claims

1. An environmentally friendly management method for the food industry based on a large model, characterized by, The method comprises the following steps: acquiring various types of real-time data of each sensor according to a preset period, and filling the data into a preset data template to obtain a real-time data report; extracting prompt words from the real-time data report based on a large language model to obtain a first prompt word set, and obtaining a plurality of first knowledge segments from the automatically updated knowledge graph based on the first prompt word set; extracting prompt words from each first knowledge segment based on a large language model to obtain a second prompt word set, and merging the first prompt word set and the second prompt word set to form an optimal prompt word set; a large language model is indexed by a preset evaluation item, and a first initial evaluation text is generated based on the optimal prompt word set; wherein the evaluation item comprises a one-to-one correspondence of data type sub-item, pollution level sub-item and optimization scheme sub-item, and the large language model introduces a self-attention mechanism to dynamically allocate weights to each data type sub-item according to the latest environmental protection regulations; wherein the large language model is periodically iteratively updated at a preset frequency, and the latest pollution-related data of each industry is used to construct data samples during initial training and each iteration update; segmenting the first initial evaluation text according to a preset evaluation report template, and filling it into the corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report.

2. The large model-based environmental management method for the food industry according to claim 1, characterized by, After the large language model is indexed by a preset evaluation item and a first initial evaluation text is generated based on the optimal prompt word set; comprising: inputting the first initial evaluation text into the dynamically updated knowledge graph to obtain a plurality of second knowledge segments; comparing the first initial evaluation text and the second knowledge segment based on a large language model, and when part of the content in the first initial evaluation text is opposite in semantics to the corresponding content in the second knowledge segment, replacing the part of the content in the first initial evaluation text with the corresponding content in the second knowledge segment to obtain an updated evaluation text.

3. The large model-based environmental management method for the food industry according to claim 1, characterized by, After segmenting the first initial evaluation text according to a preset evaluation report template, and filling it into the corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report; comprising: synchronously updating the real-time evaluation report to the government affairs system; when there is an emission abnormality, issuing a punishment instruction to the enterprise corresponding to the real-time evaluation report, and adjusting the acquisition period of various types of real-time data according to the rectification period to obtain real-time rectified emission data.

4. The large model-based environmental management method for the food industry according to claim 3, characterized by, The method comprises the following steps: acquiring the number of punishment instructions received by any enterprise within a first preset period; dynamically adjusting the acquisition period of various types of real-time data of the corresponding enterprise based on the number of instructions; wherein the acquisition period and the number of instructions are inversely proportional.

5. The large model-based eco-friendly management method for the food industry according to claim 1, characterized in that, The method comprises the following steps: acquiring all real-time evaluation reports of any enterprise within a second preset period; analyzing and processing the real-time evaluation report based on a large language model to obtain a second initial report text; The second initial evaluation text is segmented according to a preset summary report template, and is filled into a corresponding position of the summary report template to obtain an evaluation summary report corresponding to the all real-time evaluation reports.

6. An environment-friendly management system for the food industry based on a large model, characterized by, Comprise: The first acquisition module is used for acquiring various types of real-time data of each sensor according to a preset period, and filling into a preset data template to obtain a real-time data report; The second acquisition module is used for extracting first prompt words from the real-time data report based on a large language model to obtain a first knowledge fragment based on the first prompt word set and the automatically updated knowledge graph; The third acquisition module extracts second prompt words from each first knowledge fragment based on a large language model to obtain a second prompt word set, and merges the first prompt word set and the second prompt word set to form an optimal prompt word set; The text generation module is used for a large language model to generate a first initial evaluation text based on a preset evaluation item, wherein the evaluation item comprises a one-to-one corresponding data type sub-item, a pollution level sub-item and an optimization scheme sub-item, and the large language model introduces a self-attention mechanism to dynamically allocate weights to each data type sub-item according to the latest environmental protection regulations; wherein the large language model is periodically iteratively updated according to a preset frequency, and each time the initial training and each iteration update are performed, the latest pollution related data of each industry is used to construct a data sample; The report generation module is used for segmenting the first initial evaluation text according to a preset evaluation report template, and filling into a corresponding position of the evaluation report template to obtain a real-time evaluation report corresponding to the real-time data report.

7. The large model-based environmentally friendly management system for the food industry according to claim 6, characterized by, Comprise: The fourth acquisition module is used for inputting the first initial evaluation text into a dynamically updated knowledge graph to obtain a second knowledge fragment; The text correction module is used for comparing the first initial evaluation text and the second knowledge fragment based on a large language model, and when part of the content in the first initial evaluation text is opposite in semantics to the corresponding content in the second knowledge fragment, replacing the part of the content in the first initial evaluation text with the corresponding content in the second knowledge fragment to obtain an updated evaluation text.

8. The large model-based environmentally friendly management system for the food industry according to claim 6, characterized by, Comprise: The synchronous update module is used for synchronously updating the real-time evaluation report to a government affairs system; The periodic update module is used for issuing a penalty instruction to an enterprise corresponding to the real-time evaluation report when there is an emission abnormality, and adjusting the acquisition period of each type of real-time data according to a rectification period to obtain real-time rectified emission data.

9. An electronic device, comprising: The processor is coupled with a memory, and the memory stores a computer program configured to be executed by the processor to perform the method of any one of claims 1-5.

10. A computer-readable storage medium, characterized in that, The computer program is stored on the computer and is executed by the computer to implement the method of any one of claims 1-5.

Images