A large model-based grain storage history grain condition intelligent data manual system

Abstract

The application discloses a kind of based on big model's grain storage history grain condition intelligent data manual system, comprising: intelligent query service subsystem, which receives the query content input by user, retrieves and calls matching data according to the query content, sorts the matching data obtained by retrieval according to matching degree, time correlation and data integrity, obtains grain condition retrieval data and outputs;Big model intelligent agent service subsystem is connected with the intelligent query service subsystem, receives the grain condition retrieval data, assembles big model input context, generates and outputs grain condition comprehensive analysis result.The application provides based on big model's grain storage history grain condition intelligent data manual system, can break through multi-source data, break data island, realize grain condition intelligent retrieval and automatic analysis and research.

Description

Technical Field

[0001] This invention belongs to the technical field of interdisciplinary application of computer information technology and grain storage safety management, and specifically relates to an intelligent data manual system for historical grain conditions based on a large model. Background Technology

[0002] With the deep integration of computer, Internet of Things and artificial intelligence technologies into the grain storage industry, various grain storage monitoring equipment, offline ledgers and annual inspection records have continuously accumulated massive amounts of historical grain storage data. Various monitoring terminals and storage management software have been widely used in daily grain storage monitoring, ledger archiving and safety supervision. Computer systems have become the core support tool for daily grain storage management, grain condition traceability and safety assessment.

[0003] my country's grain reserve system has accumulated a large amount of historical grain condition data over its long-term operation, including data on temperature and humidity in grain warehouses, regional rainfall, stratified temperature of grain piles, meteorological data, and pest data. This data typically originates from different years, different grain warehouses, and different monitoring equipment, resulting in problems such as inconsistent field formats, inconsistent sampling frequencies, scattered data storage, and a lack of unified indexes. These data silos are formed during computer storage and data transfer. Traditional management methods rely heavily on Excel spreadsheets, paper manuals, or static electronic documents, requiring staff to manually search, filter, summarize, and write monitoring reports, making comprehensive analysis across grain warehouses, years, and indicators difficult.

[0004] As grain storage safety management shifts from post-event recording to proactive analysis, the value of historical grain condition data is no longer limited to archiving and querying. It needs to be used for trend analysis, anomaly tracing, risk prediction, and report generation. However, existing computer systems supporting grain depots often only provide simple data storage or keyword retrieval functions, lacking unified governance, knowledge-based organization, anomaly pattern mining, natural language interaction, and automatic report generation capabilities for historical data. Furthermore, they haven't established complete data transmission links, making it difficult for various types of data to automatically connect to analysis modules. Especially in scenarios lacking real-time data or relying primarily on historical diagnostic data, how to build an intelligent data manual system capable of integrating data, invoking algorithms, generating detection conclusions, and outputting historical grain condition reports has become a crucial technical problem that needs to be solved in the field of grain storage safety management. Summary of the Invention

[0005] The purpose of this invention is to provide a smart data manual system for historical grain conditions based on a large model, which can connect multiple data sources, break down data silos, and realize intelligent retrieval and automated analysis and judgment of grain conditions.

[0006] The technical solution provided by this invention is as follows: A large-scale model-based intelligent data manual system for historical grain storage conditions includes: The intelligent query service subsystem receives the query content input by the user, retrieves matching data according to the query content, sorts the retrieved matching data according to the matching degree, time relevance and data integrity, and outputs the grain information retrieval data. The large model intelligent agent service subsystem is connected to the intelligent query service subsystem, receives the grain condition retrieval data, assembles the large model input context, and generates and outputs the comprehensive grain condition analysis results.

[0007] Preferably, the intelligent query service subsystem includes: The query parsing unit parses the query content to obtain specific query content such as grain warehouse name, time range, grain variety, monitoring indicators, and anomaly type; The intent recognition unit identifies the query content, determines and outputs the query intent; the query intent includes: query, comparison, trend analysis, anomaly backtracking and risk prediction; The graph matching unit uses a graph neural network matching algorithm to calculate the matching degree between the query content and the retrieved matching data. The result sorting unit sorts the retrieved matching data according to matching degree, time relevance and data integrity, and obtains and outputs the grain condition retrieval data.

[0008] Preferably, the matching data consists of historical grain condition time-series data, sub-grain condition knowledge graphs, abnormal cause analysis results, and risk prediction results retrieved based on the specific query content and the query intent.

[0009] Preferably, the large-scale intelligent agent service subsystem includes: The prompt word arrangement unit receives the query intent and selects the corresponding prompt template according to the query intent; The context assembly unit integrates and assembles the grain condition retrieval data into a large model input context, and generates and outputs the comprehensive grain condition analysis results. A result verification unit is generated, which performs consistency verification on the numerical values, time, grain warehouse name, and source of reference in the comprehensive grain condition analysis results. The interactive management unit supports human-computer interaction processes and allows for multiple rounds of follow-up questions, supplementation of query conditions, and regeneration of the comprehensive analysis results of the grain situation.

[0010] Preferably, the comprehensive analysis results of the grain situation include: natural language responses, structured detection conclusions, and disposal recommendations.

[0011] Preferably, it also includes: The intelligent data manual interaction layer receives user operation commands and performs natural language queries, condition selection, data viewing, report viewing, and export operations. The intelligent data manual interaction layer is connected to the intelligent query service subsystem and sends the user-input query content to the intelligent query service subsystem.

[0012] Preferably, it also includes: The historical grain condition monitoring report generation subsystem is connected to the large model intelligent agent service subsystem and the intelligent data manual interaction layer. It receives the comprehensive analysis results of the grain condition, arranges the content according to the comprehensive analysis results, generates and exports the historical grain condition monitoring report, and transmits the historical grain condition monitoring report to the intelligent data manual interaction layer for visualization or export.

[0013] Preferably, it also includes: The data access and governance subsystem is connected to the intelligent data manual interaction layer, imports multi-source historical grain condition data and performs data governance to obtain and output standardized historical grain condition time series data. The time-series database management subsystem is connected to the intelligent data manual interaction layer, the data access and governance subsystem, and the intelligent query service subsystem. It receives and stores the historical grain condition time-series data, receives the specific query content and the query intent, and retrieves and outputs the matching historical grain condition time-series data. The grain condition knowledge graph management subsystem is connected to the intelligent data manual interaction layer, the time series database management subsystem, and the intelligent query service subsystem. It receives the historical grain condition time series data, constructs and updates the grain condition knowledge graph, receives the specific query content and the query intent, and retrieves and outputs the matching sub-grain condition knowledge graph. The anomaly analysis and risk prediction subsystem is connected to the intelligent data manual interaction layer, the time-series database management subsystem, the grain condition knowledge graph management subsystem, and the intelligent query service subsystem. It retrieves the historical grain condition time-series data and the grain condition knowledge graph, receives the specific query content and the query intent, and retrieves and outputs the matching anomaly cause analysis results and the risk prediction results.

[0014] Preferably, it also includes: The permission and operation and maintenance management subsystem is connected to the intelligent data manual interaction layer and all subsystems, and performs unified management and control over user roles and data access permissions, as well as operation logs and operation status data generated by each subsystem.

[0015] The beneficial effects of this invention are as follows: The intelligent data manual system for historical grain conditions based on a large model provided by this invention can connect multi-source isolated grain condition data and break down data silos; it achieves standardization and time-series storage of heterogeneous historical data through unified governance, and completes the knowledge-based organization of grain condition data using knowledge graphs; it forms an automated analysis link based on natural language retrieval and a large model intelligent agent, automatically parsing queries, cross-database retrieval, and outputting analysis conclusions and visual reports, reducing manual processing costs; it is equipped with hierarchical permissions and full-process log auditing, realizing full traceability of data, model calls, and report exports, and can conduct anomaly reviews and risk predictions based on historical data, deeply explore the value of existing data, improve the efficiency and credibility of grain condition analysis, and provide technical support for digital safety management and intelligent risk decision-making in grain storage. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall architecture of the intelligent data manual system for historical grain conditions based on a large model, as described in this invention. Figure 2 This is a flowchart illustrating the historical grain situation data governance and knowledge-based process described in this invention. Figure 3 The flowchart for generating intelligent query and historical grain condition monitoring reports as described in this invention; Figure 4 This is a flowchart illustrating the generation and permission maintenance process of historical grain condition monitoring reports as described in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0018] like Figure 1 As shown, this invention provides an intelligent data manual system for historical grain conditions based on a large model. The specific implementation process is as follows: This invention provides a large-scale model-based intelligent data manual system for historical grain conditions, comprising: an intelligent data manual interaction layer, a data access and governance subsystem, a time-series database management subsystem, a grain condition knowledge graph management subsystem, an anomaly analysis and risk prediction subsystem, an intelligent query service subsystem, a large-scale model intelligent agent service subsystem, a historical grain condition detection report generation subsystem, and an access control and maintenance management subsystem. Specifically: The intelligent data manual interaction layer receives user operation commands and performs a full range of operations, including natural language query, condition selection, data viewing, report viewing and export. The intelligent data manual interaction layer distributes user operation commands to each subsystem, receives the processing results returned by each subsystem and displays them visually, realizing interactive linkage between the user and each subsystem.

[0019] A data access and governance subsystem is connected to the intelligent data manual interaction layer. This subsystem imports multi-source historical grain condition data and performs data governance to obtain and output standardized historical grain condition time-series data. The multi-source historical grain condition data includes: grain warehouse environmental temperature and humidity data, regional rainfall data, grain pile stratification temperature data, meteorological data, and pest data. The data governance includes: format standardization, data cleaning, and time-series alignment.

[0020] The data access and governance subsystem includes: a data import unit, which supports batch import of multi-source historical grain condition data in Excel, CSV, and database interface formats; a format adaptation unit, which performs format standardization processing on the imported multi-source historical grain condition data: mapping field names, units of measurement, and time formats from different grain warehouses, years, and data sources to unified fields using a data dictionary; a data preprocessing unit, which cleans the multi-source historical grain condition data and performs time-series alignment based on a unified time granularity to obtain and output standardized historical grain condition time-series data; the data cleaning includes: filtering invalid records; filling missing data values ​​using linear interpolation, pre- and post-value filling, or null value retention strategies based on indicator type; and identifying and removing outliers based on business thresholds and statistical rules; and a data quality verification unit, which performs quality checks on the historical grain condition time-series data and outputs a data quality report; the data quality report includes the results of missing rate, outlier rate, duplication rate, and time continuity verification.

[0021] The time-series database management subsystem is connected to the intelligent data manual interaction layer, the data access and governance subsystem, and the intelligent query service subsystem. The time-series database management subsystem receives the historical grain condition time-series data and performs multi-dimensional partitioning storage, establishing a multi-dimensional joint index. The time-series database management subsystem receives user operation instructions issued by the intelligent data manual interaction layer, specific query content and query intent identified by the intelligent query service subsystem, and outputs the historical grain condition time-series data and source mapping that match the user's operation instructions, specific query content, and query intent.

[0022] The time-series database management subsystem includes: a partitioned storage unit, which partitions and stores the received historical grain condition time-series data into the time-series database according to the dimensions of grain warehouse, region, grain variety, monitoring indicator, and time; an index management unit, which establishes a joint index on the historical grain condition time-series data according to the dimensions of grain warehouse, region, grain variety, monitoring indicator, and time, enabling the time-series database management subsystem to quickly respond to various query requests initiated by time period, grain warehouse, indicator, and anomaly type, thereby improving data retrieval efficiency; a data version management unit, which manages the version of the historical grain condition time-series data, and saves the mapping relationship between the original data, cleaned data, and analyzed data, facilitating subsequent traceability of data sources; and a data query interface unit, which provides standardized data call interfaces to the grain condition knowledge graph management subsystem, the anomaly analysis and risk prediction subsystem, the intelligent query service subsystem, and the large model intelligent agent service subsystem, providing stable time-series data call services to each subsystem.

[0023] The grain condition knowledge graph management subsystem is connected to the intelligent data manual interaction layer, the time-series database management subsystem, and the intelligent query service subsystem. The grain condition knowledge graph management subsystem receives historical grain condition time-series data output by the time-series database management subsystem, extracts entities and relationships from the historical grain condition time-series data, constructs and updates the grain condition knowledge graph, and stores the grain condition knowledge graph in a graph database. The grain condition knowledge graph management subsystem also receives user operation instructions from the intelligent data manual interaction layer, specific query content identified by the intelligent query service subsystem, and outputs a sub-grain condition knowledge graph that matches the user's operation instructions, specific query content, and query intent.

[0024] The grain condition knowledge graph management subsystem includes: a graph construction unit, which uses a recursive semantic segmentation algorithm to divide the continuous historical grain condition time-series data into semantically complete grain condition event segments; and uses entity recognition and entity relationship extraction algorithms to extract entities such as granaries, grain, monitoring indicators, environmental factors, grain condition status, time segments, and grain layers and their relationships from the segmented grain condition event segments, converting the historical grain condition time-series data into triplet data, and using the triplet data to construct the grain condition knowledge graph. A graph update unit, which performs incremental updates or conflict resolution on the grain condition knowledge graph according to a preset period based on newly added historical grain condition data, ensuring the timeliness, accuracy, and completeness of the grain condition knowledge graph. A graph storage unit, which stores the grain condition knowledge graph in a graph database, saving entity, relationship, attribute, and data source records, realizing the solidified storage and traceability management of knowledge data. The graph retrieval unit accurately retrieves and outputs matching sub-grain information knowledge graphs in the graph database based on the user's operation instructions issued by the intelligent data manual interaction layer, the specific query content and query intent identified by the intelligent query service subsystem, and the user's operation instructions.

[0025] The anomaly analysis and risk prediction subsystem is connected to the intelligent data manual interaction layer, the time-series database management subsystem, the grain condition knowledge graph management subsystem, and the intelligent query service subsystem. The anomaly analysis and risk prediction subsystem retrieves historical grain condition time-series data and entity, event, and relational data from the grain condition knowledge graph. Combining historical grain condition anomaly cases, it uses an anomaly pattern mining algorithm based on the Transformer model to analyze the correlation between anomaly events and temperature, humidity, rainfall, grain temperature changes, and pest records. It then compares these findings with grain condition risk warning thresholds to generate anomaly cause analysis results and risk prediction results. The grain condition knowledge graph management subsystem receives user operation commands from the intelligent data manual interaction layer and specific query content and intent identified by the intelligent query service subsystem. It then outputs the anomaly cause analysis results and risk prediction results that match the user's operation commands, specific query content, and query intent.

[0026] The anomaly analysis and risk prediction subsystem includes: an anomaly case library unit, which stores historical grain condition anomaly cases, saving basic case information, anomaly characteristics, handling measures, and handling effects; each historical grain condition anomaly case includes: grain warehouse number, grain variety, anomaly time, anomaly characteristics, handling measures, and handling effects; an anomaly pattern mining unit, which combines the historical grain condition time-series data, the grain condition knowledge graph, and the historical grain condition anomaly cases, and uses an anomaly pattern mining algorithm based on the Transformer model to analyze the correlation between anomaly events and temperature and humidity, rainfall, grain temperature changes, and pest records, generating and outputting the anomaly cause analysis results; a risk threshold management unit, which maintains corresponding grain condition risk warning thresholds according to region, season, grain variety, and warehouse type; and a prediction result output unit, which completes risk assessment based on the grain condition risk warning thresholds, generating and outputting the risk prediction results; the risk prediction results include: risk level, key influencing factors, and recommended attention period.

[0027] The time-series database management subsystem, the grain condition knowledge graph management subsystem, and the anomaly analysis and risk prediction subsystem are all directly connected to the intelligent data manual interaction layer. They can directly retrieve matching data from the above systems according to the user's operation instructions and transmit it to the intelligent data manual interaction layer for visualization.

[0028] The intelligent query service subsystem is connected to the intelligent data manual interaction layer, the time-series database management subsystem, the grain condition knowledge graph management subsystem, and the anomaly analysis and risk prediction subsystem. The intelligent query service subsystem receives and parses the natural language query content sent by the intelligent data manual interaction layer to obtain specific query content, identifies the query intent in the natural language query content, retrieves matching data as candidate matching data based on the specific query content and the query intent, and sorts the retrieved candidate matching data according to matching degree, time relevance, and data completeness to obtain grain condition retrieval data. Finally, the intelligent query service subsystem outputs the specific query content, the query intent, and the grain condition retrieval data.

[0029] The candidate matching data is multi-source candidate matching data, which includes: historical grain condition time-series data, sub-grain condition knowledge graph, anomaly cause analysis results, and risk prediction results retrieved according to the specific query content and query intent. In practical applications, based on the specific query content and query intent, one or more of the following data are selected for retrieval and sorted to form the grain condition retrieval data.

[0030] The intelligent query service subsystem includes: a query parsing unit, which parses the natural language query content to obtain specific query content such as grain warehouse name, time range, grain variety, monitoring indicators, and anomaly type; an intent recognition unit, which identifies the natural language query content, determines and outputs the query intent, and clarifies the query needs of the system operator; the query intent includes: query, comparison, trend analysis, anomaly backtracking, and risk prediction; a graph matching unit, which uses a graph neural network matching algorithm to calculate the matching degree between the natural language query content and the retrieved multi-source candidate matching data; and a result sorting unit, which sorts the retrieved multi-source candidate matching data according to matching degree, time relevance, and data integrity, and obtains and outputs the grain condition retrieval data (structured candidate results).

[0031] A large-scale intelligent agent service subsystem is connected to the intelligent query service subsystem. The large-scale intelligent agent service subsystem receives the query intent and the grain condition retrieval data, selects a corresponding prompt template based on the query intent, integrates and assembles the grain condition retrieval data into a large-scale model input context, and generates and outputs a comprehensive grain condition analysis result according to the selected prompt template. The comprehensive grain condition analysis result includes: natural language response, structured detection conclusions, and disposal recommendations.

[0032] The large model intelligent agent service subsystem is also connected to the time-series database management subsystem, the grain condition knowledge graph management subsystem, and the anomaly analysis and risk prediction subsystem, and can auxiliaryly retrieve data and detailed information from each subsystem as needed to supplement the input context of the large model.

[0033] The large-scale model intelligent agent service subsystem includes: a prompt word arrangement unit, which selects the corresponding prompt template according to the query intent; a context assembly unit, which integrates and assembles the grain condition retrieval data into a large-scale model input context, generating and outputting comprehensive grain condition analysis results; and a result verification unit, which performs consistency verification on the values, time, grain warehouse names, and reference sources in the comprehensive grain condition analysis results, resolving inconsistencies between the generated comprehensive grain condition analysis results and the original data. When the result verification unit fails, the large-scale model input context is reassembled, or the user is prompted to reconfirm, supplement, or re-enter the query conditions through the interaction management unit. The interaction management unit supports human-computer interaction processes, supports multi-round follow-up questions, query condition supplementation, and regeneration of the comprehensive grain condition analysis results, adapting to complex grain condition analysis query scenarios.

[0034] The historical grain condition monitoring report generation subsystem is connected to the intelligent data manual interaction layer and the large-scale model intelligent agent service subsystem. This subsystem receives the comprehensive grain condition analysis results, arranges the content based on these results, selects the corresponding monitoring report template, and generates a traceable, reusable, and exportable historical grain condition monitoring report according to the user-selected grain warehouse, time range, monitoring indicators, and monitoring report template. This report meets the needs of daily monitoring, anomaly review, reporting and summarizing, and management decision-making. The historical grain condition monitoring report generation subsystem can also transmit the historical grain condition monitoring report to the intelligent data manual interaction layer for visualization or export.

[0035] The historical grain condition monitoring report generation subsystem includes: a template management unit, which maintains various monitoring report templates such as daily monitoring reports, monthly monitoring reports, annual monitoring reports, anomaly retrospective monitoring reports, and special monitoring reports; a content arrangement unit, which organizes the comprehensive analysis results of the grain condition into the content of the historical grain condition monitoring report according to grain warehouse archives, monitoring data, trend analysis, anomaly records, correlation patterns, risk levels, monitoring conclusions, and disposal suggestions; a chart generation unit, which automatically generates data tables, trend line charts, indicator comparison charts, and anomaly event timelines based on the content of the historical grain condition monitoring report, improving the readability of the report; and a report export unit, which exports the completed historical grain condition monitoring report into common formats such as Word, PDF, or tables for easy archiving and application.

[0036] The permission and operation and maintenance management subsystem is connected to the intelligent data manual interaction layer and all subsystems. The permission and operation and maintenance management subsystem provides unified control over user roles and data access permissions, as well as operation logs and operation status data generated by each subsystem, to achieve permission management, task scheduling, log auditing and operation monitoring, and ensure system operation security and maintainability.

[0037] The permission and operation and maintenance management subsystem includes: a user role management unit, which classifies users into three roles: administrators, business personnel, and read-only users; a data access control unit, which divides permission boundaries according to region, data type, and data type, restricting the data access scope, model call, and report export permissions of different users; a task scheduling unit, which automatically triggers system tasks such as data import, graph update, anomaly analysis, and report generation according to preset cycles, achieving automated operation and maintenance; a log auditing unit, which records all operations such as data import, user query, model call, and report export, retaining complete logs for post-event auditing and problem tracing; and a operation monitoring unit, which monitors the overall operation status, load, and anomaly information of the time-series database, graph database, algorithm service, and large model service in real time, enabling real-time early warning of system failures.

[0038] like Figure 2 As shown, the historical grain condition data governance and knowledge-based process is as follows: Step 1: Start the data access and governance subsystem: The data import unit imports multi-source historical grain condition data; the format adaptation unit performs format standardization processing on the multi-source historical grain condition data, mapping field names, units of measurement, and time formats from different grain warehouses, different years, and different data sources to unified fields through a data dictionary, thus completing format adaptation; the data preprocessing unit performs data cleaning and time series alignment on the multi-source historical grain condition data; the data quality verification unit performs quality verification on the multi-source historical grain condition data. Step 2: Run the time series database management subsystem: The partitioned storage unit partitions and stores the historical grain condition time series data according to the dimensions of grain warehouse, region, grain variety, monitoring indicators, and time; the index management unit establishes a joint index on the historical grain condition time series data based on the dimensions of grain warehouse, region, grain variety, monitoring indicators, and time. Step 3: Run the grain condition knowledge graph management subsystem: The graph construction unit uses a recursive semantic segmentation algorithm to divide the continuous historical grain condition time-series data into semantically complete grain condition event segments; it uses entity recognition and entity relationship extraction algorithms to extract entities such as granaries, grain, monitoring indicators, environmental factors, grain condition status, time segments, and grain layers and their relationships from the segmented grain condition event segments, converting the historical grain condition time-series data into triplet data, and using the triplet data to construct the grain condition knowledge graph; the graph storage unit stores the grain condition knowledge graph in the graph database. Step 4: Output results: The data access and governance subsystem outputs standardized historical grain condition time-series data and a data quality report after governance; the time-series database management subsystem outputs historical grain condition time-series data and source mapping; the grain condition knowledge graph management subsystem outputs the grain condition knowledge graph.

[0039] like Figure 3As shown, the intelligent query and historical grain condition monitoring report generation process is as follows: Step 1: The intelligent query service subsystem is invoked to process the query requests from system users. The query parsing unit receives, parses, and identifies the specific query content in the natural language query content issued by the intelligent data manual interaction layer; the intent recognition unit identifies the query intent in the natural language query content. Step 2: Each subsystem collaboratively retrieves data. Based on the specific query content and query intent identified by the intelligent query service subsystem, the time-series database management subsystem retrieves the matching historical grain condition time-series data from the time-series database, the grain condition knowledge graph management subsystem outputs the matching sub-grain condition knowledge graph, and the anomaly analysis and risk prediction subsystem outputs the matching anomaly cause analysis results and risk prediction results. Step 3: The graph matching unit uses a graph neural network matching algorithm to calculate the matching degree between the natural language query content and the retrieved matching data; the result sorting unit sorts the retrieved matching data according to the matching degree, time relevance, and data integrity, and obtains and outputs structured candidate results. Step 3: Generate comprehensive grain condition analysis results based on the large-scale model intelligent agent service subsystem: The context assembly unit integrates and assembles the grain condition retrieval data into the large-scale model input context, generates and outputs comprehensive grain condition analysis results; the generated result verification unit performs consistency verification on the values, time, grain warehouse names, and reference sources in the comprehensive grain condition analysis results. Step 4: The historical grain condition monitoring report generation subsystem generates the historical grain condition monitoring report based on the comprehensive grain condition analysis results.

[0040] like Figure 4As shown, the historical grain condition monitoring report generation and permission maintenance process is as follows: Step 1: The permission and maintenance management subsystem completes identity and permission verification: The user role management unit verifies the user's identity and determines the user's data access scope, model call, and report export permissions. Step 2: The historical grain condition monitoring report generation subsystem runs: The template management unit selects the corresponding monitoring template; the content arrangement unit organizes and arranges the comprehensive grain condition analysis results into the content of the historical grain condition monitoring report according to grain warehouse archives, monitoring data, trend analysis, abnormal records, correlation patterns, risk levels, monitoring conclusions, and disposal suggestions; the chart generation unit automatically generates data tables, trend line charts, indicator comparison charts, and abnormal event timelines based on the content of the historical grain condition monitoring report. The historical grain condition monitoring report generation subsystem generates and exports traceable, reusable, and exportable historical grain condition monitoring reports. Step 3: The permission and operation and maintenance management subsystem performs full-process operation and maintenance control: The task scheduling unit automatically triggers system tasks such as data import, graph update, anomaly analysis, and report generation according to a preset cycle to achieve automated operation and maintenance; The log auditing unit records all operations such as data import, user query, model call, and report export throughout the process, and retains complete logs for post-event auditing and problem tracing; The operation monitoring unit monitors the overall operation status, load, and anomaly information of the time series database, graph database, algorithm service, and large model service in real time to achieve real-time early warning of system faults and ensure stable system operation.

[0041] Example 1: Historical Grain Condition Data Access and Governance: The intelligent data manual system for historical grain conditions based on a large model provided by this invention receives multi-source historical grain condition data from 2023 to 2024 for Grain Storage Warehouse No. 1 in Akto County. This multi-source historical grain condition data includes warehouse type, grain variety, inventory quantity, internal temperature and humidity, external temperature and humidity, and grain pile layer temperature. The data access and governance subsystem first unifies fields such as "warehouse temperature" and "internal temperature" into "internal temperature" according to the data dictionary, and unifies different time formats into a standard timestamp. Then, it filters invalid records, completes short-term missing values, and aligns data from different sampling frequencies to a unified time granularity, obtaining standardized historical grain condition time-series data. Finally, the time-series database management subsystem writes the historical grain condition time-series data into the time-series database.

[0042] Example 2: Knowledge Graph Construction: The intelligent data manual system for historical grain conditions based on a large model provided by this invention reads the historical grain condition time-series data from the time-series database. The grain condition knowledge graph management subsystem uses a recursive semantic segmentation algorithm to divide the continuous historical grain condition time-series data into multiple grain condition event fragments.

[0043] Entity recognition and entity relationship extraction algorithms are used to extract entities such as granaries, grain, monitoring indicators, time segments, and grain condition status from the segmented grain condition event fragments, and their relationships are formed into triples such as "Warehouse No. 1 - Storage - White Hard Winter Wheat", "Upper Temperature - Belongs to - Temperature Value Set", and "June 15, 2023 to June 16, 2023 - Grain Condition Status - Temperature Slowly Rising". The grain condition knowledge graph is constructed and written into the graph database.

[0044] Example 3: Anomaly Analysis and Risk Prediction: The intelligent data manual system for historical grain conditions based on a large model provided by this invention reads the historical anomaly diagnosis forms and the corresponding time period data of temperature, humidity, rainfall, grain temperature and pests in the historical grain condition anomaly cases. The data from 30 days before the anomaly occurred to 7 days after the anomaly ended is input into the anomaly pattern mining algorithm based on the Transformer model. The anomaly analysis and risk prediction subsystem outputs the correlation weights between the anomaly event and the increase in external temperature, the change in internal humidity, the rainfall process and the change in the temperature of the upper layer of the grain pile, and generates risk warning rules for the high temperature period in summer.

[0045] Example 4: Natural Language Query and Historical Grain Condition Monitoring Report Generation: The user inputs the natural language query content as "Analyze the abnormal grain temperature in Grain Warehouse No. 1 of Akto County in the summer of 2023 and generate a monitoring report". The intelligent query service subsystem identifies the query intent of the natural language query content as anomaly analysis and report generation. The specific query content is identified as: grain warehouse name is Akto County No. 1 warehouse, time range is summer 2023, and indicator is abnormal grain temperature. Based on the specific query content and query intent, matching data is retrieved: the time-series database management subsystem retrieves matching historical grain condition time-series data from the time-series database; the grain condition knowledge graph management subsystem constructs a matching sub-grain condition knowledge graph from the grain condition knowledge graph; the anomaly analysis and risk prediction subsystem outputs matching anomaly cause analysis results and risk prediction results; the intelligent query service subsystem calculates the matching degree between the natural language query content and the retrieved matching data, and the large model intelligent agent service subsystem integrates the data information, outputting the anomaly time, changes in key indicators, related environmental factors, detection conclusions, and disposal suggestions; finally, the historical grain condition detection report generation subsystem generates a historical grain condition detection report.

[0046] The intelligent data manual system for historical grain conditions based on a large model provided by this invention can connect scattered and isolated multi-source grain condition data, effectively breaking down data silos in the industry; it achieves standardized and time-series storage of heterogeneous historical data through unified data governance, and realizes knowledge-based organization of grain condition data with the help of knowledge graphs; relying on natural language retrieval and large model intelligent agents to build an automated analysis link, it can intelligently parse specific query content, query intent, cross-database matching and retrieval data, and automatically generate standardized analysis conclusions and visualized detection reports, significantly reducing the workload of manual summarization, data comparison, and report compilation; it is equipped with hierarchical permission control and full-process log auditing, realizing full traceability of all historical data, model calls, and report exports; based on massive archived historical data, it completes anomaly backtracking and grain condition risk prediction, fully explores the value of existing historical grain condition data, improves the efficiency and reliability of grain storage data analysis, and provides complete technical support for digital management and intelligent risk decision-making for grain storage safety.

[0047] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A smart data manual system for historical grain storage conditions based on a large model, characterized in that, include: The intelligent query service subsystem receives the query content input by the user, retrieves matching data according to the query content, sorts the retrieved matching data according to the matching degree, time relevance and data integrity, and outputs the grain information retrieval data. The large model intelligent agent service subsystem is connected to the intelligent query service subsystem, receives the grain condition retrieval data, assembles the large model input context, and generates and outputs the comprehensive grain condition analysis results.

2. The intelligent data manual system for historical grain conditions based on a large model as described in claim 1, characterized in that, The intelligent query service subsystem includes: The query parsing unit parses the query content to obtain specific query content such as grain warehouse name, time range, grain variety, monitoring indicators, and anomaly type; An intent recognition unit identifies the query content, determines and outputs the query intent; the query intent includes: query, comparison, trend analysis, anomaly backtracking and risk prediction; The graph matching unit uses a graph neural network matching algorithm to calculate the matching degree between the query content and the retrieved matching data. The result sorting unit sorts the retrieved matching data according to matching degree, time relevance and data integrity, and obtains and outputs the grain condition retrieval data.

3. The intelligent data manual system for historical grain conditions based on a large model as described in claim 2, characterized in that, The matching data consists of historical grain condition time-series data, sub-grain condition knowledge graphs, abnormal cause analysis results, and risk prediction results retrieved based on the specific query content and the query intent.

4. The intelligent data manual system for historical grain conditions based on a large model as described in claim 3, characterized in that, The large-model intelligent agent service subsystem includes: The prompt word arrangement unit receives the query intent and selects the corresponding prompt template according to the query intent; The context assembly unit integrates and assembles the grain condition retrieval data into a large model input context, and generates and outputs the comprehensive grain condition analysis results. A result verification unit is generated, which performs consistency verification on the numerical values, time, grain warehouse name, and source of reference in the comprehensive grain condition analysis results. The interactive management unit supports human-computer interaction processes and allows for multiple rounds of follow-up questions, supplementation of query conditions, and regeneration of the comprehensive analysis results of the grain situation.

5. The intelligent data manual system for historical grain conditions based on a large model as described in claim 4, characterized in that, The comprehensive analysis results of the grain situation include: natural language responses, structured detection conclusions, and disposal recommendations.

6. The intelligent data manual system for historical grain conditions based on a large model as described in claim 4, characterized in that, Also includes: The intelligent data manual interaction layer receives user operation commands and performs natural language queries, condition selection, data viewing, report viewing, and export operations. The intelligent data manual interaction layer is connected to the intelligent query service subsystem, and sends the query content entered by the user to the intelligent query service subsystem.

7. The intelligent data manual system for historical grain conditions based on a large model as described in claim 6, characterized in that, Also includes: The historical grain condition monitoring report generation subsystem is connected to the large model intelligent agent service subsystem and the intelligent data manual interaction layer. It receives the comprehensive analysis results of the grain condition, arranges the content according to the comprehensive analysis results, generates and exports the historical grain condition monitoring report, and transmits the historical grain condition monitoring report to the intelligent data manual interaction layer for visualization or export.

8. The intelligent data manual system for historical grain conditions based on a large model as described in claim 6, characterized in that, Also includes: The data access and governance subsystem is connected to the intelligent data manual interaction layer, imports multi-source historical grain condition data and performs data governance to obtain and output standardized historical grain condition time series data. The time-series database management subsystem is connected to the intelligent data manual interaction layer, the data access and governance subsystem, and the intelligent query service subsystem. It receives and stores the historical grain condition time-series data, receives the specific query content and the query intent, and retrieves and outputs the matching historical grain condition time-series data. The grain condition knowledge graph management subsystem is connected to the intelligent data manual interaction layer, the time series database management subsystem, and the intelligent query service subsystem. It receives the historical grain condition time series data, constructs and updates the grain condition knowledge graph, receives the specific query content and the query intent, and retrieves and outputs the matching sub-grain condition knowledge graph. The anomaly analysis and risk prediction subsystem is connected to the intelligent data manual interaction layer, the time-series database management subsystem, the grain condition knowledge graph management subsystem, and the intelligent query service subsystem. It retrieves the historical grain condition time-series data and the grain condition knowledge graph, receives the specific query content and the query intent, and retrieves and outputs the matching anomaly cause analysis results and the risk prediction results.

9. The intelligent data manual system for historical grain conditions based on a large model as described in claim 8, characterized in that, Also includes: The permission and operation and maintenance management subsystem is connected to the intelligent data manual interaction layer and all subsystems, and performs unified management and control over user roles and data access permissions, as well as operation logs and operation status data generated by each subsystem.

Images