An edible mushroom full-industry-chain multi-agent closed-loop management and control system and method

Abstract

This invention discloses a multi-agent closed-loop management and control system and method for the entire edible fungi industry chain, relating to the field of digital management and control technology. It includes: a data acquisition module, a digital twin module, an artificial intelligence analysis module, an intelligent agent module for industrialized edible fungi cultivation, an intelligent agent module for deep processing of edible fungi, an intelligent agent collaborative control module, an execution module, a feedback update module, and a national health and medical big data intelligent agent module. Through the layered linkage of seven modules—data acquisition, digital twin, artificial intelligence analysis, intelligent agent module for industrialized edible fungi cultivation, intelligent agent collaborative control, execution, and feedback update—a complete data closed loop of forward flow and reverse feedback is formed. This opens up the data channel for the entire edible fungi process from strain cultivation to terminal sales, replacing the traditional fragmented system and manual interconnection operation mode. It effectively solves the problems of data fragmentation and insufficient management and control coordination in the industry chain, significantly improving the efficiency of integrated management and control across the entire chain.

Description

Technical Field

[0001] This invention relates to the field of digital management and control technology, specifically to a multi-agent closed-loop management and control system and method for the entire edible fungi industry chain. Background Technology

[0002] The domestic production and sales scale of edible fungi continues to expand, while the cross-border export trade of edible fungi has been growing year by year. The industrial chain extends to cover the entire process of spawn cultivation, spawn production, facility cultivation, fresh mushroom processing, deep processing, domestic cold chain, ocean cross-border transportation, community retail, and end-consumer consumption. At the same time, it has also generated supporting management needs such as quality traceability, supply and demand allocation, carbon asset accounting, government industry supervision, and agricultural financial credit.

[0003] Currently, the digital transformation of the industry generally adopts fragmented systems such as single mushroom shed monitoring, simple traceability mini-programs, general agricultural AI tools, and independent online stores. These systems are independent of each other, can only cover a single link in the industrial chain, and cannot connect the data of the entire chain. At the same time, existing digital products lack a full-domain virtual simulation base and a hierarchical autonomous intelligent agent scheduling architecture, making it difficult to achieve two-way data exchange between production, deep processing, trade, and consumption. They cannot rely on end-user demand to adjust upstream strains and cultivation capacity. There are prominent pain points in the industry such as production and sales mismatch, high fresh food loss, delayed handling of quality risks, inability to quantify and trade carbon assets, and large amount of manual operation in cross-border trade. There is an urgent need for a full-chain closed-loop management and control solution that integrates digital twins and multi-agent collaboration.

[0004] Furthermore, the existing multi-agent management system for edible fungi only covers the planting, processing, distribution, production and sales, and carbon sequestration industrial chain. It lacks an independent intelligent scheduling unit for the intensive processing of edible fungi. Processes such as drying, freeze-drying, ready-to-eat fungi products, edible fungi extracts, and medicinal cuisine processing lack dedicated autonomous intelligent agents for overall management. Processing parameters, finished product quality, processing losses, and the proportion of raw materials for deep processing cannot be coordinated and scheduled across the entire chain. It does not extend to the consumer's health end, lacking links for foodborne safety medical traceability, population dietary suitability, and food health risk early warning and management. There is a business gap between agricultural supervision and health supervision. Industrial production data and desensitized health and medical big data belong to two independent systems, forming data silos. There is no supporting health autonomous intelligent agent to achieve collaborative decision-making across agricultural and medical scenarios, and the completeness of the entire industrial chain closed loop has obvious shortcomings.

[0005] Existing digital platforms mostly adopt a single-point centralized architecture, rarely building a layered linkage system that integrates data collection, digital twins, AI analysis, intelligent agent collaboration, equipment execution, and data feedback. They generally lack a full-domain 3D virtual simulation carrier and the ability to coordinate and schedule multiple autonomous intelligent agents. Data from various links in the industrial chain is difficult to share and communicate with each other. Work such as production control, deep processing technology management, quality control, and supply and demand allocation still requires a lot of manual intervention, making it difficult to autonomously complete the linkage between virtual and real scenarios and iterative optimization of control strategies. Summary of the Invention

[0006] The purpose of this invention is to provide a multi-agent closed-loop management and control system and method for the entire edible fungi industry chain, in order to solve the problems of fragmented architecture, insufficient integrated layered linkage, digital twin and multi-agent collaborative capabilities in existing edible fungi digital management and control systems, which make it difficult to achieve autonomous closed-loop management of the entire chain.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A closed-loop management and control system and method for the entire edible fungi industry chain with multiple intelligent agents, including: a data acquisition module, a digital twin module, an artificial intelligence analysis module, an intelligent agent module for industrialized edible fungi cultivation, an intelligent agent module for intensive processing of edible fungi, an intelligent agent collaborative control module, an execution module, a feedback and update module, and a national health and medical big data intelligent agent module;

[0009] The data acquisition module serves as the raw data input unit for the entire industry chain. It collects multi-source raw data from all stages of edible fungi cultivation, substrate production, cultivation management, deep processing, quality testing, warehousing and transportation, and terminal sales. The collected real-time data is then synchronously transmitted to the digital twin module. The data acquisition module also includes a health data anonymization access unit. This unit utilizes federated privacy computing and tiered authorization interfaces to connect to the national health and medical compliance big data resource pool, acquiring anonymized data from resident health checkup labels, foodborne adverse reaction cases, nutritional testing of edible fungi as both food and medicine, and health records. This achieves bidirectional integration and governance of industry and medical data, with original medical records stored locally and not transmitted externally. Simultaneously, the data acquisition module collects processing-specific data such as equipment parameters, raw material input ratios, semi-finished product losses, and physicochemical indicators of deep-processed products from the entire deep processing production line, including drying, freeze-drying, extraction, sterilization, and packaging processes.

[0010] The digital twin module is a virtual-real mapping simulation unit that receives full-industry chain data transmitted by the data acquisition module, constructs a full-domain virtual mapping model that corresponds one-to-one with the physical edible fungus industry chain, and synchronously generates a standardized virtual simulation dataset and sends it to the artificial intelligence analysis module. The meso-level twin model adds a three-dimensional simulation layer of the deep processing production line, which displays the equipment operation status, semi-finished product inventory, and processing energy consumption data of the drying workshop, extraction workshop, and aseptic packaging workshop in real time.

[0011] The AI ​​analysis module is an intelligent judgment and calculation unit that receives real-time data and simulation datasets output by the digital twin module. Through its built-in analysis model, it completes the judgment of fungal growth, quality risk identification, supply and demand trend prediction, and nutritional and shelf-life prediction of deep-processed products. It transmits multi-dimensional decision-making and prediction information to the intelligent agent collaborative control module. The AI ​​analysis module is equipped with a health judgment sub-model group. Based on the parallel operation of the integrated industry and medical dual-source data, it outputs three types of health dimension prediction and judgment information: health risk classification, dietary matching recommendation, and food safety traceability location, and pushes them to the national health and medical big data intelligent agent module simultaneously. The AI ​​analysis module adds a deep-processing technology optimization sub-model, which automatically optimizes process parameters such as raw material ratio, drying temperature, and extraction time based on upstream fresh mushroom production, downstream deep-processing orders, and health and wellness dietary needs.

[0012] The intelligent module for industrialized edible fungi cultivation is a dedicated scheduling unit for industrialized cultivation. It is connected to a unified message bus and generates exclusive control instructions for scenarios such as closed mycelium cultivation, intelligent mushroom production workshops, and automated mushroom stick production lines, and sends them to the execution module. The priority of industrialized control instructions is lower than that of health risk disposal instructions but higher than that of conventional open-field mushroom shed control instructions.

[0013] The intelligent module for edible fungi deep processing is a dedicated autonomous scheduling unit for the deep processing production line. It is connected to a unified global message bus and coordinates the intelligent scheduling of all processes, including drying, low-temperature freeze-drying, edible fungi polysaccharide extraction, ready-to-eat fungi snack processing, medicinal food ingredient packaging, and sterilization. Based on the upstream supply of fresh / dried mushroom raw materials, cross-border deep processing orders, and customized health and wellness dietary needs, it automatically generates dedicated instructions for processing production scheduling, raw material feeding, equipment temperature control, finished product quality inspection, and semi-finished product inventory scheduling. The deep processing control instructions have a lower priority than health risk disposal instructions but a higher priority than conventional open-field cultivation control instructions, and are coordinated and scheduled at the same level as the intelligent module for edible fungi factory cultivation. The module has two built-in auxiliary decision-making sub-units: processing loss optimization and nutritional standardization management of deep-processed finished products. It can also link with health assessment models to output adjustment plans for deep-processed product formulas suitable for people with chronic diseases and those seeking health and wellness.

[0014] The intelligent agent collaborative control module is a multi-agent scheduling and computing unit. It receives the prediction information output by the artificial intelligence analysis module, drives the collaborative computing of multiple autonomous intelligent agents, generates hierarchical regulation and control strategies adapted to production, deep processing, quality inspection, logistics, and sales scenarios, and sends them to the execution module. The intelligent agent collaborative control module is configured with a unified message bus as the global collaborative interaction hub. The unified message bus is simultaneously connected to the national health and medical big data intelligent agent module, the edible fungi factory planting intelligent agent module, and the edible fungi deep processing intelligent agent module. This enables data sharing, task linkage, and instruction priority arbitration among industrial intelligent agents, factory-specific intelligent agents, deep processing-specific intelligent agents, and health-related intelligent agents. Health risk-related scheduling instructions have higher priority than conventional production, sales, planting, and deep processing regulation instructions.

[0015] The National Health and Medical Big Data Intelligent Agent Module is a dedicated management and control unit for agricultural and medical integration, sharing a digital twin simulation base, AI analysis computing power, and a unified collaborative scheduling bus with the entire system. The module embeds a health data fusion subsystem and four types of health-specific autonomous intelligent agents. The health data fusion subsystem utilizes a federated privacy computing + consortium blockchain architecture to achieve cross-departmental data anonymization and ownership confirmation. The four types of health-specific autonomous intelligent agents are: a health risk assessment intelligent agent, a dietary and health recommendation intelligent agent, a food safety and medical traceability intelligent agent, and a health supervision and reporting intelligent agent. The health risk assessment intelligent agent integrates data on pesticide residues, heavy metals, and quality of edible fungi with population health profiles to quantify health hazards from food consumption and determine dietary contraindications for people with chronic diseases or allergies. The dietary and health recommendation intelligent agent… The system provides personalized consumption matching solutions for fresh and processed edible fungi to ordinary consumers, health and wellness institutions, and medical institutions. When the food safety and medical traceability intelligent system detects excessive levels of harmful substances in products or receives foodborne complaints, it links with the health and wellness database to trace the entire chain from planting, processing, deep processing, to cold chain distribution, accurately locating the source of contamination. The health and wellness supervision reporting intelligent system automatically summarizes the quality and health risks of edible fungi and batch testing records, completing a closed-loop reporting to the agricultural and health and wellness supervision platforms. After generating instructions such as risk locking, recall, production capacity shutdown, and deep processing line shutdown, the health intelligent system synchronously pushes these instructions to the production, quality inspection, logistics, sales intelligent systems, edible fungi factory planting intelligent system module, and edible fungi deep processing intelligent system module through a global message bus to execute control actions simultaneously.

[0016] The execution module is a physical equipment driving unit that receives control strategies from the intelligent agent collaborative control module, the edible fungi factory cultivation intelligent agent module, and the edible fungi deep processing intelligent agent module. It then performs automated adjustments to the physical equipment supporting edible fungi production, deep processing lines, warehousing, and stores. The execution module also includes a new health and wellness linkage execution logic. It receives risk management instructions from the health and medical intelligent agent and automatically completes operations such as removing problematic batches of goods from shelves, cold chain interception, production restrictions at the production base, and raw material isolation and shutdown of the deep processing line. The execution module further includes a factory cultivation end execution unit and a deep processing line execution unit. The deep processing line execution unit interfaces with the drying equipment, freeze dryer, extraction kettle, and aseptic packaging line PLC, executing process temperature, feeding ratio, and line start / stop control instructions from the deep processing intelligent agent. The factory cultivation end execution unit interfaces with the workshop environmental control and line electrical control equipment, executing workshop environment and line start / stop control instructions from the factory cultivation intelligent agent.

[0017] The feedback update module is a closed-loop feedback unit that collects real-time data on the operational results of the industrial chain after the execution module has completed its control, and feeds it back to the digital twin module for iterative updates to the full-domain virtual mapping model. Simultaneously, the feedback update module collects consumer consumption feedback, health and food safety records, health and wellness dietary usage data, market feedback on processed products, and processing loss data, and feeds it back to the artificial intelligence analysis module for incremental training of sub-models for health risks, dietary recommendations, and processing technology optimization, continuously improving the accuracy of agricultural and medical integration assessments. Simultaneously, the feedback update module collects operational data from factory workshops and processing workshops, and iteratively adapts simulation and assessment models for factory cultivation and processing scenarios. Each module transmits data unidirectionally along the forward business link and forms a reverse data feedback path through the feedback update module, forming a closed-loop data transmission cycle. Relying on the national health and medical big data intelligent module, the edible fungus factory cultivation intelligent module, and the edible fungus processing intelligent module, the control link is extended to construct an integrated dual-closed-loop system for industrial control and health supervision, encompassing "strain cultivation—factory workshop production and processing—processing—cross-border circulation—terminal retail—human health feedback."

[0018] Furthermore, the data acquisition module includes an environmental parameter acquisition unit, a production process acquisition unit, a multispectral image acquisition unit, a logistics information acquisition unit, a health data anonymization access unit, and a deep processing data acquisition subunit. The environmental parameter acquisition unit acquires data on mushroom shed temperature and humidity, carbon dioxide, and substrate moisture content. The production process acquisition unit acquires data on strain propagation, mushroom stick sterilization, and processing and sorting processes. The multispectral image acquisition unit acquires images of mycelium and fruiting body diseases. The logistics information acquisition unit acquires data on cold chain temperature sensing, cross-border warehousing, and store inventory turnover. The deep processing data acquisition subunit acquires data on drying room temperature and humidity curves, freeze-drying vacuum degree, extraction solvent ratio, sterilization time, finished product packaging weight, nutritional testing indicators of deep-processed finished products, and processing waste data. The health data anonymization access unit deploys a federated privacy computing gateway, authorizing access only to anonymized tagged health and medical data, while original medical records and sensitive personal information are isolated locally and not transmitted externally. The data acquisition module simultaneously acquires equipment operation data from factory workshops, automated production lines, and deep processing production lines.

[0019] Furthermore, the digital twin module constructs a three-level hierarchical virtual mapping model. The three levels of virtual mapping models are: a microscopic twin model for simulating the growth state of single mycelial cells and spawn, a mesoscopic twin model for simulating mushroom sheds, processing plants, deep processing workshops, cold chain warehouse equipment and cultivation environment, and a macroscopic twin model for simulating overall production capacity, domestic and foreign trade supply and demand, and deep processing capacity. The three-level twin models synchronously receive real-time data streams from the data acquisition module to achieve real-time synchronization of virtual and real states. The mesoscopic twin model adds a simplified factory workshop simulation layer and a complete deep processing production line simulation layer to synchronously display the real-time operating status of the workshop, production line, and drying / extraction workshop. The macroscopic twin model is configured with a health supervision visualization layer to display the health risk level of all batches of fresh products and deep-processed products, food safety traceability trajectory, and regional dietary suitability population distribution.

[0020] Furthermore, the AI ​​analysis module incorporates four specialized sub-models: a growth prediction sub-model, a disease identification sub-model, a quality analysis sub-model, and a production and sales demand prediction sub-model. A new sub-model for optimizing deep processing technology is also added. The growth prediction sub-model predicts the mycelial development cycle based on twin simulation data. The disease identification sub-model identifies risks of contaminating bacteria, mold, and pests. The quality analysis sub-model grades and determines the quality of finished edible fungi. The production and sales demand prediction sub-model integrates production, cross-border, end-user, and deep processing order data to predict regional supply and demand fluctuations. The deep processing technology optimization sub-model automatically outputs optimal processing parameters based on raw material quality, health and wellness dietary needs, and export standards, reducing processing losses and standardizing finished product nutritional indicators. The AI ​​analysis module is equipped with a health assessment sub-model group, which, based on parallel computation of agricultural and medical dual-source data, outputs three types of assessment information: health risk classification, dietary recommendations, and traceability. The AI ​​analysis module also adds a lightweight factory-style growth prediction auxiliary model and a lightweight deep processing technology prediction model, supporting the computation of the intelligent edible fungi factory-style planting module and the intelligent edible fungi deep processing module, respectively.

[0021] Furthermore, the intelligent agent collaborative control module is configured with four types of basic autonomous intelligent agents: production management intelligent agent, quality inspection intelligent agent, logistics scheduling intelligent agent, and sales feedback intelligent agent. These intelligent agents share data and collaborate on tasks through a unified message bus. Based on the predictive information output by the artificial intelligence analysis module, they generate four hierarchical control strategies: cultivation regulation, batch interception, cold chain allocation, and store replenishment. The unified message bus synchronously connects to the four types of health-specific autonomous intelligent agents within the national health and medical big data intelligent agent module, the edible fungi factory cultivation intelligent agent module, and the edible fungi deep processing intelligent agent module. This enables cross-industry, health, factory cultivation, and deep processing multi-scenario event triggering and linkage. Health risk warning instructions have higher priority than all conventional industry, factory cultivation, and deep processing control instructions.

[0022] Furthermore, the intelligent agent collaborative control module also integrates a cross-border transaction sub-intelligent agent and a carbon sink accounting intelligent agent. The cross-border transaction sub-intelligent agent completes overseas inquiry matching, ocean cold chain scheduling, and export carbon label generation based on supply and demand forecast data. The carbon sink accounting intelligent agent calculates the carbon emission reduction of individual products and deep-processed products throughout their entire life cycle based on a digital twin model, generating carbon credit certificates that can be put on the blockchain. The carbon sink accounting intelligent agent and the cross-border transaction sub-intelligent agent can collaborate with the dietary health recommendation intelligent agent, the edible fungus factory cultivation intelligent agent module, and the edible fungus deep processing intelligent agent module to generate cross-border export product files that combine carbon labels, healthy eating suggestions, factory standardization traceability, and deep processing technology traceability.

[0023] Furthermore, the system adds a consortium blockchain evidence storage submodule; this submodule anonymizes and encrypts the data on the entire industry chain, carbon footprint, consumer feedback, and deep processing processes stored in the digital twin module, and uploads it to the blockchain, providing credible traceability archive support for each intelligent agent within the intelligent agent collaborative control module; the consortium blockchain simultaneously retains factory batch production records and deep processing batch processing records; the consortium blockchain evidence storage submodule simultaneously encrypts and uploads health and wellness integrated data, food safety disposal records, and health risk rating archives to the blockchain, achieving tamper-proof and credible evidence storage of data across the entire agricultural, medical, planting, and deep processing chain, supporting joint traceability verification across agricultural, health, and market supervision departments.

[0024] Furthermore, the execution module is configured with four types of execution units: a mushroom shed electrical control execution unit, a deep processing electrical control execution unit, a cold chain scheduling execution unit, a community store terminal execution unit, and a cross-border customs declaration execution unit. Each of these four execution units receives strategies from its corresponding intelligent agent and completes operations such as equipment adjustment, cargo transportation, disposal of near-expiry goods, and automatic generation of cross-border documents. The factory-based planting end execution unit, as a newly added sub-unit of the execution module, enables automated control of workshop temperature and humidity, and production line start / stop. A deep processing production line execution unit is added to independently manage the entire set of processing equipment, including drying, freeze-drying, extraction, and packaging. The community store terminal execution unit adds a health alert pop-up window and a mandatory locking function for removing high-risk products, simultaneously differentiating between health alerts for fresh mushrooms and deep-processed health and wellness products. The cross-border customs declaration execution unit can simultaneously export two sets of product health compliance test reports for fresh and deep-processed products for overseas customs clearance verification.

[0025] Furthermore, the system adopts a cloud-edge hierarchical deployment architecture; the digital twin macro model, global AI analysis model, and overall planning intelligent agent are deployed in the cloud; the micro twin model, lightweight disease identification model, and on-site lightweight intelligent agent are deployed at the edge terminals of production, deep processing workshops, and stores. The edge terminals of factory workshops are equipped with lightweight factory intelligent agents and simple simulation models, while the edge terminals of deep processing workshops are equipped with lightweight deep processing technology intelligent agents and lightweight nutritional risk identification models, all of which support offline basic control; the edge terminals support offline data collection and caching, and synchronously update the digital twin model after connecting to the network; the cloud uniformly deploys a health and medical big data fusion gateway, a global health judgment sub-model, and a global deep processing technology overall planning unit; the edge terminals are equipped with lightweight health risk identification models, which can quickly complete the initial assessment of basic health risks of fresh and deep-processed single products on-site, and synchronously transmit to the cloud after connecting to the network to complete a comprehensive assessment of the entire domain.

[0026] A multi-agent closed-loop management and control method for the entire edible fungi industry chain, characterized by comprising the following steps:

[0027] S1. Call the data acquisition module to collect real-time data of the entire edible fungi industry chain, and simultaneously access the de-identified national health and medical big data through the privacy computing gateway to complete the cleaning, spatiotemporal alignment, and fusion modeling of the two types of data; and simultaneously collect data on factory workshops, automated production line equipment and environment.

[0028] S2. The digital twin module builds a three-level virtual mapping model of micro, meso and macro levels and loads a health supervision visualization layer to generate an agricultural and medical integration simulation dataset; it also loads a simplified simulation layer of a factory workshop to achieve virtual-real synchronization.

[0029] S3, the artificial intelligence analysis module runs in parallel with five types of specialized sub-models: growth prediction sub-model, disease identification sub-model, quality analysis sub-model, production and sales demand prediction sub-model, and deep processing technology optimization sub-model, as well as a health assessment sub-model group. It simultaneously outputs three types of prediction and assessment information: industry, deep processing technology, and health; and simultaneously outputs lightweight factory-style growth prediction data and lightweight deep processing technology prediction data.

[0030] S4, the intelligent agent collaborative control module, schedules four types of basic autonomous intelligent agents and four types of health-specific autonomous intelligent agents—health risk assessment intelligent agent, dietary and health recommendation intelligent agent, food safety and medical traceability intelligent agent, and health supervision and reporting intelligent agent—through a unified message bus for collaborative operations. Simultaneously, it schedules the edible fungi factory cultivation intelligent agent module to complete collaborative decision-making in factory scenarios and the edible fungi deep processing intelligent agent module to generate deep processing production scheduling and process control instructions. The global unified message bus schedules instructions according to priority: health risk management instructions are issued to the execution module first, followed by factory cultivation control instructions, deep processing control instructions, and finally conventional open-field cultivation control instructions.

[0031] S5, the execution module completes the automated control of edible fungi production, deep processing production lines, warehousing, and cross-border equipment, and simultaneously executes health-related operations such as interception of high-risk goods, prohibition of sales in stores, and reporting of health materials; at the same time, it completes the automatic control of workshop and automated production line equipment through the factory planting end execution unit, and adjusts the parameters of drying, extraction, and packaging processes through the deep processing production line execution unit; high-risk batches simultaneously trigger material isolation in the factory workshop and shutdown and sealing of the entire deep processing production line.

[0032] S6, the feedback update module transmits back industry operation data, deep processing loss and finished product feedback, consumer health feedback, and health and sanitation disposal ledgers, and iteratively updates the virtual mapping model, five types of special sub-models, and health assessment sub-model; it also transmits back the operation data of the factory workshop and the deep processing workshop, and iteratively updates the factory adaptation simulation and deep processing technology auxiliary prediction model.

[0033] S7. Repeat steps S1 to S6 to continuously iterate and optimize the model and control strategies, forming a long-term integrated dual closed loop of industrial production, deep processing, and consumer health, and simultaneously realizing small-scale closed-loop iteration of continuous industrial edible fungi cultivation and continuous optimization of deep processing production line processes.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. This invention forms a complete data closed loop with forward flow and reverse feedback through the hierarchical linkage of seven modules: data acquisition, digital twin, artificial intelligence analysis, intelligent agent module for industrialized cultivation of edible fungi, intelligent agent module for deep processing of edible fungi, intelligent agent collaborative control, execution, and feedback update. It opens up the data channel for the entire process of edible fungi from strain cultivation to terminal sales, replacing the traditional fragmented system and manual operation mode. It effectively solves the problems of data fragmentation and insufficient coordination in the industrial chain, and significantly improves the efficiency of integrated management and control across the entire chain.

[0036] 2. By integrating digital twin simulation with multi-agent collaboration, proactive intelligent control is achieved. Relying on a three-level hierarchical digital twin model, multi-scale virtual-real synchronization and simulation can be realized, which can identify potential risks such as growth diseases and supply-demand gaps in advance. Combined with event-triggered collaborative scheduling of multiple autonomous agents, cross-linked operations such as cultivation regulation, quality interception, logistics allocation, and factory workshop and production line regulation are automatically completed. This transforms the traditional passive control of post-event handling into proactive control of pre-event prediction and automatic response, which significantly improves risk response speed and control accuracy while reducing labor costs.

[0037] 3. By using a feedback update mechanism to correct simulation models and algorithm parameters in reverse, the system's control capabilities can be continuously iterated and optimized over the operating cycle. At the same time, it adopts a cloud-edge hierarchical deployment architecture to adapt to the needs of production entities of different sizes and weak network scenarios. It can also be expanded with value-added capabilities such as cross-border trade, carbon sink accounting, and consortium blockchain notarization, which not only meets the basic production control needs, but also supports the long-term value of industrial factory-scale upgrading and green development.

[0038] 4. By integrating the national health and medical big data intelligent module, the edible fungus factory cultivation intelligent module, and the edible fungus deep processing intelligent module, and relying on privacy computing and consortium blockchain to achieve compliant and secure data interoperability between agriculture and medicine, breaking down data silos in the agricultural and health industries; extending the original single closed loop of the industrial chain to the consumer's human health feedback end, while taking into account the two mainstream models of open-field cultivation and factory year-round cultivation, constructing an integrated industry + health dual closed-loop management system, realizing intelligent linkage and disposal across agricultural and medical scenarios, and significantly improving the accuracy of food safety risk tracing and health hazard early warning; at the same time, it is compatible with the joint supervision of agriculture, medicine and health, and the digital implementation of edible fungus health care, covering the digital agriculture, medical big data, and multi-intelligent agent collaborative innovation track, and adapting to the application needs of high-tech, specialized and innovative, and agricultural digitalization projects. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0040] Figure 1 This is a schematic diagram of the overall structure provided for an embodiment of the present invention. Detailed Implementation

[0041] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0042] Example 1: This example is applied to a standardized shiitake mushroom cultivation base with an annual output of 500 tons in a county. It is equipped with an independent spawn cultivation room, an automated mushroom stick production workshop, 30 intelligent mushroom sheds with controllable environment, a weight sorting and processing line, a small-scale deep processing workshop (drying and simple packaging), a constant temperature cold storage at the production site, and 28 community fresh food cooperative stores. Cross-border export business has not been carried out. The system is fully deployed with modules for data acquisition, digital twins, artificial intelligence analysis, intelligent agents for industrialized edible fungi cultivation, intelligent agents for deep processing of edible fungi, intelligent agent collaborative control, execution, feedback and update, and national health and medical big data. It utilizes four types of basic autonomous intelligent agents and a three-level digital twin model to achieve autonomous closed-loop management of the entire chain from spawn cultivation to mushroom stick production, processing and sorting, deep drying and packaging, warehousing and distribution, and terminal retail. This base is only equipped with a simple mushroom stick production line and a small-scale drying and deep processing production line. The intelligent agent module for industrialized cultivation operates in a lightweight manner, and the intelligent agent module for deep processing is deployed in a lightweight manner. The execution unit at the industrialized cultivation end only manages the production line equipment, while the execution unit at the deep processing production line manages the drying unit and the packaging line.

[0043] 1. Data Acquisition Module Deployment and Data Acquisition Process

[0044] The data acquisition module includes an environmental parameter acquisition unit, a production process acquisition unit, a multispectral image acquisition unit, a logistics information acquisition unit, a health data anonymization access unit, and a deep processing data acquisition subunit. All acquired data is aggregated through the on-site edge gateway and then uploaded to the digital twin module in real time.

[0045] Environmental parameter acquisition unit: Eight sets of integrated sensors for temperature, humidity, carbon dioxide, and substrate moisture content are evenly distributed at a height of 0.5 meters from the substrate rack in each mushroom shed, with a sampling frequency of 1 time / minute, to collect real-time data on air temperature, relative humidity, CO2 volume fraction, and substrate moisture content in the shed; High-precision constant temperature and humidity sensors are installed in the spawn cultivation room, with a sampling frequency of 1 time / 5 minutes, to monitor environmental parameters during the spawn propagation stage.

[0046] Production process data acquisition unit: The mixing machine, high-pressure sterilizer, and automatic inoculation line in the mushroom stick production workshop are all connected to the PLC data interface to automatically collect core process parameters such as the ratio of sawdust to wheat bran, sterilization temperature and time, and inoculation amount; the sorting and processing line is connected to the data interface of the weight grading equipment to record the yield, grade ratio, and defect rate of shiitake mushrooms batch by batch.

[0047] Multispectral image acquisition unit: Two 4K multispectral industrial cameras are installed in each mushroom shed. They automatically take pictures of the mushroom logs and fruiting bodies at 8:00, 14:00 and 20:00 every day, covering the visible light and near-infrared bands, for subsequent disease identification and growth status determination.

[0048] Logistics information collection unit: Temperature-sensitive GPS tags are installed in cold storage facilities at the production site and in refrigerated delivery vehicles to collect real-time data on the temperature inside the storage facilities, vehicle location, and temperature of the transport compartment; RFID reading and writing devices and inventory management terminals are installed in 28 community stores to automatically record daily inbound volume, sales volume, quantity of near-expiry goods, and loss.

[0049] The health data anonymization access unit deploys a local federated privacy computing gateway, connecting to the county-level public health platform. It only retrieves anonymized population physical labels and statistical data on adverse foodborne reactions to edible fungi. Original medical records and sensitive personal information are stored locally in isolation. Data access permissions are managed at two levels: one for agricultural bases and one for health supervision. Simultaneously, it collects operating parameters from the mushroom cultivation production line equipment, supplying them to the factory-based intelligent system for computation.

[0050] Deep processing data acquisition subunit: Temperature, humidity, and hot air circulation speed sensors are installed in the small drying workshop to collect the drying curve every 30 seconds; weighing sensors and sealing machine PLCs are connected to the packaging line to collect the weight of dried mushrooms and packaging loss; the moisture content of dried shiitake mushrooms and pesticide residue retesting data are collected simultaneously, and all deep processing real-time data are uploaded to the digital twin module to supply the intelligent agent module for scheduling and calculation of edible fungi deep processing.

[0051] 2. Construction and Operation Logic of Digital Twin Module

[0052] The digital twin module receives all real-time data uploaded by the data acquisition module and constructs a three-tiered virtual mapping model at the micro, meso, and macro levels. The update delay of the entire model data does not exceed 5 seconds, achieving real-time virtual-real synchronization between the physical industrial chain and the virtual model.

[0053] Microscopic twin model: Based on the shiitake mushroom growth dynamics model, a three-dimensional simulation model of a single mushroom stick is constructed. The substrate moisture content, environmental temperature and humidity, and image texture feature data are input to simulate the mycelial growth rate, fruiting body development stage, and estimated harvesting time in real time. At the same time, the carbon fixation during the degradation process of the mushroom stick substrate is simulated simultaneously.

[0054] The mesoscopic twin model constructs a 1:1 3D scene model of 30 mushroom sheds, production workshops, cold storage facilities at the production site, and small-scale drying and deep processing workshops. It associates the real-time operating parameters of the corresponding equipment and dynamically displays the environmental status, equipment start-up and shutdown status, current inventory, real-time temperature and humidity of the drying workshop, and semi-finished product inventory of each mushroom shed. The mesoscopic twin adds a simple production line simulation layer and a drying and processing simulation layer. It supports simulation and can simulate the impact of adjusting temperature and humidity parameters on the growth cycle, disease incidence rate, and drying loss rate of shiitake mushrooms.

[0055] Macro Twin Model: Construct a data sandbox of the entire industrial chain of the base, integrate total production capacity, demand of each store, fresh mushroom inventory, dried mushroom deep-processed finished product inventory, quality pass rate and other aggregated data, visualize the supply and demand matching status of the whole chain and the overall operating efficiency. The macro twin model integrates a health supervision visualization layer, which visualizes the health risk level of each batch of fresh mushrooms and dried mushrooms, the suitable population for store diets, and historical food safety feedback records.

[0056] 3. Intelligent Analysis Module Calculation Process

[0057] The artificial intelligence analysis module calls the real-time data and simulation dataset output by the digital twin module, runs five types of special sub-models (with a newly added sub-model for optimizing deep processing technology) and a health assessment sub-model group in parallel, and after completing the state analysis and trend prediction, transmits the three-dimensional prediction information of industry, deep processing and health to the intelligent agent collaborative control module; it also outputs the production line factory-scale growth prediction data to the intelligent agent module for edible fungi factory-scale cultivation, and outputs the drying process optimization parameters to the intelligent agent module for edible fungi deep processing.

[0058] Growth prediction sub-model: Using historical growth data from the micro twin model and environmental prediction data for the next 7 days as input, the LSTM time series prediction algorithm is used to predict the accurate harvesting date and estimated yield of each mushroom shed, with the single batch yield prediction error controlled within 8%.

[0059] Disease identification sub-model: Features are extracted from multispectral images, and common diseases such as green mold, Trichoderma, and brown rot are identified based on the trained convolutional neural network model. The identification accuracy is ≥95%, and the specific location, severity level and preliminary treatment suggestions of the disease are output.

[0060] Quality analysis sub-model: Combining the weight data of the sorting equipment with the appearance characteristics of the images, the shiitake mushrooms are automatically divided into three grades: Grade 1, Grade 2, and Defective. The percentage of each grade and the pass rate are statistically analyzed for each batch.

[0061] Production and sales demand forecasting sub-model: Integrating the historical sales data of 28 stores over the past 60 days, holiday factors, and local weather factors, the time series forecasting algorithm outputs the replenishment demand for fresh and dried mushrooms and the optimal replenishment frequency for each store in the next 7 days.

[0062] The sub-model for optimizing the deep processing technology: Input the amount of fresh mushrooms harvested, the moisture content of fresh mushrooms, and the demand for dried mushrooms in stores, and automatically optimize the drying temperature gradient and drying time to reduce drying losses; Combine the needs of people with chronic diseases output by the health assessment model, adjust the packaging specifications and nutritional ratio labeling scheme of dried mushrooms.

[0063] The health assessment sub-model group uses the XGBoost multimodal model, inputting data on pesticide residues, heavy metals, and appearance quality of shiitake mushrooms, as well as physical examination indicators, allergy history, and chronic disease labels of desensitized individuals. It outputs batch-by-batch health risk classification results for fresh and dried mushrooms and appropriate dietary plans, which are simultaneously pushed to the National Health and Medical Big Data Intelligent Body module.

[0064] 4. Collaborative decision-making mechanism of the intelligent agent collaborative control module

[0065] The intelligent agent collaborative control module is configured with four types of basic autonomous intelligent agents: production management intelligent agent, quality inspection intelligent agent, logistics scheduling intelligent agent, and sales feedback intelligent agent. The module is connected to the national health and medical big data intelligent agent module, which is also connected to four types of health-specific autonomous intelligent agents, edible fungi factory cultivation intelligent agent module, and edible fungi deep processing intelligent agent module. Health-related instructions have higher priority than all conventional industries, factory production lines, and deep processing drying control. All intelligent agents achieve data sharing and event-triggered task collaboration through the unified message bus.

[0066] Production management intelligent agent: Receives yield forecast data from the growth prediction sub-model, combines it with the base's production capacity plan to automatically generate mushroom shed temperature and humidity control schemes and harvesting schedules, and directly sends them to the execution module.

[0067] Intelligent quality inspection agent: Receives disease identification and quality analysis results, automatically generates interception instructions for batches of fresh and dried mushrooms with excessive disease or defect rates, and pushes them simultaneously to the production end, deep processing workshop and store terminal.

[0068] Logistics scheduling intelligent agent: Receives production and sales demand forecasts and real-time fresh and dried mushroom inventory data, and automatically generates daily cold chain delivery routes and precise replenishment quantities of fresh and dried mushrooms for each store.

[0069] Sales Feedback Controller: This controller aggregates actual sales volume, near-expiry losses, and online consumer reviews from stores, and pushes this data back to the Production Management Controller and Deep Processing Controller modules to dynamically adjust subsequent production capacity planning, mushroom harvesting schedule, and drying and processing production schedule.

[0070] The collaborative logic of the intelligent agent module for edible fungi deep processing is as follows: It receives the optimal drying parameters output by the deep processing technology optimization sub-model, and combines them with the total daily harvest of fresh mushrooms and dried mushroom orders from stores to automatically generate start-up and shutdown scheduling for the drying unit, fresh mushroom feeding amount, and dried mushroom packaging production instructions, which are then sent to the execution unit of the deep processing production line. When the dietary health recommendation intelligent agent pushes the customized dried mushroom demand of people with chronic diseases, the deep processing intelligent agent automatically adjusts the packaging specifications and product health labels.

[0071] When the health risk assessment agent outputs a high-risk batch warning, the message bus simultaneously triggers multi-agent linkage: the quality inspection agent locks the batch of goods, the logistics scheduling agent suspends the corresponding cold chain delivery, the sales feedback agent pushes store management instructions, the health supervision reporting agent automatically generates a risk report and pushes it to the regulatory platform, the edible fungi factory cultivation agent module simultaneously issues instructions to seal materials on the production line, the edible fungi deep processing agent module issues instructions to isolate raw materials and seal finished products and shut down the drying production line, and the dietary health recommendation agent simultaneously pushes fresh and dried mushroom consumption plans for different groups to stores, all without human intervention.

[0072] 5. Automated execution process of the execution module

[0073] The execution module is equipped with corresponding execution units for mushroom shed electrical control, deep processing electrical control, cold chain scheduling, community store terminals, factory planting, and deep processing production lines. Each unit receives control strategies from the corresponding intelligent agents and executes them. The module has health and medical linkage logic and can receive risk handling instructions issued by health and medical intelligent agents to automatically complete operations such as removing problematic batches of goods from shelves, cold chain interception, production restriction at the base, and shutdown and sealing of deep processing production lines.

[0074] Mushroom shed electrical control execution unit: It interfaces with the electrical control cabinets of the ventilation fans, humidifiers, and shading systems in the mushroom shed, and automatically adjusts the operating level and start / stop duration of the equipment according to the control instructions of the production management intelligent body, so as to stabilize the temperature inside the shed within the target value ±0.5℃ range.

[0075] The deep processing electrical control execution unit connects to the main power distribution and temperature control cabinet of the deep processing workshop, and uniformly manages the power supply and constant temperature circuits of the entire production line for drying, freeze drying, extraction, and sterilization. It receives the workshop's overall control and scheduling instructions issued by the intelligent body module for edible fungi deep processing. When it detects equipment overload, excessive temperature and humidity, or health risk warning signals in the workshop, it can automatically cut off the corresponding area's electrical control circuit to achieve overall safety management of the deep processing workshop.

[0076] Cold chain scheduling execution unit: Receives the delivery plan from the logistics scheduling intelligent agent, automatically sends it to the cold chain vehicle scheduling terminal, and generates the optimal delivery route based on real-time road conditions.

[0077] Community store terminal execution unit: Receives batch interception instructions from the quality inspection intelligent agent and replenishment notifications from the sales feedback intelligent agent. Simultaneously displays fresh and dried mushroom removal notices and next-day arrival pop-ups on the store traceability interaction terminal. Also displays dietary health tips for edible fungi for different groups. The system locks and prohibits the sale of fresh products and deep-processed dried mushroom products that pose a high health risk.

[0078] The factory-scale planting end-execution unit connects to the PLC of the mushroom stick production line, receives instructions from the factory-scale planting intelligent agent to adjust the production line operating parameters, and triggers the production line to stop and seal high-risk batches.

[0079] The deep processing production line execution unit connects to the drying unit and the automatic packaging equipment PLC, and receives temperature, feeding, and packaging parameter instructions from the edible fungi deep processing intelligent module; if a high health risk is detected in the upstream fresh mushrooms, the raw material feeding channel for drying is automatically cut off, and the corresponding semi-finished and finished products are sealed.

[0080] 6. Closed-loop iteration process of feedback update module

[0081] The feedback update module automatically collects actual operational data at fixed daily intervals, including actual harvest yield, actual disease incidence, actual store sales, terminal loss rate, actual equipment energy consumption data, drying and processing losses, and sales feedback of dried mushroom products. It also simultaneously collects consumer feedback, health and food safety records, and data on health and wellness dietary use, feeding this data back to the digital twin module and the artificial intelligence analysis module for iterative updates. Simultaneously, it collects operational data from the production line and drying workshop, making minor updates to the factory-based auxiliary prediction model and the deep processing technology optimization sub-model. Based on actual output data, the digital twin module corrects the kinetic parameters of the microscopic growth model and the drying and processing simulation parameters. To reduce the discrepancy between simulation and reality, the historical risk records of the macro-twin health supervision layer are updated synchronously. The artificial intelligence analysis module conducts incremental training on the five types of industry sub-models and health assessment sub-model groups based on the newly added data of the day, continuously optimizing the accuracy of prediction, risk judgment, and deep processing technology optimization. The intelligent agent collaborative control module, the intelligent agent module for edible fungi factory planting, and the intelligent agent module for edible fungi deep processing fine-tunes the decision threshold based on the feedback results, forming a complete industrial closed loop of "data collection - virtual simulation - intelligent prediction - collaborative decision-making - entity execution - feedback update". Overlaying the two-way iterative link of agricultural and medical data, a dual closed loop integrating industrial planting, deep processing, and health supervision is constructed.

[0082] Example 2: A large-scale edible fungi industry group, covering a strain research and development center, 3 standardized mushroom stick factories, 120 intelligent mushroom sheds, 2 sorting and processing plants, a large-scale edible fungi deep processing center (drying, freeze-drying, polysaccharide extraction, and ready-to-eat medicinal fungi product production lines), bonded warehouses at the place of origin, and 120 community stores in China. At the same time, it conducts export business of fresh mushrooms, dried mushrooms, edible fungi extracts, and ready-to-eat fungi snacks to 15 countries in Southeast Asia and the European Union.

[0083] Based on an existing architecture, this embodiment fully utilizes all scheduling logic of the edible fungi deep processing intelligent agent module, configures cross-border transaction sub-intelligent agents, carbon sink accounting intelligent agents, and consortium blockchain evidence storage sub-modules, and upgrades the cross-border adaptability of the national health and medical big data intelligent agent module to achieve closed-loop management of the entire chain of domestic and foreign trade, planting, deep processing, and integrated agriculture and medicine. The group is equipped with multiple intelligent mushroom-growing workshops, fully utilizing all scheduling logic of the edible fungi factory planting intelligent agent module. The factory planting end execution unit manages all equipment in the workshop and production line, while the deep processing production line execution unit coordinates the entire set of production line equipment for freeze drying, extraction, and aseptic packaging.

[0084] The data acquisition module has been expanded to include a logistics information acquisition unit with added temperature-sensitive tags for ocean-going cold chain containers and a port customs declaration data interface. This allows for the automatic collection of cross-border end-to-end data, such as ocean transport container temperature and humidity, vessel route and location, customs declaration information, and phytosanitary certificate status. The deep processing data acquisition subunit has been expanded to collect data on freeze-drying vacuum degree, polysaccharide extraction kettle temperature and solvent ratio, sterilization curves of ready-to-eat products, nutritional testing of extracts, and carbon footprint data of deep-processed finished products. It also simultaneously and completely collects full-scale factory and deep-processing production data from workshops, production lines, and deep processing centers. The health data anonymization access unit is configured with a cross-border food health standard database interface, simultaneously importing EU and Southeast Asian standards for heavy metals, pesticide residues, and deep-processing additive limits in edible fungi, providing data support for cross-border compliance determination by the health risk assessment intelligent body.

[0085] The digital twin model expands to include a mesoscopic twin model with the addition of 3D simulation units for bonded warehouses, ocean-going containers, and a complete deep-processing center production line simulation unit. This allows for real-time simulation of bonded inventory status, container temperature and humidity changes, equipment load in the extraction workshop, and semi-finished product inventory and processing losses. The mesoscopic twin layer fully displays the workshop cluster and real-time simulation of the freeze-drying / extraction workshop. The macroscopic twin model adds a global trade data layer and a deep-processing capacity sandbox layer, integrating import tariff policies of various countries, real-time exchange rates, and historical demand data for fresh mushrooms / deep-processed products from overseas buyers. This simulates and extrapolates the efficiency and cost fluctuations of cross-border trade and deep-processing capacity turnover. The macroscopic twin model also simultaneously overlays a global food safety and health standards layer, comparing the health indicators of fresh and deep-processed extract products with the export country's access threshold.

[0086] The AI ​​analysis model expands its production and demand forecasting sub-model by adding overseas demand forecasting dimensions for fresh mushrooms and processed products. It integrates historical export data, overseas holidays, and tariff policy changes to predict overseas procurement demand and price trends over the next 30 days, while also assessing the risk of storage and transportation losses for fresh mushrooms and processed products on different shipping routes. The factory-based lightweight auxiliary model simultaneously outputs workshop capacity balancing data. The deep processing technology optimization sub-model adds a cross-border customized formula branch, automatically adjusting extraction and ready-to-eat product processing parameters based on food additives and nutritional standards in different countries. The health assessment sub-model group includes a cross-border compliance judgment branch, outputting the judgment results on whether fresh mushrooms and processed extracts meet the health access standards of the exporting country.

[0087] The intelligent agent collaborative control module expands beyond the four basic intelligent agents by adding a cross-border transaction sub-intelligent agent and a carbon sink accounting intelligent agent. Both extended intelligent agents are connected to a unified message bus and communicate with the basic intelligent agent, health intelligent agent, edible fungus factory cultivation intelligent agent module, and edible fungus deep processing intelligent agent module. The cross-border transaction sub-intelligent agent completes the matching of overseas fresh mushroom and deep-processed product inquiries, ocean cold chain scheduling, and export carbon label generation based on supply and demand forecast data. The carbon sink accounting intelligent agent calculates the carbon emission reduction of individual fresh mushrooms and deep-processed products throughout their entire life cycle using a digital twin model and generates carbon credit certificates that can be put on the blockchain. The carbon sink accounting intelligent agent, the cross-border transaction sub-intelligent agent, the dietary health recommendation intelligent agent, the edible fungus factory cultivation intelligent agent module, and the edible fungus deep processing intelligent agent module collaborate to generate cross-border export product files that combine carbon labels, healthy eating suggestions, factory workshop traceability, and deep processing full-process traceability files.

[0088] The execution module has been expanded to include a new cross-border customs declaration execution unit, which automatically connects to the International Trade Single Window system to complete the automatic declaration, status synchronization, and tax calculation of customs documents. The cross-border customs declaration execution unit can simultaneously export two independent product health and compliance test reports for fresh mushrooms and deep-processed extracts for overseas customs clearance verification. The factory-based planting end execution unit fully manages all workshop environmental control and production line speed control equipment. The deep-processing production line execution unit independently manages the freeze-drying unit, extraction reactor, and aseptic vacuum packaging of the entire deep-processing production line, supporting single-process shutdown and full-line isolation and sealing.

[0089] The consortium blockchain evidence storage submodule adopts a multi-node distributed deployment. It anonymizes and encrypts the entire industry chain, carbon footprint, consumer feedback, and batch data of all deep processing steps stored in the digital twin module and uploads them to the blockchain. Simultaneously, it completes the encryption and uploading of health and hygiene integration data, food safety disposal records, and cross-border batch health risk rating files to the blockchain. It also independently stores the factory batch production records and extraction and packaging records of each batch of deep processing. Each link generates a unique hash identifier, making the data tamper-proof and traceable, providing reliable traceability archive support for various intelligent entities across agriculture, health, market supervision, and customs departments.

[0090] Feedback from overseas customers on the quality of fresh mushrooms and processed products, data on damage during transit, customs clearance timeliness, and health complaints from overseas consumers regarding processed edible fungi products are uniformly transmitted back to the system through a feedback update module. This allows for iterative optimization of cross-border logistics simulation models, health assessment sub-models, processing technology optimization sub-models, and shiitake mushroom storage and transportation parameters. Simultaneously, it iterates on workshop factory simulation models and freeze-drying extraction processing simulation models. This provides reverse guidance for upstream strain selection, cultivation process adjustments, and processing formula optimization, enabling the transmission of overseas terminal demand for fresh and processed products and health feedback to the upstream planting and processing ends across the entire supply chain.

[0091] Example 3: The county-level agricultural and rural affairs bureau's industry supervision platform adopts a cloud-edge hierarchical deployment architecture to achieve unified monitoring, early warning and supervision of 12 edible fungi production bases and 3 deep processing workshops within the county. It does not require direct control of on-site equipment and focuses on the overall industry, food safety and health, and deep processing capacity situation perception and risk management.

[0092] 1. Cloud Deployment Content

[0093] The cloud-based unified deployment of a digital twin macro model, a global AI analysis model, a coordinated regulatory intelligent body, a health and medical big data fusion gateway, a global health assessment sub-model, a global factory-based data aggregation and analysis unit, and a global deep-processing capacity coordination unit aggregates production, quality, circulation, deep-processing technology, and health risk data uploaded from all bases and deep-processing workshops in the county. This constructs a county-wide edible mushroom industry twin sand table, enabling unified capacity statistics, county-wide disease risk early warning, quality and safety traceability, county-wide carbon aggregate, and county-wide health risk accounting for fresh mushrooms and deep-processed products.

[0094] 2. Edge deployment content

[0095] Each production base and deep processing workshop deploys an edge computing gateway equipped with a lightweight micro-twin model, a lightweight disease identification model, a lightweight on-site intelligent agent, and a lightweight health risk identification model. It supports offline data collection and caching, and synchronizes to the cloud after connecting to the network. The edge gateway of the factory base is equipped with a lightweight factory intelligent agent to complete the basic control of the workshop and production line locally. The edge gateway of the deep processing workshop is equipped with a lightweight deep processing process intelligent agent and a lightweight deep processing nutritional risk identification model to complete the basic process control of drying and packaging locally. The cloud uniformly deploys a health and medical big data fusion gateway and a global health judgment sub-model. The edge terminal is equipped with a lightweight health risk identification model, which can quickly complete the initial assessment of the basic health risks of fresh mushrooms and deep-processed single products on-site, and synchronize to the cloud after connecting to the network to complete the comprehensive assessment of the entire domain.

[0096] 3. Closed-loop process for regulatory collaboration

[0097] The integrated regulatory intelligent agent automatically generates daily industry operation reports and a comprehensive risk warning list, which are then pushed to the regulatory department's terminal. The integrated regulatory intelligent agent, in conjunction with the health intelligent agent, the cloud-based factory integrated analysis unit, and the cloud-based deep processing integrated unit, identifies risks of diseases, food safety and health issues, and excessive additives in deep processing across the entire region. It automatically issues disposal warning instructions to the edge intelligent agents at each base and deep processing workshop. After the base and deep processing workshop complete rectification and disposal, all production records, deep processing process ledgers, and health rectification ledgers are simultaneously uploaded back to the cloud for archiving. Regulatory personnel can remotely conduct sampling inspections of fresh mushrooms and deep-processed products across the entire region via a large twin screen, eliminating the need for on-site verification in rural areas, thus forming a long-term, joint regulatory closed loop involving agriculture, health, and market supervision.

[0098] Example 4: This example targets medium-sized edible fungi deep processing enterprises without their own planting bases. They purchase fresh mushroom raw materials and carry out drying, freeze-drying, polysaccharide extraction, and ready-to-eat mushroom packaging processing. They also provide raw material cold storage, finished product storage, and distribution channels to health and wellness stores. The core of the system verifies the independent scheduling, agricultural-medical linkage, and closed-loop iteration capabilities of the intelligent body module for edible fungi deep processing.

[0099] I. System Deployment Architecture

[0100] The cloud-based deployment includes a global health assessment model, a national health and medical big data intelligent body, and a collaborative control message bus; the enterprise edge gateway deploys lightweight data acquisition, a simplified digital twin micro + meso model, and a lightweight AI analysis module; the edible fungi deep processing intelligent body module is fully deployed, the edible fungi factory planting intelligent body only reserves nodes, the execution module only enables three types of execution units: deep processing production line, cold chain, and store; and the local deployment of alliance chain nodes retains processing batch files.

[0101] II. Data Acquisition and Digital Twin Support

[0102] The data acquisition module relies on the deep processing data acquisition subunit to collect data on the quality of purchased raw materials, equipment parameters for the entire process of drying / extraction / sterilization, and nutritional testing data of finished products. It also integrates data on desensitized health and wellness diets and foodborne adverse reaction data. All data is bound to batch codes and synchronized to the digital twin. The digital twin constructs a 1:1 mesoscopic simulation layer of the deep processing workshop, displaying the real-time operating conditions of each production line's equipment, semi-finished product inventory, and processing losses. It supports process simulation and provides a basis for parameter optimization simulation for the deep processing intelligent agent.

[0103] III. AI Analysis Supporting Models

[0104] The AI ​​module features a lightweight raw material quality identification model, fully utilizing the deep processing technology optimization sub-model and the lightweight nutritional risk identification model. It outputs three types of judgment results: raw material compatibility determination, optimal processing technology, and health grading of deep-processed products. All results are pushed to the edible fungi deep processing intelligent agent module.

[0105] IV. Core Scheduling Logic of the Intelligent Agent Module for Deep Processing of Edible Fungi

[0106] The module relies on a unified message bus to achieve autonomous decision-making, and has three built-in sub-units: production scheduling, process adaptive adjustment, and health and wellness formula adaptation.

[0107] The production scheduling subunit automatically generates processing plans by combining raw material inventory and store orders. When equipment fails, it automatically adjusts the production line load and pushes back the procurement demand when raw materials are insufficient.

[0108] The process control subunit dynamically adjusts the drying temperature and extraction solvent ratio based on the moisture content of the raw materials and simulation data. It automatically extends the sterilization time when there is a slight risk to the raw materials.

[0109] The health and wellness formula adaptation subunit receives health and wellness requirements, reduces the salt content of ready-to-eat products, customizes the packaging specifications of high-purity extracts, and automatically generates healthy consumption labels. Instructions are prioritized as follows: health risk management instructions > deep processing technology control instructions > raw material scheduling instructions. When raw materials exceed heavy metal limits, the intelligent agent directly issues instructions for raw material isolation and a complete production line shutdown.

[0110] V. Automated Management and Control of Execution Units

[0111] The deep processing production line execution unit directly receives instructions from the intelligent agent and automatically adjusts the parameters of drying, freeze-drying, extraction, and packaging equipment; the feeding gate identifies high-risk raw materials and automatically locks the channel; the health and wellness store terminal displays dietary matching prompts for deep-processed products in a pop-up window, and high-risk products are forcibly locked and removed from the shelves.

[0112] VI. Feedback Closed-Loop Iteration and Consortium Blockchain Notification

[0113] The feedback update module collects daily data on processing losses, finished product qualification rates, health and wellness consumption feedback, and health and sanitation handling records. This serves two purposes: firstly, it corrects the digital twin process simulation parameters; secondly, it incrementally trains the process optimization model and fine-tunes the decision thresholds of the deep-processing intelligent agent, continuously reducing processing losses. The consortium blockchain encrypts and uploads the processing technology, raw material traceability, and health rating records for each batch, supporting rapid cross-departmental traceability and verification by agriculture, health, and market supervision departments.

[0114] The closed-loop management process is as follows:

[0115] S1. Collect comprehensive data on purchased raw materials, deep processing equipment, and health and wellness needs;

[0116] S2. The twin model completes the simulation and deduction of the processing technology;

[0117] S3 and AI models output process optimization and health assessment results;

[0118] S4. The intelligent agent for the deep processing of edible fungi generates production scheduling, process, and customized health and wellness control instructions through collaborative computation.

[0119] S5. The deep processing production line execution unit automatically completes equipment control and risk isolation;

[0120] S6, processing, consumption, and health data reverse feedback iterative model and intelligent agent;

[0121] S7. The cyclical operation forms an independent small closed loop of "purchased raw materials - deep processing - health and wellness terminals - health feedback".

[0122] This embodiment relies on a sophisticated processing intelligent agent to achieve autonomous scheduling of multiple processes, reducing the workload of manual process control by 70%; it links health and wellness data to achieve customized processing of health and wellness products, intercepts food safety risks of raw materials and finished products in advance, and continuously optimizes the processing technology through closed-loop feedback, reducing overall processing losses by 8%-12%.

[0123] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A multi-agent closed-loop management and control system and method for the entire edible fungi industry chain, characterized in that, include: The system comprises a data acquisition module, a digital twin module, an artificial intelligence analysis module, an intelligent agent module for industrialized edible fungi cultivation, an intelligent agent module for intensive processing of edible fungi, an intelligent agent collaborative control module, an execution module, a feedback update module, and a national health and medical big data intelligent agent module. The data acquisition module collects multi-source raw data from all stages of edible fungi cultivation and transmits it to the digital twin module. It includes a health data anonymization access unit and an intensive processing data acquisition subunit. It relies on federated privacy computing to connect to anonymized national health and medical compliant big data, achieving bidirectional integration of industry and medical data. The digital twin module constructs a three-level virtual mapping model: micro, meso, and macro. The meso twin model adds a simulation layer of the intensive processing production line, and the macro twin model is configured with a health supervision visualization layer. The artificial intelligence analysis module has five built-in specialized sub-models: growth prediction sub-model, disease identification sub-model, quality analysis model, production and sales demand prediction model, and deep processing technology optimization sub-model. It is equipped with a health assessment sub-model group to output health dimension assessment information. The intelligent module for industrialized edible fungi cultivation is connected to a unified message bus, which generates exclusive control commands for closed mycelium cultivation, intelligent mushroom production workshops, and automated mushroom stick production lines and sends them to the execution module. The intelligent module for the deep processing of edible fungi is connected to a unified message bus to coordinate the scheduling of the entire deep processing process, including drying, freeze drying, extraction, and packaging. Based on the supply of raw materials and health and wellness needs, it generates exclusive control instructions for deep processing and sends them to the execution module. The control instructions have a lower priority than health risk instructions but are at the same level as the control instructions for factory cultivation. The intelligent agent collaborative control module is equipped with a unified message bus, which connects to the national health and medical big data intelligent agent module, the edible fungi factory cultivation intelligent agent module, and the edible fungi deep processing intelligent agent module. Health risk-related scheduling instructions have higher priority than regular production, sales, cultivation, and deep processing control instructions. The national health and medical big data intelligent agent module embeds a health data fusion subsystem, as well as four types of health-specific autonomous intelligent agents: health risk assessment intelligent agent, dietary health recommendation intelligent agent, food safety and medical traceability intelligent agent, and health supervision and reporting intelligent agent. The execution module has health and hygiene linkage execution logic, and can receive health intelligent agent instructions to complete the interception of risky goods and the locking of stores; the execution module adds a factory planting end execution unit and a deep processing production line execution unit, which are used to execute the equipment control instructions issued by the edible fungi factory planting intelligent agent module and the edible fungi deep processing intelligent agent module, respectively; the feedback update module synchronously collects consumer health feedback, health and hygiene disposal ledgers, and deep processing production loss data, and iteratively trains the health judgment sub-model and the deep processing technology optimization sub-model; Each module transmits data in a forward unidirectional manner and sends data back in the reverse direction through the feedback update module, forming a dual closed loop that integrates industry control and human health supervision.

2. The multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 1, characterized in that, The data acquisition module includes an environmental parameter acquisition unit, a production process acquisition unit, a multispectral image acquisition unit, a logistics information acquisition unit, a health data anonymization access unit, and a deep processing data acquisition subunit. The health data anonymization access unit deploys a federated privacy computing gateway, which isolates and stores original medical records and sensitive personal medical information locally, and only outputs anonymized and tagged data to the outside world. The deep processing data acquisition subunit collects equipment parameters, raw material ratios, finished product nutritional testing, and processing loss data for the entire process of drying, freeze-drying, extraction, sterilization, and packaging.

3. The multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 1, characterized in that, The health supervision visualization layer of the digital twin module is used to display the health risk level of each batch of edible fungi, food safety traceability trajectory, and regional population dietary suitability distribution data.

4. The multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 1, characterized in that, The health assessment sub-model group integrates industry data and desensitized medical data for parallel processing, and outputs three types of assessment information: health risk classification, dietary matching recommendation, and food safety pollution source location.

5. The multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 1, characterized in that, The intelligent agent collaborative control module is configured with four types of basic autonomous intelligent agents: production management, quality inspection, logistics scheduling, and sales feedback. The unified message bus enables data sharing and event linkage triggering between the four types of basic autonomous intelligent agents and the four types of health-specific autonomous intelligent agents.

6. The multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 5, characterized in that, The intelligent agent collaborative control module also integrates a cross-border transaction sub-intelligent agent and a carbon sink accounting intelligent agent; the carbon sink accounting intelligent agent and the cross-border transaction sub-intelligent agent can work together with the dietary and health recommendation intelligent agent to generate cross-border export product files that include carbon labels and healthy eating suggestions.

7. The multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 1, characterized in that, The system has a consortium blockchain evidence storage sub-module; the consortium blockchain evidence storage sub-module synchronously de-identifies and encrypts data, carbon footprints, and health records integrated with health and wellness into the blockchain, supporting cross-departmental joint traceability and verification in agriculture and health.

8. The multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 1, characterized in that, The execution module includes a mushroom shed electrical control execution unit, a deep processing electrical control execution unit, a cold chain scheduling execution unit, a community store terminal execution unit, a cross-border customs declaration execution unit, and a deep processing production line execution unit. The community store terminal execution unit is equipped with a dietary health pop-up window and locks the system for high-risk products. The cross-border customs declaration execution unit can export product health compliance test reports for customs clearance.

9. A multi-agent closed-loop management and control system for the entire edible fungi industry chain according to claim 1, characterized in that, The system adopts a cloud-edge hierarchical deployment architecture; the cloud deploys a digital twin macro model, a global AI model, a health and medical big data fusion gateway, and a global health assessment sub-model; the edge is equipped with a lightweight disease identification model and a lightweight health risk identification model, which supports offline data caching and synchronization to the cloud simulation layer after connecting to the network.

10. A multi-agent closed-loop management and control method for the entire edible fungi industry chain, applied to the system described in any one of claims 1 to 9, characterized in that, Includes the following steps: S1. Call the data acquisition module to collect real-time data of the entire edible fungi industry chain, and simultaneously access the de-identified national health and medical big data through the privacy computing gateway to complete the cleaning, spatiotemporal alignment and fusion modeling of the two types of data. S2. The digital twin module builds a three-level virtual mapping model and loads a health monitoring visualization layer to generate an agricultural and medical integration simulation dataset. S3, the artificial intelligence analysis module runs in parallel four types of special sub-models: growth prediction sub-model, disease identification sub-model, quality analysis model, and production and sales demand prediction sub-model, as well as a health assessment sub-model group, and outputs two types of prediction and assessment information: industry and health. S4. The intelligent agent collaborative control module schedules four types of basic autonomous intelligent agents and four types of health-specific autonomous intelligent agents—health risk assessment intelligent agent, dietary health recommendation intelligent agent, food safety and medical traceability intelligent agent, and health supervision and reporting intelligent agent—through a unified message bus, and prioritizes issuing scheduling instructions for health risk disposal. S5, the execution module completes the automated control of edible fungi production, deep processing, warehousing, and cross-border equipment, and simultaneously executes health-related operations such as interception of risky goods, ban on sales in stores, and reporting of health materials; S6. The feedback and update module transmits back industry operation data, consumer health feedback, and health and sanitation handling records, and iteratively updates the virtual mapping model, four types of special sub-models, and health assessment sub-model. S7. Repeat steps S1 to S6 to continuously iterate and optimize the model and control strategies, forming a long-term integrated closed loop for healthy industrial production and consumption.

Images