Multi-source planting and breeding waste resource comprehensive management and control intelligent agent system and operation method

Abstract

The invention discloses a multi-source planting and breeding waste resource comprehensive management and control agent system and an operation method. The system comprises a production system physical entity, a data acquisition agent, a data interaction agent, a twinborn model and an analysis decision agent. The data acquisition agent acquires real process data in the process; the data interaction agent is used for data transmission and processing among the agents; the analysis decision agent performs analysis and optimization through a machine learning algorithm; and the twin model performs simulation based on each optimization scheme in the optimization scheme set, so that the analysis decision agent determines a target production scheme based on each simulation result. According to the method, the big data of the technological process is quickly analyzed through a machine learning algorithm, and meanwhile, the real-time monitoring control of the data in the resource utilization technological process of the multi-source planting and breeding waste is realized by utilizing distributed collaborative optimization among the intelligent agents, so that the resource utilization efficiency of the multi-source planting and breeding waste is effectively improved, and the intelligent management and control level of a production system is improved.

Description

Technical Field

[0001] The present invention relates to the field of waste resource utilization, and more specifically, to an intelligent system for comprehensive management and control of multi-source breeding waste resources and an operation method thereof. Background Art

[0002] Currently, large-scale and intensive farming and planting generate a large amount of waste, which not only causes serious ecological pollution but also constitutes a major constraint on the further development of modern agriculture and animal husbandry. Farming and animal husbandry waste is rich in cellulose, organic matter, crude fat, and various nutrients such as nitrogen, phosphorus, potassium, calcium, and sulfur. If properly utilized, turning waste into treasure, it represents a potentially valuable agricultural resource.

[0003] In recent years, the resource utilization of agricultural and livestock waste has primarily involved anaerobic fermentation of various raw materials (such as straw, fruit peels, rotten vegetables, and various livestock and poultry manure) to produce biogas, which is then purified and refined into biogas. The remaining biogas liquid and residue are then processed into organic fertilizer and returned to the fields. While this production model achieves a certain degree of resource recovery and utilization, it still faces challenges such as a lack of effective control over equipment operating data and difficulty in adaptively adjusting to market fluctuations. Summary of the Invention

[0004] In view of this, the present invention provides an intelligent system and operation method for comprehensive management of multi-source agricultural and breeding waste resources, which is used to solve the problems of existing agricultural and breeding waste resource utilization methods, the lack of effective management and control of equipment operation data, and difficulty in adaptive adjustment.

[0005] To achieve the above objectives, the following solutions are proposed:

[0006] An intelligent system for comprehensive management and control of multi-source agricultural and livestock waste resources, including: a production system physical entity, a data collection intelligent agent, a data interaction intelligent agent, a twin model, and an analysis and decision-making intelligent agent;

[0007] The data acquisition intelligent agent collects real process data of the physical entity of the production system during the process;

[0008] The data interaction agent is used to control the transmission and processing of data between the agents in the intelligent system, pre-process the real process data, and obtain the original process data;

[0009] The analysis and decision-making intelligent agent analyzes and optimizes the original process data and historical experience data through a machine learning algorithm to obtain an optimization solution set;

[0010] The twin model performs simulation based on each optimization scheme in the optimization scheme set, so that the analysis and decision-making intelligent agent can determine the target production scheme based on each simulation result.

[0011] Preferably, the physical entity of the production system includes: production raw materials, machinery and equipment, and auxiliary tools in the process of resource utilization of multi-source breeding waste.

[0012] Preferably, the data acquisition agent includes: temperature sensors, pressure sensors, flow sensors, vibration sensors, speed sensors, state sensors, as well as video cameras, scanners, RFID, and SCADA;

[0013] The actual process data includes raw material ratio, pumping pressure, fermentation temperature, feed flow rate, biogas concentration, tank liquid level, stirring frequency, motor speed, equipment vibration, instrument noise, production video and production image.

[0014] Preferably, the data interaction agent includes: a communication module, a data access module, a data preprocessing module and a multimodal interaction module;

[0015] The communication module provides an information interface for data transmission and interaction between agents based on the communication protocol;

[0016] The data access module aggregates and accesses data from different agents through gRPC or Kafka message queues;

[0017] The data preprocessing module is used to perform preprocessing operations such as cleaning, format conversion and normalization on the collected data;

[0018] The multimodal interaction module is used to realize the interaction of structured data and unstructured data between various intelligent agents.

[0019] Preferably, the twin model includes: a geometric model, a physical model, a rule model, a behavior model, a simulation model and a business model;

[0020] The geometric model is the three-dimensional shape, spatial structure and layout distribution of the physical entities of each production system;

[0021] The physical model includes a kinetic model or a thermodynamic model of each link of the process;

[0022] The rule model includes a constraint model, a standard model and a security operation process model;

[0023] The behavior model includes a state machine model, a raw material matching operation process model and a timing coordination model;

[0024] The simulation model includes a multi-physics field coupling simulation model, a dynamic simulation model and a beat optimization simulation model;

[0025] The business model includes a raw material collection, storage and transportation planning model, a quality traceability management model, a production system energy efficiency management model and a production capacity assessment and prediction model.

[0026] Preferably, the analysis and decision-making agent includes: a collaboration mechanism module, a control decision module, a system optimization module, a task allocation and adjustment module, and a monitoring feedback module;

[0027] The collaboration mechanism module includes a distributed or hybrid collaboration mode and task collaboration process between each agent and a conflict resolution mechanism for each agent;

[0028] The control decision module includes mechanism model driven and data driven control algorithms, and dynamic decision mechanisms triggered by time or events;

[0029] The system optimization module includes optimization objectives and optimization algorithms;

[0030] The task allocation and adjustment module includes task decomposition basis, task allocation strategy and task adjustment mechanism;

[0031] The monitoring feedback module monitors the process flow in real time based on the real process data collected by the data acquisition intelligent agent in real time, and implements threshold alarms and abnormality handling based on the feedback early warning mechanism.

[0032] Preferably, the agent system further comprises: an application service agent;

[0033] The application service agent is verified through data-driven service and virtual-reality fusion.

[0034] Preferably, the services provided by the applied service agent include: dynamic production scheduling optimization service, energy efficiency management service, quality traceability service, predictive service and visualization service;

[0035] The dynamic production scheduling optimization service dynamically adjusts the production plan and scheduling scheme based on constraints and determines the time and sequence of production tasks;

[0036] The energy efficiency management service evaluates unit product energy consumption and comprehensive energy efficiency through energy flow and load curve analysis, thereby optimizing energy allocation;

[0037] The quality traceability service records and tracks data from the entire production process;

[0038] The predictive services are used to monitor the health status of production equipment and predict failures, predict market demand for products, and predict raw material supply;

[0039] The visualization service uses graphical means to intuitively display production process data, processes and status.

[0040] Preferably, the application service agent further comprises: a raw material collection and transportation planning algorithm;

[0041] The raw material collection and transportation planning algorithm uses indicators such as raw material category, raw material production and inventory, raw material distribution location, raw material collection frequency, raw material transportation method, and raw material processing cost as input parameters of the optimization algorithm, and takes the lowest total transportation cost, shortest transportation time, balanced transportation task distribution, and optimal transportation route as output parameters. It uses a machine learning algorithm to perform multi-objective optimization iteration to improve reasoning efficiency and quickly obtain the Pareto optimal solution set.

[0042] An operating method for an intelligent system for comprehensive management and control of multi-source agricultural and breeding waste resources, applied to the aforementioned intelligent system for comprehensive management and control of multi-source agricultural and breeding waste resources, includes:

[0043] S1 Raw material preprocessing: The data acquisition agent collects various raw material parameters in real time and transmits them to the analysis and decision-making agent through the data interaction agent. The analysis and decision-making agent preliminarily screens the raw material types and ratios based on daily production requirements and existing raw material inventory. Then, based on historical production data, it recommends raw material formulas. The twin model simulates the impact of different raw material formula parameters on gas production efficiency to determine the optimal raw material ratio scheme.

[0044] S2 anaerobic fermentation to produce biogas: The data acquisition agent monitors the fermentation process parameters of the fermentation tank in real time and transmits them to the analysis and decision-making agent through the data interaction agent. The analysis and decision-making agent performs analysis based on historical experience and fermentation process parameters. The twin model performs multi-physics simulation and optimization of fluid mechanics and biochemical reaction dynamics to determine the optimal fermentation process parameters.

[0045] S3 Biogas Purification: The data collection agent collects the performance parameters of purified biogas in real time; the analysis and decision-making agent uses machine learning algorithms to perform multi-objective collaborative optimization and iteratively search for the Pareto optimal solution;

[0046] S4: Drip irrigation fertilizer preparation using biogas slurry: The data collection agent collects biogas slurry component indicators in real time; the analysis and decision-making agent uses machine learning algorithms to optimize the irrigation and fertilizer preparation parameters based on component indicators, nutrient demand models, and soil testing data.

[0047] S5 Biogas Residue Preparation for Organic Fertilizer: The data acquisition agent monitors the environmental parameters of the fermentation pile in real time, and the analysis and decision-making agent adjusts the environmental parameters according to the preset fermentation process curve and target parameter range;

[0048] S6 Product Sales and Market Forecast: Analytical decision-making agents formulate sales strategies and pricing plans through machine learning algorithms.

[0049] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0050] The present invention provides an intelligent system for comprehensive management and control of multi-source breeding waste resources, including: a production system physical entity, a data acquisition intelligent entity, a data interaction intelligent entity, a twin model and an analysis and decision-making intelligent entity. The data acquisition intelligent entity collects the real process data of the physical entity of the production system during the process; the data interaction intelligent entity is used to control the transmission and processing of data between the intelligent entities in the intelligent system, and pre-process the real process data; the analysis and decision-making intelligent entity analyzes and optimizes the original process data and historical experience data through a machine learning algorithm to obtain a set of optimization solutions; the twin model simulates each optimization solution in the set of optimization solutions, so that the analysis and decision-making intelligent entity can determine the target production solution based on each simulation result. The present invention uses a machine learning algorithm to quickly analyze the big data of the process, and at the same time uses distributed collaborative optimization between intelligent entities to achieve real-time monitoring and control of data in the process of multi-source breeding waste resource utilization, effectively improving the efficiency of multi-source breeding waste resource utilization and enhancing the level of intelligent management and control of the production system.

[0051] The present invention uses machine learning algorithms and other means to perform multi-objective optimization of the process, accurately adjust the conflicts and coupling effects between multiple objectives, and quickly improve the accuracy and flexibility of optimization decisions, thereby meeting the needs of different production resources, costs, benefits and environments.

[0052] The present invention uses digital twin technology to construct multi-scale and multi-level twin models of each link. Through virtual-reality mapping, dynamic simulation and data interaction, it can simulate, analyze, verify and predict the dynamic behavior of each production link in the resource utilization process of multi-source breeding waste in a virtual environment, thereby realizing data integration and full life cycle management of the entire production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0054] Figure 1 A schematic diagram of an intelligent system for comprehensive management and control of multi-source agricultural and livestock waste resources provided by an embodiment of the present invention;

[0055] Figure 2 A schematic diagram of an intelligent system architecture provided by an embodiment of the present invention;

[0056] Figure 3 A schematic diagram of a deep learning network structure for multi-objective optimization of raw material collection and transportation provided by an embodiment of the present invention;

[0057] Figure 4 This is a schematic diagram of the spatial distribution of raw materials in an embodiment of the present invention;

[0058] Figure 5 A radar chart showing the normalized values ​​of six parameters provided in an embodiment of the present invention;

[0059] Figure 6 An improved NSGA-III iterative optimization flow chart provided by an embodiment of the present invention;

[0060] Figure 7 The iterative convergence curve of the process objective function provided by the embodiment of the present invention;

[0061] Figure 8 A schematic diagram of the production process and main control nodes provided by an embodiment of the present invention;

[0062] Figure 9 This is an analysis diagram of the prediction results of the production capacity prediction model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0064] First, combine Figure 1-2 The integrated management and control intelligent agent system for resource utilization of multi-source breeding waste provided by an embodiment of the present invention is introduced. The integrated management and control intelligent agent system includes: a production system physical entity, a data acquisition intelligent agent, a data interaction intelligent agent, a twin model and an analysis and decision-making intelligent agent. Therefore, the integrated management and control intelligent agent system can be described as ZYMA = {DAA, DIA, ADA, ASA, DATA}, wherein DAA represents the data acquisition intelligent agent connected to the physical entity of the production system; DIA represents the data interaction intelligent agent corresponding to multi-source heterogeneous data, which is used to promote data interaction and information transmission between each intelligent agent; ADA represents the analysis and decision-making intelligent agent corresponding to the twin model, which provides a dynamic management and control solution for the production process; ASA represents the application service intelligent agent corresponding to application verification, which realizes the implementation of management decisions and feedback of results; DATA represents the twin data involved in the twin model (DTMPS) modeling process, and the architectural relationship between each intelligent agent is as follows: Figure 1 shown.

[0065] (1) Production system physical entity

[0066] The physical entity of a production system is a multi-dimensional, multi-layered, integrated hardware entity comprised of six components: people, machines, materials, methods, environment, and measurement. It serves as the physical foundation for intelligent management and control of the production process. It encompasses the raw materials, machinery, equipment, and auxiliary tools involved in the resource utilization of multi-source farming and animal husbandry waste.

[0067] "People" refers to workshop personnel, including equipment operators, on-site management personnel, technical R&D personnel, etc., and it is necessary to monitor the attendance, work status, task progress and other information of workshop personnel. "Machinery" refers to machinery and equipment, including production equipment, testing equipment and auxiliary tools. Among them, production equipment includes cutting machines, fermentation tanks, mixers, desulfurization towers, grit chambers, homogenization tanks, biogas purification and purification equipment, biogas slurry and sludge solid-liquid separators, drip irrigation fertilizer preparation machines and organic fertilizer preparation machines; testing equipment includes flow meters, purity meters, etc.; auxiliary tools include various tools, auxiliary tools and conveying equipment such as loading racks, installation cards and adjustment disks to assist in production. It is necessary to monitor the operating status of machinery and equipment, equipment calibration status, failure rate and maintenance records and other information. "Materials" refers to production materials, including raw materials, semi-finished products and auxiliary materials required for production. Raw materials include crop waste such as straw, as well as livestock manure such as chicken, pig, or cow manure. Semi-finished products include biogas, biogas liquid, and biogas residue. Auxiliary materials, such as process water and added trace elements, require monitoring of material type, quality, and storage management. "Methods" refers to the process methods and procedures used in the production process, such as biogas production processes, biogas purification processes, and organic fertilizer production methods. The standards and regulations followed, such as corporate standards for biogas production, industry regulations for biogas purification, and organic fertilizer processing regulations, require monitoring of the process methods, process parameters, production planning and management methods, quality control measures, and standards and specifications. "Environment" refers to the environmental conditions at the production site, encompassing both the physical and human environments. The physical environment includes workshop equipment layout, logistics channels, workshop temperature and humidity, noise and dust levels, while the human environment includes employee status and emergency response mechanisms. "Measurement" refers to testing and measurement at every stage of production. This includes monitoring production process indicators such as biogas production, fermentation tank temperature, and gas tank flow, as well as final product quality testing such as biogas purity, organic fertilizer quality grade, and nutrient concentration of drip irrigation fertilizer. This requires traceability monitoring of all production processes and closed-loop management and detection of product quality issues to improve production efficiency, ensure product quality, and ensure production safety. The physical entities of the production system provide raw data to the data acquisition agent and receive execution instructions from the analysis and decision-making agent to complete various production tasks.

[0068] Depending on the raw material mix, production capacity plan, and product type, the corresponding sensing, detection, and analysis devices will vary. The data acquisition agent (DAA) collects information about the physical entities of the production system in real time, including equipment status data, production process data, and statistical forecast data. This data is then mapped to a digital twin model of the production system (DTMPS) via network transmission for modeling, analysis, and processing.

[0069] (2) Data Collection Agent

[0070] The data acquisition agent collects actual process data from the physical entities of the production system during the process, enabling dynamic perception and data collection of all elements and processes of the physical entities. The data acquisition agent (DAA) uses various sensors, including temperature sensors, pressure sensors, flow sensors, vibration sensors, speed sensors, and state sensors, as well as video cameras, scanners, RFID, and SCADA, to dynamically perceive multi-source heterogeneous information data during the operation of the production system in real time. This data primarily includes raw material ratios, pumping pressure, fermentation temperature, feed flow, biogas concentration, tank liquid level, stirring frequency, motor speed, equipment vibration, instrument noise, and monitoring videos and images.

[0071] Furthermore, the data acquisition agent can also include GPS sensors on transport vehicles, PROFIBUS buses, and industrial Ethernet. The data acquisition agent can communicate and interact with the data interaction agent through communication protocols, standardized interfaces, relevant data interfaces, and simulation tool interfaces. Relevant data interfaces include JDBC interfaces for interacting with MySQL databases, and simulation tool interfaces include Simulink and ANSYS interfaces.

[0072] The embodiment of the present invention uses a sensor network to dynamically perceive multi-source real-time information of the process, improves the accuracy and completeness of the collected data, and improves the real-time and stability of process data transmission through wireless networks, industrial Ethernet, cloud platforms, etc.

[0073] (3) Data Interaction Agent

[0074] The data interaction agent is used to manage the transmission and processing of data between agents in the intelligent system, preprocessing real process data to obtain raw process data. The data interaction agent includes a communication module, a data access module, a data preprocessing module, and a multimodal interaction module.

[0075] The communication module provides an information interface for data transmission and interaction with external users and other intelligent entities, supporting a variety of communication protocols and methods. For example, it can use CAN bus, PROFIBUS bus interface, RS-485 bus interface, industrial Ethernet interface, WiFi, NFC, Zigbee, 5G, etc. Through network and interface standardization, various types of data and information can be interconnected, facilitating information exchange and processing within the twin model of the production system (DTMPS).

[0076] The data access module aggregates and accesses data from different intelligent entities. It can connect to MES and SCADA through gRPC or Kafka message queues, aggregate and access data from different data sources such as databases, IoT devices, and application data, providing a foundation for subsequent data processing.

[0077] The data preprocessing module can clean, convert, and normalize the collected raw data to filter and reduce noise, fill in missing values, and remove outliers. Furthermore, automatic annotation tools can be used to further improve data quality.

[0078] The multimodal interaction module is used to enable interaction between structured and unstructured data between agents. The multimodal interaction module can support multiple forms of structured and unstructured data interaction, such as text, images, and audio, to achieve a richer and more comprehensive interactive experience.

[0079] Furthermore, to ensure privacy and security in data processing, the data interaction agent can also include a data security module. This module employs methods such as data encryption and access control to protect data. For example, TLS / SSL data encryption and X.509 identity authentication can be used. Differential privacy techniques can be used to anonymize data and ensure compliance with relevant regulations.

[0080] (4) Analytical Decision-Making Agent

[0081] The analytical decision-making agent can not only predict the effects of different decisions through simulation based on the twin model, but also drive the execution of various functions of the physical entities of the production system.

[0082] The functions that analytical decision-making agents can achieve include: 1) using statistical analysis and data mining to extract valuable information and knowledge from massive amounts of real-time data to support the agent's decision-making. 2) utilizing machine learning algorithms and deep learning models to model, analyze, and predict large-scale data and its nonlinear relationships. 3) identifying anomalies in the data to prevent them from interfering with the correct decision-making process. 4) optimizing the dynamic allocation and utilization efficiency of various production resources by developing reasonable production operation plans and procedures, ensuring the efficient execution of target tasks. Furthermore, the execution results can be fed back to facilitate timely adjustment and optimization of subsequent decisions, forming a closed-loop control process.

[0083] The analytical and decision-making agent uses machine learning algorithms to analyze and optimize raw process data and historical experience data to obtain a set of optimized solutions. The analytical and decision-making agent includes a collaborative mechanism module, a control decision module, a system optimization module, a task allocation and adjustment module, and a monitoring and feedback module.

[0084] The collaboration mechanism module includes distributed or hybrid collaboration modes and task collaboration processes among various intelligent agents, as well as conflict resolution mechanisms for resource occupation and sharing among various intelligent agents based on task priorities and multi-objective trade-offs.

[0085] The control decision module includes both mechanism-driven and data-driven control algorithms, as well as dynamic decision-making mechanisms triggered by time or events. For example, control algorithms might include improved genetic algorithms for collection and transportation route planning, or deep learning models for predicting biogas production capacity demand. Dynamic decision-making mechanisms might include fermentation mode changes triggered by sudden temperature changes, or equipment fault inspections executed at preset intervals.

[0086] The system optimization module includes optimization objectives and optimization algorithms. For example, optimization objectives can be determined by cost, efficiency, or quality, such as minimizing energy consumption, minimizing completion time, and maximizing customer satisfaction. Optimization algorithms can include linear programming for production resource allocation and nonlinear programming for dynamic adjustment of production processes.

[0087] The task allocation and adjustment module includes task decomposition criteria, task allocation strategies, and task adjustment mechanisms. For example, task decomposition criteria can be based on attributes such as task priority and timeliness; task allocation strategies can be static allocation based on pre-specified settings or dynamic allocation based on real-time status; and task adjustment mechanisms can include load balancing, production failure handling, and more.

[0088] The monitoring and feedback module monitors the process flow in real time based on real-world process data collected by the data acquisition agent. It also implements threshold alarms and exception handling based on a feedback and early warning mechanism. By monitoring the status of each agent, environmental parameters, and system operating indicators in real time, such as the agent's operating status, task progress, workshop temperature and humidity, production system energy consumption, and equipment failure rate, the monitoring and feedback module promptly identifies potential problems and abnormal conditions. Through the feedback and early warning mechanism, it implements threshold alarms, exception handling, and dynamic regulation of the production process.

[0089] The embodiments of the present invention achieve autonomous adjustment and global optimization of the production system through distributed collaborative decision-making and dynamic optimization of multiple agents, ensuring the stability and flexibility of the system in complex environments.

[0090] (5) Twin Model

[0091] Twin models are primarily used to achieve multi-dimensional, high-precision digital mapping and dynamic simulation analysis of production factor entities. Twin models simulate each optimization solution in a set of optimization solutions, allowing the analysis and decision-making agent to determine the target production solution based on the simulation results. Twin models include: geometric models, physical models, rule models, behavioral models, simulation models, and business models.

[0092] The geometric model is the three-dimensional shape, spatial structure, and layout distribution of the physical entities of each production system constructed using tools such as SolidWorks and AutoCAD;

[0093] Physical models include dynamic or thermodynamic models for each stage of the process. For example, physical models include multi-body dynamic models using Adams to simulate the motion and forces of the feed structure in the production system, CFD fluid dynamics models for the raw material flow process and biogas purification process, and thermodynamic models for temperature field changes and heat conduction analysis during the raw material fermentation process in the fermenter.

[0094] Rule models include constraint models, standard models, and safe operation process models. For example, there is a constraint model based on the empirical formula for biogas production, an enterprise standard model for biogas production, and a safe operation process model for biogas purification.

[0095] Behavioral models include state machine models, raw material matching operation process models, and timing coordination models. For example, a state machine model for the state transition of biogas purification equipment, a raw material matching operation process model based on production capacity requirements, and a timing coordination model for each production link.

[0096] Simulation models include multi-physics coupling simulation models, dynamic simulation models, and cycle time optimization simulation models. For example, there are multi-physics coupling simulation models for the biogas production process in fermentation tanks, dynamic simulation models for raw material collection and transportation route planning, and simulation models for production line cycle optimization and bottleneck reduction.

[0097] The business model includes raw material collection, storage and transportation planning model, quality traceability management model, production system energy efficiency management model and production capacity assessment and prediction model.

[0098] The present invention provides an intelligent system for comprehensive management and control of multi-source breeding waste resources, including: a production system physical entity, a data acquisition intelligent entity, a data interaction intelligent entity, a twin model and an analysis and decision-making intelligent entity. The data acquisition intelligent entity collects the real process data of the physical entity of the production system during the process; the data interaction intelligent entity is used to control the transmission and processing of data between the intelligent entities in the intelligent system, and pre-process the real process data; the analysis and decision-making intelligent entity analyzes and optimizes the original process data and historical experience data through a machine learning algorithm to obtain a set of optimization solutions; the twin model simulates each optimization solution in the set of optimization solutions, so that the analysis and decision-making intelligent entity can determine the target production solution based on each simulation result. The present invention uses a machine learning algorithm to quickly analyze the big data of the process, and at the same time uses distributed collaborative optimization between intelligent entities to achieve real-time monitoring and control of data in the process of multi-source breeding waste resource utilization, effectively improving the efficiency of multi-source breeding waste resource utilization and enhancing the level of intelligent management and control of the production system.

[0099] The embodiments of the present invention use machine learning algorithms and other means to perform multi-objective optimization of the process, accurately adjust the conflicts and coupling effects between multiple objectives, quickly improve the accuracy and flexibility of optimization decisions, and increase production efficiency by more than 10%, thereby meeting the needs of different production resources, costs, benefits and environments.

[0100] The embodiments of the present invention utilize digital twin technology to construct multi-scale, multi-level twin models of various links. Through virtual-reality mapping, dynamic simulation, and data interaction, they can simulate, analyze, verify, and predict the dynamic behavior of each production link in the multi-source agricultural waste resource utilization process in a virtual environment, thereby achieving data integration and full lifecycle management for the entire production process. This effectively overcomes the problems of data silos and system response lags in traditional management and control processes, and can provide intelligent solutions for the agricultural waste resource utilization process. Intelligent decision-making covers over 80% of the process, ensuring the smooth operation of the production system.

[0101] The embodiment of the present invention adopts multiple intelligent agents to achieve seamless connection of production site monitoring, intelligent coordination, distributed decision-making and real-time response, thereby increasing the utilization rate of agricultural waste raw materials by more than 20%, reducing the equipment failure rate by more than 50%, greatly improving the overall performance of the system, and better adapting to dynamic changes in production conditions.

[0102] Based on the above embodiments, the comprehensive management and control intelligent agent system may further include: an application service intelligent agent.

[0103] The Application Service Agent (ASA) is an intelligent agent function customized according to the usage requirements of the production process. It can realize functions such as material information management, raw material collection and transportation route planning, raw material matching optimization, production process parameter optimization, production equipment inspection and maintenance, intelligent equipment fault diagnosis and early warning, and production capacity assessment and prediction. Through the visual operation interface of the management and control system, the relevant information results are provided to users according to their responsibilities and permissions. The Application Service Agent realizes the execution and iterative optimization of production decisions through data-driven services and virtual-real integration verification. The Application Service Agent can not only receive the decision results of the analytical decision-making agent for production operations, but also call the twin model for dynamic display of the production process and feedback the new data generated during the operation process to the physical entity of the production system. The services provided by the Application Service Agent include: dynamic production scheduling optimization service, energy efficiency management service, quality traceability service, predictive service and visualization service.

[0104] Dynamic production scheduling optimizes production plans and schedules based on constraints, determining the timing and sequence of production tasks. By weighing constraints such as order demand, equipment capacity, raw material storage availability, and delivery dates, Dynamic Production Optimization dynamically adjusts production plans and schedules for products like biogas and organic fertilizers, determining the optimal timing and sequence for production tasks. This optimizes resource allocation, increases equipment utilization, and ensures on-time product delivery.

[0105] Energy efficiency management services evaluate the unit product energy consumption and comprehensive energy efficiency of key equipment and production systems through energy flow and load curve analysis, thereby optimizing energy allocation, achieving refined energy consumption control, and effectively reducing production costs.

[0106] The quality traceability service records and tracks data from the entire production process, such as raw material supply, production process parameters, equipment operating status, operator information, etc., enabling rapid identification of problem products, accountability tracing, and quality improvement, thereby improving product quality and reliability.

[0107] Predictive services are used to monitor the health of production equipment and predict failures, predict market demand for products, and predict raw material supply.

[0108] The visualization service uses graphical means to intuitively display production process data, processes, and status. For example, it can show daily biogas production trends, monthly energy consumption curves for production systems, equipment operating status traffic lights, equipment failure sound and light alarms, and organic fertilizer processing progress Gantt charts.

[0109] Furthermore, in order to better implement the raw material collection and transportation route planning, the application service agent can also include a raw material collection and transportation planning algorithm. Figure 3 To introduce, the algorithm is as follows:

[0110] The raw material collection and transportation planning algorithm uses indicators such as raw material category, raw material production and inventory, raw material distribution location, raw material collection frequency, raw material transportation method, and raw material processing cost as input parameters of the optimization algorithm, and takes the lowest total transportation cost, shortest transportation time, balanced transportation task distribution, and optimal transportation route as output parameters. It uses machine learning algorithms to perform multi-objective optimization iterations to improve reasoning efficiency and quickly obtain the Pareto optimal solution set.

[0111] The raw materials include chicken manure, duck manure, pig manure, cow manure, straw and fruit peels, etc. Take the spatiotemporal distribution of agricultural waste in the six townships of a certain agricultural cluster as an example. Figure 4 As shown. Among the raw material distribution points in the figure, point 1 represents cow dung, which is collected and transported once every 3 days; point 2 represents pig dung, which is collected and transported once a day; point 3 represents chicken and duck dung, which is collected and transported once every 3 days; point 4 represents straw, which is seasonal production. Raw material production and inventory refers to the amount of waste generated and remaining every day / week at the waste generation point, which fluctuates with the breeding season or production cycle. The frequency of raw material collection includes the allowable storage time of various types of waste (such as daily collection or collection every few days). The mode of transportation includes the type of transportation vehicle, load capacity, energy consumption, unit distance transportation cost, etc. The raw material processing cost includes the purchase price of the raw materials and government subsidies. The total transportation cost includes transportation fees, fuel costs, labor costs, maintenance fees, etc. The transportation task allocation includes the type and weight of waste that each transportation vehicle is responsible for transporting, the location of the corresponding collection point and the transportation route, the start and end working time of the vehicle, etc. The multi-objective optimization model constructed by the selected input and output quantities is as follows:

[0112]

[0113] In the formula, x1 is the raw material category, x2 is the raw material production and inventory, x3 is the raw material distribution location, x4 is the raw material collection frequency, x5 is the raw material transportation method, and x6 is the raw material processing cost. i(i = 1, 2, ..., 6) have their own ranges of values, and a, b, c, and d are the corresponding values. f1(x) is the objective function for total transportation cost, f2(x) is the objective function for transportation time, f3(x) is the objective function for transportation task allocation, and f4(x) is the objective function for transportation route. st is the abbreviation for subject to, which is the unified mathematical notation for multi-objective optimization formulas and has no special meaning here.

[0114] Since the input variable x i The value range and unit of are different. In order to eliminate the influence of different dimensions and improve the efficiency of data training, the input variables need to be normalized according to the following formula:

[0115]

[0116] Where, Normalized data, x ij is the jth sample value of the i-th class input, max i (x ij ) is the maximum value of the input of type i; min i (x ij ) is the minimum value of the i-th type input. The normalized values ​​of the six parameters are as follows Figure 5 shown.

[0117] For high-dimensional multi-objective optimization problems, the improved NSGA-III optimization neural network model can be used for iterative optimization to improve the uniformity of the Pareto solution distribution and enhance the convergence ability. The optimization strategy flow chart of the NSGA-III algorithm is as follows: Figure 6 As shown. First, the sample point data is fitted using a multi-layer neural network until the fitting accuracy meets the requirements. Then, the fitting model is brought into the improved NSGA-III for local multi-target search. The iterative search process is as follows: (1) After determining the population size N = 50 and the number of iterations T max =300, the number of reference points H is determined by the d objective functions and their number of segments p, that is: And generate the initial population P t (2) Calculate individual fitness, perform non-dominated sorting, and generate a new population Q through selection, crossover, and mutation t (3) The population P t With Q t After merging, apply non-dominated sorting and reference point sorting, and select the first 50 individuals in the sorting result to generate a new population P t +1. This cycle continues until the maximum number of iterations is reached. The iterative convergence curve of the process objective function is as follows: Figure 7Finally, the optimized individual parameters are brought into the twin model for virtual verification, and the optimization effect is evaluated based on the simulated deviation. If the external disturbance r(.) causes the system deviation e(.) to not meet the requirements, optimization is performed again, and the network weights are adaptively adjusted to continuously correct the network output to achieve the expected target value.

[0118] In addition, in order to better optimize the raw material matching, the application service agent can also include a raw material matching optimization method. The PSO algorithm with dynamic adjustment of particle distance is used for parameter optimization to improve the algorithm search efficiency and accelerate the convergence speed. The method is as follows: the amount of biogas tanks, the amount of biogas grid-connected, the output of biogas liquid drip irrigation fertilizer, the output of biogas residue organic fertilizer, and the supply of raw materials are used as particle input parameters, and the ratio of different raw materials is used as the output parameter. In the D-dimensional search space, there are M particles, and the position X of the i-th particle is i =(x i1 ,x i2 ,...,x iD ), speed V i =(v i1 ,v i2 ,...,v iD ), whose individual extreme value P best，i =(p i1 ,p i2 ,...,p iD ), global extreme value G best，g =(g g1 ,g g2 ,...,g gD ), its particle update rules are as follows:

[0119]

[0120] Where ① is the inertia weight; c1 and c2 are constants, usually 2; t is the number of iterations; r1 and r2 are random numbers in the interval [0,1]. To improve the search capability, ① is dynamically adjusted using a nonlinear decreasing method:

[0121]

[0122] In order to evaluate the fitness of each particle and reduce the amount of calculation, the method of correcting the distance between particles is adopted: d = α (x max -x min ), α represents the particle relative to the search space [x min ,x max ] coefficient factor. Optionally, according to the particle position range, set α to 0.3, 0.6, 0.9 to correct the particle distance. Thus, d i The value range of [x i -d,xi +d]. Then calculate the fitness values ​​of all particles that meet the range, and update the speed and position of the particles according to the following formula:

[0123]

[0124] The intelligent system for integrated resource management of multi-source agricultural and livestock waste, implemented in this embodiment of the present invention, enables virtual simulation, analysis, verification, and prediction of the entire production process. Practical verification has shown that the use of this intelligent system can shorten the fermentation cycle by over 10%, increase gas production per unit of raw material by approximately 10%, and reduce overall operating costs by approximately 15%, demonstrating promising application prospects.

[0125] Next, the embodiment of the present invention introduces the operation method of the multi-source breeding waste resource integrated management and control intelligent system. Figure 8 As shown, the method is as follows:

[0126] S1 Raw material preprocessing: The data acquisition agent collects various parameters of the raw materials in real time and transmits them to the analysis and decision-making agent through the data interaction agent; the analysis and decision-making agent preliminarily screens the raw material types and ratios based on daily production requirements and existing raw material stocks, and then recommends raw material formulas based on historical production data. The twin model simulates the impact of different raw material formula parameters in the raw materials on gas production efficiency and determines the optimal raw material ratio scheme.

[0127] Specifically, the data acquisition agent collects various parameters of the raw materials in real time through the sensor network, such as the solid content and vs ratio of livestock and poultry manure, C / N ratio, moisture content, length, type and other information of the straw, and transmits it to the analysis and decision-making agent. Based on the daily production demand of biogas and the existing raw material inventory, the raw material type and ratio are preliminarily screened. Then, based on historical production data, the raw material formula is recommended. The twin model is used to simulate the effects of different straw cutting sizes (such as 2cm, 5cm, 10cm, etc.) and different raw material formula parameters on gas production efficiency in the raw materials, thereby determining the optimal raw material ratio scheme and transmitting the instructions to the automatic batching equipment to achieve precise batching. For example, a machine vision-guided intelligent cutting system can be used to adaptively adjust the straw cutting size to 3-5cm. At the same time, the RGB-D100 detector is used to identify indicators such as the uniformity of raw material mixing (preferably the threshold value of meeting the standard is >95%), the preheating temperature is preferably ±1°C), the solid content of the raw material mixture (preferably 10%-15%), and whether the raw material state meets the fermentation standard, so as to automatically adjust the operating parameters of the mixing equipment and preheating equipment to ensure that the raw material parameters are consistent with the fermentation model input requirements, and provide basic conditions for the subsequent production of biogas. During the pretreatment process, the mechanical screen will intercept large-sized (>55mm) mixtures for further mixing and stirring. At the same time, the grit chamber will remove substances with particle diameters greater than 0.3mm, and adjust the pH to 6-8 to ensure that the hydraulic retention time of the mixture is around 30-35d.

[0128] S2 anaerobic fermentation to produce biogas: The data acquisition agent monitors the fermentation process parameters of the fermentation tank in real time and transmits them to the analysis and decision-making agent through the data interaction agent; the analysis and decision-making agent performs analysis based on historical experience values ​​and fermentation process parameters; the twin model performs multi-physics field simulation and optimization of fluid mechanics and biochemical reaction dynamics to determine the optimal fermentation process parameters.

[0129] Specifically, a screw pump can be used to pump the pretreated raw materials into the CSTR fermenter, and a flow meter can be used to record the processing volume. During the biogas production process, the data acquisition agent monitors the operating status of the fermenter in real time, including indicators such as temperature, pressure, liquid level, pH value, stirring rate, and biogas production. In addition, the 3D model dynamics and operating data of the equipment can be presented in real time in the application service agent; at the same time, the data is transmitted to the analysis and decision-making agent for analysis and processing. In the twin model of the equipment, based on historical experience values ​​and collected fermentation parameters, multi-physics field simulation and optimization of the fermentation process using computational fluid dynamics and biochemical reaction kinetics are performed to determine the optimal fermentation process parameters. Then, the system adaptively switches between medium-temperature fermentation (35-40°C) and high-temperature fermentation (50-55°C) modes, and automatically controls the operating frequency and time of the stirring device to maintain efficient and stable operation of anaerobic fermentation and improve gas production.

[0130] Furthermore, a fault prediction model is established for each key piece of equipment in the production process. Big data analysis is conducted using historical and real-time monitoring data to predict potential equipment failures in advance, enabling predictive maintenance and effectively reducing equipment maintenance time and operating costs. When the data acquisition agent detects that the gas production rate has fallen below 80% of the theoretical value for two consecutive hours, an abnormal fermentation tank operating condition alarm is triggered. Simultaneously, the fermentation tank's raw material data, process data, and equipment operating status data for the past two days are sent to the analysis and decision-making agent. Using a fault tree, the root cause of the problem, such as a clogged feed pump or a leaking control valve, is analyzed. A maintenance plan is then sent to the application service agent to guide the repair personnel.

[0131] S3 biogas purification and upgrading: The data acquisition agent collects the performance parameters of the purified biogas in real time; the analysis and decision-making agent uses machine learning algorithms to perform multi-objective collaborative optimization and iteratively search for the Pareto optimal solution.

[0132] Specifically, amine washing, water washing, membrane and other methods can be used to purify biogas, collect the purity, pressure, flow and other indicators of the purified biogas in real time, and transmit the data to the application service intelligent body for visual display. At the same time, based on the collected CH4 purity and the preset purification index (>98%), an intelligent algorithm is used to coordinate the optimization of desulfurizers, desiccant, CH4 purity and energy consumption, and quickly iterate to search for the Pareto optimal solution to achieve precise control, thereby improving the purification efficiency. Record the biogas quality data after each batch of purification and purification, and conduct data mining analysis in the analysis and decision-making intelligent body based on the source of raw materials, fermentation process parameters, and purification process parameters to continuously optimize the purification process and improve the stability of biogas quality. The purified biogas enters the dual-mode gas storage cabinet for temporary storage in preparation for disposal in step S6.

[0133] S4 Drip Fertilizer Preparation with Biogas Slurry: The data collection agent collects the component indicators of biogas slurry in real time; the analysis and decision-making agent optimizes the irrigation and fertilizer preparation parameters based on the component indicators, nutrient demand model and soil test data through machine learning algorithms.

[0134] Specifically, after solid-liquid separation, the biogas slurry is first fermented in a covered biogas tank for secondary fermentation. The generated biogas enters step S3 for impurity removal and then undergoes 3MPa plate and frame filtration to prepare drip irrigation fertilizer. The process is as follows. First, a data acquisition agent is used to collect information on indicators such as nutrient content, trace elements, and organic matter in the biogas slurry. Then, based on the nutrient requirement model of crops and combined with soil testing data, the analysis and decision-making agent uses a machine learning algorithm to quickly optimize parameters such as the dilution ratio of the drip irrigation fertilizer, the type and amount of added trace elements, and other parameters. The optimized process parameters are sent to the production equipment to guide the production process.

[0135] The error in trace element addition is controlled within ±0.1%. Simultaneously, production equipment data such as stirring speed, mixing time, and fertilizer concentration are monitored in real time. Equipment operating status is adjusted based on product-specific indicators such as pH, nutrient content, and heavy metal content, ensuring that the quality of drip irrigation fertilizer meets product standards and crop growth requirements. Finally, based on historical production data, real-time production status, raw material supply, market demand, and other conditions, production capacity forecasts and production strategy adjustments are performed within the integrated management and control twin platform. Optimization results are displayed to operators through a visualization service module, facilitating production management.

[0136] S5 Preparation of organic fertilizer from biogas residue: The data acquisition agent monitors the environmental parameters of the fermentation pile in real time, and the analysis and decision-making agent adjusts the environmental parameters according to the preset fermentation process curve and target parameter range.

[0137] Specifically, first, the biogas residue obtained by solid-liquid separation is subjected to a raw material physical and chemical property analysis to extract information such as its moisture content, organic matter content, nutrient composition, and particle size, providing basic data for the subsequent organic fertilizer production process. Then, during the fermentation of biogas residue to prepare organic fertilizer, the data acquisition intelligent body monitors the temperature, humidity, oxygen content, pH value and other parameters of the fermentation pile in real time, and accurately controls the production status according to the preset fermentation process curve and target parameter range to improve the maturity and quality of the organic fertilizer. Finally, the prepared organic fertilizer product is tested for indicators such as organic matter content, nutrient content, heavy metal content, and particle strength to judge the product quality and grade, providing a basis for product pricing and market sales. Preferably, the particle size of the organic fertilizer particles is 1.5-4.5mm and is rich in amino acids, humic acid (>25%), etc. When organic fertilizer is used as edible fungus base fertilizer, the humus content, pH value, sawdust, bran ratio, etc. of the organic fertilizer are adjusted for the growth model of edible fungi such as enoki mushrooms and oyster mushrooms, so that the edible fungus yield is increased by about 10%.

[0138] S6 Product Sales and Market Forecast: Analytical decision-making agents formulate sales strategies and pricing plans through machine learning algorithms.

[0139] Specifically, by collecting and integrating multi-dimensional information such as market demand, price fluctuations, competitor dynamics, and policy and regulatory impacts of products such as biogas, drip irrigation fertilizer, and organic fertilizer, and combining it with the company's own production capacity, the decision-making intelligent body uses sales forecasting models and intelligent decision-making algorithms to formulate personalized sales strategies and pricing plans to improve sales efficiency.

[0140] Purified biogas can be directly fed into the production system's gas boilers, pressurized to 25 MPa and transported via pipeline to gas stations, or canned and shipped to users. Drip irrigation fertilizer and organic fertilizer can be customized and sold according to growers' needs. Furthermore, real-time monitoring of raw material supply and market demand fluctuations is used. Intelligent algorithms such as LSTM neural networks are used to assess their impact on product production capacity and costs, enabling the development of countermeasures. Furthermore, adjustments to raw material procurement and storage plans, expansion of raw material sourcing, and revision of production schedules can be implemented to ensure the stability and continuity of production operations.

[0141] Furthermore, the process of capacity forecasting is as follows:

[0142] In view of the temporal and nonlinear characteristics of production capacity forecasting of biogas, organic fertilizer, etc., the improved long short-term memory network (LSTM) fused with extreme gradient boosting (XGBoost) method can be used to perform classification prediction and then regression prediction to achieve comprehensive production capacity prediction.

[0143] First, recent information can be used as a dataset for the prediction model. For example, multi-dimensional information such as product market demand, price fluctuations, competitor dynamics, regulatory impact, raw material costs, and daily production capacity data from the past three months can be used as the dataset for the prediction model. 80% of this information is randomly selected as the training set and 20% as the test set for model training and validation, respectively. Furthermore, the information is formatted to unify the different dimensions. For example, policy text such as "subsidy" and "production restriction" can be converted into numerical features, and the data can be normalized and preprocessed to the range [0, 1].

[0144] Next, we use the strategy of automatically adjusting the learning rate to build an improved LSTM model, and use the training set and test set to train the model. We define the forget gate, input gate, and output gate of the LSTM respectively, and finally obtain the predicted output: Among them, ω h and ω0 are weight matrices, b0 and b γ Bias term, is hidden state, is the input, σ is the activation function, The LSTM learning rate automatic adjustment method is: when the model training effect does not improve after a iteration, the learning rate L is automatically reduced to Right now Preferably, a is between 10 and 30. During model training, the learning rate L is automatically adjusted to quickly move to the optimal value in the early stages of training. In the later stages of training, the learning rate is reduced to further optimize the training effect. Preferably, L is between 0.001 and 0.01.

[0145] Secondly, the prediction effect of the LSTM model is verified using the validation set. By calculating the model prediction value x i * and the measured value x i Root mean square error RMSE and determination coefficient R 2 To evaluate the model prediction performance, N is the number of data in the dataset, is the measured mean. If the prediction effect does not meet the requirements, adjust the LSTM model hyperparameters such as the number of hidden layers, number of iterations, and attenuation factor; otherwise, terminate the model training and output the prediction results.

[0146] Finally, the output of the LSTM model is used as the input of the XGBoost regression prediction model for secondary regression analysis. The prediction results are combined according to the overall weight to obtain the integrated prediction result, thereby improving the prediction accuracy.

[0147] The model prediction results using historical production data are compared with the measured values. Figure 9 As shown, the average RMSE is 10.65, R 2 =0.98, which proves that the prediction effect is relatively accurate and can provide guidance for actual production.

[0148] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0149] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0150] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent system for comprehensive management and control of multi-source breeding waste resources, characterized by: include: Production system physical entities, data collection agents, data interaction agents, twin models, and analysis and decision-making agents; The data acquisition intelligent agent collects real process data of the physical entity of the production system during the process; The data interaction agent is used to control the transmission and processing of data between the agents in the intelligent system, pre-process the real process data, and obtain the original process data; The analysis and decision-making intelligent agent analyzes and optimizes the original process data and historical experience data through a machine learning algorithm to obtain an optimization solution set; The twin model performs simulation based on each optimization scheme in the optimization scheme set, so that the analysis and decision-making intelligent agent can determine the target production scheme based on each simulation result.

2. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 1 is characterized in that: The physical entity of the production system includes: production raw materials, machinery and equipment, and auxiliary tools in the process of resource utilization of multi-source breeding waste.

3. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 1 is characterized in that: The data acquisition intelligent agent includes: temperature sensors, pressure sensors, flow sensors, vibration sensors, speed sensors, state sensors, as well as video cameras, scanners, RFID, and SCADA; The actual process data includes raw material ratio, pumping pressure, fermentation temperature, feed flow rate, biogas concentration, tank liquid level, stirring frequency, motor speed, equipment vibration, instrument noise, production video and production image.

4. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 1 is characterized in that: The data interaction agent includes: a communication module, a data access module, a data preprocessing module and a multimodal interaction module; The communication module provides an information interface for data transmission and interaction between agents based on the communication protocol; The data access module aggregates and accesses data from different agents through gRPC or Kafka message queues; The data preprocessing module is used to perform preprocessing operations such as cleaning, format conversion and normalization on the collected data; The multimodal interaction module is used to realize the interaction of structured data and unstructured data between various intelligent agents.

5. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 1 is characterized in that: The twin model includes: a geometric model, a physical model, a rule model, a behavior model, a simulation model, and a business model; The geometric model is the three-dimensional shape, spatial structure and layout distribution of the physical entities of each production system; The physical model includes a kinetic model or a thermodynamic model of each link of the process; The rule model includes a constraint model, a standard model and a security operation process model; The behavior model includes a state machine model, a raw material matching operation process model and a timing coordination model; The simulation model includes a multi-physics field coupling simulation model, a dynamic simulation model and a beat optimization simulation model; The business model includes a raw material collection, storage and transportation planning model, a quality traceability management model, a production system energy efficiency management model and a production capacity assessment and prediction model.

6. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 1 is characterized in that: The analysis and decision-making intelligent agent includes: a collaboration mechanism module, a control decision module, a system optimization module, a task allocation and adjustment module, and a monitoring feedback module; The collaboration mechanism module includes a distributed or hybrid collaboration mode and task collaboration process between agents and a conflict resolution mechanism for each agent; The control decision module includes mechanism model driven and data driven control algorithms, and dynamic decision mechanisms triggered by time or events; The system optimization module includes optimization objectives and optimization algorithms; The task allocation and adjustment module includes task decomposition basis, task allocation strategy and task adjustment mechanism; The monitoring feedback module monitors the process flow in real time based on the real process data collected by the data acquisition intelligent agent in real time, and implements threshold alarms and abnormality handling based on the feedback early warning mechanism.

7. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 1 is characterized in that: The agent system further includes: an application service agent; The application service agent is verified through data-driven service and virtual-reality fusion.

8. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 7 is characterized in that: The services provided by the applied service agent include: dynamic production scheduling optimization service, energy efficiency management service, quality traceability service, predictive service and visualization service; The dynamic production scheduling optimization service dynamically adjusts the production plan and scheduling scheme based on constraints and determines the time and sequence of production tasks; The energy efficiency management service evaluates unit product energy consumption and comprehensive energy efficiency through energy flow and load curve analysis, thereby optimizing energy allocation; The quality traceability service records and tracks data from the entire production process; The predictive services are used to monitor the health status of production equipment and predict failures, predict market demand for products, and predict raw material supply; The visualization service uses graphical means to intuitively display production process data, processes and status.

9. The intelligent system for comprehensive management and control of multi-source breeding waste resources according to claim 8 is characterized in that: The application service agent also includes: a raw material collection and transportation planning algorithm; The raw material collection and transportation planning algorithm uses indicators such as raw material category, raw material production and inventory, raw material distribution location, raw material collection frequency, raw material transportation method, and raw material processing cost as input parameters of the optimization algorithm, and takes the lowest total transportation cost, shortest transportation time, balanced transportation task distribution, and optimal transportation route as output parameters. It uses a machine learning algorithm to perform multi-objective optimization iteration to improve reasoning efficiency and quickly obtain the Pareto optimal solution set.

10. A method for operating an intelligent system for comprehensive management and control of multi-source breeding waste resources, characterized in that: The intelligent system for comprehensive management and control of multi-source breeding waste resources according to any one of claims 1 to 9 is applied, and the operation method includes: S1 Raw material preprocessing: The data acquisition agent collects various raw material parameters in real time and transmits them to the analysis and decision-making agent through the data interaction agent. The analysis and decision-making agent preliminarily screens the raw material types and ratios based on daily production requirements and existing raw material inventory. Then, based on historical production data, it recommends raw material formulas. The twin model simulates the impact of different raw material formula parameters on gas production efficiency to determine the optimal raw material ratio scheme. S2 anaerobic fermentation to produce biogas: The data acquisition agent monitors the fermentation process parameters of the fermentation tank in real time and transmits them to the analysis and decision-making agent through the data interaction agent. The analysis and decision-making agent performs analysis based on historical experience and fermentation process parameters. The twin model performs multi-physics simulation and optimization of fluid mechanics and biochemical reaction dynamics to determine the optimal fermentation process parameters. S3 Biogas Purification: The data collection agent collects the performance parameters of purified biogas in real time; the analysis and decision-making agent uses machine learning algorithms to perform multi-objective collaborative optimization and iteratively search for the Pareto optimal solution; S4: Drip irrigation fertilizer preparation using biogas slurry: The data collection agent collects biogas slurry component indicators in real time; the analysis and decision-making agent uses machine learning algorithms to optimize the irrigation and fertilizer preparation parameters based on component indicators, nutrient demand models, and soil testing data. S5 Biogas Residue Preparation for Organic Fertilizer: The data acquisition agent monitors the environmental parameters of the fermentation pile in real time, and the analysis and decision-making agent adjusts the environmental parameters according to the preset fermentation process curve and target parameter range; S6 Product Sales and Market Forecast: Analytical decision-making agents formulate sales strategies and pricing plans through machine learning algorithms.

Images