A closed-loop operation management method for biogas fertilizer based on organic waste resource utilization

Abstract

This invention discloses a closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste, relating to the field of fertilizer manufacturing technology. The method includes: acquiring customized biogas slurry application plans; collecting full-chain data and real-time monitoring data; constructing a causal inference model, inputting the full-chain data, real-time monitoring data, and customized biogas slurry application plans into the causal inference model to identify the root causes of unsatisfactory biogas slurry quality; generating anaerobic fermentation process adjustment strategies and modification suggestions for customized biogas slurry application plans based on the root causes; collecting task execution process data, new monitoring data, and incentive settlement results, and correlating them back to the full-chain data; iteratively optimizing the causal inference model and multi-agent game theory. This invention, by constructing a causal inference model and inputting full-chain data, real-time monitoring data, and customized biogas slurry application plans, achieves deep and accurate diagnosis of problems where biogas slurry quality or application effects do not meet expectations.

Description

Technical Field

[0001] This invention relates to the field of fertilizer manufacturing technology, specifically a closed-loop operation and management method for biogas fertilizer based on the resource utilization of organic waste, with its main IPC classification number involving C05G. It improves soil through the resource utilization of organic waste and precision fertilization. Background Technology

[0002] In recent years, the Internet of Things (IoT), big data, and artificial intelligence (AI) technologies have been widely applied in the resource utilization of agricultural organic waste, particularly in anaerobic fermentation for fertilizer production and precision agronomic management. By deploying sensor networks, real-time monitoring of raw material composition, fermentation process parameters, biogas fertilizer quality indicators, and soil moisture and crop growth in farmland has become possible. Simultaneously, path optimization algorithms and production execution tools have been used to improve collection and transportation efficiency and stabilize production processes, thus initially forming a data-driven operation and management model.

[0003] However, existing technical solutions have shortcomings in achieving closed-loop optimization across the entire chain. The data and decision-making models of each link (collection, transportation, processing, and application) are independent, forming "data silos" and "decision-making breaks." When the quality of biogas fertilizer or the field application effect does not meet expectations, existing methods struggle to penetrate the complex cross-link interactions, quickly and accurately pinpoint the root cause of the problem, and cannot automatically translate diagnostic conclusions into coordinated instructions that drive adjustments in front-end collection, mid-stage processing, and back-end application. This results in operational adjustments relying on trial and error based on experience, leading to slow responses and a lack of systematic optimization capabilities. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a closed-loop operation and management method for biogas fertilizer based on the resource utilization of organic waste to solve the problems of inaccurate diagnosis of root causes and delayed response of cross-link collaborative scheduling in the existing closed-loop management of organic waste fertilizer production and application due to the separation of data and decision-making.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste. The method includes: acquiring a customized biogas slurry application plan and collecting full-chain data and real-time monitoring data; constructing a causal inference model, inputting the full-chain data, real-time monitoring data, and customized biogas slurry application plan into the causal inference model to identify the root causes of unsatisfactory biogas slurry quality; generating anaerobic fermentation process adjustment strategies and modification suggestions for the customized biogas slurry application plan based on the root causes; converting the new requirements for organic waste raw material collection and distribution in the modification suggestions into pending task orders with dynamic incentive clauses; setting collection vehicles, fertilization service providers, and treatment stations as agents, and using multi-agent game theory to enable each agent to bid and match based on the pending task orders to form a real-time service plan; executing the real-time service plan and anaerobic fermentation process adjustment strategy, verifying the effectiveness associated with the pending task orders after execution, and settling the corresponding dynamic incentive clauses; collecting task execution process data, new monitoring data, and incentive settlement results, and feeding them back to the full-chain data to iteratively optimize the causal inference model and multi-agent game theory.

[0007] As a preferred embodiment of the closed-loop operation and management method for biogas fertilizer based on the resource utilization of organic waste described in this invention, the customized biogas fertilizer application plan is generated based on soil testing results of farmers' land and crop growth needs.

[0008] As a preferred embodiment of the closed-loop operation and management method for biogas fertilizer based on the resource utilization of organic waste described in this invention, the full-chain data includes data on the source, type, collection amount and time of organic waste, process parameters such as temperature, pH and material residence time during anaerobic fermentation, biogas fertilizer output, nitrogen, phosphorus and potassium content and decomposition quality index data, and data on the application amount, application method, application time and location of biogas fertilizer. The real-time monitoring data includes soil moisture, pH value and key nutrient content monitoring data, as well as crop growth status monitoring data.

[0009] As a preferred embodiment of the closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste described in this invention, the steps for constructing the causal inference model are as follows: A multi-source heterogeneous data fusion layer is built based on feature engineering and data fusion; a causal discovery layer is built based on causal structure learning; and a causal effect estimation layer is built based on counterfactual reasoning and effect estimation. By hierarchically connecting the multi-source heterogeneous data fusion layer, the causal discovery layer, and the causal effect estimation layer, a causal inference model is constructed.

[0010] As a preferred embodiment of the closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste described in this invention, the steps for identifying the root cause of unsatisfactory biogas slurry quality are as follows: The multi-source heterogeneous data fusion layer standardizes and fuses the full-chain data, real-time monitoring data, and customized biogas fertilizer application schemes to generate a unified spatiotemporal feature tensor. The spatiotemporal feature tensor is input into the causal discovery layer to analyze the causal dependencies between variables and construct a causal structure graph. For each candidate causal variable pointing to the target variable of biogas fertilizer quality in the causal structure diagram, counterfactual reasoning is performed through the causal effect estimation layer to calculate the causal effect value of each candidate causal variable on the target variable; All candidate causal variables were ranked according to their causal effect values, and the candidate causal variable ranked first was determined to be the root cause of the failure of biogas fertilizer quality to meet expectations.

[0011] As a preferred embodiment of the closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste described in this invention, the steps for generating the anaerobic fermentation process adjustment strategy and the modification suggestions for the customized biogas slurry application scheme are as follows: Based on the root causes, a quantitative response relationship is established between the associated process and adjustable fertilization parameters and multiple objectives such as biogas fertilizer quality, application effect, economy and environment, and corresponding constraints are defined. Based on the quantitative response relationship and corresponding constraints, the optimal combination of process parameters and fertilization parameters is solved through multi-objective optimization and simulation verification, generating modification suggestions for anaerobic fermentation process adjustment strategies and customized biogas fertilizer application schemes.

[0012] As a preferred embodiment of the closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste described in this invention, the steps of converting the biogas slurry into a task order to be executed with dynamic incentive clauses are as follows: The new requirements for the collection and distribution of organic waste materials in the revised proposals are analyzed and broken down into several basic tasks. For each basic task, obtain the state entropy of the real-time task market and the matching degree of available service providers, and calculate the dynamic incentive amount for each basic task. Each basic task is bound to its corresponding dynamic incentive quota and packaged into a task order to be executed.

[0013] As a preferred embodiment of the closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste described in this invention, the steps for forming a real-time service scheme are as follows: The collection vehicles, fertilizer service providers, and processing stations are initialized as intelligent agents, and the task orders to be executed are broadcast to each intelligent agent. Each agent reports its own state and conditional commitments to the coordinator, which then summarizes and constructs a task conflict relationship graph. Based on the task conflict relationship graph, it initiates a multi-round iterative game with conditional commitments as the strategy. When the iterative game converges, the final determined allocation relationship between the agent and the task orders to be executed is used as a real-time service solution.

[0014] As a preferred embodiment of the closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste described in this invention, the dynamic incentive clauses corresponding to the settlement include the following steps: During and after the implementation of the real-time service plan and anaerobic fermentation process adjustment strategy, multi-source process data and effect monitoring data are collected to form the original dataset to be verified. The original dataset to be verified is input into the decentralized verification network for anti-interference processing, and the comprehensive verification index of each task order to be executed is calculated. Based on the comprehensive verification index, the smart contract deployed on the blockchain is triggered to automatically calculate and execute the payment of the dynamic incentive terms according to the preset settlement rules.

[0015] As a preferred embodiment of the closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste described in this invention, the iterative optimization of the causal inference model and the multi-agent game involve the following steps: The collected task execution process data, new monitoring data, and incentive settlement results are correlated and structured to form feedback experience tuples. Each feedback experience in the feedback experience tuples is then evaluated and ranked to construct a course-based learning sequence. Based on the curriculum-based learning sequence, parameter update directions are generated for the policy network of causal inference models and multi-agent games, while dynamically allocating differentiated adaptive learning rates. The policy network of the causal inference model and multi-agent game is incrementally updated using parameter update direction and adaptive learning rate.

[0016] The beneficial effects of this invention are as follows: By constructing a causal inference model and inputting full-chain data, real-time monitoring data, and customized biogas fertilizer application schemes, it achieves in-depth and accurate diagnosis of problems where the quality or application effect of biogas fertilizer does not meet expectations; it identifies the root cause variables that lead to the problems, providing a reliable basis for generating accurate process adjustment and fertilization correction suggestions, effectively guiding the resource utilization of organic waste and precise fertilization. This not only lays the decision-making foundation for increasing soil organic matter content, improving soil structure, and achieving synergistic effects of carbon reduction, pollution reduction, and greening in ecological governance of saline-alkali land, but also reduces crop planting risks through precise fertilization schemes and reduces fertilizer costs for farmers through mechanisms such as waste-to-fertilizer swaps, thereby promoting the large-scale implementation of the crop-livestock cycle model. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a closed-loop operation and management method for biogas fertilizer based on the resource utilization of organic waste.

[0019] Figure 2 A flowchart for constructing a causal inference model and identifying root causes.

[0020] Figure 3 A flowchart for task order generation and multi-agent game.

[0021] Figure 4 A flowchart for execution and feedback optimization.

[0022] Figure 5 This is a surface plot showing the effects of fertilizer application and temperature on crop yield and soil quality.

[0023] Figure 6 A graph showing the changing trends of crop growth under different service plans. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-4 This is one embodiment of the present invention, which provides a closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste, including the following steps: S1. Obtain customized biogas fertilizer application plans and collect data from the entire chain and real-time monitoring data.

[0028] S1.1: Customized biogas fertilizer application plans are generated based on soil testing results and crop growth needs of farmers' land.

[0029] Specifically, the process involves obtaining soil testing results and crop growth requirements from farmers' land, comparing the nutrient content data from the soil testing results with the target nutrient data from the crop growth requirements to determine nutrient difference data, querying a pre-set biogas fertilizer nutrient characteristic database based on the nutrient difference data, filtering out candidate biogas fertilizer types that match the nutrient difference data, calculating the matching degree score for each candidate biogas fertilizer type based on the application history effect records of the candidate biogas fertilizer types, selecting the candidate biogas fertilizer type with the highest matching degree score as the target biogas fertilizer type, determining the total application amount of the target biogas fertilizer type based on the unit nutrient content and nutrient difference data of the target biogas fertilizer type, and allocating the total application amount to specific plots based on the area data and topographic distribution characteristics of farmers' land, forming a customized biogas fertilizer application plan that includes plot location, target biogas fertilizer type, and total application amount.

[0030] It should be noted that the pre-set biogas fertilizer nutrient characteristic database refers to a database that pre-stores unit nutrient content data and historical average application utilization rate data corresponding to various types of biogas fertilizer.

[0031] The matching score for each candidate biogas fertilizer type is calculated using the following expression: ; In the formula, Indicates the first Matching score for each candidate biogas fertilizer type; Indicates the index of candidate biogas fertilizer types; This indicates the total number of different types of nutrients. Indicates the index of nutrient element types; Indicates the first Target nutrient data for each nutrient element in crop growth requirements; Indicates the first Nutrient content data of various nutrients in soil testing results; Indicates the first Among the candidate biogas fertilizer types, the first... The nutrient content per unit of a nutrient element; This indicates the historical average application and utilization rate of the same type of biogas fertilizer in this region.

[0032] S1.2: The full-chain data includes the source, type, collection amount and time data of organic waste, the temperature, pH and material residence time process parameters during anaerobic fermentation, the output of biogas fertilizer, nitrogen, phosphorus and potassium content and decomposition quality index data, and the application amount, application method, application time and location data of biogas fertilizer. It should be noted that the source of organic waste refers to the specific place or unit that generates organic waste, such as farmers, farms, food processing plants or catering establishments; The type of organic waste refers to the specific material classification of organic waste, such as livestock and poultry manure, crop straw, kitchen waste or fruit and vegetable residues; Organic waste collection volume and time data refers to the actual weight of organic waste collected from various sources and the corresponding recording time at a specific point in time or time period. The process parameters of temperature, pH and material residence time in anaerobic fermentation refer to the temperature, pH and material residence time of the reaction system monitored in real time during the operation of the anaerobic fermentation reactor. The data on biogas fertilizer output, nitrogen, phosphorus and potassium content and composting quality indicators refer to the total weight of biogas fertilizer produced after the anaerobic fermentation process, the specific concentration values ​​of the three nutrients nitrogen, phosphorus and potassium, and the composting test values ​​that characterize the degree of stabilization of biogas fertilizer. The data on the amount, method, time and location of biogas fertilizer application refer to the weight of biogas fertilizer applied to the land in actual agricultural production, the specific application method, the date and time of the application, and the geographical coordinates of the application location.

[0033] S1.3: Real-time monitoring data includes soil moisture, pH value and key nutrient content monitoring data, as well as crop growth status monitoring data.

[0034] It should be noted that the soil moisture, pH value and key nutrient content monitoring data refer to the percentage of moisture, pH value and concentration of key nutrients such as nitrogen, phosphorus and potassium in the farmer's land at a specific time, obtained through soil sensors. Crop growth monitoring data refers to specific values ​​of crop height, leaf area index, chlorophyll content, and biomass accumulation at a specific growth stage, obtained through field observation or remote sensing equipment.

[0035] Figure 5 This diagram illustrates the impact of fertilizer application and temperature on crop yield and soil quality. Presented as a 3D surface, different colors represent different crop yield values, and color gradients indicate the range of variation in crop yield and soil quality. The color changes show the trend of crop yield and soil quality from low to high. The purple-to-red gradient area represents lower crop yield and soil quality, while the green-to-yellow gradient area represents higher crop yield and soil quality. The diagram demonstrates a relationship between crop yield and soil quality as fertilizer application and temperature change, illustrating the role of these two factors in optimizing crop growth and improving soil quality in agricultural management.

[0036] S2. Construct a causal inference model by inputting full-chain data, real-time monitoring data, and customized biogas fertilizer application schemes into the causal inference model to identify the root causes of unsatisfactory biogas fertilizer quality; based on the root causes, generate anaerobic fermentation process adjustment strategies and modification suggestions for customized biogas fertilizer application schemes.

[0037] S2.1: A multi-source heterogeneous data fusion layer is built based on feature engineering and data fusion; a causal discovery layer is built based on causal structure learning; and a causal effect estimation layer is built based on counterfactual reasoning and effect estimation. Specifically, the temporal and statistical features of the entire data chain and real-time monitoring data are extracted. These features are then aligned according to timestamps and merged into a unified format full-chain fusion dataset, forming a multi-source heterogeneous data fusion layer. Based on the full-chain fusion dataset in the multi-source heterogeneous data fusion layer, a constrained causal structure learning algorithm is used to search for conditional independence between variables, constructing a causal discovery layer that represents causal relationships. Based on the causal structure generated by the causal discovery layer, a virtual intervention scenario is set up. The difference between the actual observation results and the virtual intervention results is compared to quantify the net impact of specific process parameter adjustments or raw material changes on the quality and application effect of biogas fertilizer. The causal inference results, including root cause localization and effect quantification, are output, forming a causal effect estimation layer.

[0038] It should be noted that the constrained causal structure learning algorithm is an existing method for inferring the causal structure between variables based on statistical independence tests. The algorithm determines whether there is a direct causal connection between variables by calculating the conditional independence between pairs of variables in the full-chain fused dataset. If two variables are independent under the condition of a given set of other variables, it is determined that there is no direct causal edge between them; if they are not independent, it is determined that there is a direct causal edge.

[0039] S2.2: Connect the multi-source heterogeneous data fusion layer, causal discovery layer and causal effect estimation layer hierarchically to construct a causal inference model.

[0040] Specifically, the full-chain fused dataset generated by the multi-source heterogeneous data fusion layer is directly transmitted to the causal discovery layer. The causal discovery layer uses the full-chain fused dataset to execute a constrained causal structure learning algorithm to generate a causal structure graph representing the causal relationship between variables. The causal structure graph and the full-chain fused dataset are then transmitted to the causal effect estimation layer. Based on the causal links determined by the causal structure graph, the causal effect estimation layer performs counterfactual reasoning and effect estimation operations on the full-chain fused dataset, quantifies the net impact of a specific intervention on the outcome, and forms a causal inference result that includes root cause localization and effect quantification. The causal inference result is fed back to the multi-source heterogeneous data fusion layer to update the feature weights of the full-chain fused dataset, completes multi-level closed-loop connections, and constructs a causal inference model with continuous optimization capabilities.

[0041] It should be noted that the pre-training process of the causal inference model is based on collecting historical full-chain data and real-time monitoring data to form a historical full-chain fusion dataset. The constrained causal structure learning algorithm is executed using the historical full-chain fusion dataset to generate an initial causal structure graph. A structural equation network is constructed based on the initial causal structure graph. The mean squared error is set as the loss function to measure the deviation between the predicted effect and the actual observed value. The adaptive moment estimate is set as the optimizer to update the network weights. The learning rate (e.g., 0.01) is set to control the step size of weight adjustment. By iteratively minimizing the value of the loss function, the range of expected effect changes under different process parameter interventions is calibrated to form a pre-trained causal inference model.

[0042] It should be noted that the expected range of change refers to the fluctuation range of the biogas fertilizer quality index or crop growth index predicted by the causal inference model relative to the baseline state under the intervention of specific process parameters.

[0043] S2.3: Standardize and fuse the multi-source heterogeneous data fusion layer with the input of full-chain data, real-time monitoring data and customized biogas fertilizer application schemes to generate a unified spatiotemporal feature tensor; The entire chain of data, real-time monitoring data, and customized biogas fertilizer application plans are aligned with a unified timestamp and geographic coordinates to eliminate differences in temporal granularity and spatial resolution. The aligned entire chain of data and real-time monitoring data are numerically normalized, mapping all physical quantities to a dimensionless range of zero to one. The normalized multi-source data are stacked according to the time series dimension and the spatial distribution dimension to construct a three-dimensional data structure containing time steps, geographic grids, and feature channels, forming a unified spatiotemporal feature tensor.

[0044] S2.4: Input the spatiotemporal feature tensor into the causal discovery layer, analyze the causal dependencies between variables, and construct a causal structure graph; Specifically, the unified spatiotemporal feature tensor is decomposed into a set of time-series variables. A constrained causal structure learning algorithm is used to perform conditional independence tests on the time-series variable set. By comparing the statistical dependence of variable pairs under the given conditions of other variables, edges without direct correlation are eliminated, and variable pairs with statistical dependence are retained to form a skeleton graph. Based on the chronological order and collision rules, the causal flow is identified, and the undirected edges in the skeleton graph are transformed into directed edges, generating a causal structure graph that represents the causal relationship between raw material characteristics, process parameters, biogas quality, and application effects. The causal structure graph clearly shows the complete causal chain from the type of organic waste to the process parameters of temperature, pH, and material residence time in the anaerobic fermentation process, to the biogas output, nitrogen, phosphorus, and potassium content, and decomposition quality indicators, ultimately affecting the monitoring data of crop growth status.

[0045] It should be noted that the collision rule refers to the logical criterion in causal structure learning that when two variables both point to a third variable and the two variables are unconditionally independent, the two variables and the third variable constitute a "collision structure" and the causal arrow points to the intermediate variable.

[0046] S2.5: For each candidate causal variable pointing to the target variable of biogas fertilizer quality in the causal structure diagram, counterfactual reasoning is performed through the causal effect estimation layer to calculate the causal effect value of each candidate causal variable on the target variable; Specifically, for each candidate causal variable pointing to the biogas yield, nitrogen, phosphorus, and potassium content, and decomposition quality index data in the causal structure diagram, a counterfactual scenario without intervention is constructed. The actual observed sources, types, collection amounts, and time data of organic waste, as well as the process parameters of temperature, pH, and material residence time during anaerobic fermentation, are substituted into the counterfactual scenario to simulate the theoretical results when each candidate causal variable remains at the baseline level. The actual observed biogas yield, nitrogen, phosphorus, and potassium content, and decomposition quality index data are compared with the theoretical results under the counterfactual scenario. The difference between the two is determined as the causal effect value of each candidate causal variable on the target variable. The causal effect value quantifies the net impact of changes in a single process parameter or raw material characteristic on biogas quality, directly identifying the root cause of quality fluctuations.

[0047] S2.6: Sort all candidate causal variables according to the causal effect value, and determine the candidate causal variable ranked first as the root cause of the failure of biogas fertilizer quality to meet expectations.

[0048] Specifically, the absolute values ​​of the causal effects of each candidate causal variable on the target variable are extracted. All candidate causal variables are then arranged in descending order of their absolute causal effect values ​​to form a causal effect ranking list. The candidate causal variable that ranks first in the causal effect ranking list is selected as the root cause of the failure of biogas fertilizer output, nitrogen, phosphorus and potassium content, and decomposition quality indicators to meet expectations. The root cause corresponds to the process parameters or raw material characteristics that have the greatest impact on biogas fertilizer quality, directly guiding the formulation of adjustment strategies.

[0049] It should be noted that "not meeting expectations" refers to the pre-set quality threshold range of biogas fertilizer. When the actual measured output, nitrogen, phosphorus and potassium content and decomposition degree of biogas fertilizer are lower than the lower limit of the quality threshold range or exceed the allowable fluctuation range, it is judged as "not meeting expectations". The quality threshold range for biogas slurry is set based on organic fertilizer industry standards, target crop growth requirement data, and historical records of the effects of high-quality biogas slurry application. The specific steps are as follows: First, review the mandatory indicators regarding maturity, seed germination index, and heavy metal limits in organic fertilizer industry standards to determine the basic safety lower limit. Second, extract the optimal nutrient concentration range for the target crop's growth requirements as the core reference range for nitrogen, phosphorus, and potassium content. Third, statistically analyze the biogas slurry quality values ​​corresponding to the highest-yielding batches from historical integrated data across the entire supply chain, calculate the average value, and set upper and lower fluctuation ratios to form the biogas slurry quality threshold range, including the maturity index threshold range, the total nitrogen content threshold range, and the upper limit for lead content. For example, the maturity index threshold range is set to 0.8 to 1, based on the standard requirement of a seed germination index greater than 80% and the historical average of high-yielding batches being 0.9. The total nitrogen content threshold range is set to 2.5% to 3.5%, based on the optimal nitrogen requirement concentration for corn crops and the statistical average of historical high-quality biogas slurry applications. The upper limit for lead content is set to 50 mg / kg, based on the mandatory limits stipulated in organic fertilizer standards.

[0050] S2.7: Based on the root cause, establish a quantitative response relationship between the associated process and adjustable fertilization parameters and multiple objectives such as biogas fertilizer quality, application effect, economy and environment, and define the corresponding constraints. Specifically, the root causes identified as leading to substandard biogas quality are mapped to specific adjustable parameters in the anaerobic fermentation process, such as temperature, pH, and material residence time, or in the data on biogas application rate, application method, application time, and location. Historical effect data stored in the pre-trained causal inference model are used to establish a correspondence table between the range of adjustable parameter changes and biogas output, nitrogen, phosphorus, and potassium content, composting quality indicators, crop growth monitoring data, operating costs, and carbon emissions, forming a quantitative response relationship. The physical limit value of the adjustable parameters is set as the first constraint, the biogas quality threshold range as the second constraint, and the upper limit of operating costs and carbon emission limits as the third constraint. The quantitative response relationship and the three types of constraints together constitute the solution boundary of the multi-objective optimization problem, ensuring that the adjustment strategy is implemented within a safe, economical, and environmentally friendly range.

[0051] S2.8: Based on the quantitative response relationship and corresponding constraints, the optimal combination of process parameters and fertilization parameters is solved through multi-objective optimization and simulation verification, and suggestions for adjusting the anaerobic fermentation process and the customized biogas fertilizer application scheme are generated.

[0052] Specifically, based on the quantitative response relationship and corresponding constraints, all combinations of process parameters such as temperature, pH, and material residence time in the anaerobic fermentation process, along with data on the application rate, method, time, and location of biogas slurry, are traversed within the feasible domain defined by the constraints. The quantitative response relationship is used to predict the biogas slurry yield, nitrogen, phosphorus, and potassium content, composting quality indicators, crop growth monitoring data, operating costs, and carbon emissions for each combination. Combinations that simultaneously meet the biogas slurry quality threshold range, operating cost upper limit, and carbon emission limit are screened. From the screened combinations, the combination with the best crop growth monitoring data and the lowest operating cost is selected as the optimal combination of process parameters and fertilization parameters. An anaerobic fermentation process adjustment strategy is generated based on the optimal combination of process parameters, and a modification suggestion for a customized biogas slurry application plan is generated based on the optimal combination of fertilization parameters.

[0053] It should be noted that the anaerobic fermentation process adjustment strategy refers to the specific operational instructions formulated to optimize the process parameters of temperature, pH and material residence time in the anaerobic fermentation process, which are based on the root cause. These include the target temperature setpoint, the target pH adjuster addition amount and the target material residence time extension. The modification suggestion for the customized biogas fertilizer application plan refers to the updated instructions on the application amount, application method, application time and location data of the original customized biogas fertilizer application plan based on the optimal combination of fertilization parameters. It includes the adjusted application weight per plot, recommended application machinery type, latest application date and precise geographical coordinates.

[0054] S3. Transform the new requirements for the collection and distribution of organic waste materials in the proposed amendments into pending task orders that specify dynamic incentive clauses.

[0055] S3.1: Analyze the new requirements for the collection and distribution of organic waste materials in the revised proposal and break them down into several basic tasks; Specifically, data on the adjusted application weight per plot and the extended residence time of the target material are extracted from the revised suggestions of the customized biogas slurry application plan. This data is then used to deduce the type and total weight of organic waste required for replenishment. Based on the required type and total weight of organic waste, and considering the source distribution of the organic waste and the geographical location of the treatment station, transportation routes from each source site to the treatment station are planned. These routes are divided into several independent segments according to geographical regions and time windows, with each segment corresponding to a basic task. Each basic task includes a clearly defined start and end point, a specified type of organic waste, a required collection weight, and a specified completion deadline.

[0056] S3.2: For each basic task, obtain the state entropy of the real-time task market and the matching degree of available service providers, and calculate the dynamic incentive amount for each basic task. The state entropy of the real-time task market is obtained by the following expression: ; In the formula, It represents the state entropy of the real-time task market, characterizing the degree of disorder in the distribution of tasks in the market; This indicates the total number of types of basic tasks that have not been assigned at the current moment; Indicates the type index of the basic task; Indicates the first The proportion of basic tasks to the total number of unassigned basic tasks; express The natural logarithm of .

[0057] To obtain the matching degree of available service providers, the expression is: ; In the formula, Indicates the first Each available service provider is rated on its suitability for a specific basic task; Indicates the starting coordinates of the basic task; Indicates the first The current geographic coordinates of the available service providers; This indicates the historical average speed of vehicles collected in this area; This indicates the specified time limit for completing the basic task; Indicates the first The rated load capacity of the vehicles of each available service provider; This indicates the required weight to collect for the basic task.

[0058] ; In the formula, Indicates the first Dynamic incentive limits for each basic task; Indicates the first Standard benchmark pricing for each basic task; The natural logarithm of the total number of types of unassigned basic tasks at the current moment; Indicates the first The available service provider is targeting the first... Matching score for each basic task; This indicates that among all available service providers, the one targeting the first... The highest matching score for the basic task.

[0059] S3.3: Bind each basic task to its corresponding dynamic incentive quota and encapsulate it as a task order to be executed.

[0060] Specifically, the specified organic waste type, required collection weight, start location, end location, specified completion time limit and initial priority are extracted from each basic task, and associated with the corresponding dynamic incentive amount. The bound data is then packaged in a unified data format to form a task order to be executed, which includes task identifier, material attributes, spatiotemporal constraints, priority level and dynamic incentive amount value.

[0061] S4. The collection vehicles, fertilization service providers and processing stations are set as intelligent agents. Through multi-agent game, each intelligent agent bids and matches based on the task orders to be executed, forming a real-time service plan.

[0062] S4.1: Initialize the collection vehicle, fertilizer service provider, and processing station as intelligent agents, and broadcast the task orders to be executed to each intelligent agent; Specifically, the collection vehicles, fertilizer service providers, and processing stations will be registered as intelligent agents with independent decision-making capabilities. Each intelligent agent will be assigned a unique identity and configured with its current geographical location, load capacity, or service capacity status information. A first reward rule will be set for the collection vehicle intelligent agent, a second reward rule will be set for the fertilizer service provider intelligent agent, and a third reward rule will be set for the processing station intelligent agent. A communication network covering all intelligent agents will be established to broadcast the pending task orders, which include task identifiers, material attributes, spatiotemporal constraints, priority levels, and dynamic incentive amounts, to each intelligent agent.

[0063] The first reward rule expression for collecting vehicle intelligent agents is: ; In the formula, This represents the net revenue generated from collecting data on vehicle intelligence agents. This represents the estimated total travel path length collected by the vehicle's intelligent agent. This represents the fuel consumption cost per unit route. This represents the positive time difference between the expected arrival time and the specified completion time. This indicates the penalty standard for delays per unit of time.

[0064] The expression for the second reward rule of the fertilizer service provider's intelligent agent is: ; In the formula, This represents the overall utility value of the fertilizer service provider's intelligent agent; This indicates the expected improvement in crop growth monitoring data; This represents the converted value of long-term credit points corresponding to a unit increase in the growth index. This indicates the expected operation path length of the fertilizer service provider's intelligent agent; This represents the cost of manpower and machinery wear and tear per unit path.

[0065] The expression for the third reward rule of the processing station agent is: ; In the formula, This represents the operational optimization values ​​of the processing station's intelligent agent; The score indicates the degree of matching between the type of organic waste to be received and the required parameters of the anaerobic fermentation process. This indicates the basic value of raw material processing corresponding to a unit degree of matching; This indicates the current backlog of goods in the processing station; This represents the penalty cost coefficient per unit of warehouse backlog.

[0066] It should be noted that the specific steps for setting the penalty cost coefficient per unit of storage backlog are as follows: calculate the daily fixed maintenance cost and unit volume occupation cost of the processing station; determine the time decay weight based on the tolerance limit of the raw material storage time according to the process parameters of temperature, pH and material residence time in the anaerobic fermentation process; and use the product of the daily average cost and the time decay weight as the penalty cost coefficient per unit of storage backlog; the example value is 15; the basis for the value is to cover the additional energy consumption, space occupation opportunity cost and potential fermentation gas production loss caused by the decline in raw material quality due to storage backlog.

[0067] S4.2: Each agent reports its own state and conditional commitments to the coordinator, which summarizes and constructs a task conflict relationship graph, and initiates a multi-round iterative game based on the task conflict relationship graph with conditional commitments as the strategy. Specifically, each agent sends its bidding strategy, which includes the bid price and the service commitment time, to the coordinator. The coordinator extracts the geographical location and time window data from all bidding strategies, compares the overlap of spatiotemporal resources occupied by different bidding strategies, and connects bidding strategies with resource conflicts as nodes and edges to construct a task conflict graph. Based on the task conflict graph, the coordinator sends conflict warning information to agents with conflicting connections, triggering multiple rounds of iterative game. Each agent adjusts the bid price or service commitment time in its own bidding strategy according to the conflict warning information, regenerates an updated bidding strategy containing the new bid price and service commitment time, and feeds it back to the coordinator. The coordinator refreshes the task conflict graph according to the updated bidding strategy, repeating the conflict detection and strategy adjustment process until there are no conflicting connections in the task conflict graph.

[0068] S4.3: When the iterative game converges, the final determined allocation relationship between the agent and the task order to be executed is used as the real-time service solution.

[0069] Specifically, when there are no more conflicting edges in the task conflict relationship graph, the convergence of the multi-round iterative game is determined, and the correspondence between the unique identity identifier and the task identifier contained in the updated bidding strategies fed back by all agents at this time is extracted; each unique identity identifier is bound to the corresponding task identifier to form a conflict-free strategy combination that includes a clear allocation relationship between the collection vehicle, the fertilizer service provider and the processing station and the task order to be executed; the bid value and service commitment time in the conflict-free strategy combination are confirmed as the final execution standard, and a real-time service plan containing the specific execution entity, execution object, execution price and execution time limit is generated.

[0070] It should be noted that if the conflict cannot be eliminated after multiple iterations, the service solution with the least conflict will be forcibly selected or the task will be re-released.

[0071] Figure 6 The graph shows the trend of crop growth rate with fertilizer application rate under different service plans (blue and red curves). The red dashed box in the graph represents a magnified window, providing more detailed local data analysis. By magnifying the graph, the fluctuation of crop growth rate and the response under different fertilizer application rates can be clearly seen. The graph shows the range of fertilizer application rates from 12 to 18 kg / mu, and marks the characteristic peaks and the points with the greatest differences between the two curves, so as to better understand the impact of fertilizer application rate on crop growth.

[0072] S5. Implement the real-time service plan and anaerobic fermentation process adjustment strategy. After implementation, verify the achievement of the effect associated with the pending task orders and settle the corresponding dynamic incentive terms.

[0073] S5.1: During and after the implementation of the real-time service plan and anaerobic fermentation process adjustment strategy, collect multi-source process data and effect monitoring data to form the original dataset to be verified. Specifically, during the execution of the real-time service plan and the anaerobic fermentation process adjustment strategy, the vehicle-mounted terminal records and collects the actual driving route, actual arrival time, and actual collected weight. The processing station controller records the actual operating values ​​of process parameters such as temperature, pH, and material residence time during the anaerobic fermentation process, as well as the output of biogas slurry, nitrogen, phosphorus, and potassium content, and decomposition quality indicators. After execution, soil sensors acquire monitoring data on soil moisture, pH value, and key nutrient content, and field observations acquire monitoring data on crop growth. The actual driving route, actual arrival time, actual collected weight, actual operating values, biogas slurry output, nitrogen, phosphorus, and potassium content, decomposition quality indicators, soil moisture, pH value, key nutrient content monitoring data, and crop growth monitoring data are aligned and merged according to timestamps and geographical coordinates to form a raw dataset to be verified, containing the entire process execution record and the final effect indicators.

[0074] S5.2: Input the original dataset to be verified into the decentralized verification network for anti-interference processing, and calculate the comprehensive verification index of each task order to be executed; Specifically, the original dataset to be verified is distributed to multiple verification nodes in the decentralized verification network. Each verification node cross-compares the actual driving path, actual arrival time, actual collection weight, actual running values, biogas output, nitrogen, phosphorus and potassium content and decomposition quality index data, soil moisture, pH value and key nutrient content monitoring data, and crop growth status monitoring data in the original dataset to be verified, and removes abnormal data that are inconsistent with the records of other nodes to complete the anti-interference processing.

[0075] It should be noted that the pre-training process of the decentralized verification network is based on collecting complete execution records of historical organic waste collection and fertilization services, forming a historical raw dataset containing historical driving routes, historical arrival times, historical collection weights, historical operational values, historical biogas quality data, historical soil monitoring data, and historical crop growth data. Each simulated verification node is assigned an initial voting weight of one and initial anomaly detection logic is set (a rule that marks a data entry as an anomaly when the absolute value of the difference between the recorded value of a specific historical data entry by a simulated verification node and the recorded values ​​of the majority of other simulated verification nodes exceeds a preset tolerance range). Multiple copies of the historical raw dataset are then made and distributed to multiple simulated verification nodes. Each simulated verification node cross-references the same task records in the original historical dataset, identifying and marking historical data entries with numerical deviations or logical conflicts. It then calculates the consistency ratio between the marking results of each simulated verification node and the known real historical data benchmark. When the consistency ratio falls below a preset standard, the voting weight of that simulated verification node is reduced (e.g., by 10%), and the tolerance range for anomaly judgments of that simulated verification node is narrowed (e.g., by 10%). This process of cross-referencing, marking, consistency statistics, and weight and judgment logic adjustment is repeated until the consistency judgment rate of all simulated verification nodes with the known real historical data reaches the preset standard, forming a pre-trained decentralized verification network with anti-interference capabilities.

[0076] It should be noted that the specific steps for setting the preset tolerance range are as follows: calculate the standard deviation of various monitoring data in the known real records, count the maximum measurement error of the sensor equipment used, and take the sum of three times the standard deviation and the maximum measurement error as the tolerance range; the exemplary value ranges are ±5 for driving path deviation, ±0.05 for collected weight deviation, and ±0.5 for temperature deviation; the basis for the value is to cover the natural dispersion of data caused by normal environmental fluctuations and the inherent accuracy limitations of the equipment, so as to ensure that only abnormal data that deviates from the true distribution are marked.

[0077] The preset standard refers to the quantitative indicator set during the pre-training process of the decentralized verification network to determine whether the network has sufficient anti-interference ability. Specifically, it is defined as the minimum percentage (e.g., 95%) of the number of records that reach a consistent judgment result after all simulated verification nodes cross-compare the known real historical records.

[0078] The comprehensive verification index for each pending task order is calculated using the following expression: ; In the formula, Indicates the first The comprehensive verification index of a number of pending task orders; This represents the index of pending task orders; Indicates the first The actual processing time of each pending task order; Indicates the first The specified time limit for each pending task order; Indicates the first The actual number of tasks completed for each pending task order; Indicates the first The specified demand quantity for each pending task order; Indicates the first Actual performance values ​​of the pending task orders; Indicates the first The expected target effect value of each pending task order.

[0079] S5.3: Based on the comprehensive verification index, trigger the smart contract deployed on the blockchain to automatically calculate and execute the payment of dynamic incentive terms according to the preset settlement rules.

[0080] Specifically, the comprehensive verification index of each pending task order is transmitted to a smart contract deployed on the blockchain. The smart contract reads the comprehensive verification index value and compares it with the range in the preset settlement rules. When the comprehensive verification index value is in the high-confidence range, the smart contract locks the full amount of the dynamic incentive for the b-th basic task as the final dynamic incentive amount for execution. When the comprehensive verification index value is in the medium-confidence range, the smart contract reduces the dynamic incentive amount of the b-th basic task according to a preset ratio as the final dynamic incentive amount for execution. When the comprehensive verification index value is below the low-confidence range, the smart contract triggers a freeze command and suspends the settlement operation. Based on the comparison results, the smart contract generates a transfer command and automatically transfers the final dynamic incentive amount to the digital wallet address of the corresponding smart agent through the blockchain network, completing the automatic settlement of the dynamic incentive terms.

[0081] It should be noted that the preset settlement rules refer to the logical judgment clauses pre-written in the smart contracts deployed on the blockchain, which are used to map the comprehensive verification index value to different settlement execution strategies. Specifically, they stipulate the operation instructions for full allocation corresponding to the high confidence interval, proportional reduction corresponding to the medium confidence interval, and freezing corresponding to the low confidence interval and below. The preset ratio refers to a fixed coefficient value that is explicitly defined in the preset settlement rules and is used to reduce the dynamic incentive amount of the b-th basic task when the comprehensive verification index value is in the medium confidence range.

[0082] S6. Collect task execution process data, new monitoring data and incentive settlement results, and correlate them with the data in the whole chain to iteratively optimize the causal inference model and multi-agent game.

[0083] S6.1: The collected task execution process data, new monitoring data and incentive settlement results are correlated and structured to form feedback experience tuples. The value of each feedback experience in the feedback experience tuples is evaluated and ranked to construct a course-based learning sequence. Specifically, task execution process data, new monitoring data, and incentive settlement results are aligned and bound according to task identifiers and spatiotemporal coordinates to form feedback experience tuples containing triples of state, action, and result. The comprehensive verification index value and the final dynamic incentive amount value of the final execution are extracted from the feedback experience tuples, and the product of the two is used as the experience value score characterizing the quality of a single task execution. Based on the experience value score, all feedback experience tuples are arranged from high to low to generate an experience value ranking list. The feedback experience tuples in the experience value ranking list are divided into several learning stages in order of increasing difficulty and decreasing value. Each learning stage contains a fixed number of feedback experience tuples, forming a course-based learning sequence from easy to difficult and from good to bad.

[0084] S6.2: Based on the curriculum-based learning sequence, it generates parameter update directions for the policy network of causal inference model and multi-agent game, and dynamically allocates differentiated adaptive learning rates. Specifically, based on the curricular learning sequence, feedback experience tuples are extracted sequentially from each learning stage. The state and action data in the feedback experience tuples are fed into a causal inference model consisting of three fully connected layers and a policy network for multi-agent games consisting of a bidirectional long short-term memory network and a policy head for forward inference. The difference between the inference results and the actual results in the feedback experience tuples is compared. The inference error is quantified using the mean squared error loss function and the gradient is backpropagated to determine the weight adjustment direction of the causal inference model and the policy network for multi-agent games. Based on the order of the feedback experience tuples in the curricular learning sequence and their experience value scores, the largest adaptive learning rate is assigned to high-value feedback experience tuples in the early stages, and the smallest adaptive learning rate is assigned to low-value feedback experience tuples in the advanced stages.

[0085] For example, when the experience value score is higher than 0.8 and the learning stage is in the first 30%, an adaptive learning rate of 0.2 is assigned to the feedback experience tuples that are high value and in the early stage; when the experience value score is lower than 0.5 or the learning stage is in the last 50%, an adaptive learning rate of 0.01 is assigned to the feedback experience tuples that are low value or in the advanced stage; and an intermediate adaptive learning rate of 0.05 is assigned to the feedback experience tuples that fall between the two cases.

[0086] S6.3: Use parameter update direction and adaptive learning rate to incrementally update the policy network of causal inference model and multi-agent game.

[0087] Specifically, the determined weight adjustment direction and the corresponding adaptive learning rate are transmitted to the policy network of the causal inference model and the multi-agent game. Based on the step size determined by the adaptive learning rate, the existing connection weights of the policy network of the causal inference model and the multi-agent game are fine-tuned along the weight adjustment direction to complete the incremental update.

[0088] In summary, this invention achieves in-depth and accurate diagnosis of problems where the quality or effect of biogas fertilizer does not meet expectations by constructing a causal inference model and inputting full-chain data, real-time monitoring data, and customized biogas fertilizer application schemes. It identifies the root cause variables that lead to the problems, providing a reliable basis for generating precise process adjustments and fertilization correction suggestions. This effectively guides the resource utilization of organic waste and precise fertilization, laying a decision-making foundation for increasing soil organic matter content, improving soil structure, and achieving synergistic effects of carbon reduction, pollution reduction, and greening in ecological governance of saline-alkali land. It also reduces crop planting risks through precise fertilization schemes and reduces fertilizer costs for farmers through mechanisms such as waste-to-fertilizer swaps, thereby promoting the large-scale implementation of the crop-livestock cycle model.

[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste, characterized in that, include: Obtain customized biogas fertilizer application plans and collect data from the entire chain and real-time monitoring data; A causal inference model is constructed, and the whole chain data, real-time monitoring data and customized biogas fertilizer application plan are input into the causal inference model to identify the root causes of the failure of biogas fertilizer quality to meet expectations; based on the root causes, the anaerobic fermentation process adjustment strategy and the modification suggestions of the customized biogas fertilizer application plan are generated. The new requirements for the collection and distribution of organic waste materials in the revised proposal will be transformed into task orders to be executed with dynamic incentive clauses. The collection vehicles, fertilizer service providers and processing stations are set as intelligent agents. Through multi-agent game, each intelligent agent bids and matches based on the task orders to be executed, forming a real-time service plan. Implement real-time service plans and anaerobic fermentation process adjustment strategies. After implementation, verify the achievement of the effects associated with the pending task orders and settle the corresponding dynamic incentive terms. Collect task execution process data, new monitoring data, and incentive settlement results, and correlate them with the entire chain of data to iteratively optimize the causal inference model and multi-agent game.

2. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 1, characterized in that, The customized biogas fertilizer application plan is generated based on soil testing results and crop growth needs of the farmers' land.

3. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 1, characterized in that, The full-chain data includes data on the source, type, collection volume and time of organic waste, process parameters such as temperature, pH and material residence time during anaerobic fermentation, biogas output, nitrogen, phosphorus and potassium content and decomposition quality indicators, as well as data on the application amount, application method, application time and location of biogas. The real-time monitoring data includes soil moisture, pH value and key nutrient content monitoring data, as well as crop growth status monitoring data.

4. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 1, characterized in that, The steps for constructing the causal inference model are as follows: A multi-source heterogeneous data fusion layer is built based on feature engineering and data fusion; a causal discovery layer is built based on causal structure learning; and a causal effect estimation layer is built based on counterfactual reasoning and effect estimation. By hierarchically connecting the multi-source heterogeneous data fusion layer, the causal discovery layer, and the causal effect estimation layer, a causal inference model is constructed.

5. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in any one of claims 2-4, characterized in that, The steps for identifying the root cause of unsatisfactory biogas fertilizer quality are as follows: The multi-source heterogeneous data fusion layer standardizes and fuses the full-chain data, real-time monitoring data, and customized biogas fertilizer application schemes to generate a unified spatiotemporal feature tensor. The spatiotemporal feature tensor is input into the causal discovery layer to analyze the causal dependencies between variables and construct a causal structure graph. For each candidate causal variable pointing to the target variable of biogas fertilizer quality in the causal structure diagram, counterfactual reasoning is performed through the causal effect estimation layer to calculate the causal effect value of each candidate causal variable on the target variable; All candidate causal variables were ranked according to their causal effect values, and the candidate causal variable ranked first was determined to be the root cause of the failure of biogas fertilizer quality to meet expectations.

6. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 5, characterized in that, The proposed adjustments to the anaerobic fermentation process and the suggested modifications to the customized biogas fertilizer application scheme are as follows: Based on the root causes, a quantitative response relationship is established between the associated process and adjustable fertilization parameters and multiple objectives such as biogas fertilizer quality, application effect, economy and environment, and corresponding constraints are defined. Based on the quantitative response relationship and corresponding constraints, the optimal combination of process parameters and fertilization parameters is solved through multi-objective optimization and simulation verification, generating modification suggestions for anaerobic fermentation process adjustment strategies and customized biogas fertilizer application schemes.

7. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 1, characterized in that, The steps to convert the task order into an execution order containing dynamic incentive terms are as follows: The new requirements for the collection and distribution of organic waste materials in the revised proposals are analyzed and broken down into several basic tasks. For each basic task, obtain the state entropy of the real-time task market and the matching degree of available service providers, and calculate the dynamic incentive amount for each basic task. Each basic task is bound to its corresponding dynamic incentive quota and packaged into a task order to be executed.

8. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 1, characterized in that, The steps to form a real-time service solution are as follows: The collection vehicles, fertilizer service providers, and processing stations are initialized as intelligent agents, and the task orders to be executed are broadcast to each intelligent agent. Each agent reports its own state and conditional commitments to the coordinator, which then summarizes and constructs a task conflict relationship graph. Based on the task conflict relationship graph, it initiates a multi-round iterative game with conditional commitments as the strategy. When the iterative game converges, the final determined allocation relationship between the agent and the task orders to be executed is used as a real-time service solution.

9. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 8, characterized in that, The dynamic incentive clauses corresponding to the settlement are implemented through the following steps: During and after the implementation of the real-time service plan and anaerobic fermentation process adjustment strategy, multi-source process data and effect monitoring data are collected to form the original dataset to be verified. The original dataset to be verified is input into the decentralized verification network for anti-interference processing, and the comprehensive verification index of each task order to be executed is calculated. Based on the comprehensive verification index, the smart contract deployed on the blockchain is triggered to automatically calculate and execute the payment of the dynamic incentive terms according to the preset settlement rules.

10. The closed-loop operation and management method for biogas slurry based on the resource utilization of organic waste as described in claim 1, characterized in that, The iterative optimization of the causal inference model and the multi-agent game involve the following steps: The collected task execution process data, new monitoring data, and incentive settlement results are correlated and structured to form feedback experience tuples. Each feedback experience in the feedback experience tuples is then evaluated and ranked to construct a course-based learning sequence. Based on the curriculum-based learning sequence, parameter update directions are generated for the policy network of causal inference models and multi-agent games, while dynamically allocating differentiated adaptive learning rates. The policy network of the causal inference model and multi-agent game is incrementally updated using parameter update direction and adaptive learning rate.

Images