A digital-twin-based simulation and decision assistance method for the whole process of pasture production

Abstract

The present application relates to the field of smart agriculture technology, and more particularly to a pasture production whole-process simulation and decision assistance method based on digital twinning. The method comprises collecting pasture environment parameters, animal physiological behavior parameters and production performance parameters and preprocessing; constructing a physiological-environment coupled digital twinning model, including an environment field simulation submodel, an animal individual physiological simulation submodel and a bidirectional coupling mechanism therebetween; using real-time sensor data to calibrate the twinning model online; when the production performance index is abnormal, calling a pre-constructed structural causal diagram, locating the root cause through counterfactual reasoning, outputting the contribution degree ranking of each upstream factor and the complete causal transmission path; and simulating and verifying the intervention scheme in the twinning model according to the located root cause and outputting decision suggestions. The present application realizes closed-loop interactive simulation of the environment field and animal physiology, and solves the defect that the abnormality cannot be traced in the prior art.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture technology, and more specifically, to a method for simulating and supporting the entire process of ranch production based on digital twins. Background Technology

[0002] Digital twin technology, by creating digital mirror images of physical objects, enables the mapping and interaction between the physical and information worlds, and is considered one of the key technologies driving the development of smart agriculture. In recent years, digital twin technology has seen initial exploration and application in the field of ranching and animal husbandry, with related technical solutions mainly including the following categories:

[0003] Category 1: Digital Twin Modeling of Aquaculture Environments. Existing technologies integrate multi-source sensor data with 3D modeling techniques to construct 3D environmental field models of aquaculture farms, simulating and visualizing the spatiotemporal distribution of environmental parameters such as temperature, humidity, wind speed, and pollutant concentrations; or they use computational fluid dynamics methods to numerically simulate the flow field, temperature field, and concentration field within aquaculture buildings, obtaining refined environmental distribution data. While this type of method achieves the digitization and visualization of the aquaculture environment, it lacks a physiological-level correlation mechanism between the environmental model and individual animals; how changes in environmental parameters affect the animals themselves is not incorporated into the modeling system.

[0004] The second category: Animal individual behavior and physiological modeling. Existing technologies simulate individual animal behavior and group activity patterns by constructing animal agent models; or calculate core body temperature changes in animals based on physiological energy equations. Although this type of method incorporates some physiological mechanisms, the relationship between the environmental field and the animal physiological model is one-way data transfer or completely independent operation. The impact of environmental changes on animal physiology is described through empirical parameters rather than mechanistic equations, lacking a true coupling relationship.

[0005] The third category: Digital twin-based aquaculture systems and control. Existing technologies construct physical farms into online 3D visualization platforms to achieve remote monitoring of farm conditions; or they use digital twin models combined with reinforcement learning methods to achieve automatic control of the aquaculture environment. This type of approach focuses on system integration and control decision-making, but the control logic is mostly based on data-driven black-box models, lacking explicit expression of the complete causal link between environment, physiology, and production. Summary of the Invention

[0006] The purpose of this invention is to provide a method for simulating and supporting the entire process of ranch production based on digital twins, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention aims to provide a method for simulating and supporting the entire production process of a ranch based on digital twins, including: a data acquisition step: collecting environmental parameters, individual animal physiological and behavioral parameters, and production performance parameters through a sensor network deployed in various areas of the ranch, and cleaning, interpolating, and aligning the collected raw data in time.

[0008] Twin model construction steps: Construct a physiological-environment coupled digital twin model. This model includes an environmental field simulation sub-model for simulating the distribution of environmental parameters across the entire pasture, an individual animal physiological simulation sub-model for calculating the core body temperature and stress response of each animal, and a two-way coupling mechanism that enables the downward transmission of environmental parameters and the upward feedback of metabolic heat production and moisture dissipation between the two, thereby forming a closed-loop interactive simulation between the environmental field and animal physiology.

[0009] Online calibration steps: Using real-time sensor data, the overall state of the twin model is updated online through an ensemble data assimilation method, while the uncertain parameters in the model are continuously learned and adjusted through a recursive probability estimation method.

[0010] Anomaly tracing steps: When abnormal fluctuations in production performance indicators are detected, a pre-constructed causal graph of the ranch production structure is invoked. This causal graph, in the form of a directed acyclic graph, defines the causal direction and influence intensity between environmental factors, management measures, physiological states, health levels, and production indicators. For all upstream factors in the causal graph that may affect the abnormal target variable, counterfactual reasoning simulations are performed one by one. That is, assuming that the factor is at a normal level and other conditions remain unchanged, the coupled twin model is rerun to observe whether the anomaly still exists. Based on this, the actual contribution ratio of each upstream factor to the anomaly is calculated and sorted from largest to smallest contribution. At the same time, the complete causal transmission path from the root cause factor to the abnormal target variable is output.

[0011] Decision support steps: Based on the identified root causes, different intervention schemes are set in the coupled twin model and simulations are performed respectively. The comprehensive impact of each scheme on the overall environment, animal physiology and production performance is compared, and the optimal intervention scheme is selected for output.

[0012] As a further improvement to this technical solution, the environmental parameters in the data acquisition step include air temperature, relative humidity, wind speed, light intensity, ammonia concentration, and carbon dioxide concentration; the animal individual physiological and behavioral parameters include core body temperature, activity level, feeding time, and rumination time; and the production performance parameters include milk production, feed intake, and daily weight gain.

[0013] As a further improvement to this technical solution, the twin model construction steps specifically include: environmental field simulation sub-model construction: discretizing the pasture building space into a three-dimensional grid, and solving the temperature, wind speed and pollutant concentration values ​​of each grid node based on the conservation of mass, momentum, energy and the mass diffusion equation to obtain the distribution of environmental parameters of the entire pasture.

[0014] Animal individual physiological simulation sub-model construction: A heat balance calculation relationship is established for each animal. The core body temperature changes over time based on the animal's body weight, tissue heat capacity, metabolic heat production, and sensible heat, latent heat, and conduction heat dissipation pathways. At the same time, a stress response calculation relationship is established. The degree of deviation of each environmental parameter from the thermoneutral zone is transformed by a nonlinear threshold function and then weighted and summed to obtain the comprehensive stress intensity. This stress intensity is used to inhibit normal feed intake and accelerate the decline of production performance.

[0015] Two-way coupling mechanism construction: Within each simulation time step, the environmental parameters of each region output by the environmental field simulation sub-model are used as the input boundary conditions of the physiological simulation sub-model. The metabolic heat production and moisture dissipation of individual animals calculated by the physiological simulation sub-model are used as source terms to feed back into the energy equation and material equation of the environmental field simulation sub-model. This process is repeated iteratively until the change in environmental state and physiological state of all individuals between two adjacent iterations is less than the preset tolerance, at which point the calculation of the current step ends.

[0016] As a further improvement to this technical solution, the triggering condition for the online calibration step is: when the absolute value of the deviation between the model prediction value and the real-time sensor measured value of any key variable exceeds a preset threshold and the out-of-tolerance state continues for more than a preset duration, the calibration process is automatically started.

[0017] As a further improvement to this technical solution, the nodes in the causal graph include at least ambient temperature, ambient humidity, wind speed, ammonia concentration, feed intake, ventilation volume, core body temperature, stress intensity, feed intake, activity level, health score, disease events, milk production, and daily weight gain; directed edges point from cause nodes to result nodes. The causal strength weight of each edge is determined through historical data learning or expert experience assignment, with a value range of 0 to 1.

[0018] As a further improvement to this technical solution, the counterfactual reasoning in the anomaly tracing step is specifically as follows: extract all upstream factors that may affect the anomaly target variable from the causal graph as a set of candidate root causes; for each candidate root cause variable, replace the actual observed value of the variable at the time of the anomaly with the expected value of the variable under normal conditions, keep all other variables unchanged, and rerun the coupled twin model to perform a backtracking simulation from before the anomaly occurred to after the anomaly occurred; and then calculate the causal contribution of the candidate root cause.

[0019] As a further improvement to this technical solution, the output method of the complete causal transmission path is as follows: starting from the root cause node, each intermediate variable passed along the path is listed sequentially along the causal graph direction, and the actual change direction and magnitude of each intermediate variable during the anomaly are marked, until the anomaly target variable is reached.

[0020] As a further improvement to this technical solution, the intervention schemes in the decision support steps include one or more combinations of adjusting environmental control equipment parameters, changing feeding strategies, adjusting ventilation modes, and changing milking or regrouping plans; the simulation results of each intervention scheme are presented in the form of prediction curves to show the environmental recovery time, the speed at which animal physiological indicators return to normal, and the final degree of recovery of production performance.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] This invention constructs an animal individual physiological simulation sub-model containing a heat balance equation and a stress response function, and connects it to an environmental field simulation sub-model through a two-way coupling mechanism. Environmental parameters serve as the input boundary of the physiological model, while animal metabolic heat production and moisture dissipation serve as source terms fed back to the energy and material equations of the environmental field. The model is iteratively solved alternately within each time step until convergence. Compared to existing technologies where the environmental and physiological models are independent or only involve unidirectional data transfer, the coupled model of this invention can realistically reflect the entire process of the animal's physiological response and its reverse influence on the environment when environmental parameters change drastically, significantly improving prediction accuracy under extreme conditions.

[0023] This invention innovatively introduces structural causal graphs and counterfactual reasoning into the field of digital twins for ranches. When production anomalies occur, the system does not search for "variables related to the anomaly" at the data level, but instead verifies one by one through counterfactual simulation whether the anomaly would still occur if a certain upstream factor had not deviated from the normal level, thereby truly distinguishing between causal relationships and statistical correlations. Simultaneously, this invention outputs the complete causal transmission path from the root cause to the target variable of the anomaly and the contribution ranking of each node, enabling ranch managers to accurately locate the source of the problem, understand the mechanism of the anomaly, directly guide targeted interventions, and significantly reduce trial-and-error costs. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0025] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example: Please refer to Figure 1 As shown, a method for simulating and supporting the entire production process of a ranch based on digital twins is provided, including: data acquisition steps: collecting environmental parameters, individual animal physiological and behavioral parameters, and production performance parameters through a sensor network deployed in various areas of the ranch, and cleaning, interpolating, and aligning the collected raw data in time.

[0028] In one specific embodiment, the environmental parameters in the data acquisition step include air temperature, relative humidity, wind speed, light intensity, ammonia concentration, and carbon dioxide concentration; the individual animal physiological and behavioral parameters include core body temperature, activity level, feeding time, and rumination time; and the production performance parameters include milk production, feed intake, and daily weight gain.

[0029] Twin model construction steps: Construct a physiological-environment coupled digital twin model. This model includes an environmental field simulation sub-model for simulating the distribution of environmental parameters across the entire pasture, an individual animal physiological simulation sub-model for calculating the core body temperature and stress response of each animal, and a two-way coupling mechanism that enables the downward transmission of environmental parameters and the upward feedback of metabolic heat production and moisture dissipation between the two, thereby forming a closed-loop interactive simulation between the environmental field and animal physiology.

[0030] In one specific embodiment, the twin model construction steps specifically include: environmental field simulation sub-model construction: discretizing the ranch building space into a three-dimensional grid, and solving the temperature, wind speed and pollutant concentration values ​​of each grid node based on the conservation of mass, momentum, energy and the mass diffusion equation to obtain the distribution of environmental parameters of the entire ranch area.

[0031] Temperature field Velocity field and pollutant concentration field Solve using the following equations respectively:

[0032]

[0033]

[0034]

[0035] in, Where is the thermal diffusivity, air density, For isobaric specific heat, For kinematic viscosity, The pollutant diffusion coefficient, For heat source items, For pollutant source items, This is the acceleration due to gravity.

[0036] Animal individual physiological simulation sub-model construction: A heat balance calculation relationship is established for each animal. The core body temperature changes over time based on the animal's body weight, tissue heat capacity, metabolic heat production, and sensible heat, latent heat, and conduction heat dissipation pathways. At the same time, a stress response calculation relationship is established. The degree of deviation of each environmental parameter from the thermoneutral zone is transformed by a nonlinear threshold function and then weighted and summed to obtain the comprehensive stress intensity. This stress intensity is used to inhibit normal feed intake and accelerate the decline of production performance.

[0037] The nonlinear threshold function adopts an S-shaped curve form. When the environmental parameters are within the animal's thermoneutral zone, the output value of the function approaches zero. When the environmental parameters exceed the upper or lower limit of the thermoneutral zone, the output value increases nonlinearly with the degree of deviation until it approaches saturation.

[0038] The individual animal physiological sub-models include a heat balance model and a stress-response model;

[0039] In the aforementioned thermal equilibrium model, the change in the animal's core body temperature is described by the following formula:

[0040]

[0041] in, For the animal's weight, The specific heat of animal tissue, Core body temperature For metabolic heat production, For sensible heat dissipation, To dissipate latent heat, For conduction and heat dissipation;

[0042] In the stress-response model, the stress intensity index Calculated by the following formula:

[0043]

[0044] in, For the first One environmental parameter, This is a single-factor stress response function. These are the weighting coefficients.

[0045] The stress intensity The following formula affects feed intake and production performance:

[0046]

[0047]

[0048] in, This represents the actual amount of food consumed. This is the normal amount of food intake. The stress-feed intake inhibition coefficient. For production performance indicators, This is the feed conversion efficiency coefficient. This represents the stress damage coefficient.

[0049] The single-factor stress response function The threshold effect is described using a sigmoid function:

[0050]

[0051]

[0052] in, The lower critical temperature for animals. The high critical temperature for animals, This represents the sensitivity coefficient.

[0053] Two-way coupling mechanism construction: Within each simulation time step, the environmental parameters of each region output by the environmental field simulation sub-model are used as the input boundary conditions of the physiological simulation sub-model. The metabolic heat production and moisture dissipation of individual animals calculated by the physiological simulation sub-model are used as source terms to feed back into the energy equation and material equation of the environmental field simulation sub-model. This process is repeated iteratively until the change in environmental state and physiological state of all individuals between two adjacent iterations is less than the preset tolerance, at which point the calculation of the current step ends.

[0054] Forward and reverse coupling are solved iteratively at each time step using the following formula until convergence:

[0055]

[0056]

[0057] in, Let the environmental field state vector be... Let be the physiological state vector of all individual animals. The number of iterations. For solving the environmental field sub-model, The operator for solving the physiological sub-model is convergence based on the condition that the relative change in the state vector between two adjacent iterations is less than a preset threshold.

[0058] Online calibration steps: Using real-time sensor data, the overall state of the twin model is updated online through an ensemble data assimilation method, while the uncertain parameters in the model are continuously learned and adjusted through a recursive probability estimation method.

[0059] In one specific embodiment, the online calibration step is triggered when the absolute value of the deviation between the model prediction value and the real-time sensor measurement value of any key variable exceeds a preset threshold and the out-of-tolerance state continues for more than a preset duration, the calibration process is automatically started.

[0060] In the online calibration step, the ensemble data assimilation method uses an ensemble Kalman filter algorithm to weight and fuse real-time observation data with model prediction values ​​to update the model's state variables; the recursive probability estimation method uses a sequential Monte Carlo method to recursively update the posterior probability distribution of the model's unknown parameters.

[0061] Anomaly tracing steps: When abnormal fluctuations in production performance indicators are detected, a pre-constructed causal graph of the ranch production structure is invoked. This causal graph, in the form of a directed acyclic graph, defines the causal direction and influence intensity between environmental factors, management measures, physiological states, health levels, and production indicators. For all upstream factors in the causal graph that may affect the abnormal target variable, counterfactual reasoning simulations are performed one by one. That is, assuming that the factor is at a normal level and other conditions remain unchanged, the coupled twin model is rerun to observe whether the anomaly still exists. Based on this, the actual contribution ratio of each upstream factor to the anomaly is calculated and sorted from largest to smallest contribution. At the same time, the complete causal transmission path from the root cause factor to the abnormal target variable is output.

[0062] In one specific embodiment, the nodes in the causal graph include at least ambient temperature, ambient humidity, wind speed, ammonia concentration, feed intake, ventilation volume, core body temperature, stress intensity, feed intake, activity level, health score, disease events, milk production, and daily weight gain; directed edges point from cause nodes to result nodes. The causal strength weight of each edge is determined through historical data learning or expert experience assignment, with a value ranging from 0 to 1.

[0063] In one specific embodiment, the counterfactual reasoning in the anomaly tracing step is as follows: extract all upstream factors that may affect the target variable of the anomaly from the causal graph as a set of candidate root causes; for each candidate root cause variable, replace the actual observed value of the variable at the time of the anomaly with the expected value of the variable under normal conditions, keep all other variables unchanged, and rerun the coupled twin model to perform a backtracking simulation from before the anomaly occurred to after the anomaly occurred; and then calculate the causal contribution of the candidate root cause.

[0064]

[0065] in, These are actual observed values. This is a normal predicted value. This is a counterfactual result.

[0066] In one specific embodiment, the complete causal transmission path is output as follows: starting from the root cause node, each intermediate variable passed along the path is listed sequentially along the causal graph direction, and the actual change direction and magnitude of each intermediate variable during the anomaly are marked, until the anomaly target variable is reached.

[0067] Decision support steps: Based on the identified root causes, different intervention schemes are set in the coupled twin model and simulations are performed respectively. The comprehensive impact of each scheme on the overall environment, animal physiology and production performance is compared, and the optimal intervention scheme is selected for output.

[0068] In one specific embodiment, the intervention plan in the decision support step includes one or more combinations of adjusting environmental control equipment parameters, changing feeding strategies, adjusting ventilation modes, and changing milking or regrouping plans; the simulation results of each intervention plan are presented in the form of prediction curves to show the environmental recovery time, the speed at which animal physiological indicators return to normal, and the final degree of recovery of production performance.

[0069] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for simulating and supporting the entire production process of a ranch based on digital twins, characterized in that, include: Data acquisition steps: Collect environmental parameters, individual animal physiological and behavioral parameters, and production performance parameters, and clean, interpolate, and align the collected raw data in time series. Twin model construction steps: Construct a physiological-environment coupled digital twin model, which includes an environmental field simulation sub-model for simulating the distribution of environmental parameters across the entire pasture, an individual animal physiological simulation sub-model for calculating the core body temperature and stress response of each animal, and a two-way coupling mechanism to realize the downward transmission of environmental parameters and the upward feedback of metabolic heat production and moisture dissipation between the two. Online calibration steps: The overall state of the twin model is updated online using real-time sensor data, while the uncertain parameters in the model are continuously learned and adjusted. Anomaly tracing steps: When abnormal fluctuations in production performance indicators are detected, the pre-built cause-effect diagram of the ranch production structure is invoked. Counterfactual reasoning simulation is performed one by one for all upstream factors that may affect the abnormal target variable in the cause-effect diagram. Based on this, the actual contribution ratio of each upstream factor to the anomaly is calculated and sorted from largest to smallest contribution. At the same time, the complete causal transmission path from the root cause factor to the abnormal target variable is output. Decision support steps: Based on the identified root causes, different intervention schemes are set in the coupled twin model and simulations are performed respectively. The comprehensive impact of each scheme on the overall environment, animal physiology and production performance is compared, and the optimal intervention scheme is selected for output.

2. The method for simulating and supporting the entire process of ranch production based on digital twins according to claim 1, characterized in that: The environmental parameters in the data acquisition steps include air temperature, relative humidity, wind speed, light intensity, ammonia concentration, and carbon dioxide concentration; the individual animal physiological and behavioral parameters include core body temperature, activity level, feeding time, and rumination time. Production performance parameters include milk production, feed intake, and daily weight gain.

3. The method for simulating and supporting the entire process of ranch production based on digital twins according to claim 1, characterized in that: The twin model construction steps specifically include: Environmental field simulation sub-model construction: The ranch building space is discretized into a three-dimensional grid, and the temperature, wind speed and pollutant concentration values ​​of each grid node are solved to obtain the distribution of environmental parameters of the entire ranch area; Animal individual physiological simulation sub-model construction: A thermal balance calculation relationship is established for each animal to calculate the change of core body temperature over time; At the same time, a stress response calculation relationship is established, and the degree of deviation of each environmental parameter from the thermoneutral zone is converted by a nonlinear threshold function and then weighted and summed to obtain the comprehensive stress intensity. This stress intensity is used to inhibit normal feed intake and accelerate the decline of production performance. Two-way coupling mechanism construction: Within each simulation time step, the environmental parameters of each region are used as the input boundary conditions of the physiological simulation sub-model, and the metabolic heat production and moisture dissipation of individual animals are used as source terms to feed back to the energy equation and material equation of the environmental field simulation sub-model. This process is repeated iteratively until the change in environmental state and physiological state of all individuals between two adjacent iterations is less than the preset tolerance, at which point the calculation of the current step ends.

4. The method for simulating and supporting the entire process of ranch production based on digital twins according to claim 1, characterized in that: The online calibration step is triggered when the absolute value of the deviation between the model prediction value and the real-time sensor measurement value of any key variable exceeds a preset threshold and the deviation continues for more than a preset duration.

5. The method for simulating and supporting the entire process of ranch production based on digital twins according to claim 1, characterized in that: The nodes in the causal graph include at least ambient temperature, ambient humidity, wind speed, ammonia concentration, feed amount, ventilation volume, core body temperature, stress intensity, feed intake, activity level, health score, disease events, milk production, and daily weight gain; directed edges point from cause nodes to result nodes.

6. The method for simulating and supporting the entire process of ranch production based on digital twins according to claim 1, characterized in that: The counterfactual reasoning in the anomaly tracing step specifically includes: Extract all upstream factors that may affect the abnormal target variable from the causal graph as a set of candidate root causes. For each candidate root cause variable, replace the actual observed value of the variable at the time of the anomaly with the expected value of the variable under normal conditions, keep all other variables unchanged, and rerun the coupled twin model to perform a backtracking simulation from before the anomaly occurred to after the anomaly occurred; then calculate the causal contribution of the candidate root cause.

7. The method for simulating and supporting the entire process of ranch production based on digital twins according to claim 1, characterized in that: The complete causal transmission path is output as follows: starting from the root cause node, each intermediate variable passed along the path is listed sequentially along the causal graph, and the actual change direction and magnitude of each intermediate variable during the anomaly are marked, until the anomaly target variable is reached.

8. The method for simulating and supporting the entire process of ranch production based on digital twins according to claim 1, characterized in that: The intervention schemes in the decision support steps include one or more combinations of adjusting environmental control equipment parameters, changing feeding strategies, adjusting ventilation modes, and changing milking or regrouping plans; the simulation results of each intervention scheme are presented in the form of prediction curves to show the environmental recovery time, the speed at which animal physiological indicators return to normal, and the final degree of recovery of production performance.

Images