Laying hen breeding simulation system based on digital twinning

Abstract

The invention discloses a laying hen breeding simulation system based on digital twinning, and relates to the technical field of intelligent breeding, and the system comprises a data collection unit which is used for obtaining multi-dimensional state parameters and outputting the multi-dimensional state parameters to a biological state quantification unit; the biological state quantification unit is used for resolving an individual metabolic rate index and a group aggregation entropy; the time delay characteristic analysis unit is used for quantifying and generating a biological response time delay complexity index; the regulation and control instruction generation unit is used for calculating a temperature demand reference value and further generating a final temperature set target value; and the model self-optimization unit is used for predicting a biological state at the next moment, comparing the biological state with an actual observation state after regulation and control execution, generating a prediction error, and carrying out iterative updating on preset weights and coefficients in the system. According to the method, by accurately quantifying the biological state and response characteristics of the laying hens and combining model self-optimization, accurate, flexible and continuous self-perfect intelligent environment regulation and control are achieved, and the breeding benefits and animal welfare are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent breeding technology, and in particular to a laying hen breeding simulation system based on digital twins. Background Art

[0002] Modern egg-laying chicken farming is rapidly developing towards scale, intensification, and automation. Precise control of the farming environment is crucial to ensuring the health and productivity of laying hens. Existing technical solutions primarily rely on deploying sensor networks within farming facilities to monitor macro-physical environmental parameters such as temperature, humidity, and key gas concentrations in real time. Based on this monitoring data, environmental control systems (such as fans, wet curtains, and heaters) can execute pre-set control logic. For example, they can automatically start and stop equipment when the temperature falls below or rises above a static threshold, thereby maintaining environmental indicators within a relatively stable range. This automated environmental monitoring and control technology has, to a certain extent, replaced some of the management practices that relied on manual experience, improving the efficiency of farming management and becoming an indispensable technical foundation for current intensive farming.

[0003] Existing technical solutions have significant limitations. First, their core basis for regulation is physical environmental parameters, essentially treating complex biological populations as passive responders to environmental changes, while ignoring the inherent physiological state, behavioral needs, and diversity of laying hens as living organisms. Management methods are often based on a one-size-fits-all approach based on age, failing to address the individual environmental needs of the same flock due to differences in weight, health, and egg production. Second, their control logic is often a rigid "set-and-go" model that lacks consideration for the response characteristics of biological systems. The regulatory actions themselves can become new stressors due to their abruptness. Furthermore, monitoring the stress state of flocks has traditionally relied on sampling and testing individual individuals. This approach not only has a lag effect and interferes with the animals, but also fails to reflect the overall real-time status of the flock. Finally, if existing systems include models, these models are typically static and immutable once established. They are unable to adapt to dynamic changes in the flock due to growth, health changes, or seasonal changes. This causes the models to become inaccurate over time, impacting the long-term effectiveness of regulation. Summary of the Invention

[0004] The purpose of the present invention is to provide a laying hen breeding simulation system based on digital twins, which solves the problems existing in the background technology.

[0005] To solve the above technical problems, the present invention provides a laying hen breeding simulation system based on digital twins, comprising: a data acquisition unit for acquiring multidimensional state parameters in the breeding facility in real time and outputting them to a biological state quantification unit, wherein the multidimensional state parameters include physiological information of individual laying hens, the overall spatial position distribution of the flock, and physical environment parameters; The biological state quantification unit is used to calculate the individual metabolic rate index and group aggregation entropy based on the physiological information and overall spatial position distribution obtained by the data acquisition unit; A time-lag characteristic analysis unit is used to quantify and generate a biological response time-lag complexity index based on the change process of group aggregation entropy after environmental regulation in historical data; The control instruction generation unit is used to calculate the temperature requirement baseline value based on the individual metabolic rate index and group aggregation entropy generated by the biological state quantification unit, and further generate the final temperature setting target value in combination with the biological response time lag complexity index generated by the time lag characteristic analysis unit; The model self-optimization unit is used to predict the biological state at the next moment based on the temperature set target value and the current biological state, and compare it with the actual observed state after the control is executed to generate a prediction error. The preset weights and coefficients in the system are iteratively updated based on the prediction error.

[0006] Preferably, the biological state quantification unit calculates the individual metabolic rate index by: The collected body weight, egg production rate and age data are standardized to generate dimensionless physiological index values; the dimensionless physiological index values ​​are weighted and summed with their corresponding preset weights to generate an individual metabolic rate index.

[0007] Preferably, the process of calculating the population aggregation entropy by the biological state quantification unit includes: Multiple grid areas are virtually divided within the breeding space. An image analysis algorithm is used to identify and count the proportion of the total number of chickens distributed in each grid area to generate a regional distribution ratio. The regional distribution ratio is then substituted into a preset information entropy formula for calculation to generate group aggregation entropy.

[0008] Preferably, the process of generating the biological response time lag complexity index by the time lag feature analysis unit includes: Obtain the time-varying curve of the group aggregation entropy after the historical environmental regulation event; determine the moment when the absolute value of the derivative of the time-varying curve is continuously lower than the preset threshold to solve the lag time of the biological response; normalize the lag time to generate a dimensionless lag coefficient; and perform weighted summation of the dimensionless lag coefficients to generate a biological response time lag complexity index.

[0009] Preferably, the calculation process of the temperature requirement reference value by the control instruction generation unit includes: Obtaining preset standard temperature values, ideal metabolic rate index, and ideal aggregation entropy; Calculate the deviation between the current average metabolic rate index and the ideal metabolic rate index, and generate a metabolic rate temperature compensation value based on the preset temperature sensitivity coefficient; Calculate the deviation between the current group aggregation entropy and the ideal aggregation entropy, and generate the aggregation entropy temperature compensation value by combining the preset entropy sensitivity coefficient; The standard temperature value is corrected using the metabolic rate temperature compensation and the aggregate entropy temperature compensation to generate a temperature requirement reference value.

[0010] Preferably, the process of generating the temperature setting target value by the control instruction generating unit includes: The biological response time-lag complexity index is substituted into the preset exponential decay model to generate a control intensity factor; the difference between the temperature demand baseline value and the actual temperature currently collected is calculated to generate a temperature target deviation; the temperature target deviation is multiplied by the control intensity factor and summed with the actual temperature to generate the temperature setting target value for the next control time step.

[0011] Preferably, the prediction process of the biological state at the next moment by the model self-optimization unit is: Construct a state transition prediction model; take the individual metabolic rate index and group aggregation entropy observed at the current moment, and the temperature setting target value generated by the control instruction generation unit as input; perform calculations through the state transition prediction model, and output the predicted values ​​of the individual metabolic rate index and group aggregation entropy at the next moment.

[0012] Preferably, the iterative update process of the model self-optimization unit includes: After the control is completed, the individual metabolic rate index and group aggregation entropy of the next time step are actually observed; the actual observation value is compared with the predicted value, and the weighted deviation is calculated to generate the prediction error; the prediction error is used as the loss function, and the optimization algorithm is adopted with the goal of minimizing the loss function, and all preset adjustable weights and coefficients in the system are periodically updated.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention deepens the understanding of farmed animals by constructing a new quantitative indicator of biological status. It is no longer limited to monitoring the external environment. Instead, it integrates key physiological information such as body weight, egg production rate, and age that affect energy demand into a computable quantitative indicator through the individual metabolic rate index, thereby gaining insight into and responding to the real needs of different individuals. At the same time, the introduced group aggregation entropy can convert the macroscopic spatial distribution pattern of the chicken flock into a dynamic indicator for evaluating its overall comfort and stress level through non-contact image analysis, thus achieving real-time and global perception of the group welfare status.

[0014] 2. The present invention establishes a dynamic and flexible closed-loop control decision-making mechanism. The control target is no longer a fixed industry standard, but a temperature reference value that closely fits the current biological needs of the chicken flock based on real-time individual metabolic rate and group aggregation entropy feedback. It also innovatively quantifies the hysteresis and complexity of the biological population in responding to environmental changes, and dynamically adjusts the intensity and rhythm of the control instructions based on this, realizing a progressive flexible compensation strategy, effectively avoiding secondary stress on the chicken flock caused by sudden environmental changes, and improving the stability and safety of the control process.

[0015] 3. Through the built-in state transition prediction model, the present invention can preview the possible biological consequences of the control decision before it is executed. After obtaining the real observation data after the control, the error between the prediction and reality will be calculated. Driven by this, the optimization algorithm is used to reversely update all key model parameters within the system. This "prediction-verification-correction" closed-loop mechanism enables the digital twin model to continuously learn and evolve, dynamically approaching the real conditions of the physical farm, thereby ensuring the long-term adaptability, accuracy and robustness of the system in different growth cycles, different health conditions and changing external environments of the chicken flock. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can derive other drawings based on these drawings without inventive effort. Figure 1 It is a logic block diagram of the system of the present invention. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 The present invention provides a laying hen breeding simulation system based on digital twins, comprising: a data acquisition unit for acquiring multidimensional state parameters in the breeding facility in real time and outputting them to a biological state quantification unit, wherein the multidimensional state parameters include physiological information of individual laying hens, the overall spatial position distribution of the flock, and physical environment parameters; The biological state quantification unit is used to calculate the individual metabolic rate index and group aggregation entropy based on the physiological information and overall spatial position distribution obtained by the data acquisition unit; A time-lag characteristic analysis unit is used to quantify and generate a biological response time-lag complexity index based on the change process of group aggregation entropy after environmental regulation in historical data; The control instruction generation unit is used to calculate the temperature requirement baseline value based on the individual metabolic rate index and group aggregation entropy generated by the biological state quantification unit, and further generate the final temperature setting target value in combination with the biological response time lag complexity index generated by the time lag characteristic analysis unit; The model self-optimization unit is used to predict the biological state at the next moment based on the temperature set target value and the current biological state, and compare it with the actual observed state after the control is executed to generate a prediction error. The preset weights and coefficients in the system are iteratively updated based on the prediction error.

[0019] This embodiment provides a layer chicken farming simulation system based on digital twins. The system runs on a computing device equipped with a processor and memory. It aims to achieve accurate perception, dynamic simulation, and closed-loop feedback control of the layer chicken farming environment through a series of tightly coupled computing units. The system as a whole forms a self-consistent technical closed loop, including a data acquisition unit, a biological state quantification unit, a time-delay feature analysis unit, a control instruction generation unit, and a model self-optimization unit. The data acquisition unit, whose purpose is to provide real-time, multi-dimensional data input for the entire digital twin system, is the starting point for mapping the physical world to the digital space. In this embodiment, this unit is implemented through a sensor network deployed within the breeding facility. Specifically, it uses an overhead machine vision system combined with radio frequency identification tags worn by individuals to capture and record the physiological information of selected sample laying hens. In this embodiment, the physiological information refers to core data reflecting the health and production status of individuals, mainly including weight, age, and egg production rate. At the same time, overhead imaging equipment is used for continuous shooting to obtain the overall spatial distribution of the chickens. In addition, a standard environmental sensor array is deployed to continuously monitor the physical environmental parameters in the breeding space, covering temperature, humidity, and the concentration of key gases (such as ammonia and carbon dioxide). All collected raw data is structured and processed and output to subsequent units in real time. The biological state quantification unit is designed to convert the collected raw, heterogeneous data into standardized, calculable indicators that can characterize the biological state of laying hens. In this embodiment, this unit receives physiological information and spatial position distribution data from the data acquisition unit. It contains two core algorithm modules: the first calculates a comprehensive individual metabolic rate index based on the weight, egg production rate, and age of individual laying hens; the second calculates the group aggregation entropy based on the overall spatial position distribution of the flock using information entropy theory. These two indices together constitute a digital description of the current biological state of the laying hens. The time-delay characteristic analysis unit aims to quantify the hysteresis and complexity of the biological system (chicken flock)'s response to environmental control commands, providing a key basis for the subsequent generation of adaptive control strategies. This unit is not based on instantaneous data, but rather conducts an in-depth analysis of environmental control events stored in a historical database and the subsequent biological state changes. Specifically, it analyzes how long it takes for the group's aggregate entropy, representing the group's stress level, to reach a new stable state after a temperature or ventilation adjustment. This process is quantified to generate a biological response time-delay complexity index. The control instruction generation unit is designed to calculate the optimal environmental control target value based on the current biological state and the response characteristics of the system. In this embodiment, this unit is the key link in achieving adaptive control. It first calculates a theoretical temperature requirement baseline value based on the individual metabolic rate index and group aggregation entropy generated by the biological state quantification unit. This baseline value reflects the ideal temperature of the chickens in their current state. Then, it combines the biological response time lag complexity index generated by the time lag characteristic analysis unit to dynamically adjust the intensity and speed of the control, and further generates a final temperature setting target value to be executed in the next control cycle. The purpose of the model self-optimization unit is to ensure that the system model can continuously learn and self-improve, and to guarantee its long-term accuracy and adaptability. In this embodiment, this unit forms a complete prediction-feedback-correction closed loop. After the control instruction is generated and before it is executed, the unit will predict the biological state of the chickens at the next moment (i.e., individual metabolic rate index and group aggregation entropy) based on the current biological state and the upcoming temperature setting target value through a built-in state transition prediction model. After the actual control is completed, the system will compare the actual observed state at the next moment with the previously predicted value to generate a prediction error. This error is used as a loss function to drive an optimization algorithm to periodically iteratively update all preset and adjustable weights and coefficients in the system, thereby achieving self-calibration and continuous optimization of the model. Through the collaborative work of the above five units, this embodiment constructs a complete digital twin closed-loop system from perception to decision-making to optimization. Compared with traditional breeding models that rely on static feeding standards or manual experience, this system can accurately quantify the comprehensive biological status of individual and group laying hens in real time, and innovatively incorporate the hysteresis of biological responses into the decision-making model. This makes environmental control no longer a simple "set-and-execute" process, but transforms it into an intelligent and flexible process that can foresee results, adapt to objects, and self-correct. Ultimately, the system can significantly improve feed conversion rate and egg production performance while ensuring the welfare of laying hens and reducing stress reactions, thereby maximizing breeding benefits and providing solid technical support for the standardization and intelligence of the breeding process.

[0020] Example 2: The biological state quantification unit calculates the individual metabolic rate index by: The collected body weight, egg production rate and age data are standardized to generate dimensionless physiological index values; the dimensionless physiological index values ​​are weighted and summed with their corresponding preset weights to generate an individual metabolic rate index.

[0021] This example is a specific implementation of the individual metabolic rate index calculation process in the biological state quantification unit described in Example 1. This process aims to unify raw physiological data in different physical units into a dimensionless index framework to achieve standardized comparison of metabolic levels between different individuals. To achieve the above objectives, the biological state quantification unit first normalizes the three core physiological data collected from laying hens: weight, egg production rate, and age. Normalization involves a mathematical transformation designed to eliminate numerical differences in the original data due to differences in units and dimensions, placing them within a comparable range. In this embodiment, dimensionless physiological index values ​​are generated by calculating the relative deviations between each real-time physiological data item and its ideal value at the corresponding age. Specifically, the standardization process can be calculated using the following formula: ; is a dimensionless physiological index value; is the real-time physiological data collected (e.g., current body weight or egg production rate); It is the ideal physiological data value corresponding to the current age. This ideal value is preset in the system and can be derived from authoritative feeding management guidelines or data statistics of historical best production batches. Based on this, the system performs a weighted summation of these dimensionless physiological index values ​​with their corresponding preset weights to generate the final individual metabolic rate index. This calculation process is defined by a preset individual metabolic rate index model. The individual metabolic rate index model is a linear weighted model constructed based on the recognized principles of animal energy metabolism. It considers the total metabolic level of an organism to be a weighted linear combination of its basic physiological indicators. Its mathematical expression is: ; M is the individual metabolic rate index, which is a dimensionless comprehensive value that quantifies the comprehensive energy metabolism level of a single individual in the current physiological state; m, r, and a are dimensionless physiological index values, corresponding to the standardized body weight, egg production rate, and age, respectively. Their data are derived from the results of standardization of the raw data obtained by the data acquisition unit. are preset weights, which are dimensionless weight coefficients corresponding to body weight, egg production rate, and age. They represent the relative contribution of each physiological indicator to the overall metabolic level. The initial values ​​of these weights are set based on published research results in poultry physiology and nutrition, and their sum is constrained to be 1. They can be iteratively updated during subsequent model self-optimization. Through the above-mentioned implementation mode, the present invention provides a refined means of quantifying the physiological state of laying hens; compared with the traditional method of roughly grouping and management based solely on age, the individual metabolic rate index can more comprehensively and dynamically reflect the actual energy demand differences caused by individual weight differences and egg production levels; this gain effect makes the subsequent environmental temperature control more targeted and can meet the real needs of individuals with different metabolic levels, thereby improving the uniformity of the group while avoiding energy waste or production performance loss caused by one-size-fits-all management.

[0022] Example 3: The biological state quantification unit calculates the population aggregation entropy by: Multiple grid areas are virtually divided within the breeding space. An image analysis algorithm is used to identify and count the proportion of the total number of chickens distributed in each grid area to generate a regional distribution ratio. The regional distribution ratio is then substituted into a preset information entropy formula for calculation to generate group aggregation entropy.

[0023] This embodiment is a specific implementation of the group aggregation entropy solution process in the biological state quantification unit of Example 1. This process is intended to convert the descriptive spatial distribution image of the chicken flock captured by the camera equipment into a quantitative numerical indicator that can indicate the overall comfort and stress level of the flock. To achieve this goal, the biological state quantification unit's calculation of the aggregation entropy of the population begins with virtual gridding of the two-dimensional plan view of the breeding space within the system's digital twin model. Virtual gridding involves dividing the bird's-eye view of the entire chicken house into multiple equally sized, non-overlapping grid areas at the software level. This provides discrete spatial units for subsequent chicken population statistics. The total number of grid cells, N, is a preset parameter that strikes a balance between spatial resolution and computational efficiency. Subsequently, the system processes the collected real-time images of the chicken flock through a preset image analysis algorithm; in this embodiment, the image analysis algorithm refers to a target recognition and counting model based on computer vision technology, such as YOLO or a similar algorithm, which is trained with chicken feature data and can accurately identify each chicken in the image and determine the grid area where it is located; through this algorithm, the system can identify and count the number of chickens distributed in each grid area, and calculate the proportion of this number to the total number of chickens, thereby generating the regional distribution ratio of each grid ; Finally, the system distributes the area ratio of all grids Substitute a preset information entropy formula for calculation to generate the group aggregation entropy; this calculation process is defined by the group aggregation entropy model; the group aggregation entropy model is a model that borrows the concept of entropy in information theory to quantify the degree of disorder in the system; in this embodiment, the more uniform the distribution of the chickens, the higher the disorder, and the higher the entropy value, which means that the chickens are more comfortable and relaxed; conversely, if the chickens gather in a local area, the distribution is more orderly and the entropy value is lower, which means that there may be environmental stress; its mathematical expression is: ; H is the aggregation entropy of the flock, a dimensionless value used as an ordinal indicator in this system; its function is to quantify the uniformity of the spatial distribution of the flock, thereby indirectly evaluating the macro-stress state of the flock; this value is calculated by this formula; N is the total number of grid areas, a dimensionless integer; its value comes from the preset value based on the size of the breeding space and the required accuracy when the system is initialized; is the regional distribution ratio, which is the ratio of the number of chickens distributed in the i-th grid area to the total number of chickens. It is a dimensionless probability value. Its data comes from the statistical calculation results of real-time images through image analysis algorithms. The added benefit of this implementation is that it provides a non-contact, low-cost, and highly real-time means of assessing group stress status. Traditional stress assessments often rely on sampling and measuring blood physiological indicators, which are subject to lags, stress interference, and inability to cover the entire population. However, group aggregation entropy, through the quantification of macro-behavioral patterns, can capture signals early in the occurrence of environmental stress (such as aggregation behavior caused by excessive local wind speed or unsuitable temperature), enabling the system to conduct more forward-looking and proactive environmental interventions, effectively preventing the occurrence of large-scale stress events and ensuring animal welfare and production stability.

[0024] Example 4: The process of generating the biological response time lag complexity index by the time lag feature analysis unit includes: Obtain the time-varying curve of the group aggregation entropy after the historical environmental regulation event; determine the moment when the absolute value of the derivative of the time-varying curve is continuously lower than the preset threshold to solve the lag time of the biological response; normalize the lag time to generate a dimensionless lag coefficient; and perform weighted summation of the dimensionless lag coefficients to generate a biological response time lag complexity index.

[0025] This embodiment is a specific implementation of the biological response time lag complexity index generation process in the time lag feature analysis unit of Example 1. The purpose of this process is to quantify the degree of sluggishness or stickiness exhibited by the behavior pattern of a flock of chickens as a complex biological system when it returns to a new steady state after being stimulated by the external environment.

[0026] To achieve this quantification process, the system has a built-in control event log module to record key control events. The actual temperature The absolute value of the difference Exceeding the preset trigger threshold (For example ), and when the control instruction is executed continuously for at least one control cycle, the system marks this as a valid temperature control event and records the start time of the event, the control type (temperature increase / temperature decrease), and the relevant state parameters before and after the control.

[0027] The time-lag feature analysis unit is activated at a preset analysis period (e.g., every 24 hours) or after a sufficient number of similar events (e.g., 10) have accumulated in the event log, and performs a retrospective analysis of the historical data in the log. The unit first selects specific types of control events (e.g., temperature control) from the log and extracts the time-varying curve of the group aggregation entropy H recorded continuously thereafter. Then, in order to objectively determine the moment when the flock behavior reaches a new steady state, the system calculates the derivative of this time-varying curve. The absolute value of a preset threshold It is introduced here, and it is a very small positive number. Its technical principle is that after the group aggregation entropy reaches a steady state, its value fluctuates naturally only within a very small range, and its rate of change should be close to zero. The specific value of the threshold can be set based on the statistical analysis of a large number of chicken aggregation entropy time series data under normal conditions, and the 95% or 99% quantile of the absolute value of the change rate is taken to ensure that only statistically significant changes are regarded as the end of state transition. Therefore, the system continuously monitors , when the value is within the preset steady state judgment time The value is always below the preset threshold When the time is right, it is determined that the chicken group behavior has reached a new steady state. A system parameter is preset to a value greater than one control cycle, for example, the total duration of three control cycles, to ensure that the observed stability is not a random transient phenomenon; the time from the moment the environmental control command begins to the moment of steady state is calculated as the lag time of the biological response; This process is performed separately for different types of environmental disturbances (such as temperature, humidity, and air quality) to obtain their respective lag times. These lag times are then normalized to eliminate the magnitude differences in lag times caused by different disturbance events and convert them into a unified, dimensionless lag coefficient. Specifically, the normalization process can adopt the maximum and minimum normalization method, and its calculation formula is: ; is the dimensionless hysteresis coefficient corresponding to a specific disturbance (such as temperature, humidity or air quality); is the biological response lag time measured under the disturbance; and These are preset reference values ​​for the maximum and minimum lag times that may be caused by this type of disturbance, derived from a large amount of historical data. This method allows lag times of different magnitudes to be uniformly mapped to the range of 0 to 1. To enable those skilled in the art to reproduce, a method for determining and Statistical calibration method: Also use the effective control event log accumulated during the system initialization calibration period or early operation stage (for example, the first 30 days). Extract the biological response lag time of all similar disturbances (such as temperature disturbance) from the log After statistical analysis of the data set and elimination of outliers, the 95th percentile of the data set can be set as , set the 5th percentile to These two values ​​should also be updated regularly based on more historical data as the system runs over a long period of time.

[0028] Finally, these dimensionless hysteresis coefficients are weighted and summed to generate a comprehensive biological response time-lag complexity index. This calculation is defined by a preset biological response time-lag complexity index model. The biological response time-lag complexity index model is a linear weighted model designed to integrate the response hysteresis of multiple environmental factors. Its mathematical expression is: ; L is the biological response time lag complexity index, which is a dimensionless comprehensive index. Its function is to quantify the overall response sensitivity of the current chicken flock to environmental changes. are dimensionless hysteresis coefficients, representing the normalized response lag times caused by temperature, humidity, and air quality disturbances, respectively. Their data are derived from the results of normalizing the corresponding lag times in historical data. is the preset weight, which is the dimensionless weight corresponding to the three hysteresis coefficients mentioned above; its function is to characterize the importance of different environmental factors on the overall status of the chicken flock; its initial value is set according to the general understanding of the threat of each environmental factor to the health of the chicken flock, such as the weight of air quality Usually assigned high values; these weights sum to 1 and can be adjusted by the model self-optimization unit; The innovation of this implementation lies in that it allows the system to no longer view the chicken flock as an ideal object that responds instantaneously to control commands, but instead quantifies its biological inertia for the first time. The resulting benefit is that the system can adjust the aggressiveness of its control strategy based on the complexity index L. When the chicken flock is in a state of complex stress and exhibits high response lag (high L value), the system will adopt a gentler and more gradual control method, effectively avoiding violent fluctuations in environmental parameters caused by excessive control, thereby improving the robustness and stability of the entire closed-loop control system.

[0029] Example 5: The calculation process of the temperature requirement reference value by the control instruction generation unit includes: Obtaining preset standard temperature values, ideal metabolic rate index, and ideal aggregation entropy; Calculate the deviation between the current average metabolic rate index and the ideal metabolic rate index, and generate a metabolic rate temperature compensation value based on the preset temperature sensitivity coefficient; Calculate the deviation between the current group aggregation entropy and the ideal aggregation entropy, and generate the aggregation entropy temperature compensation value by combining the preset entropy sensitivity coefficient; The standard temperature value is corrected using the metabolic rate temperature compensation and the aggregate entropy temperature compensation to generate a temperature requirement reference value.

[0030] This embodiment is a specific implementation of the temperature requirement baseline value calculation process in the control instruction generation unit of Example 1. The core purpose of this calculation process is to use the recommended temperature in the industry standard feeding guide as a basis and make personalized and dynamic adjustments based on the current real-time biological status of the chicken flock to obtain a theoretically optimal temperature value. To achieve this calculation, the control instruction generation unit needs to obtain a series of preset reference parameters for the calculation process of the temperature requirement reference value; the standard temperature value The source is the recommended ambient temperature obtained from the authoritative feeding and management guide based on the age of the laying hen; the ideal metabolic rate index and ideal aggregation entropy The sources of these two values ​​are determined by finding the statistical averages of metabolic rate and aggregation entropy associated with the highest egg production rate and the lowest mortality rate under strictly controlled experimental conditions. They represent the biological status indicators of the chicken flock at the best production performance; In order to enable those skilled in the art to reproduce, a method for determining an ideal biological state index is provided. and System initialization calibration method: At the beginning of system deployment, set a period of The calibration cycle is 14 days (for example, 14 days). During this period, the environmental control adopts the traditional static feeding standard based on age, and the system continuously collects and records the physiological information of all laying hens and the group aggregation entropy of the flock. After the calibration period, the day or days with the best comprehensive production performance (such as the highest average egg production rate and the lowest mortality rate) during this period are analyzed, and the statistical average of the average individual metabolic rate index and the average group aggregation entropy measured during these days is used as the ideal metabolic rate index preset by the system. and ideal aggregation entropy .

[0031] Next, the system compares the real-time feedback of biological status indicators with the ideal value, calculates the deviation, and generates the corresponding temperature compensation amount based on the preset sensitivity coefficient; specifically, it includes: Calculate the average metabolic rate index of the current flock Ideal metabolic rate index and compares the deviation with a preset temperature sensitivity coefficient. Combined, they generate a temperature compensation for metabolic rate; Similarly, calculate the current group aggregation entropy H and the ideal aggregation entropy The deviation between the two, combined with the preset entropy sensitivity coefficient , generating the aggregate entropy temperature compensation; Finally, the system uses these two calculated temperature compensation values ​​to perform a double correction on the standard temperature value to generate the final temperature requirement reference value. This calculation logic is defined by a preset temperature requirement reference value model. The temperature requirement reference value model is a control model based on feedback correction. Its physical meaning is that when the actual state of the chicken group deviates from the optimal state, it compensates by adjusting the temperature. Its mathematical expression is: ; is the temperature requirement benchmark value, in degrees Celsius (°C); its function is to provide a theoretical target for the next stage of dynamic regulation; this value is calculated by this formula; It is a standard temperature value in degrees Celsius (°C); its source is a preset industry standard or feeding guide; are the current biological state indicators, namely the average metabolic rate index and group aggregation entropy of the current chicken flock, both of which are dimensionless; their data come from the real-time calculation results of the biological state quantification unit; are indicators of ideal biological state, namely the ideal metabolic rate index and the ideal aggregation entropy, both dimensionless; their values ​​are derived from the optimal values ​​calibrated through previous experiments and preset in the system; is the sensitivity coefficient, and the unit is degrees Celsius (℃) to ensure the consistency of the physical dimensions on both sides of the formula. The physical meaning of is the amount of temperature compensation required for each unit deviation of the average metabolic rate index from the ideal value; The physical meaning of is the same; their initial values ​​are set based on thermodynamics and animal behavior knowledge, and can be adjusted through the model self-optimization unit; Through this implementation method, the core basis of environmental regulation has changed from static, universal feeding standards to dynamic, targeted biological needs; the resulting benefit is a greatly improved level of refined environmental management; for example, when the average metabolic rate of the chicken flock is higher than the ideal value (possibly due to large feed intake and high activity level), the system will automatically calculate a slightly lower temperature requirement baseline value to help dissipate heat; when the group aggregation entropy is lower than the ideal value (aggregation behavior occurs), the system will calculate a higher temperature requirement baseline value to alleviate cold stress; this real-time, feedback-based correction ensures a precise match between environmental supply and biological needs, thereby improving energy utilization efficiency and animal production performance.

[0032] Example 6: The process of generating the temperature setting target value by the control instruction generating unit includes: The biological response time-lag complexity index is substituted into the preset exponential decay model to generate a control intensity factor; the difference between the temperature demand baseline value and the actual temperature currently collected is calculated to generate a temperature target deviation; the temperature target deviation is multiplied by the control intensity factor and summed with the actual temperature to generate the temperature setting target value for the next control time step.

[0033] This embodiment is a specific implementation of the process for generating the final temperature setting target value in the control instruction generation unit. This process, after calculating the theoretical temperature demand reference value, combines the system's response characteristics to generate a practical and executable specific temperature setting value for the next control time step. It embodies a prudent and intelligent flexible control strategy. The process first introduces a control intensity factor , whose purpose is to dynamically adjust the amplitude of the control action according to the sluggishness of the biological system's response. This factor is generated by a preset exponential decay model based on the biological response delay complexity index L calculated in the previous step. The technical principle of the exponential decay model is that when the system response is complex and sluggish (L value is high), the control intensity should be significantly reduced to avoid excessive intervention. Its mathematical expression is: ; is the control intensity factor, which is a dimensionless factor between 0 and 1; it acts as a scaling factor for the control step size; L is the biological response time-delay complexity index, which is the dimensionless input value from the time-delay characteristic analysis unit; k is the decay constant, a dimensionless positive number. Its function is to adjust the severity of the impact of the L value on the regulation intensity. Its value is determined by regression analysis of historical regulation effect data and can be updated by the model self-optimization unit. After obtaining the control intensity factor, the system calculates the temperature demand benchmark value The actual temperature collected by the current sensor The difference between the two is defined as the temperature target deviation; Finally, the temperature target deviation is combined with the control intensity factor Multiplying them together, we get the actual adjustment amount, which is then summed with the current actual temperature to generate the temperature setting target value for the next control time step. This calculation process is defined by the final control instruction model: ; is the target temperature setting value in degrees Celsius (°C). This is the final output of this unit and the execution instruction sent to the underlying environmental control hardware (such as air conditioners and heaters). is the actual temperature in degrees Celsius (°C); its data comes from the real-time reading of the temperature sensor in the data acquisition unit; is the control intensity factor, which is the dimensionless value calculated in the previous step; is the temperature requirement reference value, in degrees Celsius (℃); The technical solution of this embodiment integrates the biological response time lag characteristics and the dynamic demand benchmark, thereby achieving a synergistic gain effect; it not only determines the ultimate goal of regulation (by ), and more importantly, the process of achieving that goal (determined by and L); its technical effect is to achieve a progressive flexible compensation strategy; when the flock is stable and responds quickly (L value is low, Approaching 1), the system will quickly reach the ideal temperature Close together to ensure the efficiency of control; when the flock is fragile and slow to respond (L value is high, approaching 0), the system only makes small changes based on the current temperature in a gradual adjustment manner, thereby effectively avoiding secondary stress on the chickens caused by sudden environmental changes; this adaptive regulation rhythm greatly enhances the safety and stability of the system.

[0034] Example 7: The prediction process of the model self-optimization unit for the biological state at the next moment is: Construct a state transition prediction model; take the individual metabolic rate index and group aggregation entropy observed at the current moment, and the temperature setting target value generated by the control instruction generation unit as input; perform calculations through the state transition prediction model, and output the predicted values ​​of the individual metabolic rate index and group aggregation entropy at the next moment.

[0035] This embodiment is a specific implementation of the process of predicting the biological state at the next moment in the model self-optimization unit described in Example 1. This prediction function is the logical prerequisite for the entire system to perform self-optimization and closed-loop learning. Its purpose is to preview the possible biological consequences of the decision based on the current known state and the upcoming control decision. To achieve this prediction, a state transition prediction model is built into the system. Its core function is to fit the complex dynamic relationship between environmental inputs (temperature regulation) and biological state outputs (changes in metabolic rate and aggregate entropy). This model can be trained based on historical data and can be implemented using a multivariate linear regression model or more complex machine learning models such as recurrent neural networks (RNNs), which are particularly adept at processing time series data. The input of the model includes three key variables: the average individual metabolic rate index of the chicken group observed at the current time t and group aggregation entropy , and the temperature setting target value just calculated and generated by the control instruction generation unit and to be executed in the next time step ; Substitute these three input variables into the state transition prediction model for operation, and the model will output a set of predicted values, that is, the next moment The predicted values ​​of individual metabolic rate index and group aggregation entropy; this prediction process can be expressed by the following functional relationship: ; Is to predict the biological state, respectively, the model for the next moment Dimensionless predicted values ​​of individual metabolic rate index and group aggregation entropy; is the current biological state, which are the dimensionless biological state indicators actually observed at the current time t, provided by the biological state quantification unit; is the control instruction, which is the currently calculated temperature setting target value (°C) to be executed in the next time step, provided by the control instruction generation unit; It is the state transition prediction function, which represents the specific mathematical or algorithmic structure of the state transition prediction model and is used to calculate the output; In a preferred embodiment, the state transition prediction function and The multiple linear regression model can be used for construction, and its mathematical expression is: ; and is the predicted value; and Input for the model; These are all trainable parameters of the model, whose initial values ​​are determined by regression analysis of historical data and are iteratively updated in subsequent model self-optimization units; is a dimensionless parameter, and the coefficient The dimension of is the reciprocal of temperature (e.g. 1 / °C) to ensure the consistency of the dimensions on both sides of the equation; The added benefit of this implementation is that it gives the system the ability to make forward-looking predictions. Without this predictive function, the system can only know the effectiveness of adjustments after they have been made. However, through the built-in predictive model, the system has a reasonable expectation of the consequences of the decision at the same time as it makes the decision. This provides a precise quantitative target for subsequent error calculations and self-correction of model parameters, and is the core technical link in achieving the transition from simple open-loop control or hysteresis feedback control to truly predictive and adaptive control.

[0036] Example 8: The iterative update process of the model self-optimization unit includes: After the control is completed, the individual metabolic rate index and group aggregation entropy of the next time step are actually observed; the actual observation value is compared with the predicted value, and the weighted deviation is calculated to generate the prediction error; the prediction error is used as the loss function, and the optimization algorithm is adopted with the goal of minimizing the loss function, and all preset adjustable weights and coefficients in the system are periodically updated.

[0037] This embodiment is a specific implementation of the iterative update process in the model self-optimization unit. This process is the closed-loop endpoint for the entire digital twin system to achieve long-term autonomous learning, adapt to environmental changes, and maintain high-precision control capabilities. Its core is to quantitatively compare the model's predicted performance with real-world results and use this deviation to reversely drive the model's self-improvement. The iterative update process is triggered after the control of each control cycle is completed; the system first observes and calculates the next time step through the data acquisition unit and the biological state quantification unit. The true biological state, that is, the actual observed individual metabolic rate index and the actual observed group aggregation entropy ; The system then compares these two actual observations with the predicted values ​​generated by the model at the previous moment. and Compare and calculate the weighted deviation between them to generate a comprehensive prediction error; this error is calculated by a preset prediction error model; the prediction error model is used to quantify the accuracy of the model's prediction, and its output value is usually called a loss function in the field of machine learning; its mathematical expression is: ; is the prediction error, a dimensionless value that represents the overall deviation of the model prediction from the actual result; are the actual observed values, which are the metabolic rate index and entropy value actually observed in the next time step after the control is executed, both of which are dimensionless; is the model predicted value, which is the corresponding value predicted by the model based on the control instructions, and both are dimensionless; is the error weight, which is a dimensionless preset weight coefficient. Its function is to balance the importance of the two indicators, metabolic rate and aggregation entropy, in the total error calculation. The sum of them may not be 1 and can be adjusted by the model self-optimization unit. Finally, the system uses this calculated prediction error As a loss function, a preset optimization algorithm is used to iteratively update all preset adjustable weights and coefficients in the system with the goal of minimizing the loss function; the optimization algorithm usually adopts the gradient descent method or its variants (such as Adam optimizer); the algorithm calculates the loss function The partial derivatives (i.e., gradients) of all preset adjustable parameters in the system are adjusted slightly in the opposite direction of their gradients. The adjustable parameters in principle include all preset weights, coefficients, and non-physical constants defined in the various embodiments of this specification, such as the weights in the individual metabolic rate index model. ; Weights in the time-delay complexity index model of biological response ; Sensitivity coefficient in the temperature demand reference value model ; the decay constant k in the regulatory intensity factor model; and the error weight in the prediction error model etc., so that the next prediction error This process is repeated periodically; This iterative update mechanism is the core guarantee for the present invention to achieve long-term robustness and high precision. It cooperates with the prediction function of the model self-optimization unit to form a complete learning closed loop. The significant technical effect it brings is the adaptability of the system. As time goes by, whether the chickens enter a new growth cycle (physiological parameter changes), encounter sudden disease challenges (stress pattern changes), or the breeding environment undergoes seasonal changes, the system can automatically adjust its internal model parameters through a continuous "prediction-verification-correction" cycle to continuously fit the latest biological-environmental dynamic relationship. This makes the technical solution of the present invention different from a static expert system, and constructs a dynamic model that can continuously iterate and optimize itself according to the measured data, so as to always maintain the effectiveness and advancement of the regulation strategy under various complex and changeable actual breeding conditions.

[0038] This technical solution achieves a fundamental breakthrough in the perception, decision-making and control levels of traditional farming models by constructing a laying hen farming simulation system based on digital twins; its beneficial effects and technological progress are not the result of improvements in a single link, but are derived from a complete and self-consistent closed-loop logic that deeply integrates biology, information theory and cybernetics.

[0039] Existing technologies are usually limited to the monitoring and passive regulation of macroscopic physical environmental parameters such as temperature and humidity within aquaculture facilities; this approach ignores the intrinsic state and diversity of needs of organisms as the core of regulation, and essentially simplifies complex biological communities into passive responders to the environment.

[0040] This solution achieves a cognitive leap by introducing two core quantitative indicators of biological status: First, the establishment of the individual metabolic rate index has the physical significance of constructing a dynamic energy requirement model for each individual laying hen, non-dimensionalizing and linearly superimposing key physiological information that directly affects energy consumption (weight, egg production level, age). This formula is not a simple numerical addition, but is based on the basic principles of animal energy metabolism, uniformly mapping physiological data of different dimensions into a scalar field that can represent individual energy requirements. This enables the system to evolve from extensive management based on group averages to refined management that focuses on individual differences, providing a calculable physical basis for meeting the unique environmental needs of different individuals at different life cycles. Secondly, the introduction of group aggregation entropy, whose theoretical basis comes from the description of system disorder in statistical mechanics; it is innovatively used here to describe the macroscopic order parameter of the spatial distribution of chicken flocks; a uniform and loosely distributed chicken flock corresponds to a high entropy state, which physically means that the system is in a lower energy, more stable and comfortable state; on the contrary, the aggregation behavior of the chicken flock leads to a probability distribution The appearance of spikes in local areas and a decrease in the system entropy indicate the formation of an ordered structure, which is often driven by external environmental stress (such as local low temperatures or airflow) and is a direct macroscopic manifestation of group stress. With this indicator, the system can detect the overall welfare level and stress state of the entire group in real time and without interference through non-contact image analysis. Compared with traditional stress judgment methods that rely on individual sampling detection, this is a huge improvement in real-time, globality and non-destructiveness.

[0041] Traditional environmental control logic relies on a set-and-execute model based on static thresholds. For example, if the temperature falls below 20°C, the heating turns on. This control approach is rigid and lags behind, and it fails to consider the biological response characteristics of the control action itself. The control action itself may become a new stressor. This solution builds a multi-level adaptive compensation and control mechanism, which is progressive in the following aspects: First, the control target is no longer fixed, but a dynamically generated temperature demand benchmark value; the essence is a feedback control law, which uses the industry standard temperature As a benchmark, using the current biological state ( ) and the ideal state ( ) as the error signal, through the sensitivity coefficient and Perform negative feedback correction; this ensures that the regulatory target always closely follows the true biological needs of the flock; Secondly, the regulation process is flexible, which introduces the quantification of the inertia of the biological system. The construction of the biological response time lag complexity index L is to quantify the response function characteristics of the biological population as a complex system. On this basis, the regulation intensity factor is modeled as an exponential decay form. The physical connotation of this design is that when the system response is slow and the complexity is high (L value is large), Approaching 0, the control behavior will become extremely gentle; the final temperature setting target value calculation formula is like an intelligently adjusted damper; when When it approaches 1 (the system responds quickly), the system will quickly move towards the target Convergence; when When approaching 0 (the system response is slow), the adjustment step size is extremely small each time, thus achieving the optimal state smoothly and stably without causing violent oscillation of the system; this is essentially different from the step-type and impact-type control of the existing technology.

[0042] The effectiveness of any model-based system depends on the model's ability to accurately reflect the real world. Existing models are typically static, immutable once established. However, biological populations themselves are dynamic (growth, aging, disease), and the environment is constantly changing. Fixed models will inevitably become inaccurate over time. The core advancement of this scheme lies in the design of its model self-optimization unit, which gives the system the ability to self-evolve; this mechanism previews the consequences of regulatory actions through a state transition prediction model and obtains real observation values. Then, the prediction error between the two is calculated; this error It is used as a loss function to optimize all adjustable parameters in the system (including metabolic rate model weights) through optimization algorithms (such as gradient descent). , temperature correction coefficient , time-lag model weight etc.) for periodic iterative updates; This means that this solution does not build a fixed expert system, but an intelligent entity that can continuously carry out the scientific cycle of "hypothesis-verification-learning"; it can continuously self-correct to make its internal digital twin model increasingly close to the real dynamics of the physical entity, thereby ensuring the accuracy, robustness and adaptability of the system to unknown changes during long-term operation; this is an advanced feature that is completely different from existing technologies and has essential differences.

[0043] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A laying hen breeding simulation system based on digital twins, characterized in that: include: The data acquisition unit is used to obtain multi-dimensional state parameters in the breeding facility in real time and output them to the biological state quantification unit. The multi-dimensional state parameters include physiological information of individual laying hens, the overall spatial position distribution of the flock, and physical environment parameters; The biological state quantification unit is used to calculate the individual metabolic rate index and group aggregation entropy based on the physiological information and overall spatial position distribution obtained by the data acquisition unit; A time-lag characteristic analysis unit is used to quantify and generate a biological response time-lag complexity index based on the change process of group aggregation entropy after environmental regulation in historical data; The control instruction generation unit is used to calculate the temperature requirement baseline value based on the individual metabolic rate index and group aggregation entropy generated by the biological state quantification unit, and further generate the final temperature setting target value in combination with the biological response time lag complexity index generated by the time lag characteristic analysis unit; The model self-optimization unit is used to predict the biological state at the next moment based on the temperature set target value and the current biological state, and compare it with the actual observed state after the control is executed to generate a prediction error. The preset weights and coefficients in the system are iteratively updated based on the prediction error.

2. The laying hen breeding simulation system based on digital twin according to claim 1, characterized in that: The biological state quantification unit calculates the individual metabolic rate index by: The collected body weight, egg production rate and age data are standardized to generate dimensionless physiological index values; the dimensionless physiological index values ​​are weighted and summed with their corresponding preset weights to generate an individual metabolic rate index.

3. The laying hen breeding simulation system based on digital twin according to claim 1, characterized in that: The biological state quantification unit calculates the population aggregation entropy by: Multiple grid areas are virtually divided within the breeding space. An image analysis algorithm is used to identify and count the proportion of the total number of chickens distributed in each grid area to generate a regional distribution ratio. The regional distribution ratio is then substituted into a preset information entropy formula for calculation to generate group aggregation entropy.

4. The laying hen breeding simulation system based on digital twin according to claim 1, characterized in that: The process of generating the biological response time lag complexity index by the time lag feature analysis unit includes: Obtain the time-varying curve of the group aggregation entropy after the historical environmental regulation event; determine the moment when the absolute value of the derivative of the time-varying curve is continuously lower than the preset threshold to solve the lag time of the biological response; normalize the lag time to generate a dimensionless lag coefficient; and perform weighted summation of the dimensionless lag coefficients to generate a biological response time lag complexity index.

5. The laying hen breeding simulation system based on digital twin according to claim 1, characterized in that: The calculation process of the temperature requirement reference value by the control instruction generation unit includes: Obtaining preset standard temperature values, ideal metabolic rate index, and ideal aggregation entropy; Calculate the deviation between the current average metabolic rate index and the ideal metabolic rate index, and generate a metabolic rate temperature compensation value based on the preset temperature sensitivity coefficient; Calculate the deviation between the current group aggregation entropy and the ideal aggregation entropy, and generate the aggregation entropy temperature compensation value by combining the preset entropy sensitivity coefficient; The standard temperature value is corrected using the metabolic rate temperature compensation and the aggregate entropy temperature compensation to generate a temperature requirement reference value.

6. The laying hen breeding simulation system based on digital twin according to claim 1 or 5, characterized in that: The process of generating the temperature setting target value by the control instruction generating unit includes: The biological response time-lag complexity index is substituted into the preset exponential decay model to generate a control intensity factor; the difference between the temperature demand baseline value and the actual temperature currently collected is calculated to generate a temperature target deviation; the temperature target deviation is multiplied by the control intensity factor and summed with the actual temperature to generate the temperature setting target value for the next control time step.

7. The laying hen breeding simulation system based on digital twin according to claim 1, characterized in that: The prediction process of the model self-optimization unit for the biological state at the next moment is: Construct a state transition prediction model; take the individual metabolic rate index and group aggregation entropy observed at the current moment, and the temperature setting target value generated by the control instruction generation unit as input; perform calculations through the state transition prediction model, and output the predicted values ​​of the individual metabolic rate index and group aggregation entropy at the next moment.

8. The laying hen breeding simulation system based on digital twin according to claim 1 or 7, characterized in that: The iterative update process of the model self-optimization unit includes: After the control is completed, the individual metabolic rate index and group aggregation entropy of the next time step are actually observed; the actual observation value is compared with the predicted value, and the weighted deviation is calculated to generate the prediction error; the prediction error is used as the loss function, and the optimization algorithm is adopted with the goal of minimizing the loss function, and all preset adjustable weights and coefficients in the system are periodically updated.

Images