An adaptive regulation method and system for laying hen breeding environment

Abstract

The application discloses a kind of laying hen breeding environment self-adapting regulation and control method and system, method includes: according to current laying hen breeding stage belongs to normal growth stage or transition growth stage, when for normal growth stage, obtain prediction result by calling corresponding growth stage prediction model, and first reinforcement learning input feature is generated in combination with current environment data;When for transition growth stage, then the prediction model of adjacent front and rear stage is called, and the result output by two prediction models is smoothed by coupling correction mode to obtain coupling result, then second reinforcement learning input feature is generated in combination with current environment data, finally first reinforcement learning input feature or second reinforcement learning input feature is input into reinforcement learning controller to obtain environmental regulation and control action instruction, to regulate the environmental parameter in laying hen breeding environment, satisfy the self-adapting environmental regulation and control demand of laying hen whole cycle smooth, high precision.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture technology, and relates to, but is not limited to, an adaptive control method and system for the egg-laying hen farming environment. Background Technology

[0002] The parameters of the laying hen breeding environment, such as temperature, relative humidity, light, ventilation, and concentration of harmful gases, have a decisive impact on the growth and development, physical health, egg production performance, and breeding efficiency of laying hens. Reasonable parameter settings are the core link to achieve large-scale, high-yield, low-consumption, low-stress, and high-stability breeding. Therefore, it is crucial to implement precise adaptive environmental control throughout the entire life cycle of laying hens.

[0003] Current methods for regulating the environment of laying hen houses mainly employ fixed threshold control or single-model predictive control. Single-model predictive control is designed only for a single stage of peak egg production. While this method can reduce energy consumption and environmental fluctuations to some extent, it does not cover the entire life cycle of laying hens—the brooding period, rearing period, pre-laying period, peak laying period, and post-laying period—nor does it consider the relationship between the environment and egg production rate and quality at different growth stages. Its applicability and control precision are significantly limited, failing to meet the needs of stable, efficient, and high-precision adaptive environmental control throughout the entire life cycle of laying hens. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide an adaptive control method and system for the laying hen breeding environment, which aims to overcome the defects of the prior art, provide a breeding environment that is precisely adapted to the physiological needs of laying hens throughout their entire life cycle, effectively improve the laying rate and quality of laying hens, and at the same time ensure the health of the flock.

[0005] The specific technical solutions of this invention are as follows: This invention provides a method for adaptive control of the laying hen farming environment, comprising: Real-time environmental data of the chicken coop is collected, and the collected data is preprocessed to obtain the model input features. The current age and current egg production rate of laying hens are obtained in real time, and the current laying hen breeding stage is determined based on the current age and current egg production rate of the laying hens; If the current egg-laying hen breeding stage is a normal growth stage, then the model input features are input into the prediction model corresponding to the current egg-laying hen breeding stage for prediction, and a real-time prediction result list is obtained; the current environmental data and the real-time prediction result list are used as the first reinforcement learning input features; If the current stage of egg-laying hen farming is a transitional growth stage, then a target model is determined based on the current age of the egg-laying hen. The target model includes a pre-stage model and a post-stage model. The model input features are simultaneously input into the previous stage model and the next stage model for prediction, and the previous stage prediction results output by the previous stage model and the next stage prediction results output by the next stage model are coupled to obtain a coupling result; the current environment data and the coupling result are used as the second reinforcement learning input features. The first reinforcement learning input feature or the second reinforcement learning input feature is input into the reinforcement learning controller to obtain environmental regulation action instructions, so as to regulate the environmental parameters in the egg-laying hen breeding environment.

[0006] This invention provides an adaptive control system for egg-laying hen farming environment. The system includes a data preprocessing module, a stage determination module, a first model prediction module, a target model determination module, a second model prediction module, and an environmental control module, wherein: The data preprocessing module is used to collect real-time environmental data of the chicken house and preprocess the real-time collected environmental data to obtain the model input features; The stage determination module is used to obtain the current age and current egg production rate of the laying hens in real time, and determine the current laying hen breeding stage based on the current age and current egg production rate of the laying hens. The first model prediction module is used to input the model input features into the prediction model corresponding to the current egg-laying hen breeding stage if the current egg-laying hen breeding stage is a normal growth stage, and to obtain a real-time prediction result list; and to use the current environmental data and the real-time prediction result list as the first reinforcement learning input features. The target model determination module is used to determine a target model based on the current age of the laying hen if the current laying hen breeding stage is a transitional growth stage. The target model includes a pre-stage model and a post-stage model. The second model prediction module is used to simultaneously input the model input features into the previous stage model and the subsequent stage model for prediction, and to couple the previous stage prediction results output by the previous stage model and the subsequent stage prediction results output by the subsequent stage model to obtain a coupling result; the current environment data and the coupling result are used as the second reinforcement learning input features. The environmental control module is used to input the first reinforcement learning input feature or the second reinforcement learning input feature into the reinforcement learning controller to obtain environmental control action instructions, so as to control the environmental parameters in the egg-laying hen breeding environment.

[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, based on the physiological differences in the different growth stages of laying hens, the laying hen farming stage is divided into a normal growth stage and a transitional growth stage for environmental prediction and regulation. In the normal growth stage, the farming stage is accurately determined based on the age and egg production rate of the laying hens, and the corresponding prediction model is matched to output the prediction results. In the transitional growth stage, models of adjacent preceding and following stages are simultaneously invoked, and the prediction results are smoothed through a coupling correction method. This effectively solves the problems of a single model being difficult to adapt to the entire life cycle and the potential for environmental parameter jumps during stage transitions. Real-time environmental data, prediction results, or real-time environmental data and coupling results are input into a reinforcement learning controller, which autonomously learns the multi-parameter coupling relationships and outputs environmental regulation action commands. Without relying on manually set fixed thresholds, it can adaptively match the temperature, humidity, ammonia, ventilation, and light requirements of laying hens at different stages. This improves the accuracy of environmental parameter prediction and regulation, eliminates environmental abrupt changes and flock transfer stress caused by stage transitions, effectively reduces mortality during the transition period, balances equipment operating energy consumption and farming production performance, and achieves stable, accurate, and adaptive intelligent environmental regulation throughout the entire life cycle of laying hens, making it suitable for large-scale laying hen farming applications. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments 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, wherein: Figure 1 A flowchart illustrating an adaptive control method for the laying hen farming environment provided in an embodiment of the present invention; Figure 2 A flowchart illustrating the training process of the prediction model provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an adaptive control system for egg-laying hen farming environment provided in an embodiment of the present invention. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all embodiments. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention. 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.

[0010] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0011] It should be noted that the terms "first, second, and third" used in the embodiments of the present invention are only used to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present invention described herein can be implemented in an order other than that illustrated or described herein.

[0012] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which these embodiments of the invention pertain. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0013] Figure 1 This is a flowchart illustrating an adaptive control method for the laying hen farming environment provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes at least the following steps: Step S110: Collect the current environmental data of the chicken house in real time, and preprocess the collected current environmental data to obtain the model input features.

[0014] Specifically, real-time environmental data of the chicken coop is collected. After acquiring the collected environmental data, it is first aggregated according to the sampling interval (e.g., 15 minutes) to obtain data for each time step. Then, missing value checks are performed on all time step data. If one or two consecutive time steps of current environmental data are found to be missing, linear interpolation is used to fill in the missing values ​​to obtain valid data. If three or more consecutive time steps of current environmental data are found to be missing, the environmental data of the corresponding historical time step (the same time step in the past k days) is selected to calculate the average value, and the missing values ​​are filled based on the calculated average value of the environmental data to obtain valid data. Finally, the valid data is normalized (that is, the data of different dimensions in the valid data are converted into dimensionless data) to generate the model input features.

[0015] Furthermore, this embodiment collects real-time environmental data through sensors and monitoring equipment deployed within the chicken house. The current environmental data refers to the environmental data of the chicken house at the current time step. In this embodiment, the environmental data includes, but is not limited to, temperature, relative humidity, ammonia concentration, CO2 concentration, light intensity, and ventilation volume. The ventilation volume is obtained by identifying temperature, relative humidity, ammonia concentration, CO2 concentration, light intensity, feed intake, water intake, and activity level of the chickens using a ventilation volume calculation model. The ventilation volume calculation model in this embodiment employs a pre-trained LightGBM regression model.

[0016] The sensors and monitoring devices in this embodiment include, but are not limited to, temperature and humidity sensors, ammonia sensors, carbon dioxide sensors, light sensors, electronic tank weighing equipment, water meters, infrared thermal imagers, or depth cameras. Temperature and humidity sensors are used to collect the temperature (°C) and relative humidity (%RH) of the chicken house. Ammonia sensor, used to collect ammonia concentration (ppm). Carbon dioxide sensor, used to collect CO2 concentration (ppm); A light sensor is used to collect light intensity (lux). Electronic feeder weighing equipment is used to record the feed intake (kg) of chickens. A water meter is used to record drinking water volume (L). Infrared thermal imagers or depth cameras are used to estimate the activity level of chicken flocks (rate of change of moving pixels).

[0017] Step S120: Obtain the current age and current egg production rate of the laying hens in real time, and determine the current laying hen breeding stage based on the current age and current egg production rate.

[0018] In this context, "laying hen farming stage" refers to parameters used to differentiate the stages of a laying hen's entire life cycle. This embodiment, based on the biological developmental patterns of laying hens and large-scale farming technical regulations, divides the entire life cycle of laying hens into five consecutive stages: brooding period, rearing period, pre-laying period, peak laying period, and post-laying period. Furthermore, to improve the accuracy of the laying hen farming stage division, this embodiment determines the laying hen farming stage based on the age range of the hens and their egg production rate. Specifically, egg production rate is not considered during the brooding and rearing periods; the corresponding egg production rate requirements are left blank.

[0019] The specific biological developmental patterns of laying hens used in this embodiment are shown in Table 1: Table 1 The specific division of the pre-laying period, peak laying period, and post-laying period is as follows: Pre-laying period: The current age is ≥127 days and meets the corresponding laying rate requirements (laying rate ≥5% for 3 consecutive days); Peak egg production period: The current age is ≥176 days and meets the corresponding egg production rate requirements (egg production rate ≥90% for 7 consecutive days and egg production change rate ≤3%). Late laying period: The current age is ≥316 days and the corresponding laying rate requirement is met (the cumulative decrease in laying rate is ≥5% over 5 consecutive days).

[0020] Furthermore, if the above conditions for peak egg production are not met by age ≥190 days, the system will be forcibly triggered to enter peak egg production, and an abnormality warning mechanism will be activated to alert farm managers to check the health status of the flock. This forced triggering mechanism prevents the system from remaining in the previous stage for an extended period due to individual developmental differences, ensuring that environmental control is basically synchronized with physiological age.

[0021] The formula for calculating the rate of change in egg production is as follows: In the formula, Indicates the first Daily rate of change in egg production Indicates the first The daily overall egg production rate, Indicates the first The daily total egg production rate of the entire flock.

[0022] The age range and egg production rate requirements set above are merely illustrative examples of the present invention and do not constitute a limitation of the present invention. In practical applications, they can be dynamically adjusted according to the individual development status of the laying hens.

[0023] Furthermore, to adapt to the continuity of laying hen growth and development, reduce the stress response of laying hens during regrouping, and avoid model output jumps, this embodiment divides each laying hen growth stage into a normal growth stage and a transitional growth stage. The transitional growth stage refers to the last m days of the preceding laying hen growth stage and the first n days of the following laying hen growth stage. Based on the laying hens' adaptation requirements to environmental changes, the sum of m and n in this embodiment generally ranges from 4 to 12 days, with m and n being positive integers. Their specific values ​​can be the same or different.

[0024] If no transitional growth stage is set, for example, if the temperature output at time step t on day 42 of the brooding period is 28℃, while the temperature output at time step t+1 on day 43 of the rearing period is 22℃, a direct switch will result in a step change of 6℃, which, combined with the stress of relocation, will have an adverse effect on the flock. Therefore, in this embodiment, the brooding period is divided into 0-39 days (normal growth stage of the brooding period) and 40-42 days (transitional growth stage of the brooding period), and the rearing period is divided into 43-45 days (transitional growth stage of the rearing period) and 46-126 days (normal growth stage of the rearing period).

[0025] If the number of days in the transitional growth phase is set too short, less than 4 days, it will not cover the entire stress period, resulting in insufficient smoothing. If the number of days in the transitional growth phase is set too long, greater than 12 days, the transition will be too slow, causing the later-stage model to intervene in the earlier stage too early, resulting in environmental parameters deviating from the actual needs of the earlier stage. This embodiment sets the number of days in the transitional growth phase to 4-12 days, balancing stress period coverage and transition efficiency.

[0026] Step S130: If the current egg-laying hen breeding stage is a normal growth stage, then input the model input features into the prediction model corresponding to the current egg-laying hen breeding stage for prediction, and obtain a real-time prediction result list; use the current environmental data and the real-time prediction result list as the first reinforcement learning input features.

[0027] Specifically, after determining that the current stage of egg-laying hen farming is a normal growth stage, the model input features are input into the prediction model corresponding to the current stage of egg-laying hen farming for prediction, and a list of real-time prediction results is obtained.

[0028] Furthermore, if the current stage of egg-laying hen farming is the normal growth stage of the brooding period, the model input features are input into the prediction model corresponding to the brooding period for prediction, resulting in a first real-time prediction result list; if the current stage of egg-laying hen farming is the normal growth stage of the rearing period, the model input features are input into the prediction model corresponding to the rearing period for prediction, resulting in a second real-time prediction result list; if the current stage of egg-laying hen farming is the normal growth stage of the pre-laying period, the model input features are input into the prediction model corresponding to the pre-laying period for prediction, resulting in a third real-time prediction result list; if the current stage of egg-laying hen farming is the normal growth stage of the peak laying period, the model input features are input into the prediction model corresponding to the peak laying period for prediction, resulting in a fourth real-time prediction result list; if the current stage of egg-laying hen farming is the normal growth stage of the late laying period, the model input features are input into the prediction model corresponding to the late laying period for prediction, resulting in a fifth real-time prediction result list.

[0029] In this embodiment, the prediction models for each stage of egg-laying hen farming are obtained by training an LSTM model.

[0030] The prediction result list in this embodiment includes prediction results for multiple time steps. The prediction results for each time step include the predicted temperature value, predicted relative humidity, predicted light intensity value, and predicted ventilation volume.

[0031] After obtaining the list of real-time prediction results, the current environmental data and the list of real-time prediction results are used as the first reinforcement learning input features. Here, the first reinforcement learning input features refer to the feature set formed by the current environmental data and the list of real-time prediction results.

[0032] Step S140: If the current stage of egg-laying hen farming is a transitional growth stage, then the target model is determined based on the current age of the egg-laying hen. The target model includes a pre-stage model and a post-stage model.

[0033] The target model refers to the prediction model corresponding to the transitional growth stage, including prediction models for two adjacent laying hen growth stages. The prediction model corresponding to the preceding laying hen growth stage is the previous stage model, and the prediction model corresponding to the following laying hen growth stage is the subsequent stage model. In this embodiment, the transitional growth stages include the first transitional growth stage (m days after the brooding period + n days before the rearing period), the second transitional growth stage (m days after the rearing period + n days before the pre-laying period), the third transitional growth stage (m days after the pre-laying period + n days before the peak laying period), and the fourth transitional growth stage (m days after the peak laying period + n days before the late laying period).

[0034] Specifically, when the current age of the laying hen is m days after the brooding period or n days before the growing period, and the current egg production rate meets the corresponding egg production rate requirement, then the current laying hen breeding stage is determined to be the first transitional growth stage. The corresponding model for the previous stage is the prediction model for the brooding period, and the model for the later stage is the prediction model for the growing period.

[0035] When the current age of the laying hen is m days after the rearing period or n days before the pre-laying period, and the current laying rate meets the corresponding laying rate requirement, then the current laying hen breeding stage is determined to be the second transitional growth stage. The corresponding model for the previous stage is the prediction model for the rearing period, and the model for the later stage is the prediction model for the pre-laying period.

[0036] If the current age of the laying hen is m days after the pre-laying stage or n days before the peak laying stage, and the current laying rate meets the corresponding laying rate requirement, then the current laying hen breeding stage is determined to be the third transitional growth stage; the corresponding model for the previous stage is the prediction model for the pre-laying stage, and the model for the later stage is the prediction model for the peak laying stage.

[0037] If the current age of the laying hen is m days after the peak laying period or n days before the late laying period, and the current laying rate meets the corresponding laying rate requirement, then the current laying hen breeding stage is determined to be the fourth transitional growth stage; the corresponding model for the previous stage is the prediction model for the peak laying period, and the model for the later stage is the prediction model for the late laying period.

[0038] Step S150: Input the model input features into both the previous stage model and the next stage model for prediction, and couple the previous stage prediction results output by the previous stage model and the next stage prediction results output by the next stage model to obtain the coupling result; use the current environment data and the coupling result as the second reinforcement learning input features.

[0039] Specifically, when the current time step is in the first transitional growth stage, the prediction result of the previous stage is the prediction result in the first real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the second real-time prediction result list; when the current time step is in the second transitional growth stage, the prediction result of the previous stage is the prediction result in the second real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the third real-time prediction result list; when the current time step is in the third transitional growth stage, the prediction result of the previous stage is the prediction result in the third real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the fourth real-time prediction result list; when the current time step is in the fourth transitional growth stage, the prediction result of the previous stage is the prediction result in the fourth real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the fifth real-time prediction result list.

[0040] The prediction results in this embodiment include predicted temperature, predicted relative humidity, predicted ammonia concentration, and predicted ventilation volume. Understandably, since light intensity is fixed by staff according to the physiological stages of laying hens (e.g., 8 hours for rearing, 16 hours for laying), it does not dynamically change with temperature, relative humidity, ammonia concentration, and ventilation volume. Therefore, light intensity does not need to be included in model training. However, because light intensity affects the flock's activity level, egg production rate, and metabolic heat, which in turn affects temperature, relative humidity, ammonia, and ventilation volume, light intensity needs to be input during the model input stage; that is, the model input features need to include light intensity.

[0041] After determining the prediction results for the preceding and following stages, a coupling function is used to couple these results, yielding a coupled result. The coupling function is as follows: In the formula, Indicates the current time step. Indicates the current time step The corresponding coupling result, Indicates the current time step The corresponding fusion weights, Indicates the current time step The corresponding prediction results from the previous stage, Indicates the current time step The corresponding subsequent prediction results, Indicates the current time step The corresponding coupling correction term.

[0042] Since the coupling result of this embodiment is to solve the special problems corresponding to the transition growth stage, the drastic changes in ventilation volume and temperature during stage switching will cause a large fluctuation in ammonia concentration. Therefore, the coupling correction term of this embodiment includes an ammonia concentration coupling correction component to avoid the contradiction of "increasing ventilation to cool down, resulting in a sudden drop in ammonia concentration" or "reducing ventilation to keep warm, resulting in ammonia concentration exceeding the standard".

[0043] In addition, since light intensity is an independently adjustable parameter, it usually adopts a fixed time program (such as 16 hours of light per day, intensity 20-30 lux), and is not affected by the coupling of other environmental parameters such as temperature, relative humidity and ventilation. Even in the transitional growth stage, the light intensity remains stable and does not require smooth switching. Therefore, this embodiment does not set a light intensity coupling correction component.

[0044] In summary, the coupling correction terms in this embodiment include temperature coupling correction components, relative humidity coupling correction components, ammonia concentration coupling correction components, and ventilation volume coupling correction components.

[0045] in, In the formula, Indicates the current time step The corresponding age is the age during the transitional growth stage, i.e., the current transitional growth age; if the current transitional growth age belongs to the first transitional growth stage, then... Indicates the end date of the brooding period. This indicates the starting age of the developmental period; if the current transitional growth age belongs to the second transitional growth stage, then... Indicates the end date of the growth period. This indicates the starting age before egg production; if the current transitional growth age belongs to the third transitional growth stage, then... Indicates the end date of the pre-laying period. This indicates the starting age of peak egg production; if the current transitional growth age belongs to the fourth transitional growth stage, then... Indicates the age at which the peak egg-laying period ends. Indicates the starting age of the later stages of egg production; .

[0046] Furthermore, the coupling correction term in this embodiment is specifically implemented through the following steps: 1) Calculate the deviation term between the prediction results of the previous stage and the prediction results of the subsequent stage. In the formula, Indicates the current time step The corresponding temperature deviation, Indicates the current time step The temperature corresponding to the previous stage of prediction results. Indicates the current time step The temperature corresponding to the prediction results in the later stage; Indicates the current time step The relative humidity deviation corresponding to the previous stage of prediction results Indicates the current time step The relative humidity corresponding to the previous stage of prediction results Indicates the current time step The relative humidity corresponding to the later stage prediction results; Indicates the current time step The ammonia concentration deviation in the previous stage prediction results Indicates the current time step The ammonia concentration corresponding to the previous stage prediction results Indicates the current time step The corresponding ammonia concentration in the later-stage prediction results; Indicates the current time step The corresponding ventilation volume deviation in the previous stage of prediction results Indicates the current time step The ventilation volume corresponding to the previous stage of forecast results Indicates the current time step The corresponding ventilation volume in the later stage prediction results.

[0047] 2) Determine the coupling weight coefficients Based on historical data statistics, the correlation coefficients between various parameters were obtained, including: Temperature-relative humidity correlation coefficient Temperature-ammonia concentration correlation coefficient Temperature-ventilation volume correlation coefficient Correlation coefficient between relative humidity and ammonia concentration Relative humidity-ventilation volume correlation coefficient Correlation coefficient between ammonia concentration and ventilation volume The aforementioned correlation coefficients constitute the weighting coefficients for the coupling correction.

[0048] 3) Calculate the coupling correction components of each parameter. For temperature, relative humidity, ammonia concentration, ventilation volume, and light intensity, corresponding coupled correction components are constructed; among them, The temperature coupling correction component is: The relative humidity coupling correction component is: The ammonia concentration coupling correction component is: The ventilation volume coupling correction component is: In the formula, This indicates the temperature-relative humidity coupling correction weights. This indicates the temperature-ammonia concentration coupling correction weight. This indicates the temperature-ventilation volume coupling correction weight. This indicates the relative humidity-ammonia concentration coupling correction weight. This indicates the relative humidity-ventilation volume coupling correction weight. Indicates the ammonia concentration-ventilation volume coupling correction weight. To avoid overcorrection, this embodiment sets upper and lower limits for the correction components (e.g., temperature correction not exceeding ±0.5℃, relative humidity correction not exceeding ±3%, and ventilation volume correction not exceeding ±50m). 3 / h), to ensure that the corrected parameters remain within the physiological tolerance range of laying hens.

[0049] Finally, the current time step is generated. Corresponding coupling correction term After obtaining the coupling result, the current environment data and the coupling result are used as the input features for the second reinforcement learning. The second reinforcement learning input features refer to the feature set formed by the current environment data and the coupling result.

[0050] Step S160: Input the first reinforcement learning input feature or the second reinforcement learning input feature into the reinforcement learning controller to obtain environmental regulation action instructions, so as to regulate the environmental parameters in the egg-laying hen breeding environment.

[0051] Among them, the Reinforcement Learning Controller (RL Controller) is an autonomous decision-making system based on trial-and-error learning. Through continuous interaction with the environment, it constantly optimizes its control strategy, ultimately maximizing long-term cumulative rewards. Unlike traditional rule-based controllers such as PID and fuzzy control, the RL controller does not require a pre-established precise mathematical model of the environment. It can autonomously learn the optimal control laws of complex, multi-parameter coupled systems, making it more suitable for dynamic, multi-objective, and strongly coupled environmental control scenarios such as egg-laying hen farming.

[0052] Specifically, when the current egg-laying hen breeding stage is the normal growth stage, the first reinforcement learning input feature is input into the reinforcement learning controller to obtain environmental regulation action instructions, so as to regulate the environmental parameters in the egg-laying hen breeding environment; if the current egg-laying hen breeding stage is the transitional growth stage, the second reinforcement learning input feature is input into the reinforcement learning controller to obtain environmental regulation action instructions, so as to regulate the environmental parameters in the egg-laying hen breeding environment.

[0053] Before inputting the first reinforcement learning input feature or the second reinforcement learning input feature into the reinforcement learning controller, the data in the first reinforcement learning input feature or the second reinforcement learning input feature needs to be aligned in the time dimension to obtain aligned data; then the aligned data is normalized.

[0054] After obtaining the normalized and aligned data, the data is input into the reinforcement learning controller to obtain the target action vector; then the target action vector is decoded to obtain the environmental regulation action command; finally, the environmental parameters in the egg-laying hen breeding environment are regulated through the environmental regulation action command.

[0055] Furthermore, the normalized aligned data is input into the reinforcement learning controller to obtain the target action vector, specifically including the following steps: Step 1: Input the normalized aligned data into the reinforcement learning controller to obtain the initial action vector. The state vector corresponding to the normalized aligned data is as follows: In the formula, This represents the first reinforcement learning input feature corresponding to the current time step. This represents the second reinforcement learning input feature corresponding to the current time step. Indicates the current time step. This indicates the temperature at the current time step. This indicates the relative humidity at the current time step. This indicates the ammonia concentration at the current time step. This indicates the light intensity at the current time step. This indicates the ventilation volume corresponding to the current time step. This indicates the predicted temperature at the current time step in the real-time prediction results list. This indicates the predicted relative humidity at the current time step in the real-time forecast results list. This indicates the predicted ammonia concentration at the current time step in the real-time prediction results list. This indicates the predicted ventilation volume corresponding to the current time step in the real-time prediction results list. This indicates the coupling result corresponding to the current time step.

[0056] It should be noted that if the current time step corresponds to an age within the normal growth stage, then... It does not exist; if the age corresponding to the current time step is the age of the transitional growth stage, then , , , It does not exist.

[0057] The initial action vector is: In the formula, This indicates the initial opening degree of the ventilation equipment. This represents the power coefficient of the heating equipment at the initial moment. This represents the power coefficient of the refrigeration equipment at the initial moment. This represents the light intensity adjustment coefficient at the initial moment.

[0058] Step 2: Decode the initial motion vector into equipment control commands. Based on these commands, the sensors collect new environmental parameters and production data to obtain the state vector for the next time step. .

[0059] The specific decoding process is as follows: Step 3: Based on the state vector of the next time step, calculate the immediate reward value using the reward function. The reward function is: In the formula, These are the preset weighting coefficients. This indicates the preset temperature deviation reference value. This indicates the preset relative humidity deviation reference value. This indicates the preset safe upper limit for ammonia concentration. express , This indicates the baseline value for ventilation volume deviation.

[0060] Step 4: Calculate the delayed reward and update the cumulative reward based on the immediate reward value and the historical action sequence.

[0061] In this embodiment, the historical action sequence refers to the action vector and state vector stored in the historical time step.

[0062] Specifically, daily action sequences are recorded, and after a historical cycle (e.g., 14 days), delayed rewards are calculated based on actual changes in egg production rate; then, these delayed rewards are distributed through Eligibility Traces. The corresponding historical actions within the day; finally, the cumulative reward is updated based on the delayed reward.

[0063] The formula for calculating the delayed reward is as follows: In the formula, Indicates delayed reward, This represents the weighting coefficient of the reward item for the rate of change in egg production. Indicates the current time step The rate of change in egg production rate This indicates the rate of change in maximum egg production.

[0064] The formula for calculating cumulative rewards is: In the formula, Indicates the current time step Cumulative rewards As a discount factor, Indicates the maximum number of historical time steps. Represents the first in a historical period Each time step Indicates the first The discount factor corresponding to each time step Indicates the first The discount factor corresponding to each time step Indicates the first The instant reward value corresponding to each time step. This indicates a delayed reward.

[0065] Step 5: Calculate the advantage function value based on the updated cumulative reward and state value estimate.

[0066] Specifically, state value estimates are output through a value network. And according to the formula Calculate the dominance function value.

[0067] Step 6: Update the policy network parameters using the PPO algorithm based on the advantage function value.

[0068] Specifically, the policy network parameters are updated by calculating the ratio of action probabilities between the old and new policies, the pruning coefficient, and the dominance function value through the PPO pruning objective function.

[0069] The objective function for PPO pruning is: In the formula, This represents the pruning objective function of the PPO algorithm. Describe the objective function A cropping operation was used; This represents the ratio of the probability of actions using the new and old strategies. This represents the pruning factor (in this embodiment, the pruning factor is 0.2 to prevent training instability caused by excessively large single update steps). This represents the value of the dominance function. Wherein, This represents the ratio of the probability of the new strategy to the probability of the old strategy, with the numerator being the probability of the new strategy (the updated strategy) at the current time step. The corresponding first reinforcement learning input feature Take action vector The probability is given by the first factor, and the denominator is the probability of the old policy (the policy before the update) under the same state action pair. This ratio measures the magnitude of the policy update; the greater the deviation of the ratio from 1, the larger the policy change. In the formula, Indicates the current time step The corresponding ventilation equipment opening degree, Indicates the current time step The corresponding power coefficient of the heating equipment, Indicates the current time step The corresponding power factor of the refrigeration equipment, Indicates the current time step The corresponding light intensity adjustment coefficient.

[0070] Step 7: Based on the updated policy network parameters Output the target action vector.

[0071] The target action vector refers to the action vector output based on the updated policy network parameters.

[0072] Repeat steps one through seven to continuously optimize the action vector and ultimately maximize the long-term cumulative reward.

[0073] Furthermore, such as Figure 2 As shown in the figure, the adaptive control method for egg-laying hen farming environment provided in this embodiment of the invention also includes a prediction model training process. The prediction model training process specifically includes the following steps: Step S210: Preprocess the sample data of different egg-laying hen breeding stages to obtain the baseline sample data and boundary transition sample data of each egg-laying hen breeding stage.

[0074] Among them, the baseline sample data refers to the environmental data corresponding to the normal growth age used for training and testing the LSTM model, and the boundary transition sample data refers to the environmental data corresponding to the transition growth age used for training and testing the LSTM model.

[0075] The preprocessing steps in this embodiment are the same as those in step S120, and will not be repeated here to avoid repetition.

[0076] Step S220: Divide the benchmark sample data into benchmark training data and benchmark test data, and divide the boundary transition sample data into transition training data and transition test data.

[0077] Among them, benchmark training data refers to the environmental data used to train the LSTM model in the benchmark sample data, and benchmark test data refers to the environmental data used to test the trained LSTM model in the benchmark sample data; boundary transition training data refers to the environmental data used to train the LSTM model in the boundary transition sample data, and boundary transition test data refers to the environmental data used to test the trained LSTM model in the boundary transition sample data.

[0078] Specifically, the benchmark sample data is divided into benchmark training data and benchmark test data according to a preset ratio (such as 8:2 or 7:3), and the boundary transition sample data is divided into transition training data and transition test data.

[0079] Step S230: Input the benchmark training data of different egg-laying hen breeding stages into the corresponding LSTM models for training to obtain the basic models of each egg-laying hen breeding stage.

[0080] The base model refers to the model trained based on environmental data from normal growth stages.

[0081] Furthermore, to prevent overfitting, a dropout layer is added after each LSTM layer with a dropout rate of 0.2, and an early stopping strategy is adopted during training, stopping training when the validation set loss no longer decreases for 10 consecutive rounds.

[0082] Step S240: Input the benchmark test data of different egg-laying hen breeding stages into the corresponding basic model for testing, obtain the model output results, calculate the model confidence based on the model output results, and obtain the training model when the model confidence is greater than the preset confidence threshold.

[0083] Step S250: Add transitional training data to the baseline training data to construct a fine-tuned training dataset.

[0084] Specifically, based on the proportion of the transition phase in the entire cycle, a data fusion ratio is set (e.g., 80% baseline data and 20% transition data). Transition training data is then added to the baseline training data according to this ratio to construct a fine-tuning training dataset. It should be noted that the amount of transition training data added to the baseline training data should not be excessive, so that the trained model retains its overall stable control capability while also possessing smooth control capability during the transition phase.

[0085] Step S260: Input the fine-tuning training datasets for different egg-laying hen breeding stages into the corresponding training model for fine-tuning to obtain the adjusted model.

[0086] Among them, adjusting the model refers to adding transitional training data to the baseline training data to form a fine-tuned training dataset, and then inputting the fine-tuned training dataset into the training model to obtain the model.

[0087] Step S270: Input the transitional test data into the two associated training models simultaneously for testing to obtain the test results of the previous stage and the test results of the next stage.

[0088] In this embodiment, the two associated training models refer to the training models of two adjacent egg-laying hen growth stages.

[0089] Step S280: Perform coupling processing on the test results of the previous stage and the test results of the subsequent stage to obtain the coupling processing result, and calculate whether the error between the coupling processing result and the corresponding label value is within the preset error range.

[0090] The coupling process in this step is the same as that in step S150, and will not be repeated here to avoid repetition.

[0091] Step S290: When the error between the coupling processing result and the corresponding label value is within the preset error range, it indicates that the model training is complete and a prediction model for different egg-laying hen breeding stages is obtained.

[0092] Furthermore, an adaptive control method for the laying hen farming environment also includes: The number of newly added sample data is counted in real time as the new sample quantity. When the number of new samples reaches the preset sample increment, the incremental training mechanism of the model is triggered based on the new sample data, and the training process of the prediction model is re-executed. The incremental training mechanism in this embodiment adopts the mini-batch gradient descent method, with a learning rate of one-tenth of that at the initial training, and only updates the parameters of the last two layers of the model to maintain the model's adaptability to the individual characteristics of the current batch of laying hens.

[0093] This application's embodiments divide the egg-laying hen breeding process into five stages: brooding period, rearing period, pre-laying period, peak laying period, and post-laying period. Independent LSTM prediction models are constructed for each stage. Missing value imputation and normalization preprocessing ensure input data quality. Overfitting is suppressed by combining a dropout layer and an early termination strategy. A fine-tuning method combining baseline training data with transitional training data and an incremental training mechanism are employed to continuously adapt to individual hen growth patterns and changes in the breeding environment. Simultaneously, a transitional growth stage is set between adjacent egg-laying hen growth stages. A linear weighted fusion function based on age is constructed, incorporating temperature, relative humidity, ammonia concentration, and ventilation volume coupling correction terms to smoothly couple and fuse the prediction results of the models before and after the transition, effectively avoiding loops during stage transitions. When the target environmental value experiences a step change, a multi-objective reward function is constructed that includes environmental deviation, ammonia concentration, equipment energy consumption, and the rate of change in egg production. Relying on the PPO reinforcement learning controller, the control strategy is iteratively optimized based on real-time environmental status, prediction results, and coupling results. Adaptive action commands for ventilation, heating, cooling, and lighting equipment are output to achieve closed-loop intelligent control throughout the entire cycle. This invention breaks through the limitations of traditional fixed threshold and single-model control, significantly improves the prediction accuracy of environmental parameters at each stage and the stability of stage switching, greatly reduces the decrease in feed intake and mortality rate during the transition period caused by flock transfer stress, and optimizes equipment operating energy consumption while accurately matching the physiological environmental needs of laying hens at each stage, effectively improving egg production stability and the overall economic benefits of large-scale laying hen farming.

[0094] Figure 3 This is a system block diagram of an adaptive control system for egg-laying hen farming environment provided in an embodiment of the present invention. Figure 3 As shown, this embodiment provides an adaptive control system for egg-laying hen farming environment. The system includes a data preprocessing module, a stage determination module, a first model prediction module, a target model determination module, a second model prediction module, and an environmental control module, wherein: The data preprocessing module is used to collect real-time environmental data of the chicken house and preprocess the real-time collected environmental data to obtain the model input features; The stage determination module is used to obtain the current age and current egg production rate of laying hens in real time, and determine the current laying hen breeding stage based on the current age and current egg production rate. The first model prediction module is used to input the model input features into the prediction model corresponding to the current egg-laying hen breeding stage if the current egg-laying hen breeding stage is a normal growth stage, and to obtain a real-time prediction result list; the current environmental data and the real-time prediction result list are used as the first reinforcement learning input features. The target model determination module is used to determine the target model based on the current age of the laying hens if the current stage of laying hen farming is a transitional growth stage. The target model includes a previous stage model and a subsequent stage model. The second model prediction module is used to simultaneously input the model input features into the previous stage model and the subsequent stage model for prediction, and to couple the previous stage prediction results output by the previous stage model and the subsequent stage prediction results output by the subsequent stage model to obtain the coupling result; the current environment data and the coupling result are used as the second reinforcement learning input features. The environmental control module is used to input the first reinforcement learning input feature or the second reinforcement learning input feature into the reinforcement learning controller to obtain environmental control action instructions, so as to regulate the environmental parameters in the egg-laying hen breeding environment.

[0095] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of the invention, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the invention. The sequence numbers of the above-described embodiments of the invention are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0096] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0097] In the several embodiments provided by this invention, it should be understood that the disclosed methods can be implemented in other ways. The methods disclosed in the several method embodiments provided by this invention can be arbitrarily combined to obtain new method embodiments without conflict. The features disclosed in the several method embodiments provided by this invention can be arbitrarily combined to obtain new method embodiments without conflict.

[0098] The above description is merely an embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for adaptive control of the laying hen farming environment, characterized in that, include: Real-time environmental data of the chicken coop is collected, and the collected data is preprocessed to obtain the model input features. The current age and current egg production rate of laying hens are obtained in real time, and the current laying hen breeding stage is determined based on the current age and current egg production rate of the laying hens; If the current egg-laying hen breeding stage is a normal growth stage, then the model input features are input into the prediction model corresponding to the current egg-laying hen breeding stage for prediction, and a real-time prediction result list is obtained; the current environmental data and the real-time prediction result list are used as the first reinforcement learning input features; If the current stage of egg-laying hen farming is a transitional growth stage, then a target model is determined based on the current age of the egg-laying hen. The target model includes a pre-stage model and a post-stage model. The model input features are simultaneously input into the previous stage model and the next stage model for prediction, and the previous stage prediction results output by the previous stage model and the next stage prediction results output by the next stage model are coupled to obtain a coupling result; the current environment data and the coupling result are used as the second reinforcement learning input features. The first reinforcement learning input feature or the second reinforcement learning input feature is input into the reinforcement learning controller to obtain environmental regulation action instructions, so as to regulate the environmental parameters in the egg-laying hen breeding environment.

2. The adaptive control method for the laying hen breeding environment according to claim 1, characterized in that, The preprocessing of the real-time collected current environmental data to obtain model input features includes: The real-time environmental data is aggregated according to the sampling interval to obtain data for each time step, and missing values ​​are checked for all time step data. When missing environmental data is detected in one time step or two consecutive time steps, linear interpolation is used to fill in the missing values ​​to obtain valid data. When missing environmental data is detected for three or more consecutive time steps, the environmental data of the corresponding historical time step is used to calculate the average value, and the missing values ​​are filled based on the calculated average value of the environmental data to obtain valid data. Normalize the valid data to generate the input features for the model.

3. The adaptive control method for the laying hen breeding environment according to claim 1, characterized in that, If the current egg-laying hen breeding stage is a normal growth stage, then the model input features are input into the prediction model corresponding to the current egg-laying hen breeding stage for prediction, and a real-time prediction result list is obtained, including: If the current stage of egg-laying hen breeding is a normal growth stage of the brooding period, then the model input features are input into the prediction model corresponding to the brooding period for prediction, and a first real-time prediction result list is obtained. If the current egg-laying hen breeding stage belongs to the normal growth stage of the rearing period, then the model input features are input into the prediction model corresponding to the rearing period for prediction, and a second real-time prediction result list is obtained. If the current egg-laying hen breeding stage is a normal growth stage in the pre-laying stage, then the model input features are input into the prediction model corresponding to the pre-laying stage for prediction, and a third real-time prediction result list is obtained. If the current egg-laying hen breeding stage is a normal growth stage during the peak egg-laying period, then the model input features are input into the prediction model corresponding to the peak egg-laying period for prediction, and a fourth real-time prediction result list is obtained. If the current stage of egg-laying hen farming is a normal growth stage in the late egg-laying period, then the model input features are input into the prediction model corresponding to the late egg-laying period for prediction, and a fifth real-time prediction result list is obtained.

4. The adaptive control method for the laying hen breeding environment according to claim 3, characterized in that, The adaptive control method for the egg-laying hen farming environment also includes: If the current age of the laying hen is m days after the brooding period or n days before the growing period, and the current egg production rate meets the corresponding egg production rate requirement, then the current laying hen breeding stage is determined to be the first transitional growth stage. If the current age of the laying hen is m days after the rearing period or n days before the pre-laying period, and the current laying rate meets the corresponding laying rate requirement, then the current laying hen breeding stage is determined to be the second transitional growth stage. When the current age of the laying hen is m days after the pre-laying period or n days before the peak laying period, and the current laying rate meets the corresponding laying rate requirement, then the current laying hen breeding stage is determined to be the third transitional growth stage. If the current age of the laying hen is m days after the peak laying period or n days before the late laying period, and the current laying rate meets the corresponding laying rate requirement, then the current laying hen breeding stage is determined to be the fourth transitional growth stage.

5. The adaptive control method for the laying hen breeding environment according to claim 4, characterized in that, The coupling process of the preceding stage prediction result output by the preceding stage model and the following stage prediction result output by the following stage model to obtain the coupling result includes: The prediction results of the previous stage and the prediction results of the subsequent stage are coupled using a coupling function to obtain a coupling result; wherein the coupling function is: In the formula, Indicates the current time step. Indicates the current time step The corresponding coupling result, Indicates the current time step The corresponding fusion weights, Indicates the current time step The corresponding prediction results from the previous stage, Indicates the current time step The corresponding subsequent prediction results, Indicates the current time step The corresponding coupling correction term; where, In the formula, Indicates the current time step The corresponding age is the age during the transitional growth stage, i.e., the current transitional growth age; if the current transitional growth age belongs to the first transitional growth stage, then... Indicates the end date of the brooding period. This indicates the starting age of the developmental period; if the current transitional growth age belongs to the second transitional growth stage, then... Indicates the end date of the growth period. This indicates the starting age before egg production; if the current transitional growth age belongs to the third transitional growth stage, then... Indicates the end date of the pre-laying period. This indicates the starting age of peak egg production; if the current transitional growth age belongs to the fourth transitional growth stage, then... Indicates the age at which the peak egg-laying period ends. Indicates the starting age of the later stages of egg production; .

6. The adaptive control method for the laying hen farming environment according to claim 5, characterized in that, The adaptive control method for the egg-laying hen farming environment also includes: When the current time step is in the first transitional growth stage, the prediction result of the previous stage is the prediction result in the first real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the second real-time prediction result list. When the current time step is in the second transitional growth stage, the prediction result of the previous stage is the prediction result in the second real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the third real-time prediction result list. When the current time step is in the third transitional growth stage, the prediction result of the previous stage is the prediction result in the third real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the fourth real-time prediction result list. When the current time step is in the fourth transitional growth stage, the prediction result of the previous stage is the prediction result in the fourth real-time prediction result list, and the prediction result of the subsequent stage is the prediction result in the fifth real-time prediction result list.

7. The adaptive control method for the laying hen farming environment according to claim 1, characterized in that, The step of inputting the first reinforcement learning input feature or the second reinforcement learning input feature into the reinforcement learning controller to obtain environmental control action instructions includes: Data alignment is performed on the first reinforcement learning input features or the second reinforcement learning input features to obtain aligned data; The alignment data is normalized, and the normalized alignment data is input into the reinforcement learning controller to obtain the target action vector; The target action vector is decoded to obtain the environmental control action command.

8. The adaptive control method for the laying hen breeding environment according to claim 1, characterized in that, The training process of the prediction model includes: Preprocess the sample data of different egg-laying hen breeding stages to obtain the baseline sample data and boundary transition sample data of each egg-laying hen breeding stage; The benchmark sample data is divided into benchmark training data and benchmark test data, and the boundary transition sample data is divided into transition training data and transition test data. The baseline training data for different egg-laying hen breeding stages are input into the corresponding LSTM models for training to obtain the basic models for each egg-laying hen breeding stage. Benchmark test data from different egg-laying hen breeding stages are input into the corresponding basic model for testing, and the model output results are obtained. The model confidence is calculated based on the model output results. When the model confidence is greater than the preset confidence threshold, the training model is obtained. Transitional training data is added to the baseline training data to construct a fine-tuned training dataset; The fine-tuning training datasets for different egg-laying hen breeding stages are input into the corresponding training models for fine-tuning to obtain the adjusted models. The transition test data is simultaneously input into two related adjustment models for testing, to obtain the test results of the previous stage and the test results of the next stage. The test results of the previous stage and the test results of the subsequent stage are coupled to obtain a coupling result. The error between the coupling result and the corresponding label value is calculated to see if it is within a preset error range. When the error between the coupling processing result and the corresponding label value is within the preset error range, it indicates that the model training is complete and a prediction model for different egg-laying hen breeding stages is obtained.

9. The adaptive control method for the laying hen breeding environment according to claim 8, characterized in that, After obtaining the prediction model, the adaptive control method for the egg-laying hen farming environment further includes: The number of newly added sample data is counted in real time as the number of new samples. When the number of new samples reaches the preset sample increment, the model incremental training mechanism is triggered based on the new sample data, and the training process of the prediction model is re-executed.

10. An adaptive control system for the laying hen farming environment, characterized in that, The system includes a data preprocessing module, a stage determination module, a first model prediction module, a target model determination module, a second model prediction module, and an environmental control module, wherein: The data preprocessing module is used to collect real-time environmental data of the chicken house and preprocess the real-time collected environmental data to obtain the model input features; The stage determination module is used to obtain the current age and current egg production rate of the laying hens in real time, and determine the current laying hen breeding stage based on the current age and current egg production rate of the laying hens. The first model prediction module is used to input the model input features into the prediction model corresponding to the current egg-laying hen breeding stage if the current egg-laying hen breeding stage is a normal growth stage, and to obtain a real-time prediction result list; and to use the current environmental data and the real-time prediction result list as the first reinforcement learning input features. The target model determination module is used to determine a target model based on the current age of the laying hen if the current laying hen breeding stage is a transitional growth stage. The target model includes a pre-stage model and a post-stage model. The second model prediction module is used to simultaneously input the model input features into the previous stage model and the subsequent stage model for prediction, and to couple the previous stage prediction results output by the previous stage model and the subsequent stage prediction results output by the subsequent stage model to obtain a coupling result; the current environment data and the coupling result are used as the second reinforcement learning input features. The environmental control module is used to input the first reinforcement learning input feature or the second reinforcement learning input feature into the reinforcement learning controller to obtain environmental control action instructions, so as to control the environmental parameters in the egg-laying hen breeding environment.

Images