A system and method for coordinated environmental control in stacked cage chicken houses
By constructing a cage-level physiological and environmental coupled tension field and an adaptive airflow response surface, the problems of heat stress and airflow nonuniformity in poultry flocks in stacked cage chicken houses were solved, achieving real-time and stable environmental coordinated regulation and improving the thermal environment stability and air quality in the chicken house.
Patent Information
- Application Number
- CN202611569999.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-12
- Publication Date
- 2026-11-10
AI Technical Summary
Existing environmental control systems for stacked cage chicken houses lack the ability to directly perceive the real-time physiological state of poultry flocks, resulting in control decisions lagging behind physiological needs. They are unable to effectively cope with the non-uniformity of airflow organization and thermal inertia delay caused by the stacked structure, leading to heat stress and air quality problems in poultry flocks.
By constructing a cage-level physiological and environmental coupled tension field, analyzing the acoustic characteristics of flock behavior, and combining real-time microenvironmental data, a non-uniform spatial weight distribution is generated. Adaptive airflow response surface correction is performed, and airflow dead zones and turbulent hot spots are mapped as virtual control anchor points. Multivariate decoupling operations are performed to generate a collaborative driving parameter set. A time-series predictive compensator is introduced to offset thermal inertia delay and output environmental collaborative control commands.
It enables real-time integration of poultry flock physiological stress feedback into environmental regulation, eliminates pneumatic interference from multiple actuators, reduces temperature overshoot and fluctuations during regulation, improves the stability and uniformity of the thermal environment inside the house, and ensures the continuity of regulation and reliability of the system under extreme conditions in the event of sensor failure.
Smart Images

Figure CN123067641A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of environmental regulation technology, specifically to a synergistic environmental regulation system and method for stacked cage chicken houses. Background Technology
[0002] Cascade cage rearing has become the mainstream method for modern large-scale layer and broiler chicken farming due to its advantages such as high land utilization and high stocking density. However, with the increase in the number of cage layers and the increase in stocking density, the control of the microenvironment inside the cage faces unprecedented challenges. Existing environmental control systems for cascade cage chicken houses mostly rely on threshold feedback from physical sensors such as temperature, humidity, and ammonia, and use PID or fuzzy control algorithms to independently or simply link and adjust fans, wet curtains, and deflectors. This traditional control strategy has the following technical defects: First, existing environmental control systems lack the ability to directly perceive the real-time physiological state of poultry flocks. Meeting environmental parameter standards does not equate to the flock feeling comfortable. Due to individual differences, age variations, and the cumulative effect of heat stress, even if sensors show temperature and humidity within the set range, the flock may still be in a state of latent heat stress. Existing technologies fail to deeply integrate the flock's biological feedback with environmental parameters, resulting in regulatory decisions lagging behind physiological needs. For example, in a case study of an 8-layer stacked cage laying hen house with 100,000 birds operating in summer, when the outside temperature suddenly rose to 35°C, the traditional environmental control system, relying on temperature sensors located in the middle of the cages, failed to respond to the actual temperature changes. The system determined that the environment met the standard with a feedback value of 26.5℃ and maintained the fan frequency at 40Hz. However, the actual temperature in the upper cages had reached 29℃ due to the accumulation of hot air, and the high humidity made the chickens feel stuffy. The flock exhibited obvious heat stress behaviors such as panting and high-frequency chirping. Because the system could not detect the physiological stress signals represented by the acoustics, it did not trigger the fan frequency to increase until the temperature sensor reading in the upper cages exceeded the alarm threshold of 28℃ after 30 minutes. By this time, some laying hens had already experienced acute heat stress, resulting in a 3.2% decrease in the daily egg production rate and an increase in the rate of broken eggs. This exposed the lag in the feedback of a single physical parameter in perceiving the true physiological needs of poultry flocks.
[0003] Secondly, the layered structure leads to highly non-uniform and strongly coupled airflow organization within the shed. Traditional control methods treat fans, wet curtains, and deflectors as independent actuators, neglecting the aerodynamic interference effects of multiple actuators in complex three-dimensional space. When attempting to eliminate airflow dead zones in a certain area, new turbulent hotspots are often induced in other areas, creating a control oscillation phenomenon where pressing down one problem only creates another. In the aforementioned case of the same laying hen shed, to eliminate the airflow dead zone with wind speeds below 0.2 m / s in the middle cages of the third row of cages, the maintenance personnel manually adjusted the opening of the deflector in the corresponding area from 30° to 5°. 5°; Although this operation increased the wind speed in the target cage to 0.4 m / s, it did not consider the deflection effect of the change in the angle of the deflector on the downstream airflow, causing the wind speed in the lower cage of the fourth adjacent row of cages to surge to 1.8 m / s instantly and form strong turbulence, causing panic among the chickens in that area and a decrease in feed intake; Subsequently, the system automatically adjusted the angle of the deflector, which caused the original dead zone problem to recur. This repeated oscillation up to 4 times before stabilizing, and the whole process took nearly 2 hours. During this period, the chickens were in a state of discomfort, which shows the failure of the traditional decoupling control method in dealing with the complex aerodynamic coupling of stacked cages.
[0004] Meanwhile, the stacked cages themselves have significant thermal inertia, and there is a time lag between changes in environmental parameters and the actual perception of the poultry. Existing systems lack a time-series compensation mechanism for this structural thermal inertia, further exacerbating overshoot and fluctuations in regulation. In the case of switching winter heating in this laying hen house, when the outdoor temperature dropped to -5℃ at night, the system detected that the air inlet temperature was too low and immediately increased the output power of the hot air furnace to 80%. However, due to the combined thermal inertia time constant of the 8-layer metal cages and the poultry organisms being approximately 240s, there was a significant lag in heat transfer to the middle cages. While waiting for feedback, the system continued to increase heating. By the time the middle temperature sensor responded, the actual air supply temperature had exceeded the target value by 4℃, causing the poultry to experience alternating hot and cold shocks in a short period of time. The next day, the incidence of respiratory diseases was 1.8 percentage points higher than normal. This case demonstrates that traditional control strategies lacking thermal inertia time-series compensation are unable to match the dynamic thermal response characteristics of stacked structures, which is an important reason for environmental fluctuations and health risks to poultry. Summary of the Invention
[0005] To solve the above-mentioned technical problems, a collaborative control system and method for the environment of stacked cage chicken houses is provided. This technical solution solves the above-mentioned problems.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for coordinated environmental control in a stacked cage chicken house includes: Acquire real-time microenvironment sensor data streams, poultry flock behavior acoustic feature sequences, and equipment actuator feedback status for each cage in the stacked cage chicken house; Based on the analysis of the acoustic feature sequence of poultry flock behavior, the heat stress voiceprint index of poultry flock is analyzed, and combined with the temperature and humidity gradient in the real-time microenvironment sensing data stream, a cage-level physiological and environmental coupling tension field is constructed. Based on the dynamic generation of the cage-level physiological and environmental coupled tension field, a non-uniform spatial weight distribution is generated, and the pre-constructed standard airflow organization model is adaptively deformed and corrected accordingly to obtain the actual airflow response surface. The airflow dead zone and turbulence hot spot region are extracted from the actual airflow response surface, and the airflow dead zone and turbulence hot spot region are mapped as virtual control anchor points; Using the virtual control anchor point as the constraint boundary, a multivariate decoupling operation is performed on the feedback state of the device actuator to generate a set of collaborative drive parameters including fan speed, wet curtain opening degree and guide vane angle; The collaborative driving parameter set is input into a preset timing prediction compensator to offset the inherent thermal inertia delay of the stacked structure and output the final environmental collaborative control command.
[0007] Furthermore, the construction of the cage-level physiological-environment coupled tension field includes: A short-time Fourier transform was performed on the acoustic feature sequence of the flock behavior to extract the acoustic energy spectral density representing the frequency of panting and calling. The instantaneous stress deviation is obtained by performing a difference operation between the voiceprint energy spectral density and the preset reference resting voiceprint. The temperature difference and humidity difference between adjacent cages are obtained from the real-time microenvironment sensing data stream, and the local environmental non-uniformity is calculated. Using the instantaneous stress deviation as a scalar weight and the local environmental inhomogeneity as a vector basis, the scalar weight is used to perform scalar multiplication modulation on the vector basis to generate the cage-level physiological and environmental coupling tension field.
[0008] Furthermore, the adaptive deformation correction of the pre-constructed standard airflow organization model to obtain the actual airflow response surface includes: The standard airflow organization model is discretized into a three-dimensional voxel mesh; The coupled tension field between the physiological and environmental conditions at the cage level is projected onto each node of the three-dimensional voxel mesh to obtain the driving force for node deformation. Based on the node deformation driving force, elastic deformation iteration is performed on the boundary coordinates of each voxel mesh until the system potential energy converges. Smooth interpolation reconstruction is performed on the deformed voxel mesh to generate the actual airflow response surface.
[0009] Furthermore, mapping the airflow dead zone and turbulent hotspot region as virtual control anchor points includes: On the actual airflow response surface, regions where the wind speed is lower than a preset threshold and the duration exceeds a first duration threshold are identified and marked as airflow dead zones. On the actual airflow response surface, regions where the wind speed fluctuation variance is higher than a preset threshold and the duration exceeds a second duration threshold are identified and marked as turbulent hotspot regions. Calculate the geometric centroid coordinates of the airflow dead zone and the turbulent hot spot region, respectively; Centered on the geometric centroid coordinates, a spherical virtual control anchor point is generated according to a preset influence radius, and the priority of the virtual control anchor point is set proportionally to the volume of the airflow dead zone or turbulent hot spot region on which the virtual control anchor point is generated.
[0010] Furthermore, using the virtual control anchor point as a constraint boundary, a multivariate decoupling operation is performed on the feedback state of the device actuator to generate a set of coordinated drive parameters including fan speed, wet curtain opening, and guide vane angle, including: Establish a multi-dimensional actuator state space that includes fan speed, wet curtain opening, and guide vane angle; The position coordinates and priorities of the virtual control anchor points are transformed into inequality constraints in the state space of the multidimensional actuator. Under the inequality constraints, a quadratic programming solution is performed on the feedback state of the device actuator to obtain the optimal control vector that satisfies the constraints; Perform physical dimension denormalization on the optimal control vector to generate the cooperative driving parameter set.
[0011] Furthermore, before acquiring the real-time microenvironment sensor data stream for each cage within the stacked cage chicken house, the following steps are also included: Periodically send heartbeat detection packets to each sensor and record the response delay; When the response delay of three consecutive heartbeat detection packets exceeds the preset communication timeout threshold, the corresponding sensor is determined to be in an abnormal state. By using real-time data from adjacent normal sensors and a preset spatial correlation model, the missing data from abnormal sensors is virtually reconstructed, and the reconstructed data is assigned a confidence weight value lower than that of the measured data.
[0012] Furthermore, after outputting the final environmental coordination and control instructions, it also includes: Acoustic feature sequences of poultry flock behavior within a preset time window after the execution of control commands are collected as feedback verification data. Calculate the improvement rate of the poultry flock heat stress vocalization index corresponding to the feedback verification data and the poultry flock heat stress vocalization index before regulation; When the improvement rate is lower than a preset effectiveness threshold, the online model fine-tuning mechanism is triggered; Only the internal parameters of the time-series prediction compensator are updated, and the generation logic of the cage-level physiological and environmental coupling tension field is frozen to prevent overfitting.
[0013] Furthermore, after outputting the final environmental coordination and control instructions, it also includes: Monitor the cumulative rate of ammonia concentration at each of the cages; When the ammonia concentration accumulation rate in a certain cage exceeds the preset sewage discharge warning threshold, a local directional ventilation enhancement command is generated. The localized directional ventilation enhancement command is superimposed on the final environmental coordinated control command to prioritize ensuring that the air quality of cages where the ammonia concentration accumulation rate exceeds the preset pollution discharge early warning threshold meets the standards.
[0014] Furthermore, the initialization stage of constructing the cage-level physiological-environment coupling tension field also includes: Construct a synthetic data generator that includes three operating conditions: extreme high temperature, extreme high humidity, and equipment failure; The synthetic data generator is used to generate labeled virtual stress-environment coupled samples; The virtual stress and environment coupled samples are mixed with the actual collected historical verification data according to the configured mixing ratio to form the initial training set; When performing the initialization of the generation logic of the cage-level physiological and environmental coupling tension field, a phased progressive training method is adopted. First, the virtual stress and environmental coupling samples are used for training, and then the proportion of the historical verification data is gradually increased.
[0015] A multi-tiered cage chicken house environmental coordinated control system, the system comprising: The multi-source heterogeneous data acquisition module acquires real-time microenvironmental sensor data streams, poultry flock behavior acoustic feature sequences, and equipment actuator feedback status of each cage in the stacked cage chicken house. The physiological and environmental coupling field construction module analyzes the heat stress voiceprint index of the poultry flock based on the acoustic feature sequence of the flock's behavior, and combines the temperature and humidity gradient in the real-time microenvironment sensing data stream to construct a cage-level physiological and environmental coupling tension field. The airflow model adaptive correction module dynamically generates a non-uniform spatial weight distribution based on the coupled tension field of the physiological and environmental factors at the cage level, and adaptively deforms and corrects the pre-constructed standard airflow organization model accordingly to obtain the actual airflow response surface. The virtual control anchor point generation module extracts the airflow dead zone and turbulence hot spot region from the actual airflow response surface, and maps the airflow dead zone and turbulence hot spot region as virtual control anchor points; The multivariable decoupling drive module uses the virtual control anchor point as the constraint boundary to perform multivariable decoupling operations on the feedback state of the device actuator, generating a set of collaborative drive parameters including fan speed, wet curtain opening degree and guide vane angle; The timing compensation and command output module inputs the collaborative driving parameter set into a preset timing prediction compensator to offset the inherent thermal inertia delay of the stacked structure and outputs the final environmental collaborative control command.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: By analyzing the acoustic feature sequence of poultry behavior to obtain the thermal stress acoustic index, and using it as a scalar weight to modulate with the temperature and humidity gradient vector basis, a cage-level physiological and environmental coupling tension field is constructed. This tension field directly integrates the real-time physiological stress feedback of the poultry flock into the environmental control model, transforming the control objective from simply meeting environmental parameters to a collaborative optimization that also considers the actual physical comfort of the poultry flock. This solves the problem of control decisions lagging behind physiological needs due to the lack of biological feedback in environmental control systems. By discretizing the standard airflow organization model into a three-dimensional voxel mesh and using the physiological and environmental coupling tension field as the driving force for node deformation to perform elastic deformation iteration, the abstract physiological-environment coupling relationship can be transformed into a specific airflow response surface. This method achieves dynamic response to non-uniform spatial weight distribution while ensuring computational efficiency, avoiding the excessive time consumption of traditional CFD simulations that cannot meet real-time control requirements. This invention achieves uncoupled collaborative driving of multiple actuators in complex three-dimensional space. By extracting airflow dead zones and turbulent hotspots from the actual airflow response surface and mapping them as virtual control anchor points, their position coordinates and priorities are transformed into inequality constraints in the state space of the multi-dimensional actuators. The optimal control vector is then solved through quadratic programming. This method transforms the global airflow optimization problem into a constrained optimization problem with key anchor points as boundaries, eliminating aerodynamic interference effects between fans, wet curtains, and deflectors. By introducing a timing prediction compensator before the output of the collaborative driving parameter set, timing compensation is specifically performed for the thermal inertia characteristics of the stacked cage, enabling the final output environmental collaborative control command to offset the response lag caused by structural thermal inertia in advance. This reduces the temperature overshoot and fluctuation amplitude during the control process, improving the stability and uniformity of the thermal environment inside the building.
[0017] By employing heartbeat detection and spatial correlation models, the system achieves virtual reconstruction and confidence downgrading of abnormal sensor data, ensuring the continuity of control even when some sensors fail. Simultaneously, through phased, progressive training initialization and online fine-tuning based on feedback verification data, the system can reliably initialize under extreme conditions with scarce samples, and continuously adapt to the increasing age of the poultry flock and seasonal environmental changes during operation, preventing model overfitting. By monitoring the ammonia concentration accumulation rate in each cage, when a cage exceeds a preset pollution warning threshold, a localized directional ventilation enhancement command is generated and superimposed on the final environmental coordinated control command, prioritizing the air quality compliance of that cage. This mechanism achieves rapid response to localized air quality anomalies without interfering with overall temperature and humidity coordinated control. Attached Figure Description
[0018] Figure 1 This is a flowchart outlining the steps of the present invention. Figure 2 This is a system architecture diagram of the present invention. Detailed Implementation
[0019] The following description is intended to disclose the present invention so that those skilled in the art can implement it; the preferred embodiments described below are merely examples, and other obvious variations will be apparent to those skilled in the art. Reference Figure 1 As shown, a collaborative control system and method for the environment of a stacked cage chicken house includes: step S101, acquiring real-time microenvironmental sensor data streams, poultry flock behavior acoustic feature sequences, and device actuator feedback status of each cage in the stacked cage chicken house. Temperature and humidity sensors and ammonia concentration sensors are deployed at each cage position in each row of cages in the stacked cage house, with a sampling frequency set to 1Hz to form a real-time microenvironment sensing data stream; industrial-grade broadband microphone arrays are deployed at the top and middle of each row of cages, with a sampling rate of not less than 44.1kHz, to collect the acoustic characteristic sequence of poultry behavior; the frequency of the fan inverter, the percentage of the opening of the wet curtain motor, and the angle of the guide plate servo motor are read in real time through PLC or fieldbus as feedback status of the equipment actuators; In terms of algorithm processing, the raw audio waveform data is first processed by spectral subtraction to remove background noise. The specific process is as follows: the background noise power spectrum is estimated using silent frames without bird calls, and the background noise power spectrum is subtracted from the power spectrum of the noisy frequency. Then, the denoised audio waveform is reconstructed by combining the original phase. Subsequently, a dual-threshold method combining short-time energy and zero-crossing rate is used for endpoint detection. The short-time energy threshold and zero-crossing rate threshold are dynamically set according to the current background noise of the environment. Audio segments containing valid bird calls are extracted. The real-time microenvironment sensing data stream mentioned here refers to multi-dimensional time-series data containing timestamps, spatial coordinates, and corresponding physical quantity values. The bird flock behavior acoustic feature sequence refers to the valid audio waveform data after denoising and endpoint detection. The device actuator feedback status refers to the normalized state vector reflecting the current actual operating conditions of each actuator. Step S102: Based on the acoustic feature sequence of poultry behavior, analyze the heat stress voiceprint index of poultry flocks, and combine it with the temperature and humidity gradient in the real-time microenvironment sensing data stream to construct a cage-level physiological and environmental coupling tension field. A short-time Fourier transform (SFT) with a window length of 25 ms and a step size of 10 ms was performed on the acoustic feature sequence of poultry flock behavior to extract the acoustic energy spectral density in the 300 Hz to 3000 Hz frequency band, which covers the main energy distribution range of heat panting and high-frequency calls in laying hens. The acoustic energy spectral density at the current moment was compared with the baseline resting acoustic signature of the same batch of poultry flocks pre-stored in the database to obtain the poultry flock heat stress acoustic signature index, which represents the current physiological stress level and is the instantaneous stress deviation. At the same time, the temperature difference and humidity difference between adjacent upper and lower cages and between left and right cages were calculated to construct a system based on the cage positions. The local environmental inhomogeneity vector is centered at the origin. For cages located at the edge or corner of the chicken house, the temperature and humidity differences are calculated using one-sided difference or mirror symmetric extension to ensure the integrity of the edge cage vector calculation. The scalar multiplication modulation of the local environmental inhomogeneity vector is performed using the poultry heat stress voiceprint index as a scalar weight to generate a cage-level physiological and environmental coupling tension field. The cage-level physiological and environmental coupling tension field is a vector field defined in three-dimensional cage space. Its modulus reflects the combined intensity of physiological stress and environmental deviation in the poultry flock at that location, and its direction points to the optimal gradient direction for environmental improvement. Step S103: Based on the dynamic generation of the cage-level physiological and environmental coupling tension field, a non-uniform spatial weight distribution is generated, and the pre-constructed standard airflow organization model is adaptively deformed and corrected accordingly to obtain the actual airflow response surface; the standard airflow organization model is discretized into a three-dimensional voxel mesh with a resolution of 0.2m×0.2m×0.2m, which is obtained based on CFD simulation under no-load conditions; according to the magnitude of the cage-level physiological and environmental coupling tension field, it is normalized and mapped to a non-uniform spatial weight distribution between 0 and 1; the cage-level physiological and environmental coupling tension field is projected onto each voxel mesh node through trilinear interpolation as the node deformation driving force; a spring-mass system model is adopted, with nodes... The deformation driving force is an external force, with the voxel grid edges acting as elastic springs. The spring stiffness coefficient of each node is adjusted using a non-uniform spatial weight distribution. The region with a larger weight has a smaller stiffness and is more prone to deformation. The voxel grid nodes corresponding to the chicken house walls and air inlets and outlets are set as fixed boundaries and do not participate in deformation. Iterative calculation of elastic deformation is performed until the change in the total potential energy of the system is less than the preset convergence threshold 1e-5. B-spline smoothing interpolation is performed on the coordinates of the deformed voxel grid nodes to reconstruct the actual airflow response surface. The actual airflow response surface mentioned here refers to the target airflow distribution pattern that reflects the current real control needs, obtained by topological correction of the ideal airflow field after considering the physiological needs of the poultry flock and the non-uniformity of the environment. Step S104: Extract airflow dead zones and turbulent hotspot regions from the actual airflow response surface, and map these regions as virtual control anchor points; traverse all voxel units on the actual airflow response surface, identify regions where the wind speed is lower than the dynamic wind speed threshold and the duration exceeds the first duration threshold, and mark them as airflow dead zones; calculate the variance of the time series wind speed data of each voxel unit within the second duration threshold, identify regions where the wind speed fluctuation variance is higher than the dynamic variance threshold and the duration exceeds the second duration threshold, and mark them as turbulent hotspot regions; calculate the geometric centroid coordinates of the two types of regions respectively; generate spherical virtual control anchor points with the geometric centroid as the center and a preset influence radius of 1.5m, and set the anchor point priority to a value proportional to the volume of the region; when the spheres of two virtual control anchor points overlap, retain the anchor point with higher priority, or merge the two into a new anchor point, whose centroid is the volume-weighted average of the two centroids, and whose priority is the sum of the two; the virtual control anchor points mentioned here are the spatial constraint boundaries for subsequent multivariate decoupling operations, and their priority determines the weight ranking of the region in collaborative control; The dynamic wind speed threshold and dynamic variance threshold are not fixed values, but are adaptively adjusted based on the deviation between the current average temperature of the cage and the target set temperature. Specifically, when the average temperature of the cage is higher than the target set temperature, the dynamic wind speed threshold is adaptively increased by 0.05 m / s for every 1°C increase, and the dynamic variance threshold is adaptively increased by 0.02 m / s.2 For example, assuming the target set temperature is 24℃, the basic dynamic wind speed threshold is 0.3m / s, and the basic dynamic variance threshold is 0.15 (m / s). 2 If the average temperature of a certain cage location is 27℃, which exceeds 3℃, then the wind speed compensation is 3 multiplied by 0.05, which equals 0.15 m / s. At this time, the dynamic wind speed threshold for that cage location is adaptively adjusted to 0.3 plus 0.15, which equals 0.45 m / s; the dynamic variance threshold is adjusted to 0.15 plus 3 multiplied by 0.02, which equals 0.21 (m / s). 2 This adaptive adjustment mechanism ensures that under high-temperature stress, the system has stricter criteria for judging airflow dead zones and turbulence, thereby triggering stronger ventilation compensation. Step S105: Using virtual control anchor points as constraint boundaries, perform multivariate decoupling operations on the feedback state of the equipment actuators to generate a set of collaborative drive parameters including fan speed, wet curtain opening, and guide vane angle; establish a three-dimensional actuator state space with fan speed, wet curtain opening, and guide vane angle as state variables; transform the position coordinates and priorities of each virtual control anchor point into linear inequality constraints in this state space; under the premise of satisfying all inequality constraints, with the goal of minimizing actuator energy consumption, perform quadratic programming to solve the feedback state of the equipment actuators to obtain the optimal control vector; perform physical dimension denormalization on the optimal control vector to generate a set of collaborative drive parameters; the collaborative drive parameter set mentioned here is the optimal combination of control commands that meets the control requirements of all key areas after eliminating the aerodynamic coupling interference of multiple actuators. Step S106: Input the collaborative drive parameter set into the preset time prediction compensator to offset the inherent thermal inertia delay of the stacked structure and output the final environmental collaborative control command. The time prediction compensator adopts a discrete-time transfer function model based on the Smith predictor. Its internal parameters are obtained through system identification and specifically characterize the combined thermal inertia time constant of the stacked cage metal structure and the poultry organisms. This time constant is usually 120s to 300s. The collaborative drive parameter set is used as the input of the compensator, and the compensator outputs a pre-compensation control quantity that is ahead of the current time to offset the response lag caused by thermal inertia. The pre-compensation control quantity is sent to each actuator controller as the final environmental collaborative control command. The role of the time prediction compensator mentioned here is to synchronize the control command with the actual perception of the poultry in the time dimension and avoid temperature overshoot and oscillation caused by thermal inertia.
[0020] In some embodiments, constructing a cage-level physiological-environment coupled tension field includes: When performing a short-time Fourier transform on the acoustic feature sequence of bird flock behavior, a Hanning window function is used to reduce spectral leakage. After extracting the acoustic energy spectral density in the 300Hz to 3000Hz frequency band, logarithmic compression is performed to enhance weak signal features. The calculation process of logarithmic compression is as follows: first, the original energy value P(k) of the k-th frequency point is obtained; then, a minimum constant epsilon is added to this energy value to prevent the logarithm from being negative. This minimum constant is 10 to the power of -6. Next, the logarithm of the sum is taken to the base 10. Finally, the logarithm is multiplied by 10 to obtain the logarithmically compressed acoustic energy spectral density Pl. og(k) ; The baseline resting voiceprint was a reference template obtained by continuously collecting data for 7 days under thermoneutral conditions (18°C to 24°C) and without external interference, and then taking the statistical average. The calculation process for the heat stress voiceprint index of the flock is as follows: First, calculate the current voiceprint energy spectral density vector P. current Compared with the reference resting voiceprint vector P baseline The difference vector is calculated, and then the L2 norm of this difference vector is used as the numerator. Finally, the reference resting voiceprint vector P is calculated. baseline The L2 norm is used as the denominator, and the instantaneous stress deviation Delta S is obtained by dividing the numerator by the denominator. The three components of the local environmental inhomogeneity vector are the temperature difference Delta T between adjacent cages divided by the horizontal spacing Delta x, the temperature difference Delta T between adjacent cages divided by the longitudinal spacing Delta y, and the humidity difference Delta H between adjacent cages divided by the vertical spacing Delta z. The calculation process of scalar multiplication modulation is as follows: the instantaneous stress deviation Delta S is taken as a scalar and multiplied by the three components of the local environmental inhomogeneity vector to obtain the coupled tension field vector F. couple The three corresponding components; through the above technical solution, the scalar information of poultry flock physiological stress and the vector information of environmental space are mathematically fused, so that the tension field has both the sensitivity of physiological perception and the directionality of environmental gradient.
[0021] In some embodiments, adaptive deformation correction is performed on a pre-built standard airflow organization model, including: The resolution of the three-dimensional voxel mesh is adaptively set according to the size of the chicken house to ensure that each cage is covered by at least 4 voxels; the direction of the applied nodal deformation driving force is consistent with the projection direction of the cage-level physiological and environmental coupling tension field at that node, and the magnitude is proportional to the modulus of the tension field. The elastic deformation iteration uses the conjugate gradient method to accelerate convergence, with a maximum number of iterations set to 500. In the conjugate gradient method iteration, the calculation process of the system's potential energy is as follows: First, traverse all spring connections, and for each pair of nodes i and j, calculate their current distance and natural length L. ij The difference is squared and then multiplied by the spring stiffness coefficient k.ij Half of the elastic potential energy of all springs is accumulated; then the deformation driving force F of all nodes is calculated. i and the corresponding node coordinates x i The sum of the dot products; finally, the total potential energy of the system E is obtained by subtracting the sum of the dot products from the sum of the elastic potential energy; the convergence threshold is dynamically adjusted according to the size of the chicken house, and the iteration stops when the potential energy difference Delta E between two adjacent iterations is less than 10 to the power of negative 5. Smooth interpolation reconstruction uses cubic B-spline surface fitting to ensure the C2 continuity of the deformed surface; the calculation process of its basis functions is as follows: for parameter u, first calculate the coefficient of the first term (u minus t). i ) divided by (t {i+3} Subtract t i Multiply by the second-order basis function N{i,2}(u); then calculate the coefficient of the second term (t). {i+4} Subtract u) and divide by (t) {i+4} Subtract t {i+1} ), multiplied by the second-order basis function N {i+1,2} (u); Finally, add the two terms together to obtain the third-order basis function N. {i,3} (u); A smooth actual airflow response surface is generated by weighted summation of control points; Through the above technical solution, the abstract physiological-environment coupling relationship is transformed into a computable mechanical deformation problem, and millisecond-level online correction is achieved while ensuring physical rationality.
[0022] In some embodiments, mapping airflow dead zones and turbulent hotspot regions as virtual control anchor points includes: The dynamic wind speed threshold and dynamic variance threshold are adaptively configured based on the age of the flock, breed, and real-time temperature deviation. The specific adaptive adjustment mechanism and example are shown in step S104. The first duration threshold is set to 60s, and the second duration threshold is set to 30s. Turbulent hotspot identification uses sliding window variance calculation with a window length of 10s. The geometric centroid coordinates are calculated using a volume-weighted average method to avoid centroid shift caused by irregular shapes. The calculation process is as follows: First, all voxel units within the region are traversed, and the volume v of each voxel is calculated. k Its spatial coordinates p k Multiplying the values yields weighted coordinates; summing the weighted coordinates of all voxels yields the molecule; then summing the volumes v of all voxels gives the total volume. k Obtain the denominator; finally, divide the numerator by the denominator to obtain the geometric centroid coordinates C. The influence radius is automatically adapted based on the cage spacing, typically 0.8 times the diagonal length of the cage; the priority is normalized to the 0-1 range, calculated as follows: the current region volume V... region Divide by the largest defect region volume V identified in the current cell maxThe priority is obtained by using the above technical solution, which discretizes the continuous airflow defect area into a finite number of spatial constraint points with priority, thereby reducing the dimensionality of the subsequent optimization problem.
[0023] In some embodiments, multivariate decoupling operations include: The dimension of the multidimensional actuator state space is equal to the number of controllable actuators; in this embodiment, it is 3-dimensional. The state vector u is determined by the fan speed u. fan evaporative cooling pad opening degree u pad Deflector angle u deflector It consists of three components; The inequality constraint is approximated by linearization, and the nonlinear aerodynamic relationship is expanded into a first-order linear constraint near the operating point by Taylor expansion. The form is that matrix A multiplied by the state vector u is less than or equal to vector b, where matrix A is composed of the aerodynamic sensitivity Jacobian matrix of the virtual control anchor point, and vector b is composed of the difference between the anchor point priority and the target wind speed. The objective function of the quadratic programming solution is to minimize the actuator energy consumption. The calculation process is as follows: First, calculate the transpose of the state vector u, multiply it by the positive definite energy consumption weight matrix H, and then multiply it by the state vector u again. Multiply the result by half. Then, calculate the transpose of the linear cost vector f and multiply it by the state vector u. Finally, add the two terms together to obtain the objective function value. The solver uses the OSQP algorithm, supports warm start, and the single solution time is less than 10ms. The physical dimension denormalization uses a piecewise linear mapping. The calculation process is as follows: multiply the normalized control quantity u by the actuator range (Y). max Subtract Y min ), plus the actuator lower limit Y min The physical quantity Y is obtained; in the actuator dead zone, such as when the fan frequency is less than 5Hz and in the saturation zone, such as when the wet curtain opening is greater than 95%, the output of invalid commands is truncated to avoid invalid command output; through the above technical solution, the complex multi-actuator coordination problem is transformed into a standard convex optimization problem, which ensures the global optimality and real-time performance of the solution.
[0024] In some embodiments, prior to acquiring the real-time microenvironment sensing data stream, the method further includes: The heartbeat detection packets are sent using the UDP protocol, with a timeout threshold set to 500ms; after three consecutive timeouts, an anomaly is identified, and data reconstruction is immediately triggered. The spatial correlation model is trained on historical normal data and uses Kriging interpolation to reconstruct the data using data from three adjacent normal sensors. The reconstruction calculation process is as follows: First, a system of Kriging equations is constructed based on the spatial variogram and solved to obtain three weighting coefficients lambda1, lambda2, and lambda3, the sum of which is 1. Then, each weighting coefficient lambda1 is compared with the measured value Z(x) of the corresponding adjacent normal sensor.i Multiply the three products together; finally, add the three products together to obtain the reconstructed value Ztar(x0) of the abnormal sensor location; The confidence weight of the reconstructed data is set to 0.6, and the confidence of the measured data is 1.0. These values are used as weighting factors in the subsequent tension field construction. Through the above technical solution, the basic functions of the system are maintained when the sensor fails, avoiding the collapse of the overall control due to a single point of failure.
[0025] In some embodiments, after outputting the final environmental coordinated control command, the method further includes: The feedback verification data acquisition window is 60s to 180s after the command is executed, avoiding the transient transition process; the improvement rate is calculated as follows: first, the improvement amount is obtained by subtracting the heat stress voiceprint index of the flock after the control from the heat stress voiceprint index of the flock before the control S_before; then, the improvement amount is divided by the index before the control S_before; finally, it is multiplied by 100% to obtain the improvement rate eta. The effectiveness threshold is set at 15%. If the value is lower than this, the control effect is considered insufficient, triggering the online fine-tuning mechanism. Online fine-tuning updates only the time constant and gain parameters of the timing prediction compensator, using recursive least squares for online identification. The parameter update calculation process is as follows: First, the transpose of the input regression vector phi(k) is multiplied by the current parameter vector theta(k-1) to obtain the prediction output; then, the prediction error is obtained by subtracting the prediction output from the actual environmental response y(k); next, the Kalman gain K(k) is multiplied by the prediction error to obtain the parameter correction amount; finally, the parameter correction amount is added to the current parameter vector theta(k-1) to obtain the updated parameter vector theta(k). The generation logic of the cage-level physiological and environmental coupling tension field includes STFT parameters, differential operations, and scalar multiplication modulation, which are completely frozen and do not participate in online updates. Through the above technical solution, while ensuring the stability of the core physiological perception model, the system can adapt to the dynamic characteristic drift caused by equipment aging and seasonal environmental changes.
[0026] In some embodiments, after outputting the final environmental coordinated control command, the method further includes: The ammonia concentration accumulation rate was calculated using a sliding window linear regression slope, with a window length of 300 s. The calculation process was as follows: first, the number of sampling points n and the sum of timestamps within the window were counted. t The sum of ammonia concentrations C The sum of the products of timestamp and concentration tC Sum of squared timestamps t2 Then calculate the numerator as n multiplied by sum. tC Subtract sum t Multiply by sum C; Calculate the denominator n multiplied by sum t2 Subtract sum t The square of the fraction is used; finally, the numerator is divided by the denominator to obtain the ammonia concentration accumulation rate (Rate). The pollution discharge warning threshold is dynamically set based on the age of the poultry flock and the stocking density, and is usually 5 ppm / min; Localized directional ventilation enhancement commands are implemented by increasing the opening of the deflector plate and the speed of the local fan in the corresponding area, with the increment proportional to the degree of exceedance; the calculation process is as follows: subtract the emission warning threshold Rate from the ammonia concentration accumulation rate Rate. threshold The excess amount is obtained, and the excess amount is multiplied by the proportional coefficient Kp to obtain the control increment Delta u; The superposition method is vector addition, that is, the basic collaborative driving parameter set u base The final instruction u is obtained by adding the control increment Delta u. final When the superimposed instructions exceed the physical limits of the actuator, the low-priority instructions are truncated according to priority. Through the above technical solution, rapid and accurate handling of local air quality anomalies can be achieved without disrupting the overall temperature and humidity coordinated control.
[0027] In some embodiments, the initialization phase of constructing the cage-level physiological-environment coupling tension field further includes: The synthetic data generator is built based on a hybrid mechanism model and GAN. It simulates an environment of 35°C to 40°C under extreme high temperature conditions, an environment of 80% to 90%RH under extreme high humidity conditions, and simulates fan stoppage and wet curtain blockage under equipment failure conditions. The initial mixing ratio of virtual samples to historical validation data is 9:1. After each training epoch, the ratio of historical data increases by 5% until it reaches 1:1. The training process is phased and progressive, consisting of 20 epochs. The first 10 epochs use only virtual samples, while the proportion of real data is gradually increased in the latter 10 epochs. The loss function is a weighted sum of mean squared error and adversarial loss. Through the above technical solution, the problem of model initialization failure caused by the scarcity of samples under extreme conditions is solved, and the robustness of the system under rare and abnormal conditions is improved.
[0028] Example 2: Refer to Figure 2 As shown, a multi-tiered cage chicken house environmental collaborative control system includes: The multi-source heterogeneous data acquisition module consists of a sensor network, an audio acquisition array, and a PLC communication interface. It is responsible for acquiring real-time micro-environment sensor data streams, acoustic feature sequences of flock behavior, and feedback status of equipment actuators. The physiological and environmental coupling field construction module, implemented by DSP or FPGA, performs STFT, differential operation and scalar multiplication modulation to construct the cage-level physiological and environmental coupling tension field; The airflow model adaptive correction module, implemented by an embedded GPU or high-performance MCU, performs voxel mesh discretization, elastic deformation iteration, and smooth interpolation reconstruction to generate the actual airflow response surface. The virtual control anchor point generation module, implemented by the main control CPU, performs region identification, centroid calculation, and anchor point generation. The multivariable decoupling driver module, implemented by the main control CPU, performs state-space modeling, constraint transformation, and quadratic programming solution. The timing compensation and instruction output module is implemented by a dedicated timer and DAC or communication interface, performing Smith prediction compensation and instruction issuance; The modules interact with each other via a high-speed internal bus or shared memory, with an overall control cycle of no more than 200ms, meeting the real-time requirements for coordinated control of the environment in stacked cage chicken houses. Through the above technical solutions, this invention achieves deep integration and perception of poultry flock physiological state and environmental parameters, transforming the control objective from achieving a single environmental parameter standard to a collaborative optimization that also considers the actual physical comfort of the poultry flock; it adaptively corrects the airflow model through a mechanical deformation mechanism, achieving millisecond-level online calculation while ensuring physical rationality; it reduces control oscillations caused by aerodynamic interference from multiple actuators by decoupling virtual control anchor points and secondary planning; it ensures accurate triggering of ventilation compensation under high-temperature stress by introducing a temperature deviation-driven threshold adaptive upward adjustment mechanism; it reduces temperature overshoot and fluctuation amplitude by offsetting the thermal inertia delay of the stacked structure through a time-series predictive compensator; it improves the robustness and adaptability of the system under complex operating conditions through sensor anomaly tolerance, online fine-tuning, and ammonia-directed ventilation enhancement mechanisms; and it alleviates the problem of scarce samples under extreme operating conditions through phased progressive training initialization. In summary, this invention improves the technical problems of lack of physiological perception, airflow coupling interference, and thermal inertia lag in existing stacked cage chicken house environmental control systems, thereby enhancing poultry welfare and breeding production efficiency.
[0029] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely the principles of the present invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for coordinated environmental control in a stacked chicken coop, characterized in that, The method includes: Acquire real-time microenvironment sensor data streams, poultry flock behavior acoustic feature sequences, and equipment actuator feedback status for each cage in the stacked cage chicken house; Based on the analysis of the acoustic feature sequence of poultry flock behavior, the heat stress voiceprint index of poultry flock is analyzed, and combined with the temperature and humidity gradient in the real-time microenvironment sensing data stream, a cage-level physiological and environmental coupling tension field is constructed. Based on the dynamic generation of the cage-level physiological and environmental coupled tension field, a non-uniform spatial weight distribution is generated, and the pre-constructed standard airflow organization model is adaptively deformed and corrected accordingly to obtain the actual airflow response surface. The airflow dead zone and turbulence hot spot region are extracted from the actual airflow response surface, and the airflow dead zone and turbulence hot spot region are mapped as virtual control anchor points; Using the virtual control anchor point as the constraint boundary, a multivariate decoupling operation is performed on the feedback state of the device actuator to generate a set of collaborative drive parameters including fan speed, wet curtain opening degree and guide vane angle; The collaborative driving parameter set is input into a preset timing prediction compensator to offset the inherent thermal inertia delay of the stacked structure and output the final environmental collaborative control command.
2. The method for coordinated environmental control in a stacked cage chicken house according to claim 1, characterized in that, The construction of the cage-level physiological and environmental coupled tension field includes: A short-time Fourier transform was performed on the acoustic feature sequence of the flock behavior to extract the acoustic energy spectral density representing the frequency of panting and calling. The instantaneous stress deviation is obtained by performing a difference operation between the voiceprint energy spectral density and the preset reference resting voiceprint. The temperature difference and humidity difference between adjacent cages are obtained from the real-time microenvironment sensing data stream, and the local environmental non-uniformity is calculated. Using the instantaneous stress deviation as a scalar weight and the local environmental inhomogeneity as a vector basis, the scalar weight is used to perform scalar multiplication modulation on the vector basis to generate the cage-level physiological and environmental coupling tension field.
3. The method for coordinated environmental control in a stacked chicken coop according to claim 1, characterized in that, The adaptive deformation correction of the pre-constructed standard airflow organization model to obtain the actual airflow response surface includes: The standard airflow organization model is discretized into a three-dimensional voxel mesh; The coupled tension field between the physiological and environmental conditions at the cage level is projected onto each node of the three-dimensional voxel mesh to obtain the driving force for node deformation. Based on the node deformation driving force, elastic deformation iteration is performed on the boundary coordinates of each voxel mesh until the system potential energy converges. Smooth interpolation reconstruction is performed on the deformed voxel mesh to generate the actual airflow response surface.
4. The method for coordinated environmental control in a stacked chicken coop according to claim 1, characterized in that, The step of mapping the airflow dead zone and turbulent hotspot region as virtual control anchor points includes: On the actual airflow response surface, regions where the wind speed is lower than a preset threshold and the duration exceeds a first duration threshold are identified and marked as airflow dead zones. On the actual airflow response surface, regions where the wind speed fluctuation variance is higher than a preset threshold and the duration exceeds a second duration threshold are identified and marked as turbulent hotspot regions. Calculate the geometric centroid coordinates of the airflow dead zone and the turbulent hot spot region, respectively; Centered on the geometric centroid coordinates, a spherical virtual control anchor point is generated according to a preset influence radius, and the priority of the virtual control anchor point is set proportionally to the volume of the airflow dead zone or turbulent hot spot region on which the virtual control anchor point is generated.
5. The method for coordinated environmental control in a stacked chicken coop according to claim 1, characterized in that, The process involves using the virtual control anchor point as a constraint boundary to perform multivariate decoupling operations on the feedback state of the device actuator, generating a set of coordinated drive parameters including fan speed, wet curtain opening, and guide vane angle, including: Establish a multi-dimensional actuator state space that includes fan speed, wet curtain opening, and guide vane angle; The position coordinates and priorities of the virtual control anchor points are transformed into inequality constraints in the state space of the multidimensional actuator. Under the inequality constraints, a quadratic programming solution is performed on the feedback state of the device actuator to obtain the optimal control vector that satisfies the constraints; Perform physical dimension denormalization on the optimal control vector to generate the cooperative driving parameter set.
6. The method for coordinated environmental control in a stacked chicken coop according to claim 1, characterized in that, Before acquiring the real-time microenvironment sensor data stream for each cage in the stacked cage chicken house, the following steps are also included: Periodically send heartbeat detection packets to each sensor and record the response delay; When the response delay of three consecutive heartbeat detection packets exceeds the preset communication timeout threshold, the corresponding sensor is determined to be in an abnormal state. By using real-time data from adjacent normal sensors and a preset spatial correlation model, the missing data from abnormal sensors is virtually reconstructed, and the reconstructed data is assigned a confidence weight value lower than that of the measured data.
7. The method for coordinated environmental control in a stacked chicken coop according to claim 1, characterized in that, After outputting the final environmental coordination and control instructions, it also includes: Acoustic feature sequences of poultry flock behavior within a preset time window after the execution of control commands are collected as feedback verification data. Calculate the improvement rate of the poultry flock heat stress vocalization index corresponding to the feedback verification data and the poultry flock heat stress vocalization index before regulation; When the improvement rate is lower than a preset effectiveness threshold, the online model fine-tuning mechanism is triggered; Only the internal parameters of the time-series prediction compensator are updated, and the generation logic of the cage-level physiological and environmental coupling tension field is frozen to prevent overfitting.
8. The method for coordinated environmental control in a stacked chicken coop according to claim 1, characterized in that, After outputting the final environmental coordination and control instructions, it also includes: Monitor the cumulative rate of ammonia concentration at each of the cages; When the ammonia concentration accumulation rate in a certain cage exceeds the preset sewage discharge warning threshold, a local directional ventilation enhancement command is generated. The localized directional ventilation enhancement command is superimposed on the final environmental coordinated control command to prioritize ensuring that the air quality of cages where the ammonia concentration accumulation rate exceeds the preset pollution discharge early warning threshold meets the standards.
9. The method for coordinated environmental control in a stacked chicken coop according to claim 1, characterized in that, The initialization stage of constructing the cage-level physiological-environment coupling tension field also includes: Construct a synthetic data generator that includes three operating conditions: extreme high temperature, extreme high humidity, and equipment failure; The synthetic data generator is used to generate labeled virtual stress-environment coupled samples; The virtual stress and environment coupled samples are mixed with the actual collected historical verification data according to the configured mixing ratio to form the initial training set; When performing the initialization of the generation logic of the cage-level physiological and environmental coupling tension field, a phased progressive training method is adopted. First, the virtual stress and environmental coupling samples are used for training, and then the proportion of the historical verification data is gradually increased.
10. A system for coordinated environmental control of a stacked cage chicken house, used to implement the method for coordinated environmental control of a stacked cage chicken house as described in any one of claims 1 to 9, characterized in that, The system includes: The multi-source heterogeneous data acquisition module acquires real-time microenvironmental sensor data streams, poultry flock behavior acoustic feature sequences, and equipment actuator feedback status of each cage in the stacked cage chicken house. The physiological and environmental coupling field construction module analyzes the heat stress voiceprint index of the poultry flock based on the acoustic feature sequence of the flock's behavior, and combines the temperature and humidity gradient in the real-time microenvironment sensing data stream to construct a cage-level physiological and environmental coupling tension field. The airflow model adaptive correction module dynamically generates a non-uniform spatial weight distribution based on the coupled tension field of the physiological and environmental factors at the cage level, and adaptively deforms and corrects the pre-constructed standard airflow organization model accordingly to obtain the actual airflow response surface. The virtual control anchor point generation module extracts the airflow dead zone and turbulence hot spot region from the actual airflow response surface, and maps the airflow dead zone and turbulence hot spot region as virtual control anchor points; The multivariable decoupling drive module uses the virtual control anchor point as the constraint boundary to perform multivariable decoupling operations on the feedback state of the device actuator, generating a set of collaborative drive parameters including fan speed, wet curtain opening degree and guide vane angle; The timing compensation and command output module inputs the collaborative driving parameter set into a preset timing prediction compensator to offset the inherent thermal inertia delay of the stacked structure and outputs the final environmental collaborative control command.
