Method, device and system for guiding livestock and poultry by cooperating and regulating environmental parameters by multiple fans

Abstract

The application discloses a livestock and poultry guiding method, device and system based on multi-fan collaborative regulation of environmental parameters, and relates to the technical field of livestock and poultry environmental control. The method comprises the following steps: establishing a mapping relationship library; dividing a ventilation area based on a real-time position of a guiding task and livestock and poultry, and dividing the ventilation area into a target environment maintenance area, a repelling environment construction area and an air flow parameter gradual change channel connecting the target environment maintenance area and the repelling environment construction area; calculating running parameter sets of fan execution units based on air parameter indexes set for each ventilation area and the mapping relationship library, and adjusting air supply quantity and air supply angle of each fan execution unit; monitoring displacement changes of the livestock and poultry in the air flow parameter gradual change channel; when monitoring data meet area switching conditions, adjusting spatial ranges of the repelling environment construction area and the air flow parameter gradual change channel, and returning to the previous step until the livestock and poultry enter the target environment maintenance area; and when detecting that the number of the livestock and poultry in the target environment maintenance area reaches a preset threshold, switching the fan execution units to a global uniform air supply mode.

Description

Technical Field

[0001] This invention relates to the field of livestock and poultry environmental control technology, and more specifically, to a livestock and poultry guidance method, equipment, and system for multi-fan coordinated regulation of environmental parameters. Background Technology

[0002] In large-scale livestock farming, operations such as grouping, transferring, weighing, and disease prevention often require guiding livestock to specific areas. These operations demand highly efficient and low-stress guidance methods to ensure livestock health and improve farming efficiency. There is an urgent need for automated and intelligent guidance systems designed to reduce human intervention while avoiding fright or harm to animals, thus meeting modern animal welfare standards.

[0003] Currently, common solutions include traditional manual herding methods and automated herding devices based on sound or electric shocks. Manual herding relies on on-site operator command, which is inefficient and easily triggers stress responses in livestock. While automated herding devices can reduce the burden on manpower, they still violate animal welfare principles by creating unpleasant stimuli to drive the animals.

[0004] In addition, farms commonly use environmental control equipment such as fans and wet curtains, but these systems usually adopt a centralized control mode, with fans starting and stopping synchronously and running at a fixed speed. This can only maintain the overall environment in a uniform and stable manner, and cannot achieve precise spatial control.

[0005] Existing solutions have significant shortcomings. Both manual herding and automated devices can easily cause stress in livestock and poultry, affecting their growth and development and posing safety hazards. Traditional environmental control systems lack the ability to sense the real-time location of livestock and poultry, making it impossible to identify guidance needs or assess effectiveness. Furthermore, due to their coarse control logic, the system struggles to proactively create differentiated and gradient environmental distributions within a space, thus preventing the upgrading of environmental control equipment into dynamic behavioral intervention tools and limiting its widespread application in smart farming. Summary of the Invention

[0006] This invention provides a method, device, and system for guiding livestock and poultry by coordinating the regulation of environmental parameters using multiple fans, in order to improve at least one of the aforementioned technical problems.

[0007] Firstly, the present invention provides a livestock and poultry guidance method based on the coordinated regulation of environmental parameters by multiple fans, which includes S1 to S5.

[0008] S1. A pre-established mapping database of the airflow field of the ventilation system. The database defines the correspondence between the distribution field of environmental physical quantities in the target space and the set of operating parameters of the fan actuator. The mapping database includes an inverse mapping model trained from paired data generated by computational fluid dynamics simulation.

[0009] S2. Based on the guidance task and the real-time location of livestock and poultry, the ventilation area is divided into the controlled space into the target environment maintenance area, the repulsion environment construction area, and the airflow parameter gradual change channel connecting the two.

[0010] S3. Based on the air parameter indicators set for each ventilation area, generate the target environmental physical quantity distribution field in the target space. Then, input the inverse mapping model to solve the operating parameter set of the fan actuator based on the mapping relationship library, adjust the air volume and air angle of each fan actuator, and construct a differentiated airflow organization distribution with spatial gradient in the controlled space to drive livestock and poultry to move by utilizing the environmental pressure difference.

[0011] S4. Monitor the displacement changes of livestock and poultry in the airflow parameter gradual change channel. When the monitoring data meets the area switching conditions, adjust the spatial range of the repulsion environment construction zone and the airflow parameter gradual change channel, and return to S3 until the livestock and poultry enter the target environment maintenance zone.

[0012] S5. When the number of livestock and poultry in the target environment maintenance zone is detected to reach a preset threshold, the control fan execution unit switches to the global uniform air supply mode.

[0013] Secondly, the present invention provides a livestock and poultry guidance device for coordinated regulation of environmental parameters by multiple fans, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor. The computer program can be executed by the processor to implement the livestock and poultry guidance method for coordinated regulation of environmental parameters by multiple fans as described in any paragraph of the first aspect.

[0014] Thirdly, the present invention provides a livestock and poultry guidance system for coordinated regulation of environmental parameters by multiple fans, which includes an environmental sensing network, a livestock and poultry positioning module, a central controller and a fan execution cluster.

[0015] An environmental sensing network is used to collect environmental physical quantity data from multiple points within the breeding space and send it to the central controller.

[0016] The livestock and poultry positioning module is used to acquire the location coordinates and distribution information of individual livestock and poultry or groups and send them to the central controller.

[0017] The central controller includes an environmental distribution and wind turbine control mapping model unit and a dynamic partitioning and decision algorithm unit, and is configured to execute the steps of the method described in any paragraph of the first aspect.

[0018] The fan execution cluster is used to receive the fan control command set issued by the central controller and regulate the environmental field distribution of the breeding space.

[0019] By adopting the above technical solution, the present invention can achieve the following technical effects:

[0020] Compared with existing technologies, this invention significantly improves the execution efficiency and processing capacity of related processes, while effectively reducing system resource consumption, demonstrating excellent technical effects and application value. By optimizing the method and device structure, this invention enhances operational convenience, reduces manual intervention, strengthens system stability and reliability, and expands the functional scope of existing technologies, thus exhibiting significant advantages in practicality, compatibility, and scalability. Attached Figure Description

[0021] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the specific embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some specific embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a livestock and poultry guidance method based on the coordinated regulation of environmental parameters by multiple fans.

[0023] Figure 2 This is a signal flow diagram for supervised learning training using a convolutional neural network (CNN).

[0024] Figure 3 This is a diagram illustrating the initial partitioning process performed by the central controller.

[0025] Figure 4 This is a diagram illustrating the partitions that trigger the central controller's execution after the update.

[0026] Figure 5 This is a simplified structural diagram of a livestock and poultry guidance system based on the coordinated regulation of environmental parameters by multiple fans. Detailed Implementation

[0027] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention.

[0028] To address the shortcomings of existing technologies, this invention aims to provide a livestock and poultry guidance method and system based on the coordinated regulation of environmental parameters by multiple fans. Its core objective is to establish a mapping relationship between the distribution of the livestock and poultry living space environment and the multi-dimensional regulation parameters of the fan execution cluster, and to drive the fan execution cluster to work collaboratively, thereby constructing a comfortable gradient environmental field capable of smooth movement and change. This utilizes the livestock and poultry's instinct to seek advantage and avoid harm, guiding them to spontaneously and smoothly move along a preset path towards the target area, while simultaneously achieving precise guidance and system energy conservation.

[0029] Example 1, please refer to Figures 1 to 5The first embodiment of the present invention provides a livestock and poultry guidance method based on the coordinated regulation of environmental parameters by multiple fans. It can be executed by a livestock and poultry guidance system based on the coordinated regulation of environmental parameters by multiple fans (hereinafter referred to as: livestock and poultry guidance system) to realize steps S1 to S5.

[0030] S1. A pre-established mapping database of the airflow field of the ventilation system is created. This database defines the correspondence between the distribution field of environmental physical quantities within the target space and the set of operating parameters for the fan actuators. The mapping database includes an inverse mapping model trained using paired data generated from computational fluid dynamics simulations. The inverse mapping model takes the distribution data of temperature, wind speed, and humidity fields within the target space as input, and the set of operating parameters consisting of the on / off state, speed, and airflow angle of each fan actuator as output. It is trained using a loss function that includes physical constraint terms.

[0031] S2. Based on the guidance task and the real-time location of livestock and poultry, the ventilation area is divided into the controlled space into the target environment maintenance area, the repulsion environment construction area, and the airflow parameter gradual change channel connecting the two.

[0032] S3. Based on the air parameter indicators set for each ventilation area, generate the target environmental physical quantity distribution field in the target space. Then, input the inverse mapping model to solve the operating parameter set of the fan actuator based on the mapping relationship library, adjust the air volume and air angle of each fan actuator, and construct a differentiated airflow organization distribution with spatial gradient in the controlled space to drive livestock and poultry to move by utilizing the environmental pressure difference.

[0033] S4. Monitor the displacement changes of livestock and poultry in the airflow parameter gradual change channel. When the monitoring data meets the area switching conditions, adjust the spatial range of the repulsion environment construction zone and the airflow parameter gradual change channel, and return to S3 until the livestock and poultry enter the target environment maintenance zone.

[0034] S5. When the number of livestock and poultry in the target environment maintenance zone is detected to reach a preset threshold, the control fan execution unit switches to the global uniform air supply mode.

[0035] Compared with existing technologies, this invention significantly improves the execution efficiency and processing capacity of related processes, while effectively reducing system resource consumption, demonstrating excellent technical effects and application value. By optimizing the method and device structure, this invention enhances operational convenience, reduces manual intervention, strengthens system stability and reliability, and expands the functional scope of existing technologies, thus exhibiting significant advantages in practicality, compatibility, and scalability.

[0036] Based on the above embodiments, in an optional embodiment of the present invention, S1 specifically includes S11 to S15.

[0037] S11. Construct a three-dimensional simulation model that includes the geometric features of the controlled space and the structural features of the internal equipment. Preferably, S11 specifically includes S111 to S113.

[0038] S111. Build a full-size geometric model using the architectural drawings of the controlled space. The geometric model includes the internal structure of the cage, feed line, and waterline.

[0039] S112. Mesh the fluid domain and perform mesh independence verification to determine the optimal number of meshes.

[0040] S113. Set the fan actuator unit as the velocity inlet boundary and the exhaust port as the pressure outlet boundary.

[0041] S12. Based on the three-dimensional simulation model, batch computational fluid dynamics simulations are performed by adjusting the combination of operating parameters of the fan actuator to obtain multiple sets of steady-state flow field data. Preferably, S12 specifically includes S121 to S123.

[0042] S121. Enumerate the combinations of operating parameters of the fan actuator to form multiple sets of simulation conditions, and perform computational fluid dynamics simulation on each set of conditions until steady state.

[0043] S122. Record the spatial distribution data of temperature field, wind speed field and humidity field at the height of livestock and poultry activity during steady state of each working condition, and form the corresponding distribution data of environmental field physical quantities.

[0044] S123. Record the fan control instruction set corresponding to each set of operating conditions. The fan control instruction set shall include at least the on / off status, speed and air delivery angle of each fan execution unit.

[0045] S13. Extract the distribution data of environmental field physical quantities at the activity height level of livestock and poultry from the steady-state flow field data, and perform data cleaning and normalization to construct a training dataset containing "paired data of wind turbine control command set and environmental field physical quantity distribution". Preferably, S13 specifically includes S131 to S135.

[0046] S131. In the computational domain of the three-dimensional simulation model, extract a horizontal slice plane that covers the height of livestock and poultry activity.

[0047] S132. Discretize the horizontal slice plane into a regular grid point matrix, and extract the temperature, wind speed and relative humidity at each grid point.

[0048] S133. Generate temperature field feature map, wind speed field feature map and humidity field feature map respectively, and perform dimensionless normalization on the physical quantities in the feature maps to obtain the distribution data of environmental field physical quantities.

[0049] S134. Obtain the set of operating parameters of the corresponding fan actuator during simulation. The set of operating parameters includes the on / off state, speed and air delivery angle of each fan actuator, and then vectorize them.

[0050] S135. Establish a correspondence between the distribution data of the environmental field physical quantities and the set of operating parameters to obtain the training dataset.

[0051] S14. Construct a neural network solver. Preferably, the neural network solver in S14 is a deep convolutional neural network model. Preferably, the network includes an input layer, hidden layers, and an output layer.

[0052] The input layer receives normalized environmental field physical quantity distribution data, and the dimension of the input data is... The number 3 represents the three characteristic channels: temperature, wind speed, and humidity. and These represent the number of rows and columns of the regular grid point matrix, respectively.

[0053] The hidden layer contains multiple alternating convolutional layers, pooling layers, and fully connected layers, which are used to extract the spatial features of the environmental flow field and perform dimensionality reduction mapping.

[0054] The size of the output layer is consistent with the total number of controllable parameters of the wind turbine actuator, and the output vector corresponds to the set of operating parameters of the wind turbine actuator.

[0055] S15. Design a loss function that includes physical constraints, and iteratively train the neural network solver using the training dataset until the loss function converges, obtaining an inverse mapping model from the distribution field of environmental physical quantities to the set of wind turbine operating parameters. Preferably, the loss function in S15 is as follows.

[0056] .

[0057] .

[0058] .

[0059] In the formula, This is the total loss function. This represents the mean square error. Temperature penalty. for The total dimensions of typhoon fan on / off status, speed, and wind direction angle control parameters. To predict the instruction vector. This is the actual instruction vector. is the regularization coefficient. This indicates taking the larger value. For the prediction of the mapping model Temperature of the coordinates. This represents the upper limit of the comfortable temperature corresponding to the current livestock and poultry species and growth stage.

[0060] The purpose of step S1 is to establish the breeding environment. Temperature field distribution function under various working conditions Wind speed field distribution function Humidity field distribution function Switching status of the fan control Rotation speed Wind direction angle The mapping relationship between parameters.

[0061] Based on this mapping relationship, the distribution of physical quantities in the target environment field is obtained. To the optimal wind turbine control instruction set mapping model It is also integrated into the environmental distribution and wind turbine control mapping model unit of the central controller.

[0062] in , , These are the horizontal, vertical, and height coordinates of the three-dimensional spatial points inside the breeding shed. For wind turbine index. This is an index for operating conditions. Indicates the first The first working condition The on / off status of the typhoon generator (represented by 0 / 1, where 0 is off and 1 is on). Indicates the first The first working condition The rotational speed of the typhoon fan. Indicates the first The first working condition The wind direction angle of the typhoon generator.

[0063] First, based on the actual structure of the aquaculture farm, a three-dimensional computational fluid dynamics (CFD) simulation model was established. Then, using CFD simulation, the fan execution cluster was enumerated. In the dimensional control parameter space Various working conditions, among which It is a positive integer, representing the total number of operating conditions of the simulated control combination.

[0064] Definition of the first Fan control instruction set for group operating conditions as follows.

[0065] .

[0066] .

[0067] in, Indicates the first Typhoon machine in The state under various working conditions. This is an index for operating conditions. For wind turbine index. and All are positive integers. , The number of operating conditions. This refers to the number of wind turbines.

[0068] Record the distribution of environmental field physical quantities at various points within the breeding shed when steady state is reached under this operating condition. Then, paired data of the wind turbine control command set and the distribution of environmental field physical quantities generated by computational fluid dynamics (CFD) simulation are used. As training samples, a convolutional neural network (CNN) is used for supervised learning training to build a network that achieves... With the goal of avoiding high-temperature stress in livestock and poultry, the distribution of physical quantities in the target environmental field is optimized based on this principle. To the optimal wind turbine control instruction set The mapping model. Defined as a function. It is also integrated into the environmental distribution and fan control model unit of the central controller.

[0069] .

[0070] in, Indicates inclusion Paired data of wind turbine control command set for each operating condition and distribution of environmental field physical quantities. A set of. For temperature field characteristic map, For wind speed field characteristic map and This is a characteristic diagram of the humidity field.

[0071] like Figure 2 As shown, this embodiment uses a convolutional neural network (CNN) for supervised learning training, specifically including the following steps.

[0072] Paired data of wind turbine control command set and environmental field physical quantity distribution generated by computational fluid dynamics (CFD) simulation. As training samples. At high levels of livestock and poultry activity. Discretize the continuous space into a horizontal section with a number of rows. Column number After the rule network is established, the distribution of each environmental field physical quantity is... Mesh sampling is performed on this horizontal cross section, i.e. .in Represents the high level of livestock and poultry activity Two-dimensional distribution coordinates on this specific horizontal cross section ( (This is a fixed value).

[0073] After sampling, three two-dimensional feature maps (mesh maps), i.e., temperature field feature maps, are generated respectively. Wind speed field characteristic map With humidity field characteristic map .in, For its grid points Temperature value corresponding to the location. For its grid points Wind speed value corresponding to the location. For its grid points The humidity value corresponding to the location. This is the row index of the rule network, and its value range is... to . This is the column index for the rule network, and its value range is... to .

[0074] The three two-dimensional feature maps are stacked into a three-channel two-dimensional feature tensor, which is used as input data for a convolutional neural network (CNN) mapping model. .in It is the field of real numbers.

[0075] The three-channel two-dimensional feature tensor The characteristic is that when the channel index is 1, the corresponding temperature data, that is, a slice of the three-channel two-dimensional feature tensor 1, is the temperature field feature map. When the channel index is 2, the corresponding wind speed data, that is, a slice of the two-dimensional feature tensor 2 of the three channels, is the wind speed field feature map. When the channel index is 3, the corresponding humidity data, that is, a slice of the three-channel two-dimensional feature tensor 3, is the humidity field feature map. Furthermore, the two-dimensional feature tensor of the three channels... In position The three channel values ​​at that location are, in turn, equal to the values ​​of the three feature maps at the same position. The value at that location, i.e. , , .

[0076] The two-dimensional feature tensor of the above three channels Each spatial location channel vector It represents the combined state of physical quantities such as temperature, wind speed and humidity at this coordinate point.

[0077] The first Fan control instruction set Vectorization yields the actual instruction vector output by the convolutional neural network (CNN). .in for The overall dimensions of typhoon fan on / off status, speed, and wind direction angle control parameters. That is... The continuous parameters of rotational speed and wind direction angle are processed to the range of [0,1].

[0078] Then the three-channel two-dimensional feature tensor The input is fed into the input layer of a convolutional neural network (CNN).

[0079] Preferably, the convolutional layers of the convolutional neural network (CNN) are configured with A learnable convolutional kernel, through... Convolution operations are performed to extract the spatial distribution features of the temperature, wind speed, and humidity fields, which are closely related to the fan control effect. Specifically, the convolution kernel can learn to recognize regional textures representing temperature uniformity, local flow direction features representing wind speed orientation, and regional features representing humidity distribution uniformity. The convolution operation is characterized by each convolution kernel acting as a weight matrix within the two-dimensional feature tensor. The convolution kernel moves position by position along the row and column directions with a preset stride. At each local region, it calculates the weighted sum of the values ​​of the kernel and all channels within that local region, and outputs this sum to the corresponding position in the newly generated feature map, thus obtaining... A new feature map.

[0080] These feature extraction capabilities stem from the pairing of wind turbine control command sets and environmental field physical quantity distributions using convolutional neural networks (CNNs). End-to-end learning enables the convolution kernel parameters to converge to a state that can effectively respond to the distribution of environmental field physical quantities under different control objectives.

[0081] The pooling layers of a convolutional neural network (CNN) respectively handle the above... The new feature map is downsampled, characterized by dividing each new feature map into multiple non-overlapping local grid cells, and performing aggregation operations on the values ​​within each cell. The aggregation operations include, but are not limited to, taking the maximum value of all values ​​within the local cell (max pooling) or calculating the average value of all values ​​within the local cell (average pooling). This operation reduces the spatial size of the new feature map by aggregating adjacent feature responses in the space, thereby achieving data dimensionality reduction while retaining the key spatial features extracted by convolution, and enhancing the stability of the model to slight changes in the spatial location of the input environment field.

[0082] The fully connected layer of a convolutional neural network (CNN) integrates and compresses all the new feature maps after downsampling into a one-dimensional global feature vector. And map it to the wind turbine control command space. .

[0083] This process can be formally expressed as: ,in and and are the weight matrix and bias vector, respectively, which are the trainable parameters of the convolutional neural network (CNN). They are not preset but are determined during the CNN training phase by using the weight matrix and bias vector. Paired data of wind turbine control command set under various operating conditions and distribution of environmental field physical quantities The signal is automatically learned through iterative optimization, with the goal of monitoring signals, minimizing prediction errors, and avoiding high-temperature stress on livestock and poultry.

[0084] After training, and The values ​​are thus determined and fixed, collectively encoding the optimal mapping law from global environmental field features to specific wind turbine control parameters. This layer learns and stores the wind turbine control parameter combinations required to achieve various target environmental fields. The convolutional neural network (CNN) ultimately outputs a predicted command vector. Each dimension corresponds to a specific control parameter for a wind turbine. Its learning objective is to make the predicted instruction vector Infinitely close to the real instruction vector .

[0085] The paired data after featureization and vectorization processing according to The proportion is divided into training sets Validation set and test set .in , , Training set Validation set Test set Proportion, and , , All are percentages and satisfy and The partitioning was performed on all simulation data, with the aim of training a general mapping model applicable to different environmental targets. Preferably, the training involves a general mapping model. Specifically, the steps include the following:

[0086] Using the training set The trainable parameters of the convolutional neural network (CNN) are trained iteratively. In each iteration, for the predicted instruction vector... The mean squared error (MSE) is used as the main loss function to calculate the predicted instruction vector. With the actual instruction vector The gap between them. Meanwhile, to guide the model in accurately achieving environmental goals while avoiding high-temperature stress on livestock and poultry, a penalty term for high-temperature areas is added to the loss function, penalizing areas exceeding the upper limit of comfortable temperatures corresponding to the current livestock and poultry species and growth stage. The region is subject to L2 norm penalty, which can be expressed as: This penalty term incurs a squared penalty (L2 norm) on predicted temperatures exceeding the comfort temperature upper limit to punish the formation of high-temperature zones, thereby mitigating the risk of heat stress in livestock and poultry and improving their welfare. This results in the total loss function. .

[0087] Minimize through trainable parameters By training set Repeat the above process to make the mapping model In the training set The optimal input-output relationship is approximated, and the input-output relationship is the model parameter.

[0088] During iterative training, each completed After one training cycle, the validation set is used. The currently trained model parameters are evaluated. The specific process is as follows: the current model parameters are temporarily fixed, and the validation set... Sample input mapping model Calculate the validation set The loss value on the validation set is saved. The model parameters with the minimum loss are selected as candidate optimal models.

[0089] After training, the test set was used. The candidate optimal model was independently tested to evaluate its accuracy and reliability in outputting fan control commands for simulated operating conditions not used in training.

[0090] By solidifying the network structure and model parameters of the candidate optimal model, a stable and universal mapping model is obtained. This solidified model is then integrated into the environmental distribution and fan control mapping model unit of the central controller.

[0091] Based on the above embodiments, in an optional embodiment of the present invention, S2 specifically includes S21 to S23.

[0092] S21. A preset fixed endpoint area is set as the target environment maintenance zone, and its environmental field is set as the livestock and poultry comfort range. Preferably, the air parameter index corresponding to the livestock and poultry comfort range is set as the first target air parameter index that meets the livestock and poultry comfort range.

[0093] S22. The initial gathering area of ​​livestock and poultry is set as a repulsion environment construction area, and its environmental field is set as an unpleasant environment that prompts livestock and poultry to leave. Preferably, the air parameter index corresponding to the unpleasant environment that prompts livestock and poultry to leave is set as a second target air parameter index that deviates from the comfort range of livestock and poultry, so as to form a mildly unpleasant environment for repelling livestock and poultry.

[0094] S23. Establish a guiding trajectory connecting the edge of the repulsive environment construction zone and the entrance of the target environment maintenance zone, and set the area covered by this trajectory as a gradual airflow parameter channel, with its environmental field set as a continuous gradient field from an unsuitable environment to a comfortable environment. Preferably, the air parameter index corresponding to the gradual airflow parameter channel is set as a gradient air parameter index that changes continuously and smoothly from the second target air parameter index to the first target air parameter index.

[0095] Specifically, based on the guidance task, the central controller performs initial partitioning and dynamically divides and moves the partitions according to the real-time location of livestock and poultry fed back by the livestock and poultry positioning module. The partitioning system includes the target area, the eviction area, and the dynamically changing corridor connecting the two.

[0096] Figure 3 As shown, the target area is a pre-defined fixed endpoint region, and its environmental field is denoted as... The initial range of the eviction zone is set based on the principle of livestock and poultry seeking profit, and coincides with the initial aggregation area of ​​livestock and poultry identified by the livestock and poultry positioning module. This environmental field is denoted as... The principle of livestock and poultry avoidance is utilized to create a slightly uncomfortable environment that encourages livestock and poultry to leave. The dynamic, gradually changing corridor environment field is denoted as... It was set to be a mildly uncomfortable environment. To a comfortable environment A continuous, smooth gradient field, and this region is not a fixed space, but a preset guiding trajectory connecting the expulsion zone to the target zone.

[0097] In this invention, the zoning system consisting of the target area, the deportation area, and the dynamically evolving corridor is dynamically evolving to achieve continuous, smooth, and non-regressive guidance. Its position and spatial range are not fixed at the start of the task, but are periodically replanned and updated by the central controller based on real-time position information fed back by the livestock positioning module. This mechanism ensures the continuous spatial progression of the guidance environment.

[0098] Based on the above embodiments, in an optional embodiment of the present invention, S3 specifically includes S31 to S33. Specifically, the central controller determines the target area environment field corresponding to the current dynamic partition. , Demolition Area Environmental Site and dynamic gradient corridor environment In conjunction with real-time data collected by the environmental sensing network, the central controller generates the current distribution of target environmental field physical quantities. Then will Input the mapping model The current optimal set of wind turbine control commands is calculated. This instruction set precisely specifies the on / off status, speed, and wind direction angle of each wind turbine in the turbine execution cluster. The wind turbine execution cluster receives and executes the optimal wind turbine control instruction set. Precisely control the environmental field of the target area in physical space , Demolition Area Environmental Site and dynamic gradient corridor environment .

[0099] S31. Based on the air parameter indices corresponding to the target environment maintenance zone, the repulsion environment construction zone, and the airflow parameter gradual change channel, generate the target environment physical quantity distribution field within the target space.

[0100] S32. Input the target environmental physical quantity distribution field into the inverse mapping model corresponding to the mapping relationship library, solve for the operating parameter set of the fan actuator, and generate the fan control command set. Specifically, the operating parameter set is subjected to tolerance threshold constraints and adjacent fan control quantity smoothing processing to ensure that the output commands meet preset safe ventilation / air conditioning constraints and reduce sudden changes in spatial airflow. Preferably, the operating parameter set includes at least the on / off state, speed, and air supply angle of each fan actuator. The core logic features of the control are as follows;

[0101] Preferably, the air parameters in the target environmental maintenance zone meet the comfort range constraints for livestock and poultry. Specifically, in the target zone, the fans are adjusted to maintain a stable and comfortable environment, meaning that the environmental physical quantities meet the preset comfort range for livestock and poultry.

[0102] . . .

[0103] In the formula, , and These are the set values ​​for temperature, wind speed, and humidity in the target environment maintenance zone, respectively. , and These represent the lower limits of temperature, wind speed, and humidity for the comfort of livestock and poultry. , and These are the upper limits for temperature, wind speed, and humidity, respectively, for the comfort of livestock and poultry.

[0104] Preferably, the air parameters in the eviction environment construction zone meet the constraint of mild discomfort and not exceeding the tolerance threshold. Specifically, in the eviction zone, the controlled fans create continuous, mild discomfort stimulation, that is, the environmental physical quantities exceed the comfort range of livestock and poultry but do not exceed the tolerance threshold of livestock and poultry;

[0105] . . .or . . .

[0106] In the formula, , and These are the set values ​​for temperature, wind speed, and humidity in the repellent environment construction zone. , and These are the upper limits of temperature, wind speed, and humidity that livestock and poultry can tolerate. , and These are the lower limits of temperature, wind speed, and humidity that livestock and poultry can tolerate.

[0107] Preferably, the gradient air parameter indices of the gradually changing airflow channel satisfy a gradient function that continuously changes along the channel path. Specifically, in the dynamically changing corridor, the fans are adjusted to create a temperature, wind speed, and humidity along the dynamically changing corridor path. (From the eviction zone end) To the target area A continuously changing gradient environment field, the changes in its physical quantities satisfy a preset gradient function.

[0108] .

[0109] .

[0110] .

[0111] In the formula, The path position is along the airflow parameter gradually changing channel from the end of the dispersive environment construction zone to the end of the target environment maintenance zone. This represents the channel length. , and Positions Temperature, wind speed, and humidity at the location. , and These are the set values ​​for temperature, wind speed, and humidity in the repellent environment construction zone. , and These are the set values ​​for temperature, wind speed, and humidity in the target environment maintenance zone, respectively. , and They are respectively in The gradient functions of temperature, wind speed, and humidity that change monotonically within the interval.

[0112] The gradient function can be a linear function. .

[0113] S33. The control fan execution unit adjusts the air supply volume and air supply angle according to the fan control instruction set to construct a differentiated airflow organization distribution with spatial gradient within the controlled space.

[0114] Preferably, during livestock migration, the central controller compares the actual movement trajectory of the livestock with the preset dynamic gradient corridor in real time. If the livestock movement lags behind, the controller fine-tunes the stimulation of the expulsion zone or optimizes the gradient of the dynamic gradient corridor. If the movement is smooth, the expulsion zone and the dynamic gradient corridor are moved forward according to the plan, so that the livestock gradually move to the target area based on their tendency to seek advantages and avoid disadvantages.

[0115] Based on the above embodiments, in an optional embodiment of the present invention, S4 specifically includes S41 to S43.

[0116] S41. Determine the displacement or distribution changes of livestock and poultry in the airflow parameter gradual change channel based on livestock and poultry positioning data.

[0117] S42. When the displacement or distribution change meets the region switching conditions, the spatial range of the repulsion environment construction area is reduced or moved forward, and the spatial range of the airflow parameter gradient channel is adjusted simultaneously, so that the starting end of the airflow parameter gradient channel is connected to the adjusted repulsion environment construction area. The region switching conditions include: within the update cycle, the proportion of grid cells "occupied by livestock" in the grid units divided by the gradient channel is counted as the channel occupancy rate. And calculate the displacement of the centroid of the livestock group along the guiding direction. and the advancing distance of the forward front on the side of the passage entrance .when and and the duration is not less than ,or and the duration is not less than At this time, a forward update of the disengagement environment construction area is triggered. The forward shift distance is a preset value. , or by Obtained. Among them, The channel occupancy threshold represents the minimum channel occupancy percentage required to trigger a regional update. The centroid displacement threshold for the livestock and poultry population represents the minimum centroid displacement that triggers the region update. The threshold for the forward advance distance represents the minimum forward advance distance required to trigger a region update. The duration threshold represents the shortest time that the triggering condition must be met continuously.

[0118] Preferably, the central controller performs the following determination process in each update cycle (e.g., 1s to 5s).

[0119] Calculation of forward movement distance: The projected positions of livestock and poultry located in the dynamic gradient corridor along the guidance direction are sorted from smallest to largest, and the high percentile (e.g., 90 percentile) closest to the target area is taken as the "forward position" for that cycle; the difference between the forward positions of two adjacent update cycles is defined as the forward movement distance.

[0120] Calculation of the number of livestock entering the dynamic gradient corridor: Based on the spatial boundary of the dynamic gradient corridor, the number of livestock and poultry inside it in the current period is counted, and can be further converted into the proportion of the total number of livestock and poultry, which is used to characterize "the number of livestock and poultry entering the dynamic gradient corridor".

[0121] Triggering conditions and de-jittering constraints: When the distance the leading edge moves forward along the dynamic gradient corridor exceeds a first preset threshold, or the number of livestock and poultry entering the dynamic gradient corridor exceeds a second preset threshold, the central controller triggers a partition update calculation. To avoid false triggering caused by livestock and poultry turning back briefly, positioning noise, or local congestion, preferably, the central controller can also set a de-jittering constraint: the partition update calculation is only triggered after any of the above conditions are met and continue for a preset duration T (e.g., 10s to 30s).

[0122] To further illustrate, in a chicken coop guidance scenario, the first preset threshold can be 8m (i.e., the chicken flock's front line advances more than 8m towards the target area, triggering an update); the second preset threshold can be 30% to 50% of the total number of livestock and poultry (i.e., the proportion of chickens entering the dynamic gradient corridor reaches a preset ratio, triggering an update). After triggering, the central controller executes the update according to step S42: updating the core area of ​​the deportation zone to the currently densely populated unguided area of ​​the livestock and poultry flock, and removing the areas where the livestock and poultry have left from the deportation zone; at the same time, the physical path of the dynamic gradient corridor is replanned so that its starting end connects to the updated edge of the deportation zone, and its ending end remains connected to the entrance of the target area, thereby forming a guidance field that dynamically shifts with the movement of livestock and poultry.

[0123] S43. Return to execute S3 to update the fan control instruction set until livestock enter the target environmental maintenance zone.

[0124] During the guidance process, the central controller monitors the forward movement distance of the livestock herd in real time. When the distance the front of the livestock herd moves forward along the dynamically varying corridor is monitored. Exceeding the first preset threshold or the number of livestock and poultry entering the dynamic gradient corridor Exceeding the second preset threshold At that time, the central controller triggers a partition update calculation.

[0125] Figure 4 As shown, the core area of ​​the eviction zone is updated to the unguided area where the livestock and poultry are currently densely concentrated, and areas where the livestock and poultry have left are removed from the eviction zone. Simultaneously, the physical path of the dynamic gradient corridor is replanned, ensuring its starting point connects to the updated edge of the eviction zone and its ending point remains connected to the entrance of the target area, thus forming a new guiding corridor that dynamically shifts with the movement of livestock and poultry. In short, the dynamic gradient corridor is spatially always connected between the latest eviction zone and the fixed target area, and it advances synchronously with the displacement of the eviction zone.

[0126] From the perspective of livestock herd perception, there is always an uncomfortable environment behind them, corresponding to the updated relocation zone, while in front of them there is always a comfort gradient constructed by the updated dynamic gradient corridor, pointing towards the target area, thus guiding the overall environmental field forward step by step. This mechanism ensures that the livestock herd is continuously driven away from the latest uncomfortable zone and attracted by the constantly extending and updated comfort gradient in front, thus being guided to the target area in a smooth and coherent manner, effectively preventing stagnation or backtracking. When the proportion of livestock in the target area is... When the third preset threshold is reached or exceeded, the guidance is considered complete.

[0127] Preferably, the third preset threshold can be 95%. The central controller sends global uniform environmental control commands to the wind turbine cluster based on seasonal and overall environmental parameters. This allows all fans to switch to a low-power, uniform ventilation normal operating mode.

[0128] Livestock and poultry guidance systems based on multi-fan coordinated regulation of environmental parameters, such as Figure 5 As shown, the system includes an environmental sensing network, a livestock and poultry positioning module, a central controller, and a fan execution cluster.

[0129] The environmental sensing network consists of distributed temperature, humidity, wind speed, and ammonia concentration sensors, used to collect real-time environmental physical quantity data from multiple points within the breeding space.

[0130] The livestock and poultry positioning module uses a device based on a visual camera combined with image recognition algorithms or ultra-wideband (UWB) positioning technology to obtain the location coordinates and distribution information of individual livestock and poultry or groups in real time.

[0131] The central controller is the core processing unit of the system. Internally, it stores and runs a livestock comfort model library, an environmental distribution and fan control mapping model unit, and a dynamic zoning and decision-making algorithm unit. The livestock comfort model library contains suitable environmental parameter ranges for different species and growth stages of livestock. The environmental distribution and fan control mapping model unit integrates data from the distribution of physical quantities in the target environmental field. To the optimal wind turbine control instruction set mapping model As an online phase for real-time solution of the optimal wind turbine control instruction set The core module. The dynamic partitioning and decision algorithm is used to perform online computation and decision-making in step S4.

[0132] The wind turbine execution cluster is composed of It consists of a variable frequency fan that can be independently controlled to start and stop, with stepless speed regulation and adjustable wind direction (fixed wind direction can be regarded as a specific wind angle), and receives and executes the fan control command set from the central controller via network. or .

[0133] Based on the above embodiments, in an optional embodiment of the present invention, S5 specifically includes S51 to S52.

[0134] S51. Control multiple fan actuators to operate in a coordinated manner to ensure uniform ventilation, thereby making the air parameter distribution in the controlled space more uniform.

[0135] S52. Under the premise of meeting the requirement of uniform ventilation, reduce the speed of at least some fan actuators or shut down at least some fan actuators to achieve low power consumption operation.

[0136] As can be seen from the above technical solution, compared with the prior art, the present invention has at least the following beneficial effects (and there is a synergistic enhancement relationship between the various beneficial effects).

[0137] This invention upgrades the ventilation / air conditioning system from "overall uniform control" to "spatial distribution control that can be used for behavior guidance": It uses a target area, a repulsion area, and a dynamically evolving corridor connecting the two to form a dynamic evolution zoning system. It creates a continuous and smooth comfort gradient environment within the breeding space, enabling livestock and poultry to actively migrate along a preset guidance trajectory driven by their instinct to seek advantages and avoid disadvantages. This overcomes the fundamental defects of traditional manual driving, such as low efficiency and high stress, as well as the inability of traditional environmental control systems to be used for behavior guidance.

[0138] The guidance process is less stressful and more controllable: the deportation zone is constructed as an environmental field that causes mild discomfort but does not exceed the tolerance threshold, and the dynamic gradient corridor is constructed as a continuous gradient field from mild discomfort to comfort; compared with strong stimulation methods such as sound and electric shock, this invention guides livestock and poultry to move through the "spatial gradient + closed-loop movement" method, which can significantly reduce the stress response of livestock and poultry, improve animal welfare and the safety of the guidance process.

[0139] The real-time performance and accuracy of multi-fan collaborative control are higher: This invention establishes a mapping model between the distribution of physical quantities of the target environment field and the optimal fan control command set, enabling the central controller to quickly generate high-dimensional control commands such as the on / off status, speed and wind direction angle of multiple fans during the online phase. This allows for differentiated and gradient airflow organization and environmental parameter distribution in the target area, the evacuation area and the dynamic gradient corridor, thereby improving the spatial resolution and control accuracy of ventilation / air conditioning control.

[0140] Enhanced optimization and control capabilities within safety constraints: This invention introduces a high-temperature penalty mechanism (with the optimization principle of avoiding high-temperature stress on livestock and poultry) into the training of the mapping model aimed at accurately realizing the target environmental field. This enables the system to actively suppress the formation of unfavorable areas that may cause high-temperature stress on livestock and poultry when generating control commands, thereby balancing guidance effectiveness and safety during the guidance process.

[0141] Unifying guidance efficiency and energy saving: This invention uses a dynamic zoning and shifting mechanism to ensure that the precise control of the fan cluster is mainly concentrated in the currently effective eviction zone and the dynamic gradient corridor; unnecessary fan loads are promptly removed from areas where livestock and poultry have left, reducing ineffective air supply and energy consumption; after guidance is completed, the system can switch to the normal operation mode of global uniform ventilation and enter low power consumption operation, thereby taking into account guidance efficiency, operation stability and energy utilization efficiency.

[0142] To facilitate understanding of the present invention, the application of this embodiment will be illustrated below using a specific application scenario.

[0143] First, temperature, humidity, and wind speed sensors are evenly deployed inside the shed to form an environmental sensing network. An infrared thermal imaging-based livestock positioning module is installed on the roof for nighttime flock location. The fan cluster consists of 10 longitudinal variable frequency fans (VF1-VF10) and 8 circulating variable frequency fans (CF1-CF8). The central controller uses an industrial server. Specific details are as follows.

[0144] Thirty integrated temperature and humidity sensors and 25 ultrasonic wind speed sensors were evenly deployed inside the chicken house, and ammonia and carbon dioxide sensors were deployed near the ventilation windows on the side walls to form a high-density environmental monitoring network with a data refresh rate of ≥1Hz.

[0145] A livestock positioning module is installed on the roof of the shed, which employs a vision system based on a top-mounted infrared thermal imaging camera array. This positioning module utilizes the significant temperature difference between chickens and the ground at night to clearly identify individuals through thermal imaging. Combined with a multi-camera visual fusion algorithm, it calculates the flock's distribution density cloud map and the position of the flock's centroid in real time, achieving a positioning accuracy better than 0.5 meters.

[0146] A longitudinal ventilation main fan wall is installed at one end of the chicken coop, consisting of 10 large-diameter variable frequency fans (VF1-VF10). Eight independently swingable variable frequency circulating fans (CF1-CF8) are evenly installed on the side walls, and all fans can receive commands via industrial Ethernet. In this embodiment, the speed adjustment range of the large-diameter variable frequency fans (VF1-VF10) is 300... Up to 1200 The speed adjustment range of the variable frequency circulating fan (CF1-CF8) is 450. Up to 1500 .

[0147] An industrial server with edge computing capabilities is used as the central controller, which runs mapping models, machine learning inference, and real-time control programs.

[0148] Based on precise architectural drawings of the chicken coop, a three-dimensional finite element analysis (FEA) model was established, including the internal structure such as cages (if caged), feed lines, and water lines, serving as the geometric basis for fluid simulation. Non-isothermal flow field simulations were performed on the model using professional computational fluid dynamics (CFD) software. The system systematically simulated the fan execution cluster (total... (Taiwan) Temperature field, wind speed field and humidity field under different typical working conditions.

[0149] Each operating condition corresponds to a set of randomly generated fan control instructions. Record the specific on / off status, speed, and wind direction angle of each fan. After the simulation reaches a steady state, record the height at which the chickens are active. The distribution data of temperature, wind speed, and humidity at various points in space on a 1-meter plane constitute the distribution of environmental field physical quantities. .

[0150] Paired data of wind turbine control command set and environmental field physical quantity distribution generated by computational fluid dynamics (CFD) simulation. As training samples. For each environmental field physical quantity distribution. At high altitude On a horizontal cross section of meters Grid sampling generates three two-dimensional feature maps for temperature, wind speed, and humidity fields, which are then stacked into a three-channel two-dimensional feature tensor. As input to the model.

[0151] Each wind turbine control instruction set Vectorization, forming a one-dimensional vector. As the true instruction vector output by the Convolutional Neural Network (CNN) mapping model, its total dimension is... It contains the on / off status, speed, and wind direction angle parameters of 18 wind turbines.

[0152] The preprocessed two-dimensional feature tensor The input layer of a Convolutional Neural Network (CNN) performs three levels of feature parsing and mapping. A CNN processes the input two-dimensional feature tensor through a set of convolutional kernels in its convolutional layers. Convolutional operations are performed to learn and extract physical field features related to the regulation effect, such as temperature, wind speed, and humidity. The feature maps extracted by the convolutional layers of a Convolutional Neural Network (CNN) are downsampled through its pooling layers to aggregate adjacent features and reduce dimensionality. Subsequently, these features are integrated and compressed into a one-dimensional global feature vector through its fully connected layers. Finally, the fully connected layer combines the global feature vectors. Linear mapping to A dimensional wind turbine control command space. This mapping is formally expressed as... ,in and The weight matrix and bias vector obtained during training are the trainable parameters of the convolutional neural network (CNN). The CNN ultimately outputs the predicted instruction vector. .

[0153] Paired data after featureization and vectorization preprocessing The training set was divided into 70%, 15%, and 15% portions. Validation set and test set The partitioning was performed on all simulation data, with the aim of training a general mapping model applicable to different environmental targets. .

[0154] Training a general mapping model Specifically, the steps are as follows.

[0155] Using the training set The trainable parameters of the convolutional neural network (CNN) are trained iteratively. In each iteration, for the predicted instruction vector... Mean squared error (MSE) is used as the main loss function. Calculate the predicted instruction vector With the actual instruction vector The gap between them.

[0156] Meanwhile, to guide the model in accurately achieving environmental goals while avoiding high-temperature stress on the chickens, a penalty term for high-temperature regions was added to the loss function, penalizing areas exceeding the upper limit of comfortable temperatures corresponding to the current growth stage of the chickens. The region is subject to L2 norm penalty.

[0157] The penalty item is described as follows:

[0158] .

[0159] In the formula is the regularization coefficient. The temperature distribution is predicted by a convolutional neural network (CNN) mapping model. This represents the upper limit of the comfortable temperature corresponding to the current growth stage of the chicken flock.

[0160] This approach penalizes the formation of high-temperature zones by applying a squared penalty (L2 norm) to predicted temperatures exceeding the comfort temperature upper limit, thereby mitigating the risk of heat stress in the chicken flock. The resulting total loss function is: By optimizing the trainable parameters to minimize .

[0161] By training set Repeat the above process to make the mapping model In the training set The model parameters are approximated to the optimal values.

[0162] During iterative training, the validation set is used after every 10 training cycles. The currently trained model parameters are evaluated. The specific process is as follows: the current model parameters are temporarily fixed, and the validation set... Sample input mapping model Calculate the validation set The loss value on the validation set is saved. The model parameters with the minimum loss are selected as candidate optimal models.

[0163] After training, the test set was used. The candidate optimal model was independently tested to evaluate its accuracy and reliability in outputting fan control commands for simulated operating conditions not used in training.

[0164] By solidifying the network structure and model parameters of the candidate optimal model, a stable and universal mapping model is obtained. This solidified model is then integrated into the environmental distribution and fan control mapping model unit of the central controller.

[0165] Assuming it's 10 PM, the entire flock needs to be smoothly guided to the capture area near the exit. After the operator selects the task to guide to the capture area on the HMI (Human-Machine Interface), the central controller receives the instruction and immediately reads real-time data from the poultry positioning module. The positioning system shows the flock gathered at the front of the coop. The central controller performs the initial dynamic zoning calculation: the target area is set as the last 15 meters of the coop's rear (the capture area), with environmental targets of 22℃ temperature and 0.2m / s wind speed (low wind speed to avoid disturbance). The de-escalation zone is set to overlap with the current core distribution area of ​​the flock (60 meters in front of the coop). The dynamic gradient corridor is set as a virtual channel approximately 3 meters wide, extending from the rear edge of the de-escalation zone to the front edge of the target area. The zoning update trigger logic is then set: when the flock's leading edge moves more than 8 meters along the dynamic gradient corridor, or when the flock density within the dynamic gradient corridor reaches 25% of the total stock, the zoning update is automatically triggered.

[0166] The central controller generates the distribution of physical quantities of the target environment field based on the above partitions. The target area should maintain comfort. The evacuation area should create a mildly uncomfortable environment, with a temperature set at 25°C (slightly higher than the comfortable nighttime temperature of 23°C) and a stable, directional breeze of 0.8 m / s. The dynamic gradient corridor requires the temperature to smoothly decrease from 25°C to 22°C and the wind speed to smoothly decrease from 0.8 m / s to 0.2 m / s.

[0167] The central controller distributes the physical quantities of the target environment field. Input the mapping model trained offline The model operates within milliseconds and outputs the first set of instructions. For example: "Activate VF2, VF4, VF6, CF3, CF5. VF2 speed 600". Used to maintain the starting wind speed of a dynamically varying corridor, VF4 rotation speed 360. Used for dynamic, gradual air supply in the middle section of a corridor, VF6 speed 300 For use in light winds in the target area. CF3 speed 600 The wind direction is directed towards the center of the deflection zone to create an uncomfortable breeze; CF5 is shut down and on standby. All other wind turbines are shut down. The wind turbine cluster receives and executes this instruction, rapidly and accurately synthesizing the planned initial gradient environmental field in physical space.

[0168] After a period of time, the chickens reacted to the uncomfortable environment of the continuously warm and breezy de-encroaching zone and began to move towards the cooler, calmer dynamic gradient corridor and the target area. The thermal imaging positioning system continuously tracked the outline of the chickens and the position of their leading edge. Approximately 5 minutes later, the leading edge of the chickens moved more than 8 meters, triggering an update. Based on the latest chicken position, the central controller reset the de-encroaching zone to the new area where the chickens were still densely concentrated (the front half of the original de-encroaching zone had become sparse). The starting point of the dynamic gradient corridor then moved to the rear edge of the new de-encroaching zone. The distribution of target environmental field physical quantities was generated based on the new zoning layout. The central controller invokes the mapping model again. Calculate the optimal wind turbine command for the next round. The key to energy saving lies in the mapping model. Calculating new instructions When the chickens have left the original eviction zone and the area is no longer part of the new eviction zone, the system will automatically reduce the load of the fans to a minimum or shut them down, provided that the new environment is suitable. For example, the CF3 fans that were originally used to create an unsuitable environment in that sub-area may be shut down or have their speed significantly reduced under the new instructions, achieving timely energy savings. The central controller then issues a second set of fan coordination control instructions. The gradient environment field moves forward as a whole.

[0169] The aforementioned closed-loop process of perception-decision-execution-reperception continues. During the guidance process, the deportation zone and the dynamically changing corridor, as continuously evolving adaptive units, dynamically shift their position and range according to the movement of the flock's front, always maintaining a specific spatial relationship with the flock. Simultaneously, the fan execution cluster, as a collaborative execution mechanism, strictly limits its regulatory role to the currently defined effective zones, achieving precise environmental shaping and promptly withdrawing or reducing the intensity of regulation in areas where animals have left, thereby optimizing operational energy consumption.

[0170] Approximately 35 minutes later, the system, through its positioning module, determined that over 95% of the chickens had entered the target area and were in a stable distribution. The central controller then determined that the guidance task had been successfully completed and, based on seasonal and overall environmental parameters, sent a global uniform environmental control command to the fan cluster. All fans switch to the normal operating mode of low power consumption and uniform ventilation, and the process ends.

[0171] This invention establishes the distribution of physical quantities in the target environmental field. With the optimal fan control command set Mapping model between Based on real-time livestock location information, the system dynamically divides and moves target areas, eviction zones, and dynamically transitioning corridors connecting them, thereby driving the fan cluster to perform coordinated control and precisely create a continuous comfort gradient field within the breeding space. This method utilizes the instinct of livestock to seek advantage and avoid harm, integrating real-time perception and closed-loop control. It can automatically adapt to complex site conditions and differences in animal behavior, achieving low-stress, high-efficiency active migration of livestock along a preset path. This system not only solves the problems of low efficiency and high stress associated with traditional manual or mechanical guidance methods, but also overcomes the fundamental shortcomings of existing environmental control systems, such as their inability to sense location and their coarse control, making them unsuitable for behavioral guidance. Furthermore, by introducing a high-temperature penalty mechanism in the training of a mapping model aimed at accurately achieving the target environmental field, the system fundamentally avoids the formation of high-temperature areas that may cause high-temperature stress in livestock, significantly improving the safety and animal welfare of the guidance process. Simultaneously, through a dynamic zoning mechanism, the system applies precise environmental control only to the currently effective eviction zone and the dynamically transitioning corridor area, while promptly removing unnecessary fan loads in areas where livestock have already left, achieving a balance between guidance efficiency and energy saving. Furthermore, the mapping model constructed based on fluid dynamics (CFD) simulation and convolutional neural network (CNN) learning enables accurate and optimized solution of high-dimensional control commands from environmental field to wind turbine, providing a reliable technical means for intelligent behavior management in different aquaculture scenarios.

[0172] Example 2: This invention provides a livestock and poultry guidance device for coordinated regulation of environmental parameters by multiple fans, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor. The computer program can be executed by the processor to implement the livestock and poultry guidance method for coordinated regulation of environmental parameters by multiple fans as described in any paragraph of Example 1.

[0173] Example 3: The present invention provides a livestock and poultry guidance system for coordinated regulation of environmental parameters by multiple fans, which includes an environmental sensing network, a livestock and poultry positioning module, a central controller and a fan execution cluster.

[0174] An environmental sensing network is used to collect environmental physical quantity data from multiple points within the breeding space and send it to the central controller.

[0175] The livestock and poultry positioning module is used to acquire the location coordinates and distribution information of individual livestock and poultry or groups and send them to the central controller.

[0176] The central controller includes an environmental distribution and wind turbine control mapping model unit and a dynamic partitioning and decision algorithm unit, and is configured to execute the steps of the method described in Embodiment 1.

[0177] The fan execution cluster is used to receive the fan control command set issued by the central controller and regulate the environmental field distribution of the breeding space.

[0178] Obviously, the embodiments described above are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A livestock and poultry guidance method based on the coordinated regulation of environmental parameters by multiple fans, characterized in that, Include: S1. Pre-establish a mapping database of the airflow field of the ventilation system; the database defines the correspondence between the distribution field of environmental physical quantities in the target space and the set of operating parameters of the fan actuator; The mapping library includes inverse mapping models trained from paired data generated by computational fluid dynamics simulations; S2. Based on the guidance task and the real-time location of livestock and poultry, the ventilation area is divided into the controlled space into the target environment maintenance area, the repulsion environment construction area, and the airflow parameter gradual change channel connecting the two. S3. Based on the air parameter indicators set for each ventilation area, generate the target environmental physical quantity distribution field in the target space, and then input the inverse mapping model to solve the operating parameter set of the fan execution unit, adjust the air volume and air angle of each fan execution unit, and construct a differentiated airflow organization distribution with spatial gradient in the controlled space to drive livestock and poultry to move by using environmental pressure difference. S4. Monitor the displacement changes of livestock and poultry in the airflow parameter gradual change channel. When the monitoring data meets the area switching conditions, adjust the spatial range of the repulsion environment construction zone and the airflow parameter gradual change channel, and return to S3 until the livestock and poultry enter the target environment maintenance zone. S5. When the number of livestock and poultry in the target environment maintenance area is detected to reach a preset threshold, the control fan execution unit switches to the global uniform air supply mode. S4 specifically includes: Based on livestock and poultry location data, determine the displacement or distribution changes of livestock and poultry within the airflow parameter gradual change channel; When the displacement or distribution change meets the region switching conditions, the spatial range of the repulsion environment construction area is reduced or moved forward, and the spatial range of the airflow parameter gradient channel is adjusted simultaneously, so that the starting end of the airflow parameter gradient channel is connected to the adjusted repulsion environment construction area; wherein, the region switching conditions include: within the update cycle, the proportion of grid cells "occupied by livestock and poultry" in the grid cells divided by the gradient channel is counted as the channel occupancy rate. And calculate the displacement of the centroid of the livestock group along the guiding direction. and the advancing distance of the forward front on the side of the passage entrance ;when and and the duration is not less than ,or and the duration is not less than At that time, a forward update of the eviction environment construction area is triggered; among which, This is the channel occupancy threshold. The threshold for the centroid displacement of livestock and poultry populations; The threshold for the forward advance distance; This is the duration threshold; Return to execute S3 to update the fan control instruction set until livestock enter the target environmental maintenance zone; S5 specifically includes: Multiple fan actuators are controlled to operate in a coordinated manner to ensure uniform ventilation, thereby making the air parameter distribution within the controlled space more uniform. While ensuring uniform ventilation, reduce the speed of at least some fan actuators or shut down at least some fan actuators to achieve low-power operation.

2. The livestock and poultry guidance method based on multi-fan coordinated regulation of environmental parameters according to claim 1, characterized in that, The inverse mapping model takes the distribution data of temperature field, wind speed field and humidity field in the target space as input, and the set of operating parameters consisting of the switching state, speed and air delivery angle of each fan actuator as output, and is trained using a loss function that includes physical constraint terms. S1 specifically includes: S11. Construct a three-dimensional simulation model that includes the geometric features of the controlled space and the structural features of the internal equipment; S12. Based on the three-dimensional simulation model, by adjusting the combination of operating parameters of the fan actuator, batch computational fluid dynamics simulation is performed to obtain multiple sets of steady-state flow field data; S13. Extract the distribution data of environmental field physical quantities at the activity height level of livestock and poultry from the steady-state flow field data, and perform data cleaning and normalization to construct a training dataset containing "paired data of wind turbine control command set - distribution of environmental field physical quantities". S14. Construct a neural network solver; S15. Design a loss function that includes physical constraints, and use the training dataset to iteratively train the neural network solver until the loss function converges, thereby obtaining an inverse mapping model from the distribution field of environmental physical quantities to the set of wind turbine operating parameters.

3. The livestock and poultry guidance method based on multi-fan coordinated regulation of environmental parameters according to claim 2, characterized in that, S13 specifically includes: Within the computational domain of the 3D simulation model, a horizontal slice plane covering the height of livestock and poultry activity is extracted; The horizontal slice plane is discretized into a regular grid point matrix, and the temperature, wind speed and humidity at each grid point are extracted. Temperature field feature map, wind speed field feature map and humidity field feature map are generated respectively, and the physical quantities in the feature maps are normalized without dimension to obtain the distribution data of environmental field physical quantities. The set of operating parameters for the corresponding fan actuators during simulation is obtained. The set of operating parameters includes the on / off state, speed and air delivery angle of each fan actuator, and is then vectorized. The environmental field physical quantity distribution data is correlated with the set of operating parameters to obtain the training dataset; The neural network solver of S14 is a deep convolutional neural network model; the network includes an input layer, hidden layers, and an output layer. The input layer receives normalized environmental field physical quantity distribution data, and the dimension of the input data is... The number 3 represents the three characteristic channels: temperature, wind speed, and humidity. and These represent the number of rows and columns of the regular grid point matrix, respectively. The hidden layer contains multiple alternating convolutional layers, pooling layers, and fully connected layers, used to extract spatial features of the environmental flow field and perform dimensionality reduction mapping; The size of the output layer is consistent with the total number of controllable parameters of the wind turbine actuator, and the output vector corresponds to the set of operating parameters of the wind turbine actuator.

4. The livestock and poultry guidance method based on multi-fan coordinated regulation of environmental parameters according to claim 2, characterized in that, The loss function for S15 is as follows: ； ； ； In the formula, This is the total loss function; Mean square error; As a temperature penalty; for The total dimensions of typhoon fan on / off status, speed, and wind direction angle control parameters; To predict the instruction vector; This is the actual instruction vector; The regularization coefficient is used. This indicates taking the larger value; For the prediction of the mapping model Temperature of the coordinates; This represents the upper limit of the comfortable temperature corresponding to the current livestock and poultry species and growth stage. S11 specifically includes: A full-size geometric model is created using the architectural drawings of the controlled space; the geometric model includes the internal structure of the cage, material line, and waterline. The fluid domain is meshed, and mesh independence is verified to determine the optimal number of meshes. Set the fan actuator as the velocity inlet boundary and the exhaust outlet as the pressure outlet boundary; S12 specifically includes: The operating parameter combinations of the fan actuator are enumerated to form multiple sets of simulation conditions, and computational fluid dynamics simulation is performed on each set of conditions until steady state is reached. Record the spatial distribution data of temperature, wind speed and humidity fields at the height of livestock and poultry activity during steady state of each set of working conditions to form the corresponding distribution data of environmental field physical quantities. Record the set of fan control instructions corresponding to each set of operating conditions. The set of fan control instructions shall include at least the on / off status, speed and air delivery angle of each fan actuator.

5. The livestock and poultry guidance method based on multi-fan coordinated regulation of environmental parameters according to claim 1, characterized in that, S2 specifically includes: Set the preset fixed endpoint area as the target environment maintenance zone, and set its environmental field to the comfort range of livestock and poultry. The initial gathering area of ​​livestock and poultry is set as a repulsive environment construction area, and its environmental field is set as an unsuitable environment that prompts livestock and poultry to leave. Establish a guiding trajectory connecting the edge of the destructive environment construction zone and the entrance of the target environment maintenance zone, and set the area covered by this trajectory as a gradual airflow parameter channel, with its environmental field set as a continuous gradient field from an unpleasant environment to a comfortable environment.

6. The livestock and poultry guidance method based on multi-fan coordinated regulation of environmental parameters according to claim 5, characterized in that, The air parameter index corresponding to the comfort range of livestock and poultry is set as the first target air parameter index to meet the comfort range of livestock and poultry. The air parameter index corresponding to the unpleasant environment that prompts livestock and poultry to leave is set as a second target air parameter index that deviates from the comfort range of livestock and poultry, so as to form a slightly unpleasant environment for driving away livestock and poultry. The air parameter index corresponding to the airflow parameter gradient channel is set as a gradient air parameter index that changes continuously and smoothly from the second target air parameter index to the first target air parameter index.

7. The livestock and poultry guidance method based on multi-fan coordinated regulation of environmental parameters according to claim 1, characterized in that, S3 specifically includes: Based on the air parameter indices corresponding to the target environment maintenance zone, the repulsion environment construction zone, and the airflow parameter gradual change channel, a target environment physical quantity distribution field is generated within the target space. The target environmental physical quantity distribution field is input into the inverse mapping model corresponding to the mapping relationship library, and the operating parameter set of the fan execution unit is calculated and the fan control command set is generated. Among them, the operating parameter set is subjected to tolerance threshold constraints and adjacent fan control quantity smoothing processing to ensure that the output command meets the preset safety ventilation / air conditioning constraints and reduces spatial airflow abrupt changes. The control fan actuator adjusts the air volume and air angle according to the fan control instruction set to construct a differentiated airflow organization distribution with spatial gradient within the controlled space; The set of operating parameters includes at least the on / off status, speed and air delivery angle of each fan actuator; The air parameters in the target environmental maintenance zone meet the comfort range constraints for livestock and poultry. ; ; In the formula, , and These are the set values ​​for temperature, wind speed, and humidity in the target environment maintenance zone, respectively. , and These are the lower limits of temperature, wind speed, and humidity for the comfort of livestock and poultry, respectively. , and These are the upper limits for temperature, wind speed, and humidity, respectively, for the comfort of livestock and poultry; The air parameters in the repulsion environment construction area meet the constraints of mild discomfort and not exceeding the tolerance threshold; ; ; ;or ; ; In the formula, , and These are the set values ​​for temperature, wind speed, and humidity in the repulsive environment construction zone, respectively. , and These are the upper limits of temperature, wind speed, and humidity that livestock and poultry can tolerate, respectively. , and These are the lower limits of temperature, wind speed, and humidity that livestock and poultry can tolerate, respectively. The gradient air parameter indices of the gradually changing airflow parameter channel satisfy a gradient function that changes continuously along the channel path position; ； ； ； In the formula, The path position from the end of the dispersive environment construction zone to the end of the target environment maintenance zone along the gradually changing airflow parameter channel; This refers to the channel length; , and Positions Temperature, wind speed, and humidity at the location; , and These are the set values ​​for temperature, wind speed, and humidity in the repulsive environment construction zone, respectively. , and These are the set values ​​for temperature, wind speed, and humidity in the target environment maintenance zone, respectively. , and They are respectively in The gradient functions of temperature, wind speed, and humidity that change monotonically within the interval.

8. A livestock and poultry guidance device that uses multiple fans to coordinate and regulate environmental parameters, characterized in that: Includes a processor, a memory, and a computer program stored in the memory and executable by the processor; The computer program can be executed by the processor to implement a livestock and poultry guidance method based on the coordinated regulation of environmental parameters by multiple fans as described in any one of claims 1 to 7.

9. A livestock and poultry guidance system that uses multiple fans to coordinate and regulate environmental parameters, characterized in that: include: An environmental sensing network is used to collect environmental physical quantity data from multiple points within the breeding space and send them to the central controller. The livestock and poultry positioning module is used to acquire the location coordinates and distribution information of individual livestock and poultry or groups of livestock and poultry and send them to the central controller. The central controller includes an environmental distribution and wind turbine control mapping model unit and a dynamic partitioning and decision algorithm unit, and is configured to execute the livestock and poultry guidance method based on the coordinated regulation of environmental parameters by multiple wind turbines as described in any one of claims 1 to 7. The fan execution cluster is used to receive the fan control command set issued by the central controller and regulate the environmental field distribution of the breeding space.

Images