Chicken house environment temperature intelligent monitoring system based on artificial intelligence

Abstract

The invention relates to the technical field of supervision control and data acquisition, and particularly discloses a henhouse environment temperature intelligent monitoring system based on artificial intelligence. The system comprises a space temperature field reconstruction module, a multi-source environmental parameter fusion analysis module, a temperature dynamic trend prediction module and a self-adaptive regulation and control decision module. High-precision sensing, multi-parameter coupling analysis, dynamic trend prediction and self-adaptive regulation and control of a henhouse temperature field are realized, and the accuracy, uniformity and energy efficiency of temperature control are improved.

Description

Technical Field

[0001] This invention belongs to the field of monitoring, control and data acquisition technology, and specifically relates to an intelligent monitoring system for chicken house environmental temperature based on artificial intelligence. Background Technology

[0002] In modern livestock and poultry farming, precise monitoring and control of environmental parameters are crucial for ensuring animal health and improving production efficiency. Among these, chicken house temperature is a core factor affecting flock growth performance, immune status, and feed conversion rate, and the development and application of intelligent monitoring systems for it constitute an important branch of this field.

[0003] Intelligent monitoring of chicken house environmental temperature aims to achieve real-time perception and dynamic control of the temperature field inside the house through sensing technology and data analysis, thereby creating a stable and suitable growth environment for the chickens. Existing technologies mainly rely on point monitoring using temperature sensors deployed in limited locations, and trigger ventilation or heating equipment to adjust the environment by setting preset thresholds.

[0004] Existing technologies struggle to accurately capture the uneven temperature distribution within chicken houses caused by spatial layout, flock activity, and ventilation. The monitoring data suffers from significant "single-point distortion" and cannot fully reflect the true thermal environment at different heights and in different areas.

[0005] Meanwhile, temperature changes are highly coupled with various environmental parameters such as humidity, ventilation volume and outdoor weather. Traditional systems lack the ability to model the interaction mechanism of multiple parameters, resulting in delayed control response, high energy consumption, and an inability to effectively predict and proactively intervene in temperature trends.

[0006] Therefore, given the demand for large-scale and refined aquaculture, developing a temperature monitoring system that can overcome monitoring blind spots and achieve intelligent coupling analysis and precise control of multiple parameters has become an urgent technical challenge. Summary of the Invention

[0007] The technical problem this invention aims to solve is to overcome the shortcomings of existing chicken house environmental temperature monitoring systems, such as single-point distortion, lack of multi-parameter coupling analysis capabilities, and delayed control response. This invention provides an artificial intelligence-based intelligent monitoring system for chicken house environmental temperature. This system aims to achieve high-precision sensing of the uneven distribution of the temperature field inside the chicken house, in-depth modeling of the coupling effects of multiple parameters such as temperature, humidity, ventilation volume, and outdoor weather, and accurate prediction and forward-looking control of temperature change trends, thereby providing a stable, uniform, and suitable thermal environment for the chickens.

[0008] To achieve the above objectives, this invention provides an intelligent temperature monitoring system for chicken coops based on artificial intelligence. The system includes a spatial temperature field reconstruction module, a multi-source environmental parameter fusion analysis module, a temperature dynamic trend prediction module, and an adaptive control decision-making module.

[0009] The spatial temperature field reconstruction module is used to deploy a distributed temperature sensor array in the 3D space inside the chicken house to collect raw temperature data at different heights and in different areas inside the chicken house. This module further performs spatial interpolation calculations and 3D temperature field modeling on the collected raw temperature data to generate a 3D temperature field cloud map that can characterize the overall thermal environment distribution inside the chicken house.

[0010] The spatial interpolation calculation adopts the Kriging algorithm based on inverse distance weighting, which comprehensively considers the spatial positional relationship between sensors and the spatial autocorrelation of temperature data.

[0011] The multi-source environmental parameter fusion analysis module is used to collect humidity data, carbon dioxide concentration data, ventilation system operating parameters including wind speed and air volume inside the chicken house, and meteorological data outside the chicken house including outdoor temperature, humidity, wind speed and solar radiation intensity.

[0012] This module further constructs a multimodal data fusion network based on an attention mechanism to extract and fuse features from temperature, humidity, carbon dioxide concentration, ventilation parameters, and external meteorological data, quantifying the coupling influence weights of various environmental parameters on the internal temperature field of the chicken house. This multimodal data fusion network first maps various heterogeneous data into a high-dimensional feature space through independent feature encoders, then calculates the interaction importance between features of different environmental parameters through a multi-head attention layer, ultimately outputting an environmental state feature vector that integrates multi-source information.

[0013] The temperature dynamic trend prediction module is connected to the multi-source environmental parameter fusion analysis module to receive environmental state feature vectors.

[0014] This module constructs a time-series prediction model based on a hybrid architecture of gated recurrent units and temporal convolutional networks. The model takes the environmental state feature vector of historical time series and 3D temperature field cloud map as input. It captures the long-term temporal dependence of temperature changes through gated recurrent units, and extracts the local patterns and short-term fluctuation features of temperature changes through temporal convolutional networks. Finally, it outputs the predicted change trajectory of the 3D temperature field inside the chicken house within a specific future time window.

[0015] The adaptive control decision module is connected to the temperature dynamic trend prediction module to receive the predicted trajectory of future temperature field changes.

[0016] This module first sets the target temperature range and temperature uniformity index corresponding to different growth stages of the chicken flock.

[0017] This module further constructs a control policy generator based on deep reinforcement learning. The generator takes the current 3D temperature field, the predicted temperature change trajectory, and the target temperature parameters as state inputs, and the control commands of ventilation equipment, heating equipment, and wet curtain equipment as action space. It evaluates the value of different control actions under the long-term cumulative reward function through a deep Q-network algorithm, thereby generating the optimal sequence of equipment control commands.

[0018] This cumulative reward function comprehensively considers multiple optimization objectives, such as temperature control accuracy, energy consumption cost, and temperature field uniformity.

[0019] In a preferred embodiment of the present invention, the distributed temperature sensor array deployed in the spatial temperature field reconstruction module adopts a three-dimensional grid arrangement. The sensor nodes are evenly spaced in the length, width and height directions of the chicken house to ensure no dead angle coverage of the internal space of the chicken house.

[0020] Furthermore, the sensor nodes employ digital temperature sensors with wireless self-organizing network communication capabilities. The nodes transmit data wirelessly via the Zigbee or LoRa protocol, and the aggregation node uploads the data to the central processing server via a 4G or 5G network.

[0021] As a preferred embodiment of the present invention, the attention mechanism multimodal data fusion network in the multi-source environmental parameter fusion analysis module has 8 attention heads in its multi-head attention layer, and each attention head has 64 dimensions.

[0022] Furthermore, the network employs a mean squared error loss function and the Adam optimizer for parameter optimization during training, with the initial learning rate set to 0.001 and an exponential decay strategy adopted.

[0023] In a preferred embodiment of the present invention, the time-series prediction model in the temperature dynamic trend prediction module has 128 hidden layer units in its gated recurrent unit, 3 kernels in its temporal convolutional network, and 2 convolutional layers.

[0024] The prediction time window length of this model can be configured to 30 min, 60 min or 120 min according to actual management needs.

[0025] In a preferred embodiment of the present invention, the control strategy generator in the adaptive control decision module has a deep Q-network with a double hidden layer structure, each layer containing 256 neurons, and the activation function is the ReLU function.

[0026] Furthermore, the generator employs an experience replay mechanism and target network technology during training to improve learning stability, with the initial exploration rate set to 0.1 and linearly decaying to 0.01 over time.

[0027] In a preferred embodiment of the present invention, the system further includes a system health self-diagnosis and fault tolerance module.

[0028] This module monitors the working status and data quality of each sensor node in the distributed temperature sensor array in real time. When an abnormal data or communication interruption is detected in a sensor node, the fault tolerance process is automatically initiated.

[0029] The fault-tolerant processing flow includes filling in missing data based on data from nearby sensors and recalculating the interpolation weights in the spatial temperature field reconstruction module to ensure the continuity and reliability of the 3D temperature field cloud map generation.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention constructs a 3D temperature field cloud map that can accurately reflect the uneven temperature distribution inside the chicken house by deploying a 3D distributed temperature sensor array and combining it with a spatial interpolation algorithm. This fundamentally solves the problem of single-point distortion caused by traditional point monitoring and realizes accurate perception of the global thermal environment of the chicken house.

[0031] 2. By introducing a multimodal data fusion network based on an attention mechanism, this invention can deeply explore the complex coupling relationships between multiple environmental parameters such as temperature, humidity, ventilation, and outdoor weather, and quantify the influence weight of each parameter on the temperature field, providing a solid data foundation and model support for subsequent accurate prediction and control.

[0032] 3. The present invention adopts a time-series prediction model with a hybrid architecture of gated recurrent units and temporal convolutional networks, which can simultaneously capture the long-term trend and short-term fluctuations of temperature changes, and achieve high-precision prediction of the trajectory of future temperature field changes. This enables the system to make forward-looking interventions and effectively overcomes the drawbacks of traditional system regulation lag.

[0033] 4. This invention generates an adaptive control strategy based on a deep reinforcement learning framework. This strategy can comprehensively consider multiple objectives such as temperature control accuracy, energy consumption and uniformity, and dynamically generate the optimal equipment control command sequence. It realizes intelligent, energy-saving and refined control in complex and ever-changing environments, and significantly improves the automation level and economic benefits of chicken house environmental management.

[0034] 5. The system health self-diagnosis and fault tolerance module integrated in this invention can ensure the continuous and reliable operation of the monitoring system in the event of partial sensor failure, enhance the robustness and practicality of the entire system, and meet the high requirements of large-scale farms for system stability. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the overall technical solution architecture of the intelligent monitoring system for chicken house environmental temperature based on artificial intelligence proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the attention-based multimodal data fusion network in this invention; Figure 3 This is a logical flow diagram of the temperature dynamic trend prediction module in this invention; Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow of the adaptive control decision module in this invention; Figure 5 This is a flowchart of the workflow of the system health self-diagnosis and fault tolerance module in this invention; Detailed Implementation

[0036] Please refer to the attached document. Figure 1 This embodiment details the technical implementation of an AI-based intelligent temperature monitoring system for chicken houses. This system aims to achieve comprehensive perception, in-depth analysis, accurate prediction, and intelligent control of the internal thermal environment of chicken houses through highly integrated hardware deployment and advanced algorithm models.

[0037] The core components of the system include a space temperature field reconstruction module, a multi-source environmental parameter fusion analysis module, a temperature dynamic trend prediction module, an adaptive control decision module, and a system health self-diagnosis and fault tolerance module.

[0038] The modules interact with each other via a high-speed data bus to ensure the real-time performance and consistency of information flow.

[0039] The spatial temperature field reconstruction module is the fundamental physical layer for temperature sensing in the entire system. This module deploys a three-dimensional grid-like distributed temperature sensor array in the three-dimensional space inside the chicken coop.

[0040] Specifically, in the length direction of the chicken house, sensor nodes are evenly distributed at fixed intervals of 5m; in the width direction of the chicken house, sensor nodes are evenly distributed at fixed intervals of 4m; and in the height direction of the chicken house, sensor nodes are deployed at three different height levels: 0.5m, 1.5m, and 2.5m above the ground.

[0041] This three-dimensional grid arrangement ensures comprehensive temperature monitoring coverage throughout the chicken house, from the ground to the roof. Each sensor node uses a temperature sensing chip with high-precision digital signal output capability, achieving a temperature measurement accuracy of ±0.1℃ and a measurement range covering -10℃ to 60℃.

[0042] The sensor node integrates a microprocessor and a wireless communication unit, and the nodes transmit data wirelessly to each other through a self-organizing mesh network based on the Zigbee protocol.

[0043] The network contains several aggregation nodes, which are responsible for collecting temperature data from all ordinary sensor nodes within their communication range and uploading the aggregated data packets to the central processing server located in the monitoring center via built-in 4G or 5G communication modules.

[0044] The raw temperature data received by the central processing server is time-series data with spatial coordinates. Each data point contains the sensor node number, acquisition timestamp, temperature measurement value, and the specific location coordinates of the node in the 3D space of the chicken house.

[0045] Inside the central processing server, the spatial temperature field reconstruction module initiates the 3D temperature field modeling process.

[0046] The process first cleans and preprocesses the received raw temperature data, removing obvious outliers caused by transient communication interference, and then timestamps all valid data points.

[0047] Subsequently, the module calls the Kriging spatial interpolation algorithm based on inverse distance weighting to perform spatial interpolation calculations on the discretely distributed sensor measurement points to generate a continuous 3D temperature field.

[0048] The core idea of ​​this algorithm is that the temperature value of any unmeasured point in space is a weighted average of the temperature values ​​of known measuring points around it. The weight is inversely proportional to the distance from the unmeasured point to the known measuring point, while also taking into account the autocorrelation of the temperature data itself in space.

[0049] The specific calculation process involves constructing a variogram model to quantify spatial autocorrelation and solving the Kriging equations to obtain the optimal interpolation weights.

[0050] Finally, the module outputs a high-resolution 3D temperature field cloud map, which is stored in the server memory in the form of a digital matrix. Each element of the matrix represents the temperature estimate of a virtual voxel in the 3D space of the chicken house, thus fully characterizing the distribution of the overall thermal environment inside the chicken house, including high-temperature zones, low-temperature zones, and temperature gradient distribution.

[0051] The multi-source environmental parameter fusion analysis module is responsible for collecting and deeply fusing various environmental parameters that affect the temperature field in the chicken house. Please refer to the attached document. Figure 2This module collects humidity data inside the chicken house through an independent sensor subsystem with a measurement accuracy of 3%; carbon dioxide concentration data, with a measurement range of 0 to 5000 ppm; and ventilation system operating parameters, including wind speed values ​​measured by ultrasonic wind speed sensors installed at the air inlet and outlet, and real-time air volume values ​​calculated based on fan speed and damper opening.

[0052] Meanwhile, the module collects outdoor meteorological data, including outdoor temperature, outdoor humidity, outdoor wind speed, and solar radiation intensity, through a miniature weather station deployed outside the chicken coop.

[0053] All of these multi-source heterogeneous data are collected synchronously and stamped with a unified timestamp, and then transmitted to the central processing server via wired or wireless means.

[0054] Within the server, the multi-source environmental parameter fusion analysis module constructs and runs a multimodal data fusion network based on an attention mechanism.

[0055] The network first contains five independent feature encoders, corresponding to temperature field features, humidity features, carbon dioxide concentration features, ventilation parameter features, and external meteorological features, respectively.

[0056] The temperature field feature encoder is a 3D convolutional neural network. Its input is a 3D temperature field cloud map data matrix from the spatial temperature field reconstruction module. Through three layers of 3D convolution and pooling operations, it extracts high-dimensional feature vectors that characterize the spatial distribution pattern of temperature.

[0057] The humidity feature encoder, carbon dioxide concentration feature encoder, ventilation parameter feature encoder, and external meteorological feature encoder are all fully connected neural networks. They respectively map the raw data of scalar or one-dimensional vector to a 64-dimensional feature space.

[0058] The high-dimensional feature vectors output by all feature encoders are concatenated into a comprehensive feature matrix, which is then fed into the subsequent multi-head attention layer.

[0059] The multi-head attention layer is the core of this network, with 8 attention heads and 64 four-dimensional attention heads for each attention head.

[0060] This layer dynamically evaluates the importance of interactions between different environmental parameter features by calculating the dot product attention between the query vector, key vector, and value vector.

[0061] Specifically, for each attention head, the input feature matrix is ​​transformed into a query matrix, a key matrix, and a value matrix through linear transformation.

[0062] Then, the dot product of the query matrix and the transpose of the key matrix is ​​calculated, and after scaling, the Softmax function is applied to obtain the attention weight matrix. This weight matrix is ​​then multiplied by the value matrix to obtain the output of the attention head.

[0063] The outputs of the eight attention heads are concatenated and passed through a linear projection layer to obtain the final multi-head attention output.

[0064] This process enables the network to simultaneously focus on information from different representation subspaces, thereby more accurately capturing, for example, the combined effects of humidity changes and ventilation efficiency on the temperature field, or the indirect relationship between outdoor solar radiation and indoor carbon dioxide concentration accumulation.

[0065] The network ultimately outputs a fixed-dimensional environmental state feature vector, which integrates the essence of all multi-source information and quantifies the coupling influence weight of each environmental parameter on the internal temperature field of the chicken house.

[0066] The temperature dynamic trend prediction module receives historical time series environmental state feature vectors and historical time series 3D temperature field cloud map data from the multi-source environmental parameter fusion analysis module.

[0067] Please refer to the attached document. Figure 3 This module builds and runs a time series prediction model based on a hybrid architecture of gated recurrent units and temporal convolutional networks.

[0068] The model's input is a fixed-length time window of data, such as historical data from the past 60 time steps. Each time step contains an environmental state feature vector and a 3D temperature field contour map.

[0069] The gated loop unit is responsible for capturing the long-term time dependence of temperature changes.

[0070] This section contains a hidden layer with 128 cells.

[0071] The gated loop unit, through its update gate and reset gate mechanism, can selectively remember or forget historical information, thereby effectively learning long-term patterns in time series.

[0072] The temporal convolutional network works in parallel, responsible for extracting local patterns and short-term fluctuations in temperature changes.

[0073] This part contains two causal convolutional layers to ensure that the prediction at the current time depends only on past information, with the kernel size set to 3.

[0074] Temporal convolutional networks increase the receptive field through dilated convolution to capture dependencies at different time scales.

[0075] The output features of the gated recurrent unit and the temporal convolutional network are concatenated in the model's fusion layer, and then mapped to the prediction space for a specific future time window through a fully connected output layer.

[0076] The length of the forecast time window can be flexibly configured according to the actual chicken house management needs, with typical options including 30 min, 60 min or 120 min.

[0077] The model's final output is the predicted trajectory of the 3D temperature field inside the chicken house within the future time window. This trajectory is also represented in the form of a 3D data matrix of time series, predicting the temperature value and its trend at each point inside the chicken house at each future time step.

[0078] The adaptive control decision module is the core of the system's intelligent control.

[0079] This module receives the predicted future temperature field trajectory from the temperature dynamic trend prediction module, and the current 3D temperature field data provided by the spatial temperature field reconstruction module. Please refer to the appendix. Figure 4 This module has preset target temperature ranges and temperature uniformity indicators for chickens at different growth stages.

[0080] For example, the target temperature range for the chick stage is 32 to 35°C, and the uniformity index requires that the temperature difference between any two points in the chicken house does not exceed 2°C; the target temperature range for the rearing stage is 20 to 24°C, and the uniformity index requires that the temperature difference does not exceed 3°C. These target parameters serve as the benchmark for regulation.

[0081] This module further constructs a regulation policy generator based on deep reinforcement learning.

[0082] The generator's state space includes the current 3D temperature field data matrix, the future predicted temperature change trajectory data matrix, and the current target temperature parameters.

[0083] The action space is defined as the adjustment commands for the fan speed of the ventilation equipment, the adjustment commands for the power level of the heating equipment, and the adjustment commands for the start / stop of the water pump and the water flow of the evaporative cooling pad equipment.

[0084] The regulation strategy generator uses a deep Q-network algorithm to evaluate the long-term cumulative value that can be obtained by executing various possible regulation actions under different system states.

[0085] The deep Q-network uses a dual-hidden-layer structure, with each layer containing 256 neurons, and the activation function is the ReLU function.

[0086] Its core is learning an action-value function, or Q-function. This function is approximately expressed as:

[0087] Where S represents the state. Representative action, The parameters representing the neural network, Represents from time The initial cumulative reward. The network updates its parameters by minimizing the temporal difference error.

[0088] The design of the cumulative reward function takes into account multiple optimization objectives: temperature control accuracy reward, which encourages minimizing the deviation between the actual temperature and the target temperature; energy consumption cost penalty, which is proportional to the equipment operating power; and temperature field uniformity reward, which encourages reducing the standard deviation of the temperature distribution in the chicken house.

[0089] During training, the generator employs an experience replay mechanism, storing experience samples of system interactions in a replay buffer and randomly sampling from it for batch training to break the correlation between data.

[0090] Meanwhile, a target network technique is adopted, which uses a target network with the same structure but slower parameter updates to calculate the target Q value, thereby improving the stability of learning.

[0091] The exploration rate is set to 0.1 at the beginning of training to encourage the agent to explore new control strategies. It then decays linearly to 0.01 over time, gradually shifting towards utilizing learned best strategies. After training, the deployed generator can output the optimal sequence of device control instructions based on real-time state inputs. For example, if a significant increase in afternoon temperature is predicted, it can increase ventilation volume and activate evaporative cooling pads in advance, thereby achieving proactive, adaptive, and precise environmental control.

[0092] The system's self-diagnosis and fault-tolerance module provides reliability assurance for the entire monitoring system. Please refer to the appendix. Figure 5 This module continuously monitors the working status of each sensor node in the distributed temperature sensor array.

[0093] Monitoring indicators include the frequency of data reported by nodes, the reasonableness of data values ​​(e.g., whether they are within the preset physical possibilities), and the communication signal strength and link quality between nodes and the aggregation node. This module maintains a sensor node status table, recording the health status score of each node.

[0094] When the module detects that a sensor node has not sent data for three consecutive reporting cycles, or that the temperature value it reports has a significant statistical anomaly compared to the measurement values ​​of its spatial neighbors, it determines that the node is in a data anomaly or communication interruption state and automatically triggers the fault tolerance process.

[0095] The fault-tolerant processing procedure first involves filling in missing data. Based on the location of the abnormal node in the 3D grid, the module finds its six spatially nearest valid sensor nodes, typically in the up, down, left, right, front, and back directions. Using the real-time temperature measurements from these valid nodes, an inverse distance-weighted average algorithm is employed to calculate the estimated temperature at the abnormal node's location.

[0096] Specifically, the estimated temperature of the missing points Calculated as the temperature of neighboring points The weighted sum, weight Distance between neighboring points and missing points The reciprocal The power, usually The value is set to 2. Subsequently, the module adjusts the interpolation weight assigned to this anomalous node when calculating the 3D temperature field to zero, and uses the calculated estimated temperature value as a virtual observation value at this location in subsequent spatial interpolation calculations.

[0097] This mechanism ensures that even if some sensor nodes fail, the 3D temperature field cloud map generated by the spatial temperature field reconstruction module remains continuous and reliable, and the entire system can continue to operate stably.

[0098] This embodiment constructs a complete closed-loop system from physical perception to intelligent decision-making through the precise collaborative work of the above five core modules.

[0099] The space temperature field reconstruction module provides a high-precision base map of space temperature distribution; the multi-source environmental parameter fusion analysis module deeply analyzes the coupling mechanism of multiple factors affecting temperature; the temperature dynamic trend prediction module realizes accurate prediction of future temperature changes; the adaptive control decision module generates the optimal control strategy that comprehensively considers multi-objective optimization; and the system health self-diagnosis and fault tolerance module ensures the robustness of the system's long-term operation.

[0100] This system effectively solves the problems of single-point distortion, one-sided analysis, and lagging control in traditional chicken house temperature monitoring, providing a solid technical foundation for realizing intelligent, refined, and energy-saving management of the chicken house environment.

[0101] All module algorithm models are trained and deployed on a central processing server using the TensorFlow or PyTorch deep learning framework, and exchange data and transmit instructions through well-defined application programming interfaces. The system provides a graphical user interface for real-time display of 3D temperature field cloud maps, predicted trend curves, equipment operating status, and system alarm information, facilitating monitoring and intervention by administrators.

[0102] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0103] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence, characterized in that, include: The spatial temperature field reconstruction module is used to deploy a distributed temperature sensor array in the 3D space inside the chicken house to collect raw temperature data at different heights and in different areas inside the chicken house; and to perform spatial interpolation calculation and 3D temperature field modeling on the collected raw temperature data to generate a 3D temperature field cloud map that represents the overall thermal environment distribution inside the chicken house. The multi-source environmental parameter fusion analysis module is used to collect humidity data, carbon dioxide concentration data, ventilation system operating parameters including wind speed and air volume inside the chicken house, and meteorological data outside the chicken house including outdoor temperature, humidity, wind speed and solar radiation intensity. It also constructs a multimodal data fusion network based on the attention mechanism to extract and fuse features from temperature data, humidity data, carbon dioxide concentration data, ventilation parameters and external meteorological data to quantify the coupling influence weight of each environmental parameter on the temperature field inside the chicken house. The temperature dynamic trend prediction module is connected to the multi-source environmental parameter fusion analysis module to receive environmental state feature vectors. And a temporal prediction model based on a hybrid architecture of gated recurrent units and temporal convolutional networks is constructed; The adaptive control decision module is connected to the temperature dynamic trend prediction module to receive the predicted trajectory of future temperature field changes.

2. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 1, characterized in that, The spatial interpolation calculation adopts the Kriging algorithm based on inverse distance weighting, which comprehensively considers the spatial positional relationship between sensors and the spatial autocorrelation of temperature data.

3. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 1, characterized in that, The multimodal data fusion network first maps various heterogeneous data to a high-dimensional feature space through an independent feature encoder, then calculates the interaction importance between different environmental parameter features through a multi-head attention layer, and finally outputs an environmental state feature vector that integrates multi-source information.

4. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 1, characterized in that, The time-series prediction model takes the environmental state feature vector of historical time series and the 3D temperature field cloud map as input. It captures the long-term time dependence of temperature changes through a gated recurrent unit, and extracts the local patterns and short-term fluctuation features of temperature changes through a temporal convolutional network. Finally, it outputs the predicted change trajectory of the 3D temperature field inside the chicken house within a specific future time window.

5. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 1, characterized in that, The adaptive control decision module first sets the target temperature range and temperature uniformity index corresponding to different growth stages of the chicken flock. Furthermore, a control policy generator based on deep reinforcement learning is constructed. This generator takes the current 3D temperature field, the predicted temperature change trajectory, and the target temperature parameters as state inputs, and the control commands of ventilation equipment, heating equipment, and wet curtain equipment as action space. It evaluates the value of different control actions under the long-term cumulative reward function through a deep Q-network algorithm, thereby generating the optimal sequence of equipment control commands. The cumulative reward function comprehensively considers multiple optimization objectives, including temperature control accuracy, energy consumption cost, and temperature field uniformity.

6. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 1, characterized in that, The distributed temperature sensor array deployed in the spatial temperature field reconstruction module adopts a three-dimensional grid arrangement, with sensor nodes evenly spaced along the length, width, and height of the chicken house. The sensor nodes are digital temperature sensors with wireless self-organizing network communication capabilities, and the nodes transmit data wirelessly via Zigbee or LoRa protocols. The aggregation node uploads the data to the central processing server.

7. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 5, characterized in that, The control strategy generator in the adaptive control decision module has a deep Q-network with a double hidden layer structure and the activation function is the ReLU function; the generator adopts an experience replay mechanism and target network technology during training.

8. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 1, characterized in that, The AI-based intelligent monitoring system for chicken house environmental temperature also includes a system health self-diagnosis and fault tolerance module, which is used to monitor the working status and data quality of each sensor node in the distributed temperature sensor array in real time. When abnormal sensor node data or communication interruption is detected, the fault tolerance process is automatically started. The fault tolerance process includes filling in missing data based on data from neighboring sensors and recalculating the interpolation weights in the spatial temperature field reconstruction module.

9. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 3, characterized in that, The feature encoders in the multi-source environmental parameter fusion analysis module include: a temperature field feature encoder using a 3D convolutional neural network, with a 3D temperature field cloud map data matrix as input; and humidity feature encoders, carbon dioxide concentration feature encoders, ventilation parameter feature encoders, and external meteorological feature encoders, all of which are fully connected neural networks.

10. The intelligent monitoring system for chicken coop environmental temperature based on artificial intelligence according to claim 1, characterized in that, The 3D temperature field modeling process in the spatial temperature field reconstruction module includes: cleaning and preprocessing the received raw temperature data, removing outliers and synchronizing timestamps; then calling the Kriging spatial interpolation algorithm based on inverse distance weighting to perform spatial interpolation calculations on the discretely distributed sensor measurement points; the algorithm constructs a variogram model to quantify spatial autocorrelation and solves the Kriging equations to obtain the optimal interpolation weights.

Images