Chicken house health monitoring system and method based on Internet of Things

Abstract

The invention relates to the technical field of the Internet of Things, in particular to a henhouse health monitoring system and method based on the Internet of Things, and the system comprises an environment sensing module, a livestock and poultry information collection module, a communication module, a henhouse digital twin module, a learning early warning module, a data fusion module, a collaborative decision module, and an execution feedback module. The environment sensing module is used for multi-parameter environment sensing and edge preprocessing. By constructing a full-coverage sensing network and an intelligent analysis system, real-time synchronous acquisition and multi-modal data fusion of environment and biological signals are realized, spatial and temporal changes of the environment in a henhouse are dynamically deduced by means of a digital twinning technology, a more accurate early warning model is cooperatively trained in combination with a federal learning mechanism on the premise of not sharing local data, and the early warning efficiency is improved. Finally, multi-subsystem cooperative regulation and control are achieved through a game theory method, the hysteresis quality, isolation and subjectivity of a traditional method are effectively overcome, and the real-time performance of monitoring, the accuracy of early warning and the overall collaboration of control are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, specifically to an IoT-based chicken coop health monitoring system and method. Background Technology

[0002] Chicken houses are specialized buildings for the centralized raising of poultry, providing the necessary environmental space and protective conditions for the survival, growth, and egg production of chickens. The quality of their internal environment directly affects the health and production performance of the poultry. Monitoring chicken houses is crucial because poultry are extremely sensitive to environmental changes. Abnormalities in parameters such as temperature, humidity, and the concentration of harmful gases can quickly induce fever stress, respiratory diseases, or outbreaks. Continuous monitoring allows for the timely detection of abnormalities, early warning of disease risks, and optimization of feeding strategies, thereby ensuring animal welfare, maintaining production stability, and reducing economic losses.

[0003] Traditional chicken house monitoring methods mainly rely on regular manual inspections and simple wired sensor measurements. Manual inspections have disadvantages such as strong subjectivity, long intervals, inability to achieve all-weather monitoring, and the risk of introducing pathogens. Wired sensor systems, on the other hand, are complicated to install, lack flexibility, and have high maintenance costs. Each monitoring indicator usually operates independently, forming data silos, lacking the ability to perform multi-parameter linkage analysis, and the early warning mechanism is often based on fixed threshold alarms, which cannot achieve early and accurate risk assessment.

[0004] In summary, it is necessary to propose an IoT-based chicken coop health monitoring system and method to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an Internet of Things-based chicken coop health monitoring system and method to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The IoT-based chicken coop health monitoring system includes an environmental sensing module, a livestock and poultry information collection module, a communication module, a chicken coop digital twin module, a learning and early warning module, a data fusion module, a collaborative decision-making module, and an execution feedback module.

[0008] The environmental perception module is used for multi-parameter environmental sensing and edge preprocessing; the livestock and poultry information acquisition module is used for group behavior audio acquisition and distributed feed and water intake monitoring; the communication module is used for heterogeneous network adaptation and spatiotemporal alignment caching; the chicken coop digital twin module is used for physical field virtual mapping and behavioral environment coupling analysis; the learning and early warning module is used for distributed model training and multimodal early warning inference; the data fusion module is used for multimodal data fusion and dynamic baseline management; the collaborative decision-making module is used for multi-objective optimization and Nash equilibrium strategy solving; and the execution feedback module is used for instruction collaborative distribution and closed-loop feedback learning.

[0009] Preferably, the environment perception module further includes a multi-parameter environment sensing unit and an edge preprocessing unit;

[0010] The multi-parameter environmental sensing unit performs high-frequency synchronized environmental data acquisition operations by deploying a digital sensor array consisting of temperature and humidity sensors, ammonia sensors, carbon dioxide sensors, and illuminance sensors, in order to achieve standardized and unified collection of multi-dimensional environmental data.

[0011] The edge preprocessing unit performs filtering and instantaneous outlier removal operations on the original environmental data through embedded Kalman filtering, which is used to achieve preliminary cleaning and quality improvement of the uploaded data.

[0012] Preferably, the livestock and poultry information collection module further includes a group behavior audio collection unit and a distributed feed and water intake monitoring unit;

[0013] The group behavior audio acquisition unit uses a high-fidelity noise-reducing microphone array to continuously record group sounds inside the chicken house and separate environmental noise, in order to extract the audio features of the overall calls, coughs, and sneezes of the chickens.

[0014] The distributed feeding and drinking volume monitoring unit uses flow meters and weighing sensors installed at the end of each feeding line and drinking line to measure and report the feeding and drinking volume in real time, thereby enabling accurate tracking of group consumption behavior.

[0015] Preferably, the communication module further includes a heterogeneous network adaptation unit and a spatiotemporal alignment and caching unit;

[0016] The heterogeneous network adaptation unit receives, parses, and encapsulates data from LoRaWAN, NB-IoT, and Wi-Fi protocols through a multi-protocol gateway, enabling unified access and uplink transmission of data from all sensing layer devices.

[0017] The spatiotemporal alignment and caching unit uses a built-in RTC clock and identification system to stamp the received multi-source asynchronous data with a unified timestamp and spatial tag and perform temporary storage operations, which is used to provide data packets with a consistent time base for subsequent fusion.

[0018] Preferably, the chicken coop digital twin module further includes a physical field virtual mapping unit and a behavioral environment coupling analysis unit;

[0019] The physical field virtual mapping unit performs three-dimensional visualization simulation and deduction of the temperature field, airflow field, and gas concentration field inside the chicken house through a physical model based on computational fluid dynamics and real-time environmental data, in order to realize the global dynamic visualization of the environmental state.

[0020] The behavioral-environment coupling analysis unit performs spatial correlation analysis between flock behavior and environmental parameters by overlaying flock behavior data onto a digital twin model, which is used to accurately locate environmental anomalies and their impact on the flock.

[0021] Preferably, the learning and early warning module further includes a distributed model training unit and a multimodal early warning inference unit;

[0022] The distributed model training unit trains and updates the initial early warning model on the local server using data from the chicken coop through a federated learning framework. Only the model parameters are encrypted and uploaded to the cloud platform for aggregation operations, which is used to jointly optimize the early warning model using data from multiple parties while protecting the data privacy of each farm.

[0023] The multimodal early warning inference unit integrates the latest aggregated model from the cloud platform and inputs the real-time fused multidimensional data into the model to perform early probability calculation and graded early warning operations for disease or stress risks, thereby realizing intelligent prediction based on multidimensional data fusion.

[0024] Preferably, the data fusion module further includes a multimodal data fusion unit and a dynamic baseline management unit;

[0025] The multimodal data fusion unit performs confidence calculations and fusion decision operations on environmental and behavioral data through adaptive weighted fusion, which is used to generate a unified and reliable comprehensive evaluation index for the current health status of the chicken house;

[0026] The dynamic baseline management unit performs dynamic calculations of key indicators such as feed intake and water consumption within the normal range by combining time series analysis with information on the age and breed of the flock. This provides a personalized judgment benchmark for the early warning model that changes with the growth stage.

[0027] Preferably, the collaborative decision-making module further includes a multi-objective optimization unit and a Nash equilibrium strategy solving unit;

[0028] The multi-objective optimization unit constructs multiple benefit functions, including environmental comfort indicators, energy consumption indicators, and biosafety indicators, to perform a comprehensive benefit scoring operation on different control strategies, which is used to quantify the advantages and disadvantages of different decision-making schemes.

[0029] The Nash equilibrium strategy solving unit uses game theory to treat the environmental control subsystem, feeding management subsystem, and alarm subsystem as game participants to solve for the cooperative strategy operation that can achieve the system optimum, in order to resolve the conflict of control objectives among the subsystems.

[0030] Preferably, the execution feedback module further includes an instruction coordination distribution unit and a closed-loop feedback learning unit;

[0031] The instruction coordination and distribution unit receives the coordination strategy, parses it into a set of control instructions for specific devices, performs timing scheduling and conflict verification, and then distributes the instructions to ensure that the control commands are executed in an orderly and consistent manner.

[0032] The closed-loop feedback learning unit monitors changes in environmental and flock behavior data after the strategy is implemented, compares these changes with the expected effects of the strategy, and feeds back deviation data to the data fusion module and the learning and early warning module for continuous self-optimization of the early warning model and control strategy.

[0033] The monitoring method of a chicken coop health monitoring system based on the Internet of Things includes the following steps:

[0034] S1. Environmental Data Acquisition and Preprocessing: By deploying a digital sensor array consisting of temperature and humidity sensors, ammonia sensors, carbon dioxide sensors, and illuminance sensors, multi-dimensional environmental parameters in the chicken house are collected synchronously; embedded Kalman filters are used to perform real-time filtering and outlier removal on the collected raw environmental data to complete the edge preprocessing of the environmental data;

[0035] S2. Livestock and poultry information collection: The high-fidelity noise-reducing microphone array continuously collects the group sound signals in the chicken house, separates the environmental noise, and extracts the audio characteristics of the chickens' calls, coughs, and sneezes; the flow meters and weighing sensors installed at the end of the feeding line and water line monitor and report the feed intake and water intake data in real time.

[0036] S3. Multi-source data communication and processing: Receives, parses, and encapsulates data from multiple communication protocols, including LoRaWAN, NB-IoT, and Wi-Fi, through a multi-protocol gateway to achieve unified access for sensing layer devices; adds unified timestamps and spatial tags to multi-source asynchronous data through a built-in RTC clock and identification system, and performs temporary storage to form time-space aligned data packets;

[0037] S4. Digital Twin Modeling and Analysis: Based on computational fluid dynamics models and real-time environmental data, construct three-dimensional virtual models of temperature field, airflow field, and gas concentration field inside the chicken house to achieve dynamic visualization of environmental conditions; overlay chicken behavior data onto the digital twin model to conduct spatial correlation analysis between behavior and environmental parameters, accurately locate abnormal environmental areas and their impact on the chicken flock;

[0038] S5. Intelligent Early Warning Model Training and Inference: Adopting a federated learning framework, the early warning model is trained and updated on the local server using data from the chicken house. Only encrypted model parameters are uploaded to the cloud platform for aggregation and optimization, enabling collaborative modeling across multiple farms while ensuring data privacy. The latest aggregated model is integrated, and real-time fused multi-dimensional data is input into the model to perform early probability calculations and graded early warnings of disease and stress risks.

[0039] S6. Multimodal data fusion and management: Adaptive weighted fusion is used to calculate the confidence level and make fusion decisions on environmental and behavioral data to generate a unified assessment index of chicken house health status; combined with the age and breed information of the flock, time series analysis is used to dynamically calculate the normal range of key indicators such as feed intake and water consumption, providing personalized judgment benchmarks for early warning;

[0040] S7. Multi-objective collaborative decision-making: Construct a benefit function that includes multiple indicators such as environmental comfort, energy consumption, and biosafety, and score the comprehensive benefits of different control strategies; use game theory to treat environmental control, feeding management, and alarm subsystems as game participants, and solve for the Nash equilibrium collaborative strategy that achieves the system's optimality.

[0041] S8. Strategy Execution and Optimization: Parse the collaborative strategy into a set of specific device control instructions, distribute them to the execution devices after timing scheduling and conflict verification; monitor changes in the environment and flock behavior after strategy execution, compare with the expected results, and optimize the monitoring process.

[0042] Compared with existing technologies, the beneficial effects of this invention are as follows: By constructing a fully covered sensor network and intelligent analysis system, this invention achieves real-time synchronous acquisition of environmental and biological signals and fusion of multimodal data. It uses digital twin technology to dynamically predict the spatiotemporal changes of the chicken house environment, and combines a federated learning mechanism to collaboratively train a more accurate early warning model without sharing local data. Finally, it uses game theory to achieve collaborative regulation of multiple subsystems, effectively overcoming the lag, isolation, and subjectivity of traditional methods, significantly improving the real-time performance of monitoring, the accuracy of early warning, and the overall synergy of control, and solving the problems existing in traditional monitoring processes. Attached Figure Description

[0043] Figure 1 A flowchart of the IoT-based chicken coop health monitoring method of the present invention is shown. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Example 1: This invention proposes an Internet of Things-based chicken coop health monitoring system, including an environmental sensing module, a livestock and poultry information collection module, a communication module, a chicken coop digital twin module, a learning and early warning module, a data fusion module, a collaborative decision-making module, and an execution feedback module;

[0046] It should be noted that the environmental perception module is used for multi-parameter environmental sensing and edge preprocessing; the livestock and poultry information acquisition module is used for group behavior audio acquisition and distributed feeding and drinking volume monitoring; the communication module is used for heterogeneous network adaptation and spatiotemporal alignment caching; the chicken coop digital twin module is used for physical field virtual mapping and behavioral environment coupling analysis; the learning and early warning module is used for distributed model training and multimodal early warning inference; the data fusion module is used for multimodal data fusion and dynamic baseline management; the collaborative decision-making module is used for multi-objective optimization and Nash equilibrium strategy solving; and the execution feedback module is used for instruction collaborative distribution and closed-loop feedback learning.

[0047] In this embodiment, it should also be noted that the environment perception module further includes a multi-parameter environment sensing unit and an edge preprocessing unit.

[0048] The multi-parameter environmental sensing unit performs high-frequency synchronized environmental data acquisition operations by deploying a digital sensor array consisting of temperature and humidity sensors, ammonia sensors, carbon dioxide sensors, and illuminance sensors, in order to achieve the standardization and unified collection of multi-dimensional environmental data.

[0049] The edge preprocessing unit performs filtering and instantaneous outlier removal on the raw environmental data through embedded Kalman filtering, which is used to achieve preliminary cleaning and quality improvement of the uploaded data.

[0050] In this embodiment, it should also be noted that the livestock and poultry information collection module further includes a group behavior audio collection unit and a distributed feed and water intake monitoring unit.

[0051] The group behavior audio acquisition unit uses a high-fidelity noise-reducing microphone array to continuously record group sounds in the chicken house and separate environmental noise, in order to extract the audio features of the overall calls, coughs, and sneezes of the chickens.

[0052] The distributed feeding and drinking volume monitoring unit uses flow meters and weighing sensors installed at the end of each feeding and drinking line to measure and report the feeding and drinking volume in real time, enabling accurate tracking of group consumption behavior.

[0053] In this embodiment, it should also be noted that the communication module further includes a heterogeneous network adaptation unit and a spatiotemporal alignment and caching unit;

[0054] The heterogeneous network adaptation unit receives, parses, and encapsulates data from LoRaWAN, NB-IoT, and Wi-Fi protocols via a multi-protocol gateway, enabling unified access and uplink transmission of data from all sensing layer devices.

[0055] The spatiotemporal alignment and caching unit uses a built-in RTC clock and identification system to stamp the received multi-source asynchronous data with a unified timestamp and spatial tag and perform temporary storage operations to provide data packets with a consistent time base for subsequent fusion.

[0056] In this embodiment, it should also be noted that the chicken coop digital twin module further includes a physical field virtual mapping unit and a behavioral environment coupling analysis unit;

[0057] The physical field virtual mapping unit performs three-dimensional visualization simulation and deduction of the temperature field, airflow field, and gas concentration field inside the chicken house through a physical model based on computational fluid dynamics and real-time environmental data, in order to realize the global dynamic visualization of the environmental state.

[0058] The behavioral-environment coupling analysis unit performs spatial correlation analysis between flock behavior and environmental parameters by overlaying flock behavior data onto a digital twin model, which is used to accurately locate environmental anomalies and their impact on the flock.

[0059] In this embodiment, it should also be noted that the learning and early warning module further includes a distributed model training unit and a multimodal early warning inference unit;

[0060] The distributed model training unit uses the local chicken coop data to train and update the initial early warning model on the local server through the federated learning framework. Only the model parameters are encrypted and uploaded to the cloud platform for aggregation operations, which is used to jointly optimize the early warning model with data from multiple parties while protecting the data privacy of each farm.

[0061] The multimodal early warning inference unit integrates the latest aggregated models from the cloud platform and inputs the real-time fused multidimensional data into the model to perform early probability calculations and graded early warning operations for disease or stress risks, thereby realizing intelligent prediction based on multidimensional data fusion.

[0062] In this embodiment, it should also be noted that the data fusion module further includes a multimodal data fusion unit and a dynamic baseline management unit;

[0063] The multimodal data fusion unit performs confidence calculations and fusion decision operations on environmental and behavioral data through adaptive weighted fusion, which is used to generate a unified and reliable comprehensive assessment index of the current health status of the chicken house;

[0064] The dynamic baseline management unit uses time series analysis combined with flock age and breed information to dynamically calculate the normal range of key indicators such as feed intake and water consumption, providing personalized judgment benchmarks for early warning models that change with growth stages.

[0065] In this embodiment, it should also be noted that the collaborative decision-making module further includes a multi-objective optimization unit and a Nash equilibrium strategy solving unit;

[0066] The multi-objective optimization unit constructs multiple benefit functions, including environmental comfort indicators, energy consumption indicators, and biosafety indicators, to perform a comprehensive benefit scoring operation on different regulation strategies, which is used to quantify the advantages and disadvantages of different decision-making schemes.

[0067] The Nash equilibrium strategy solving unit uses game theory to treat the environmental control subsystem, feeding management subsystem, and alarm subsystem as game participants to solve the cooperative strategy operation that can achieve the system optimum, and is used to resolve the conflict of control objectives among the subsystems.

[0068] In this embodiment, it should also be noted that the execution feedback module further includes an instruction coordination distribution unit and a closed-loop feedback learning unit;

[0069] The instruction coordination and distribution unit receives the coordination strategy, parses it into a set of control instructions for specific devices, performs timing scheduling and conflict verification, and then distributes the instructions to ensure that the control commands are executed in an orderly and consistent manner.

[0070] The closed-loop feedback learning unit monitors changes in environmental and flock behavior data after the strategy is implemented, compares these changes with the expected effects of the strategy, and feeds back deviation data to the data fusion module and the learning and early warning module for continuous self-optimization of the early warning model and control strategy.

[0071] Example 2: In practical applications, please refer to... Figure 1 The present invention relates to a detection method for a chicken coop health monitoring system based on the Internet of Things, specifically including the following steps:

[0072] S1. Environmental Data Acquisition and Preprocessing:

[0073] S1.1: Deploy a digital sensor array to synchronously collect raw data on temperature, humidity, ammonia concentration, carbon dioxide concentration, and light intensity at different locations within the chicken house;

[0074] S1.2: Perform preliminary verification on the collected raw data at the edge computing node to eliminate obvious null values ​​or jump values ​​caused by sensor communication interruption;

[0075] S1.3: The Kalman filter algorithm is used to perform real-time smoothing on the verified time series data to optimally estimate the real environmental state at the current moment;

[0076] S1.4: Based on statistical principles or sliding window detection, outlier data points that still exist after filtering are removed;

[0077] S1.5: Package the pre-treated clean environment data in a unified format and prepare for uploading;

[0078] Synchronous acquisition of sensor data: A sensor network is formed by temperature and humidity sensors, ammonia sensors, carbon dioxide sensors and light intensity sensors deployed in different physical locations in the chicken house. The raw measurement values ​​of environmental parameters are collected synchronously at a fixed sampling frequency to form a multi-dimensional time series data vector.

[0079] Preliminary data validity check: Perform a validity check on the original data vector Z_raw(t), and remove null values ​​(NULL) caused by communication interruption and outliers that are obviously outside the physical range;

[0080] The judgment condition is: Z_min≤Z_raw(t)≤Z_max, where Z_min and Z_max are the upper and lower limits of the range of each sensor;

[0081] Specifically, dynamic state estimation based on Kalman filtering: the discrete Kalman filter algorithm is used to perform optimal estimation of the effective data;

[0082] The system state equation is:

[0083] State prediction equation:

[0084] x̂ k ⁻=Ax̂ k ₋1+Bu k ₋1;

[0085] P k ⁻=AP k ₋1A^T+Q;

[0086] State update equation:

[0087] K k =P k ⁻H^T(HP k ⁻H^T+R)⁻¹;

[0088] x̂ k =x ̂k ⁻+K k (zk -Hx̂ k ⁻);

[0089] P k =(IK k H)P k ⁻;

[0090] In the formula, x̂ k ⁻ represents the prior estimate of the state at time k, i.e., the prediction based on time k-1; x̂ k P represents the posterior estimate of the state at time k, i.e., the optimal estimate after filtering; k ⁻ represents the prior estimation error covariance matrix; P k A represents the posterior estimation error covariance matrix; B represents the state transition matrix, describing how the system state evolves over time; C represents the control input matrix; D represents the state transition matrix. k Represents the control vector; Q represents the process noise covariance matrix; R represents the measurement noise covariance matrix; H represents the observation matrix; K k Represents the Kalman gain matrix; z k This represents the actual measurement value at time k;

[0091] Outlier detection and removal: The 3σ criterion is used to filter the data x̂ k Perform outlier detection: If |x̂ k If -μ|>3σ, it is considered an outlier and removed.

[0092] Where μ and σ are the mean and standard deviation of the historical data calculated using a sliding window, respectively;

[0093] Data standardization and encapsulation: Encapsulating the processed environmental data into data frames according to a unified format:

[0094] Frame_env={timeStamp,location,T_proceSSed,H_proceSSed,NH3_proceSSed,CO2_proceSSed,L_proceSSed};

[0095] S2. Collection of livestock and poultry information:

[0096] S2.1: Acquire mixed audio streams within the chicken coop using a distributed microphone array;

[0097] S2.2: Apply adaptive filters, including the LMS algorithm, to separate steady-state environmental noise generated by fans, feeders, etc.

[0098] S2.3: Extract the time domain of the chicken calls from the denoised audio, including amplitude, zero-crossing rate and frequency domain, including Mel-frequency cepstral coefficients (MFCC) features;

[0099] S2.4: Monitor the water flow rate changes of each water supply line in real time using a flow meter, and calculate the cumulative water consumption;

[0100] S2.5: Monitor the weight change of the feed box using a weighing sensor to calculate the feed intake within a specific time period;

[0101] S2.6: Report audio characteristics, feed intake, and water intake data after adding time stamps;

[0102] S3. Multi-source data communication and processing

[0103] S3.1: The gateway listens for and receives data packets from terminal devices using different protocols;

[0104] S3.2: Parse data packets according to different protocol rules and extract valid sensor data payloads;

[0105] S3.3: Convert all parsed data into a unified JSON data format for the system.

[0106] S3.4: Attach a precise UTC timestamp and device physical location tag to each data item received by the gateway;

[0107] S3.5: Store the spatiotemporal tag data in a temporary buffer, waiting for subsequent modules to extract and process it according to a unified clock cycle;

[0108] S4. Digital Twin Modeling and Analysis

[0109] S4.1: Construct a three-dimensional mesh model of the chicken coop structure as the solution domain;

[0110] S4.2: Substitute real-time environmental data as boundary conditions and source terms into the CFD governing equations for solution;

[0111] S4.3: Visualize the spatial distribution cloud map of temperature, airflow velocity, and gas concentration;

[0112] S4.4: A heatmap mapping the distribution of chicken flocks in a twin model (derived from audio source localization or image analysis);

[0113] S4.5: Calculate the spatial overlap between chicken flock gathering areas and areas of environmental abnormality, including low temperature and high ammonia levels;

[0114] S4.6: Output the precise coordinates of environmental anomalies and a quantitative assessment report of their impact on the chicken flock;

[0115] S5. Intelligent Early Warning Model Training and Inference (Detailed algorithm description follows)

[0116] S5.1: Each local farm downloads the initial global early warning model from the cloud platform;

[0117] S5.2: The local server uses historical data from the current session to train the model and update the model parameters;

[0118] S5.3: Upload the updated local model parameters to the central server of the cloud platform after encryption;

[0119] S5.4: The cloud platform aggregates the encrypted parameters of all farms and updates the global model through a federated averaging algorithm;

[0120] S5.5: Distribute the latest global model to each farm for real-time inference;

[0121] S5.6: The local system will input multi-dimensional data fused in real time into the model to obtain the probability of health risks;

[0122] Local model initialization: Download the latest global model parameters ω_global from the cloud platform and initialize the local model: ω_local ← ω_global;

[0123] Local model training: The model is trained using the local dataset D_local, and the loss function is minimized: ω_local^=argminΣL(f(x_i;ω),y_i), where L is the loss function, f is the model, x_i is the input feature, and y_i is the label;

[0124] Encrypted upload of model parameters: The updated model parameters ω_local^ are encrypted using a homomorphic encryption algorithm and then uploaded to the cloud platform: Enc(ω_local^)→Cloud;

[0125] Federated average aggregation: The cloud platform aggregates encrypted model parameters from K farms and updates the global model through a weighted average.

[0126] Federal Average Formula:

[0127] ω_global^{new}=Σ_{k=1}^K(n_k / n_total)·ω_local^{(k)};

[0128] In the formula, ω_global^{new} represents the updated global model parameters; ω_local^{(k)} represents the local model parameters of the k-th farm; n_k represents the number of samples in the local dataset of the k-th farm; n_total represents the total number of data samples of all participating farms, n_total=Σ_{k=1}^Kn_k; (n_k / n_total) represents the weight coefficient, and the larger the amount of data of the farm, the greater its contribution to the global model;

[0129] Global model distribution: Distribute the updated global model parameters ω_global^{new} to all farms participating in federated learning;

[0130] Local Inference and Early Warning: Use the latest global model to infer the real-time data X_real-time: y_prob=Softmax(f(X_real-time;ω_global^{new})), output the probability distribution of various health risks, and realize early warning;

[0131] S6. Multimodal Data Fusion and Management:

[0132] S6.1: Normalize environmental data, audio data, consumption data, etc., to eliminate the influence of dimensions;

[0133] S6.2: An adaptive weighted fusion algorithm is adopted to dynamically allocate weights based on the reliability of each data source;

[0134] S6.3: Calculate the comprehensive health status index of the chicken house after fusion;

[0135] S6.4: Based on the age and breed of the flock, establish dynamic baseline curves for each health indicator based on historical data;

[0136] S6.5: Compare the deviation of the current indicator from the dynamic baseline in real time and calculate statistics such as Z-Score;

[0137] S6.6: Input deviation information as a feature vector into the early warning model;

[0138] S7. Multi-objective collaborative decision-making:

[0139] S7.1: Define decision-making objectives: environmental comfort, energy costs, biosafety risks, etc.

[0140] S7.2: Quantify different control strategies, including turning on fan No. 1 / turning on the wet curtain / reducing the impact of feeding on each target;

[0141] S7.3: Construct a multi-objective benefit function and evaluate the overall score of each strategy;

[0142] S7.4: Model each subsystem as a game participant, and its payoff function is the overall benefit;

[0143] S7.5: Solve for the Nash equilibrium point. The strategy combination corresponding to this point is the optimal cooperative strategy in which all subsystems have no incentive to change unilaterally.

[0144] S7.6: Outputs the final set of cooperative control instructions;

[0145] S8. Strategy Execution and Optimization:

[0146] S8.1: Parse collaborative decision-making instructions and generate control commands for specific devices, including fans, motors, and alarms;

[0147] S8.2: Check for timing conflicts between commands, including simultaneous execution of manure cleaning and increased ventilation, and adjust the execution order;

[0148] S8.3: Send the verified command to the executor;

[0149] S8.4: Continuously monitor the response data of the environment and flock behavior after the strategy is implemented;

[0150] S8.5: Compare the actual response with the expected response at the time of decision-making, and calculate the performance deviation;

[0151] S8.6: Feedback the deviation data to the data fusion and early warning module for online correction of the model and strategy.

[0152] Through the above steps, this invention constructs a fully covered sensor network and intelligent analysis system to achieve real-time synchronous acquisition of environmental and biological signals and fusion of multimodal data. It uses digital twin technology to dynamically predict the spatiotemporal changes of the chicken house environment, and combines a federated learning mechanism to collaboratively train a more accurate early warning model without sharing local data. Finally, it uses game theory to achieve collaborative regulation of multiple subsystems, effectively overcoming the lag, isolation, and subjectivity of traditional methods, significantly improving the real-time performance of monitoring, the accuracy of early warning, and the overall synergy of control, and solving the problems existing in the traditional monitoring process.

[0153] 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 likenesses.

Claims

1. A chicken coop health monitoring system based on the Internet of Things, characterized in that, It includes an environmental perception module, a livestock and poultry information collection module, a communication module, a chicken coop digital twin module, a learning and early warning module, a data fusion module, a collaborative decision-making module, and an execution feedback module; The environmental perception module is used for multi-parameter environmental sensing and edge preprocessing; the livestock and poultry information acquisition module is used for group behavior audio acquisition and distributed feed and water intake monitoring; the communication module is used for heterogeneous network adaptation and spatiotemporal alignment caching; the chicken coop digital twin module is used for physical field virtual mapping and behavioral environment coupling analysis; the learning and early warning module is used for distributed model training and multimodal early warning inference; the data fusion module is used for multimodal data fusion and dynamic baseline management; the collaborative decision-making module is used for multi-objective optimization and Nash equilibrium strategy solving; and the execution feedback module is used for instruction collaborative distribution and closed-loop feedback learning.

2. The IoT-based chicken coop health monitoring system according to claim 1, characterized in that: The environment perception module also includes a multi-parameter environment sensing unit and an edge preprocessing unit; The multi-parameter environmental sensing unit performs high-frequency synchronized environmental data acquisition operations by deploying a digital sensor array consisting of temperature and humidity sensors, ammonia sensors, carbon dioxide sensors, and illuminance sensors, in order to achieve standardized and unified collection of multi-dimensional environmental data. The edge preprocessing unit performs filtering and instantaneous outlier removal operations on the original environmental data through embedded Kalman filtering, which is used to achieve preliminary cleaning and quality improvement of the uploaded data.

3. The IoT-based chicken coop health monitoring system according to claim 2, characterized in that: The livestock and poultry information collection module also includes a group behavior audio collection unit and a distributed feed and water intake monitoring unit; The group behavior audio acquisition unit uses a high-fidelity noise-reducing microphone array to continuously record group sounds inside the chicken house and separate environmental noise, in order to extract the audio features of the overall calls, coughs, and sneezes of the chickens. The distributed feeding and drinking volume monitoring unit uses flow meters and weighing sensors installed at the end of each feeding line and drinking line to measure and report the feeding and drinking volume in real time, thereby enabling accurate tracking of group consumption behavior.

4. The IoT-based chicken coop health monitoring system according to claim 3, characterized in that: The communication module also includes a heterogeneous network adaptation unit and a spatiotemporal alignment and caching unit; The heterogeneous network adaptation unit receives, parses, and encapsulates data from LoRaWAN, NB-IoT, and Wi-Fi protocols through a multi-protocol gateway, enabling unified access and uplink transmission of data from all sensing layer devices. The spatiotemporal alignment and caching unit uses a built-in RTC clock and identification system to stamp the received multi-source asynchronous data with a unified timestamp and spatial tag and perform temporary storage operations, which is used to provide data packets with a consistent time base for subsequent fusion.

5. The IoT-based chicken coop health monitoring system according to claim 4, characterized in that: The chicken coop digital twin module also includes a physical field virtual mapping unit and a behavior-environment coupling analysis unit; The physical field virtual mapping unit performs three-dimensional visualization simulation and deduction of the temperature field, airflow field, and gas concentration field inside the chicken house through a physical model based on computational fluid dynamics and real-time environmental data, in order to realize the global dynamic visualization of the environmental state. The behavioral-environment coupling analysis unit performs spatial correlation analysis between flock behavior and environmental parameters by overlaying flock behavior data onto a digital twin model, which is used to accurately locate environmental anomalies and their impact on the flock.

6. The IoT-based chicken coop health monitoring system according to claim 5, characterized in that: The learning and early warning module also includes a distributed model training unit and a multimodal early warning inference unit; The distributed model training unit trains and updates the initial early warning model on the local server using data from the chicken coop through a federated learning framework. Only the model parameters are encrypted and uploaded to the cloud platform for aggregation operations, which is used to jointly optimize the early warning model using data from multiple parties while protecting the data privacy of each farm. The multimodal early warning inference unit integrates the latest aggregated model from the cloud platform and inputs the real-time fused multidimensional data into the model to perform early probability calculation and graded early warning operations for disease or stress risks, thereby realizing intelligent prediction based on multidimensional data fusion.

7. The IoT-based chicken coop health monitoring system according to claim 6, characterized in that: The data fusion module also includes a multimodal data fusion unit and a dynamic baseline management unit; The multimodal data fusion unit performs confidence calculations and fusion decision operations on environmental and behavioral data through adaptive weighted fusion, which is used to generate a unified and reliable comprehensive evaluation index for the current health status of the chicken house; The dynamic baseline management unit performs dynamic calculations of key indicators such as feed intake and water consumption within the normal range by combining time series analysis with information on the age and breed of the flock. This provides a personalized judgment benchmark for the early warning model that changes with the growth stage.

8. The IoT-based chicken coop health monitoring system according to claim 7, characterized in that: The collaborative decision-making module also includes a multi-objective optimization unit and a Nash equilibrium strategy solving unit; The multi-objective optimization unit constructs multiple benefit functions, including environmental comfort indicators, energy consumption indicators, and biosafety indicators, to perform a comprehensive benefit scoring operation on different control strategies, which is used to quantify the advantages and disadvantages of different decision-making schemes. The Nash equilibrium strategy solving unit uses game theory to treat the environmental control subsystem, feeding management subsystem, and alarm subsystem as game participants to solve for the cooperative strategy operation that can achieve the system optimum, in order to resolve the conflict of control objectives among the subsystems.

9. The IoT-based chicken coop health monitoring system according to claim 8, characterized in that: The execution feedback module also includes an instruction coordination distribution unit and a closed-loop feedback learning unit; The instruction coordination and distribution unit receives the coordination strategy, parses it into a set of control instructions for specific devices, performs timing scheduling and conflict verification, and then distributes the instructions to ensure that the control commands are executed in an orderly and consistent manner. The closed-loop feedback learning unit monitors changes in environmental and flock behavior data after the strategy is implemented, compares these changes with the expected effects of the strategy, and feeds back deviation data to the data fusion module and the learning and early warning module for continuous self-optimization of the early warning model and control strategy.

10. The monitoring method of the IoT-based chicken coop health monitoring system according to any one of claims 1-9, characterized in that, Includes the following steps: S1. Environmental Data Acquisition and Preprocessing: By deploying a digital sensor array consisting of temperature and humidity sensors, ammonia sensors, carbon dioxide sensors, and illuminance sensors, multi-dimensional environmental parameters in the chicken house are collected synchronously; embedded Kalman filters are used to perform real-time filtering and outlier removal on the collected raw environmental data to complete the edge preprocessing of the environmental data; S2. Livestock and poultry information collection: The high-fidelity noise-reducing microphone array continuously collects the group sound signals in the chicken house, separates the environmental noise, and extracts the audio characteristics of the chickens' calls, coughs, and sneezes; the flow meters and weighing sensors installed at the end of the feeding line and water line monitor and report the feed intake and water intake data in real time. S3. Multi-source data communication and processing: Receives, parses, and encapsulates data from multiple communication protocols, including LoRaWAN, NB-IoT, and Wi-Fi, through a multi-protocol gateway to achieve unified access for sensing layer devices; adds unified timestamps and spatial tags to multi-source asynchronous data through a built-in RTC clock and identification system, and performs temporary storage to form time-space aligned data packets; S4. Digital Twin Modeling and Analysis: Based on computational fluid dynamics models and real-time environmental data, construct three-dimensional virtual models of temperature field, airflow field, and gas concentration field inside the chicken house to achieve dynamic visualization of environmental conditions; overlay chicken behavior data onto the digital twin model to conduct spatial correlation analysis between behavior and environmental parameters, accurately locate abnormal environmental areas and their impact on the chicken flock; S5. Intelligent Early Warning Model Training and Inference: Using a federated learning framework, the early warning model is trained and updated on the local server using data from the chicken coop. Only the encrypted model parameters are uploaded to the cloud platform for aggregation and optimization, enabling collaborative modeling across multiple farms while ensuring data privacy. Integrating the latest aggregation model, real-time fused multi-dimensional data is input into the model to perform early probability calculation and graded warning of disease and stress risks; S6. Multimodal data fusion and management: Adaptive weighted fusion is used to calculate the confidence level and make fusion decisions on environmental and behavioral data to generate a unified assessment index of chicken house health status; combined with the age and breed information of the flock, time series analysis is used to dynamically calculate the normal range of key indicators such as feed intake and water consumption, providing personalized judgment benchmarks for early warning; S7. Multi-objective collaborative decision-making: Construct a benefit function that includes multiple indicators such as environmental comfort, energy consumption, and biosafety, and score the comprehensive benefits of different control strategies; use game theory to treat environmental control, feeding management, and alarm subsystems as game participants, and solve for the Nash equilibrium collaborative strategy that achieves the system's optimality. S8. Strategy Execution and Optimization: Parse the collaborative strategy into a set of specific device control instructions, distribute them to the execution devices after timing scheduling and conflict verification; monitor changes in the environment and flock behavior after strategy execution, compare with the expected results, and optimize the monitoring process.

Images