Epidemic disease prevention and control method and system based on continuous monitoring of feeding environment

Abstract

The invention discloses an epidemic disease prevention and control method and system based on continuous monitoring of a feeding environment, and relates to the technical field of epidemic disease prevention and control, and the method comprises the steps: collecting short-term dynamic environment data and long-term static breeding data of a pig farm; extracting short-term environment fluctuation features and long-term epidemic disease associated features; predicting a short-term epidemic disease risk score based on the short-term environment fluctuation characteristics, and predicting a long-term epidemic disease risk score based on the long-term epidemic disease association characteristics; and fusing the short-term epidemic disease risk score and the long-term epidemic disease risk score to obtain a comprehensive risk score, and carrying out comprehensive assessment on the epidemic disease risk. According to the method, short-term dynamic environment data and long-term static breeding data of a farm are collected, short-term environment fluctuation and long-term epidemic disease associated characteristics are extracted, short-term epidemic disease risk scores and long-term epidemic disease risk scores are predicted respectively, and a comprehensive risk score is obtained through adaptive weight fusion. The key factors of epidemic diseases can be comprehensively captured by fusing the two types of data, and the influence of short-term environment mutation and long-term breeding rules can be balanced by the adaptive weight.

Description

Technical Field

[0001] This invention relates to the field of disease prevention and control technology, and more specifically to a method and system for disease prevention and control based on continuous monitoring of the breeding environment. Background Technology

[0002] Currently, in large-scale pig farming, disease prevention and control are crucial for ensuring farming efficiency and animal health. Traditional disease prevention and control methods largely rely on manual inspections and monitoring of single environmental parameters, which suffer from problems such as delayed response and incomplete assessment. Although existing technologies have introduced environmental monitoring and data analysis, they mostly focus on the immediate correlation between real-time environmental data and diseases, neglecting the impact of long-term farming patterns (such as seasonal changes and breed characteristics) on disease occurrence. This makes it difficult to comprehensively assess disease risks, resulting in insufficiently targeted prevention and control measures.

[0003] Therefore, how to propose a disease prevention and control method and system based on continuous monitoring of the breeding environment, combining short-term environmental changes with long-term breeding patterns, to improve the accuracy and dynamic response capability of disease prevention and control, and effectively reduce the risk of disease occurrence and spread is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0004] In view of this, the present invention provides a method and system for disease prevention and control based on continuous monitoring of the breeding environment. By combining short-term environmental changes with long-term breeding patterns, it improves the accuracy and dynamic response capability of disease prevention and control, and effectively reduces the risk of disease occurrence and spread. To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A method for disease prevention and control based on continuous monitoring of the breeding environment, comprising:

[0006] Collect short-term dynamic environmental data and long-term static breeding data of pig farms;

[0007] Based on short-term dynamic environmental data and long-term static aquaculture data, short-term environmental fluctuation characteristics and long-term disease association characteristics are extracted.

[0008] Predict short-term disease risk scores based on short-term environmental fluctuation characteristics, and predict long-term disease risk scores based on long-term disease association characteristics;

[0009] By integrating short-term and long-term disease risk scores, a comprehensive risk score is obtained, which is then used to comprehensively assess disease risk.

[0010] Optionally, the short-term dynamic environmental data includes real-time temperature, real-time humidity, real-time suspended particulate matter concentration, real-time noise decibels, and real-time activity levels of the pig herd; the long-term static breeding data includes historical environmental parameter records, pig breed characteristics, historical disease occurrence records, stocking density, and feed formulation.

[0011] Optionally, the steps of extracting short-term environmental fluctuation features and long-term disease association features include:

[0012] Through time-domain analysis, temperature change rate, humidity peak, suspended particulate matter concentration skewness, and noise decibel standard deviation are extracted from short-term dynamic environmental data.

[0013] Frequency domain analysis is used to extract the frequency and amplitude of environmental parameter fluctuations from short-term dynamic environmental data.

[0014] Through statistical analysis, seasonal environmental averages, environmental parameter ranges during peak disease periods, annual disease incidence rates, and the correlation between stocking density and disease were extracted from long-term static aquaculture data.

[0015] Optionally, the short-term disease risk score prediction based on short-term environmental fluctuation characteristics includes:

[0016] Among them, R s For short-term disease risk scoring, T v H is the rate of temperature change. p P represents the peak humidity level. s For the skewness of suspended particulate matter concentration, N d A represents the standard deviation of noise in decibels. a The amplitude of the environmental parameter fluctuation is represented by α and β, which are correction coefficients.

[0017] Optionally, the prediction of long-term disease risk scores based on long-term disease association characteristics includes:

[0018] Among them, R l For long-term disease risk scoring, S e To determine the degree of matching between seasonal environmental mean and environmental parameter ranges during peak disease seasons, I y D represents the average annual incidence rate of disease. c γ represents the correlation between stocking density and disease incidence, and δ represents correction coefficients.

[0019] Optionally, the fusion of short-term and long-term disease risk scores includes:

[0020] Through adaptive weight adjustment rules, when the long-term disease risk score R... l Below the short-term disease risk score R s When R is reduced s The weights; when R l Higher than R s When R is reduced l The weights;

[0021] The fusion expression is:

[0022]

[0023] Where R is the comprehensive risk score, t is time, and λ, μ, ν, and κ are adjustment factors.

[0024] Optionally, it also includes: classifying disease risk levels based on comprehensive risk scores, issuing dynamic early warnings, and generating corresponding environmental control strategies.

[0025] Optionally, the steps of classifying disease risk levels and providing dynamic early warning include:

[0026] Define a high-risk threshold θ1 and a low-risk threshold θ2;

[0027] When R≥θ1, it is judged as high risk, triggering an emergency warning and generating a comprehensive environmental adjustment and isolation strategy;

[0028] When θ2≤R<θ1, it is judged as medium risk, triggering a general warning and generating a targeted environmental adjustment strategy;

[0029] When R < θ2, it is judged as low risk, no warning is triggered, and continuous monitoring continues.

[0030] Optionally, a disease control system based on continuous monitoring of the rearing environment includes:

[0031] Data Acquisition Module: Used to collect short-term dynamic environmental data and long-term static breeding data of pig farms;

[0032] Processing module: Used to extract short-term environmental fluctuation characteristics and long-term disease association characteristics based on short-term dynamic environmental data and long-term static aquaculture data;

[0033] Scoring module: used to predict short-term disease risk scores based on short-term environmental fluctuation characteristics and long-term disease risk scores based on long-term disease association characteristics;

[0034] Assessment module: This module integrates short-term and long-term disease risk scores to obtain a comprehensive risk score, thereby providing a comprehensive assessment of disease risk.

[0035] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for disease prevention and control based on continuous monitoring of the breeding environment, which has the following beneficial effects:

[0036] This invention proposes a disease prevention and control method based on continuous monitoring of the breeding environment, comprising: collecting short-term dynamic environmental data and long-term static breeding data of pig farms; extracting short-term environmental fluctuation characteristics and long-term disease association characteristics based on the short-term dynamic environmental data and long-term static breeding data; predicting a short-term disease risk score based on the short-term environmental fluctuation characteristics and a long-term disease risk score based on the long-term disease association characteristics; and fusing the short-term disease risk score and the long-term disease risk score to obtain a comprehensive risk score for comprehensive assessment of disease risk. This invention collects short-term dynamic environmental data and long-term static breeding data from pig farms, extracts short-term environmental fluctuation characteristics and long-term disease association characteristics, predicts short-term disease risk scores and long-term disease risk scores respectively, and obtains a comprehensive risk score through adaptive weight fusion. By fusing short-term dynamic environmental data and long-term static breeding data, it comprehensively captures key factors affecting disease occurrence, and uses adaptive weight fusion of risk scores to balance the influence of short-term environmental mutations and long-term breeding patterns. Based on this score, disease risk levels are classified and dynamic early warnings are issued, while environmental regulation strategies are generated. This invention combines short-term environmental changes with long-term farming patterns, improving the accuracy and dynamic response capability of disease prevention and control, effectively reducing the risk of disease occurrence and spread, and reducing farming losses. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of a disease prevention and control method based on continuous monitoring of the breeding environment provided by the present invention.

[0039] Figure 2 This invention provides a structural framework diagram of a disease prevention and control system based on continuous monitoring of the breeding environment. Detailed Implementation

[0040] 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.

[0041] This invention discloses a method for disease prevention and control based on continuous monitoring of the breeding environment, such as... Figure 1 As shown, it includes:

[0042] Collect short-term dynamic environmental data and long-term static breeding data of pig farms;

[0043] Based on short-term dynamic environmental data and long-term static aquaculture data, short-term environmental fluctuation characteristics and long-term disease association characteristics are extracted.

[0044] Predict short-term disease risk scores based on short-term environmental fluctuation characteristics, and predict long-term disease risk scores based on long-term disease association characteristics;

[0045] By integrating short-term and long-term disease risk scores, a comprehensive risk score is obtained, which is then used to comprehensively assess disease risk.

[0046] Furthermore, the short-term dynamic environmental data includes real-time temperature, real-time humidity, real-time suspended particulate matter concentration, real-time noise decibels, and real-time activity levels of the pig herd, which are acquired in real time through a sensor network deployed within the farm; the long-term static farming data includes historical environmental parameter records, pig breed characteristics, historical disease occurrence records, stocking density, and feed formulations, which are extracted from the farming management system database.

[0047] Furthermore, the steps for extracting short-term environmental fluctuation characteristics and long-term disease association characteristics include:

[0048] Through time-domain analysis, temperature change rate, humidity peak, suspended particulate matter concentration skewness, and noise decibel standard deviation are extracted from short-term dynamic environmental data.

[0049] By using fast Fourier transform frequency domain analysis, the fluctuation frequency and amplitude of environmental parameters in short-term dynamic environmental data are extracted.

[0050] Through statistical analysis, seasonal environmental averages, environmental parameter ranges during peak disease periods, annual disease incidence rates, and the correlation between stocking density and disease were extracted from long-term static aquaculture data.

[0051] In a specific implementation, the steps for extracting the frequency and amplitude of environmental parameter fluctuations from short-term dynamic environmental data through fast Fourier transform frequency domain analysis include:

[0052] S1: Frequency domain analysis steps based on Fast Fourier Transform (FFT) (1) Environmental parameter time series data acquisition and preprocessing

[0053] Collect continuous time series data of target environmental parameters, such as temperature, humidity, suspended particulate matter concentration, and noise decibels. The sampling frequency is set according to the fluctuation characteristics of the parameters, such as sampling once every 5 minutes, and continuously collecting data for 24 hours or more to form the original time series dataset.

[0054] Preprocessing: The original data is denoised, such as by removing instantaneous outliers using the moving average method, normalizing the data, mapping the data to the range of [0, 1], eliminating the influence of units, and obtaining stable time series data.

[0055] (2) Fast Fourier Transform (FFT)

[0056] Applying FFT to the preprocessed environmental parameter time series data converts the time-domain signal of parameter values ​​that change with time into a frequency-domain signal of amplitude values ​​that change with frequency, resulting in a spectrum.

[0057] Mathematical principle: The signal is decomposed into sine / cosine components of different frequencies using FFT, as shown in the formula:

[0058]

[0059] Where x(n) is time-domain data, X(k) is frequency-domain data, N is the number of sampling points, and k is the frequency index.

[0060] (3) Extraction of fluctuation frequency and amplitude

[0061] Analyze the amplitude of each frequency component in the spectrum: the horizontal axis is the frequency, unit: Hz, reflecting the periodic characteristics of parameter fluctuation, such as the daily periodic fluctuation corresponding to a frequency ≈ 0.0417Hz, and the vertical axis is the amplitude, reflecting the intensity of parameter fluctuation at that frequency.

[0062] Extracting peak frequencies: Identify the frequency points with the largest amplitudes in the spectrum, which are the main fluctuation frequencies of environmental parameters, such as the peak frequencies of daily temperature fluctuations.

[0063] Extract the corresponding amplitude: Record the amplitude value corresponding to the peak frequency as an intensity index of parameter fluctuation at that frequency.

[0064] S2: Environmental parameter fluctuation feature extraction steps based on convolutional network (CNN) (1) Data format conversion

[0065] Convert the preprocessed environmental parameter time series data or the FFT-generated spectral data into a format suitable for input to convolutional networks:

[0066] If the input is time series data: M consecutive sampling points are combined into a time window, such as M=24, which corresponds to a 120-minute parameter sequence, forming a two-dimensional matrix [number of samples, time window length, 1].

[0067] If the input is frequency domain data: use the spectrum obtained by FFT, the frequency-amplitude distribution as a two-dimensional image input [number of samples, frequency dimension, amplitude dimension, 1].

[0068] (2) Convolutional Network Model Construction

[0069] Input layer: Receives converted time-series or frequency-domain data.

[0070] Convolutional layers: 1D convolution is used to extract local fluctuation features from time-series data, or 2D convolution is used to extract local fluctuation features from spectrograms. For example, the 1D convolution kernel size is set to 3×1 and the sliding stride is 1 to capture the fluctuation trend of 3 consecutive sampling points; multiple sets of convolution kernels, such as 16 or 32, are used to extract fluctuation features at different scales, such as high-frequency small-amplitude fluctuations and low-frequency large-amplitude fluctuations.

[0071] Pooling layer: Max pooling is used, such as a 2×1 window, to reduce the dimensionality of the convolution result, retain key features, such as the amplitude corresponding to the peak frequency, and reduce the amount of computation.

[0072] Fully connected layer: Maps the pooled feature vectors to the output layer, outputting the predicted values ​​of the fluctuation frequency and amplitude of the environmental parameters.

[0073] (3) Model training and feature extraction

[0074] Training data: The training set is constructed using manually labeled fluctuation frequencies and amplitudes, extracted manually through FFT.

[0075] Loss function: The difference between the predicted value and the label is calculated using the mean squared error (MSE). The network parameters are optimized through backpropagation, such as a learning rate of 0.001 and 100 iterations.

[0076] Feature extraction: After training, the key frequency components of parameter fluctuations are analyzed by the feature map output by the convolutional layer, the feature positions corresponding to the main fluctuation frequencies are identified by the activation value intensity, and the final fluctuation frequency and amplitude are obtained by combining the output of the fully connected layer.

[0077] S3: Step-by-Step Connection and Objectives

[0078] First, the fundamental frequency and amplitude of the environmental parameters are obtained through FFT, which serve as training labels and initial feature references for the convolutional network.

[0079] Furthermore, by learning more complex fluctuation patterns through convolutional networks, such as nonlinear fluctuations under multi-parameter coupling, the frequency and amplitude can be accurately extracted, providing high-frequency fluctuation characteristics as a basis for disease risk assessment. For example, abnormal frequency fluctuations indicate the risk of disease transmission.

[0080] Furthermore, the short-term disease risk score prediction based on short-term environmental fluctuation characteristics includes:

[0081] Among them, R s For short-term disease risk scoring, T v H is the rate of temperature change. p P represents the peak humidity level. s For the skewness of suspended particulate matter concentration, Nd A represents the standard deviation of noise in decibels. a The amplitude of the environmental parameter fluctuation is represented by α and β, which are correction coefficients.

[0082] Furthermore, the prediction of long-term disease risk scores based on long-term disease association characteristics includes:

[0083] Among them, R l For long-term disease risk scoring, S e To determine the degree of matching between seasonal environmental mean and environmental parameter ranges during peak disease seasons, I y D represents the average annual incidence rate of disease. c γ represents the correlation between stocking density and disease incidence, and δ represents correction coefficients.

[0084] Furthermore, the fusion of short-term and long-term disease risk scores includes:

[0085] Through adaptive weight adjustment rules, when the long-term disease risk score R... l Below the short-term disease risk score R s When R is reduced s The weights; when R l Higher than R s When R is reduced l The weights;

[0086] The fusion expression is:

[0087]

[0088] Where R is the comprehensive risk score, t is time, and λ, μ, ν, and κ are adjustment factors.

[0089] This fusion approach dynamically balances the impact of short-term environmental mutations and long-term aquaculture patterns, thereby improving the accuracy of assessments.

[0090] In a specific implementation, combining the continuous time-series monitoring characteristics of breeding environment parameters, such as the dynamic fluctuations of temperature and humidity over time, and the idea of ​​fusing short-term and long-term data, a temporal neural network is introduced. Taking the Long Short-Term Memory (LSTM) network as an example, the temporal dependency relationship between environmental parameters and disease risk is captured, and the original fusion expression is extended and optimized. The core of the optimization lies in: extracting the temporal features of historical risk scores through LSTM, dynamically generating adaptive weights that change over time, and enhancing the comprehensive response capability of the fusion expression to long-term trends and short-term changes in the breeding environment. Specific steps include:

[0091] (1) Temporal Feature Extraction Formula

[0092] Let the historical time window be t = 1, 2, ..., k. For example, based on monitoring data from the past 7 days, the short-term disease risk score at time t is R. s (t), Long-term disease risk score is R l (t), construct the time-series input sequence X(t) = [R s (t),R l (t)]. Temporal features are extracted using an LSTM network:

[0093] H(t) = LSTM(X(t), H(t-1)), where H(t) is the hidden state of the LSTM at time t, reflecting the R values ​​at time t and the previous k times. s With R l The association features; H(t-1) is the hidden state at time t-1, reflecting the temporal dependency.

[0094] (2) Optimized fusion expression

[0095] Dynamic weights w are generated based on the temporal features H(t) output by the LSTM. s (t) Short-term score weights and w l (t). Long-term scoring weights are combined with the original nonlinear adjustment term to obtain the comprehensive risk score R(t) at time t:

[0096] w s (t)=σ(W s ·H(t)+b s )

[0097] w l (t)=σ(W l ·H(t)+b l )

[0098]

[0099] Where σ(·) is the Sigmoid activation function, ensuring that the weights w s (t),w l (t)∈[0,1];

[0100] W s W l Let b be the weight matrix. s b l The bias term is optimized through training, and λ, μ, ν, and κ are nonlinear adjustment factors.

[0101] (3) Extended optimization steps

[0102] Step 1: Time Series Data Preparation

[0103] Collect short-term dynamic environmental data in continuous time series, such as hourly temperature and humidity fluctuations, and long-term static aquaculture data, such as weekly disease incidence statistics, and calculate Rk for historical k time points. s (t) and R l (t), t = 1, 2, ..., k, forming the time-series input X(t) = [R s (t),R l (t)].

[0104] A training dataset is constructed using the actual occurrence of an epidemic, such as whether an epidemic breaks out at time t, as the label Y(t)∈{0,1}, where 0 represents no epidemic and 1 represents an epidemic.

[0105] Step 2: LSTM Network Construction

[0106] Input layer: Receives a time sequence X(t) with dimensions [k, 2], where k is the time window length and 2 corresponds to R. s With R l .

[0107] LSTM layer: Set to 1 LSTM layer with 32 hidden units, which can be adjusted according to the amount of data, used to extract R. s With R l Based on the historical association features, output the hidden state H(t).

[0108] Fully connected layer: Maps H(t) to the weight generation layer, and activates the output w using the Sigmoid function. s (t) and w l (t).

[0109] Step 3: Model Training and Weight Optimization

[0110] Loss function: Cross-entropy loss is used, combined with the deviation between the comprehensive risk score R(t) and the label Y(t):

[0111]

[0112] Where N is the number of samples, and R(t) needs to be normalized to [0, 1].

[0113] Optimizer: The Adam optimizer is used with a learning rate of 0.001. Iterative training is performed until the loss converges to obtain the optimal parameters W. s W l b s b l And the regulating factors λ, μ, ν, κ.

[0114] Step 4: Dynamic Integration and Risk Assessment

[0115] For real-time monitoring data, calculate R at the current time t. s(t) and R l H(t) is obtained by inputting H(t) into the trained LSTM network, and w is generated. s (t) and w l (t).

[0116] Substitute the optimized fusion expression into the calculation of R(t), and perform dynamic early warning according to the original risk classification rules of high / medium / low risk.

[0117] (4) Temporal Dependency Capture: Compared with the original static fusion formula, LSTM can mine historical R. s With R l The impact of current risks, such as the correction of current scores based on historical data from peak seasonal disease outbreaks, improves the accuracy of long-term trend predictions. Dynamic weight adaptation: w s (t) and w l (t) Adjusts in real time over time, automatically increasing R in the event of short-term environmental changes, such as a sudden increase in humidity. s Weighting, increasing R when long-term trends dominate, such as persistently high stocking densities. l Weighting enhances the timeliness of prevention and control response.

[0118] Furthermore, it also includes: classifying disease risk levels based on comprehensive risk scores, issuing dynamic early warnings, and generating corresponding environmental control strategies.

[0119] Furthermore, the steps of classifying disease risk levels and conducting dynamic early warning include:

[0120] Define a high-risk threshold θ1 and a low-risk threshold θ2;

[0121] When R≥θ1, it is judged as high risk, triggering an emergency warning and generating a comprehensive environmental adjustment and isolation strategy;

[0122] When θ2≤R<θ1, it is judged as medium risk, triggering a general warning and generating a targeted environmental adjustment strategy;

[0123] When R < θ2, it is judged as low risk, no warning is triggered, and continuous monitoring continues.

[0124] The system classifies pigs into risk levels based on a comprehensive risk score and issues early warnings. It also generates environmental control strategies for different risk levels. For example, in high-risk situations, ventilation equipment is automatically activated to regulate temperature and humidity, and the frequency of air purification is increased; in medium-risk situations, feed nutrient ratios are adjusted to enhance the pig herd's resistance.

[0125] In a specific implementation, a disease prevention and control system based on continuous monitoring of the breeding environment, such as... Figure 2 As shown, it includes:

[0126] Data Acquisition Module: Used to collect short-term dynamic environmental data and long-term static breeding data of pig farms;

[0127] Processing module: Used to extract short-term environmental fluctuation characteristics and long-term disease association characteristics based on short-term dynamic environmental data and long-term static aquaculture data;

[0128] Scoring module: used to predict short-term disease risk scores based on short-term environmental fluctuation characteristics and long-term disease risk scores based on long-term disease association characteristics;

[0129] Assessment module: This module integrates short-term and long-term disease risk scores to obtain a comprehensive risk score, thereby providing a comprehensive assessment of disease risk.

[0130] In a specific implementation, we take the application of disease prevention and control in a large-scale pig farm as an example.

[0131] I. Basic Information of the Farm

[0132] A large-scale pig farm raises 1,000 Duroc-Landrace-Large White crossbred pigs in closed pig houses equipped with temperature sensors, humidity sensors, particulate matter detectors, noise meters, and cameras to monitor pig activity. Data is collected in real time through the farm's Internet of Things (IoT) system. Long-term data is stored in the farm's management system, including environmental records for the past three years, pig breed characteristics (e.g., Duroc-Landrace-Large White crossbred pigs have moderate disease resistance), historical disease records (e.g., an outbreak of porcine epidemic diarrhea in the summer of 2022), stocking density (e.g., 1.2 pigs / m²), and feed formulation (e.g., containing basic antibiotic additives).

[0133] II. Data Collection and Preprocessing, July 1st - July 7th

[0134] (1) Short-term dynamic environmental data: collected every 5 minutes for 7 days. Some key data are as follows:

[0135] Real-time temperature: 24-32℃, briefly rose to 35℃ on July 5 due to equipment failure;

[0136] Real-time humidity: 65%-85%, with higher humidity from 9:00 to 15:00 daily;

[0137] Real-time suspended particulate matter concentration: 0.3-0.8 mg / m³ 3 ;

[0138] Real-time noise level: 55-70 dB, with higher noise levels during feeding times;

[0139] Real-time activity level of the pig herd: 20-30 times per pig per day, decreasing to 15 times per pig after July 5th.

[0140] Preprocessing: Outliers at 35℃ were removed by moving average method. After normalization, the temperature range was mapped to [0.2, 0.8] and the humidity range to [0.5, 0.9].

[0141] (2) Long-term static aquaculture data:

[0142] Historical environmental records: The average temperature in July over the past three years was 28℃, and the average humidity was 70%.

[0143] Historical disease records: The peak season for swine epidemic diarrhea is July-August, with environmental parameters ranging from temperature ≥30℃ to humidity ≥80% during the peak season.

[0144] Average annual disease incidence rate: 3.5%;

[0145] The correlation between stocking density and disease incidence is 0.6. This is based on statistical analysis of data from the past three years. The higher the correlation, the greater the impact of stocking density on disease incidence.

[0146] III. Feature Extraction

[0147] 1) Short-term environmental fluctuation characteristics, based on data from July 5th:

[0148] (1) Time domain analysis: Temperature change rate Tv = (32-24)℃ / 8h = 1.0℃ / h;

[0149] Peak humidity (Hp) = 85%;

[0150] The skewness of suspended particulate matter concentration Ps = 0.7 (positive skewness, indicating that the concentration is mostly higher than the mean);

[0151] The standard deviation of noise in decibels is Nd = 5 dB (with relatively small fluctuations).

[0152] (2) FFT frequency domain analysis:

[0153] FFT was performed on 288 data points (1 point every 5 minutes) of humidity data from July 5th to obtain a spectrum: peak frequency = 0.0417Hz, corresponding to a 24-hour cycle, i.e., daily fluctuation;

[0154] The corresponding amplitude Aa = 0.15, after normalization, reflects the intensity of daily humidity fluctuations.

[0155] 2) Characteristics of long-term disease association:

[0156] Seasonal environmental averages: July average temperature 28℃, average humidity 70%;

[0157] The seasonal environmental mean matched the parameters during the peak season for the disease with a degree of agreement Se = 0.8, because the temperature and humidity in July were close to the range of the peak season.

[0158] The average annual incidence rate of disease is Iy = 3.5%;

[0159] The correlation between stocking density and disease incidence was Dc = 0.6.

[0160] IV. Risk Score Prediction

[0161] Short-term disease risk score R s Substitute into the formula With correction factors α = 0.5 and β = 0.3,

[0162] Long-term disease risk score R l Substitute into the formula:

[0163] Let the correction factors be γ = 2.0 and δ = 0.5.

[0164]

[0165] V. Comprehensive risk assessment, combined with LSTM optimization

[0166] (1) Time series data preparation: Take R data from the past 7 days (July 1st - July 7th). s and R l The time-series input X(t) = [R] is formed. s (t),R l (t)], of which July 5th R s =0.39, R l =0.91, R for the remaining dates s Fluctuating between 0.2 and 0.4, R l Fluctuating between 0.8 and 0.95; Label Y(t): No disease outbreaks from July 1 to 4 (Y=0), 3 pigs developed diarrhea on July 5 (Y=1).

[0167] (2) LSTM training and dynamic weight generation:

[0168] Construct a 1-layer LSTM with 32 hidden units and an input dimension of [7, 2]. After training, obtain the weight matrix W. s W l and bias b s b l ;

[0169] On July 5th, the hidden state H(t) = 0.82, and dynamic weights are generated:

[0170] w s (t)=σ(W s ·H(t)+b s ) = 0.3, because R l >R s Reduce short-term weighting;

[0171] w l (t)=σ(W l ·H(t)+b l = 0.7.

[0172] (3) Comprehensive risk score R: Substituting into the optimized fusion formula, the adjustment factors are λ = 0.5, μ = 0.1, ν = 0.2, and κ = 1.0.

[0173] The calculation yields:

[0174] R(t)≈0.3×0.15+0.7×0.32+0.19≈0.045+0.224+0.19=0.459.

[0175] VI. Early Warning and Adjustment Strategies

[0176] Risk level classification: Set the high risk threshold θ1 = 0.6 and the low risk threshold θ2 = 0.3. Since 0.3 ≤ 0.459 < 0.6, it is judged as medium risk.

[0177] Dynamic early warning: Triggering a general early warning will push a message to the farm management terminal.

[0178] Adjustment strategy:

[0179] Environmental control: Turn on ventilation equipment to reduce humidity from 85% to 70%, and turn on air purifier to reduce particulate matter concentration to 0.5 mg / m³. 3 the following;

[0180] Breeding management: Adjust feed formula, add vitamin C and probiotics to enhance the pig herd's resistance, and increase the frequency of inspections (once every 2 hours).

[0181] Application effect

[0182] Follow-up inspection on July 6: The activity level of the pig herd has rebounded to 25 times per head, and there are no new cases of diarrhea; on July 7, the comprehensive risk score dropped to 0.28 (low risk), and the spread of the disease has been effectively controlled.

[0183] This example verifies that by integrating short-term environmental fluctuations with long-term breeding patterns, this embodiment can accurately identify disease risks and guide prevention and control, thereby improving the efficiency of disease prevention and control in large-scale pig farms.

[0184] In a specific embodiment, this embodiment also provides a computer device applicable to a disease prevention and control method based on continuous monitoring of the breeding environment, comprising: a memory and a processor; the memory stores computer-executable instructions, and the processor executes the computer-executable instructions to implement the disease prevention and control method based on continuous monitoring of the breeding environment as proposed in the above embodiment. The computer device can be a terminal, and includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless communication can be achieved through WIFI, carrier networks, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse, etc.

[0185] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a disease prevention and control method based on continuous monitoring of the breeding environment as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0186] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0187] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A disease prevention and control method based on continuous monitoring of feeding environment, characterized in that, The application relates to a pig disease risk assessment method and device. Short-term dynamic environment data and long-term static breeding data of a pig farm are collected; Short-term environment fluctuation characteristics and long-term disease correlation characteristics are extracted based on the short-term dynamic environment data and the long-term static breeding data; Short-term disease risk scores are predicted based on the short-term environment fluctuation characteristics, and long-term disease risk scores are predicted based on the long-term disease correlation characteristics; The short-term disease risk scores and the long-term disease risk scores are fused to obtain comprehensive risk scores, and the disease risk is comprehensively evaluated.

2. The disease prevention and control method based on continuous monitoring of the feeding environment according to claim 1, characterized in that, The short-term dynamic environment data includes real-time temperature, real-time humidity, real-time suspended particulate matter concentration, real-time noise decibel and pig group real-time activity; the long-term static breeding data includes historical environment parameter records, pig group breed characteristics, historical disease occurrence records, breeding density and feed formula.

3. The disease prevention and control method based on continuous monitoring of the feeding environment according to claim 1, characterized in that, The step of extracting the short-term environment fluctuation characteristics and the long-term disease correlation characteristics comprises: Temperature change rate, humidity peak value, suspended particulate matter concentration skewness and noise decibel standard deviation are extracted from the short-term dynamic environment data through time domain analysis; Environment parameter fluctuation frequency and amplitude in the short-term dynamic environment data are extracted through frequency domain analysis; Seasonal environment mean, disease high incidence period environment parameter range, annual average disease occurrence rate and breeding density and disease correlation degree are extracted from the long-term static breeding data through statistical analysis.

4. The disease prevention and control method based on continuous monitoring of the feeding environment according to claim 1, characterized in that, The step of predicting the short-term disease risk scores based on the short-term environment fluctuation characteristics comprises: wherein R s is a short-term blight risk score, T v is a rate of temperature change, H p is a humidity peak, P s is a concentration of suspended particulate matter skewness, N d is a noise decibel standard deviation, A a is an environmental parameter fluctuation amplitude, and a, b are correction factors.

5. The disease prevention and control method based on continuous monitoring of the feeding environment according to claim 1, characterized in that, The step of predicting the long-term disease risk scores based on the long-term disease correlation characteristics comprises: wherein R l is the long-term blight risk score, S e is the degree of matching of the seasonal environmental mean with the range of environmental parameters during the blight high period, I y is the average annual blight incidence, D c is the degree of association of the culture density with the blight, and γ, δ are correction factors.

6. The disease prevention and control method based on continuous monitoring of the feeding environment according to claim 1, characterized in that, The step of fusing the short-term disease risk scores and the long-term disease risk scores comprises: By an adaptive weight adjustment rule, the weight of R l is decreased when the long-term blight risk score R s is lower than the short-term blight risk score R s ; the weight of R l is decreased when R s is higher than R l ; The fusion expression is as follows: Wherein, R is the comprehensive risk score, t is time, and lambda, mu, nu and kappa are adjustment factors.

7. The disease prevention and control method based on continuous monitoring of the feeding environment according to claim 1, characterized in that, The application further comprises: According to the comprehensive risk score, a disease risk grade is divided, dynamic early warning is carried out, and corresponding environment regulation strategies are generated.

8. The disease prevention and control method based on continuous monitoring of the feeding environment according to claim 7, characterized in that, The step of dividing the disease risk grade and carrying out dynamic early warning comprises: A high risk threshold theta1 and a low risk threshold theta2 are defined; When R is greater than or equal to theta1, it is determined as high risk, emergency early warning is triggered, and comprehensive environment regulation and isolation strategies are generated; When theta2 is less than R and R is less than theta1, it is determined as medium risk, general early warning is triggered, and targeted environment regulation strategies are generated; When R is less than theta2, it is determined as low risk, no early warning is triggered, and continuous monitoring is carried out.

9. A disease prevention and control system based on continuous monitoring of feeding environment, characterized in that, The application comprises: A collection module for collecting short-term dynamic environment data and long-term static breeding data of a pig farm; A processing module for extracting short-term environment fluctuation characteristics and long-term disease correlation characteristics based on the short-term dynamic environment data and the long-term static breeding data; A scoring module for predicting short-term disease risk scores based on the short-term environment fluctuation characteristics, and predicting long-term disease risk scores based on the long-term disease correlation characteristics; An evaluation module for fusing the short-term disease risk scores and the long-term disease risk scores to obtain comprehensive risk scores, and comprehensively evaluating the disease risk.

Images