Animal husbandry management method and system based on intelligent data analysis

Abstract

The invention relates to the technical field of animal husbandry intelligent management, and discloses an animal husbandry management method and system based on intelligent data analysis, and the method comprises the steps: collecting the basic data, metabolic characteristics and feed component data of a target individual, and determining the current metabolism stage of the target individual; and determining a key nutrition bottleneck and a key limiting factor corresponding to the current metabolism stage of the target individual according to the key metabolism path of the current metabolism stage of the target individual and the nutrient substance interaction matrix. Then generating an optimization formula corresponding to the target individual in the current metabolism stage; and further predicting a time point when the target individual is converted from the current metabolic stage to the next-generation falling stage, and generating a feeding scheme corresponding to the next-generation falling stage. Accurate nutrient formula dynamic adjustment of the whole life cycle of animals is achieved, the feed utilization rate is increased, the growth cycle is shortened, and the feed cost is reduced.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology in animal husbandry, and more specifically, it relates to an animal husbandry management method and system based on intelligent data analysis. Background Technology

[0002] There are several technical problems to be solved in current livestock feeding management. For example, traditional livestock feeding management often adopts a batch uniform formula model, which ignores the individual differences of animals. This results in some individuals receiving excessive nutrition while others receive insufficient nutrition, failing to meet the actual needs of different individuals at different growth stages, reducing feed utilization efficiency and increasing production costs.

[0003] Existing precision feeding systems lack accurate modeling of the animal's internal nutrient metabolic network and its dynamic changes. These systems typically design feeding programs based on external characteristics and cannot adjust feeding formulas in real time according to the animal's actual internal metabolic state, resulting in low nutrient utilization efficiency. Existing feeding optimization systems are mostly designed based on static metabolic models, simplifying the animal's growth process into a few fixed stages. They fail to capture the gradual changes and transitions in metabolic patterns between different growth stages, cannot adapt to metabolic pattern transitions at different growth stages, and struggle to achieve dynamic regulation throughout the entire life cycle. Furthermore, there is a lack of effective methods to identify metabolic state transition points and key metabolic pathways. Current technologies struggle to accurately predict the timing of metabolic pattern transitions, leading to delayed or premature adjustments to feeding programs, resulting in wasted nutrient resources and low growth efficiency.

[0004] The aforementioned technical problems severely restrict the improvement of livestock production efficiency and the reduction of costs, necessitating an intelligent management method that can achieve individualized and precise feeding. Summary of the Invention

[0005] This application provides a livestock management method and system based on intelligent data analysis, which solves the technical problem of precision feeding in related technologies.

[0006] This application provides a livestock management method based on intelligent data analysis, including: Collect basic data, metabolic characteristics, and feed composition data of the target individuals; Based on the target individual's basic data, metabolic characteristics, and feed composition data, the current metabolic stage of the target individual is determined; among them, the target individual has multiple corresponding metabolic stages; each metabolic stage has a corresponding metabolic state; Based on the key metabolic pathways and nutrient interaction matrix of the target individual's current metabolic stage, the key nutritional bottlenecks and key limiting factors of the target individual at the current metabolic stage are determined. Based on key nutritional bottlenecks and key limiting factors, generate an optimized formula for the target individual at the current metabolic stage. Based on the metabolic status of the target individual in the current metabolic stage and the corresponding optimized formula, predict the time point when the target individual will transition from the current metabolic stage to the next metabolic stage, and generate the feeding plan corresponding to the next metabolic stage.

[0007] In a preferred embodiment, a livestock management system based on intelligent data analysis is used to execute a livestock management method based on intelligent data analysis, comprising: The data acquisition module is used to collect basic data, metabolic characteristics, and feed composition data of the target individual. The stage determination module is used to determine the current metabolic stage of the target individual based on its basic data, metabolic characteristics, and feed composition data. Each target individual has multiple corresponding metabolic stages, and each metabolic stage has a corresponding metabolic state. The factor determination module is used to determine the key nutritional bottlenecks and key limiting factors of the target individual at the current metabolic stage based on the key metabolic pathways and nutrient interaction matrix of the target individual at the current metabolic stage. The formulation generation module is used to generate an optimized formulation for the target individual at the current metabolic stage based on key nutritional bottlenecks and key limiting factors. The transition time determination module is used to predict the time point when the target individual transitions from the current metabolic stage to the next metabolic stage based on the target individual's metabolic state in the current metabolic stage and the corresponding optimized formula, and to generate the feeding plan corresponding to the next metabolic stage.

[0008] The beneficial effects of this application are as follows: Compared to traditional feeding methods, this application improves feed utilization, shortens the growth cycle, reduces feed costs, and simultaneously reduces health problems caused by nutritional imbalances, thus improving animal welfare. It achieves individualized management by dynamically modeling metabolic networks, shifting from superficial demand satisfaction to internal metabolic optimization. This provides a scientific feeding method based on physiological mechanisms, automatically adjusting feeding strategies according to individual differences to maximize the growth potential of each individual. It also achieves an improvement from static optimization to dynamic regulation throughout the entire life cycle, adaptively optimizing feeding strategies based on growth and development patterns to ensure animals receive optimal nutritional support at different growth stages. Through time-series prediction networks and growth trajectory prediction algorithms, it can predict metabolic mode transition points in advance, enabling smooth transitions in feeding programs, avoiding metabolic stress, and reducing feeding risks. By accurately identifying key metabolic pathways and nutritional limiting factors, it avoids nutrient waste, improves resource utilization efficiency, reduces environmental pollution, and promotes the sustainable development of animal husbandry. Attached Figure Description

[0009] Figure 1 This is a flowchart of a livestock management method based on intelligent data analysis according to this application; Figure 2 This is a structural block diagram of a livestock management system based on intelligent data analysis, as described in this application. Detailed Implementation

[0010] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0011] At least one embodiment of this application discloses a livestock management method based on intelligent data analysis, such as... Figure 1 As shown, it includes the following steps: Step S100: Collect basic data, metabolic characteristics, and feed composition data of the target individual.

[0012] Specifically, basic data including animal weight, growth stage, and physiological indicators are collected, along with metabolic characteristics at different growth stages (such as juvenile, growth, and fattening periods). Differences in metabolic pathways at each stage are analyzed to establish characteristic metabolic models and create a multi-stage metabolic model library. ,in, This represents a multi-stage metabolic model library. , , They represent the first , , Metabolic models for each growth stage This indicates the total number of growth stages. Feed composition is monitored in real time using near-infrared spectroscopy.

[0013] Step S200: Based on the target individual's basic data, metabolic characteristics, and feed composition data, determine the target individual's current metabolic stage; wherein, the target individual has multiple corresponding metabolic stages; each metabolic stage has a corresponding metabolic state.

[0014] Specifically, step S200 also includes: Step S210: Based on the target individual's basic data, metabolic characteristics, and feed composition data, construct a metabolic pathway diagram for the target individual; wherein, the metabolic pathway diagram includes nodes and edges; nodes are metabolites; edges are the conversion relationships between metabolites; each edge has a corresponding reaction rate.

[0015] Here, metabolomics analysis methods are used to analyze and process basic data such as animal body weight, growth stage, and physiological indicators to obtain metabolite composition information. Based on this information, an individual metabolic pathway graph is constructed. This pathway graph contains nodes and edges, using a graph structure. It means that, among them, Represents a collection of metabolites. Represents a set of metabolic reactions. Each edge Metabolites To metabolites The transformation relationship, and has weights. Indicates the reaction rate.

[0016] Step S220: Construct a multi-stage metabolic pattern library based on the metabolic characteristics of each metabolic stage.

[0017] To analyze the differences in metabolic pathways at different growth stages (such as infancy, growth period, and fattening period), characteristic metabolic models are established, forming a multi-stage metabolic model library. ,in, This represents a multi-stage metabolic model library. , , They represent the first , , Metabolic models for each growth stage This represents the total number of growth stages.

[0018] Step S230: Based on the feed composition data, obtain a model showing the relationship between spectral characteristics and the nutrient content of the feed.

[0019] Near-infrared spectroscopy was used to monitor feed composition in real time, and a model was established to show the relationship between spectral characteristics and nutrient content. ,in, Represents a vector of nutrient content. Represents the spectral eigenvector. This represents the mapping function. Based on this model, nutrient input can be quickly and accurately assessed, providing data support for subsequent metabolic state analysis.

[0020] Step S240: Determine the current metabolic stage of the target individual based on the Hidden Markov Model.

[0021] Hidden Markov Models (HMMs) are used to automatically identify metabolic state transition points. A Hidden Markov Model is defined as follows: ; in, This represents a Hidden Markov Model. The state transition probability matrix is... For the observation probability matrix, This represents the initial state distribution.

[0022] Based on continuously acquired physiological indicator sequences: ; in, Represents the observation sequence. , , Indicates the number of collections , , A vector of physiological indicators, This indicates the total length of the sequence.

[0023] Calculate the most likely state sequence using the Viterbi algorithm: ; in, Represents the hidden state sequence, , , Indicates the number of collections , , A metabolic state, This indicates the total length of the sequence.

[0024] This method allows the system to determine the current metabolic stage of an animal, providing a basis for subsequent precise nutritional regulation.

[0025] Step S300: Based on the key metabolic pathways and nutrient interaction matrix of the target individual's current metabolic stage, determine the key nutritional bottlenecks and key limiting factors corresponding to the target individual's current metabolic stage.

[0026] Specifically, step S300 also includes: Step S310: Identify key metabolic pathways based on the metabolic state of the target individual at its current metabolic stage.

[0027] Key metabolic pathways are identified using the FluxVariability Analysis (FVA) method. This method determines the possible range of flux for each reaction by solving the following optimization problem: ; in, This is a stoichiometric matrix; This is the reaction flux vector; For the specific reaction flux that needs to be analyzed; and These are the lower and upper limits of the reaction flux, respectively; For biomass synthesis reaction flux; Threshold for biomass synthesis; constraints This indicates that the metabolic network is in a homeostatic state. Indicates the maximum value. This represents the minimum value.

[0028] The path is evaluated using a formula by calculating the impact of changes in reaction flux on the system. Systemic influence: ; in, Representing a path The average impact on the entire metabolic network. The total number of reactions in the system; Indicates reaction Changes in the reaction The degree of impact; Indicates reaction Rate constraints. Pathways with high system influence were identified as critical metabolic pathways.

[0029] Step S320: Analyze the utilization efficiency of each nutrient at its current metabolic stage.

[0030] Analyze the utilization efficiency of each nutrient under the current metabolic state and establish a nutrient metabolism efficiency function. This indicates the current state of the metabolic network. Below, nutrients In feed formulation The utilization efficiency is calculated using the following function: ; in, Indicates nutrients Utilization efficiency under the current metabolic state; Indicates feed formulation; Indicates the time of animals The metabolic network state; Indicates feed formulation Medium nutrients Input quantity; Indicates the state of the metabolic network Below, nutrients The output amount converted into effective biomass is measured in units identical to the input amount.

[0031] Step S330: Obtain the nutrient interaction matrix based on the degree of influence between each nutrient.

[0032] Constructing a nutrient interaction matrix , of which elements Indicates nutrients For nutrients The degree of impact on utilization efficiency is calculated in the following way: ; in: Represents the nutrient interaction matrix; Represents the first in the matrix Line number Column elements; Indicates nutrients Utilization efficiency; Indicates nutrients The concentration; Indicates nutrients Utilization efficiency of nutrients The partial derivative of concentration reflects the change in nutrient content. When the concentration of nutrients changes, The rate of change of efficiency is used, with positive values ​​indicating a promoting effect and negative values ​​indicating an inhibiting effect, and the magnitude of the value indicating the intensity of the effect.

[0033] Step S340: Based on the eigenvalues ​​and eigenvectors of the key metabolic pathways, utilization efficiency, and nutrient interaction matrix, determine the key nutritional bottlenecks and key limiting factors corresponding to the target individual at the current metabolic stage.

[0034] By analyzing key metabolic pathways, utilization efficiency, and matrices By analyzing eigenvalues ​​and eigenvectors, we can determine the synergistic (positive interaction) and antagonistic (negative interaction) relationships between nutrients and identify key nutritional bottlenecks and limiting factors.

[0035] Step S400: Based on key nutritional bottlenecks and key limiting factors, generate an optimized formula for the target individual at the current metabolic stage.

[0036] Specifically, step S400 also includes: Step S410: Based on the key nutritional bottlenecks and key limiting factors, a multi-objective optimization model is obtained.

[0037] Based on the key nutritional bottlenecks and key limiting factors, the feed formulation optimization problem is modeled as a multi-objective optimization problem, with the objective function including: ; ; ; in, This indicates the goal of minimizing costs. Indicates the first The unit price of the raw materials, Indicates the first The amount of each ingredient used in the formula Indicates the total quantity of optional raw materials; This indicates the goal of maximizing nutritional balance. Indicates the first in the formula The actual content of each nutrient Indicates the first The ideal content of various nutrients Indicates the total number of nutrients considered; This represents the objective of maximizing the growth rate. Indicates the rate of weight gain. Indicates the current state of the metabolic network. Lower body weight gain rate and formula The relational function.

[0038] Step S420: Solve the multi-objective optimization problem corresponding to the multi-objective optimization model according to the improved NSGA-III algorithm; wherein, the improved NSGA-III algorithm introduces adaptive crossover and mutation operators, as well as a solution evaluation mechanism based on the metabolic model.

[0039] An improved version of NSGA-III is applied to solve multi-objective optimization problems. Compared with the traditional NSGA-II, this algorithm improves the uniformity of solution distribution in high-dimensional objective space by introducing a reference point mechanism.

[0040] The algorithm process includes: Initialize population Set a set of reference points ; Perform non-dominated sorting to obtain frontier surfaces of different levels. ,in , These represent the first-order non-dominated solution set and the second-order non-dominated solution set, respectively. Select elite individuals to build the next generation of the population. ; When the termination condition is met, the optimal Pareto front is output.

[0041] In particular, this algorithm, based on the standard NSGA-III, introduces adaptive crossover and mutation operators, as well as a solution evaluation mechanism based on a metabolic model, which improves the algorithm's performance in feed formulation optimization problems.

[0042] Step S430: Based on the nutritional efficiency maximization model, obtain the optimized formula corresponding to the target individual at the current metabolic stage.

[0043] The final optimal formula is solved using the following optimization model: ; ; ; ; in: This indicates the optimal feed formulation; This indicates the feed formulation to be evaluated; This indicates the search for a recipe that maximizes the objective function. ; Indicates the total number of nutrients considered; Indicates nutrients Weighting coefficients; Indicates the current state of the metabolic network. Lower nutrients Utilization efficiency; Indicates the time of animals The metabolic network state; Indicates the total quantity of optional feed ingredients; Indicates the first The proportion of each raw material in the formula; and They represent the first The minimum and maximum allowable proportions of each ingredient in the formula; Indicates the number of limiting nutrients that need to be considered; Indicates formula The Middle The content of limiting nutrients; and They represent the first The minimum and maximum allowable amounts of each limiting nutrient.

[0044] Nutritional efficiency maximization models include: Nutrient Metabolic Pathway Module: Used to model the metabolic transformation process of each nutrient based on the metabolic pathway diagram and calculate the utilization efficiency under a given metabolic state. Nutrient Interaction Adjustment Module: Used to adjust the initial efficiency by taking into account the synergistic and antagonistic relationships between nutrients; Multi-stage metabolic transformation prediction module: used to predict the transformation pathway and efficiency of nutrients in the body by combining the metabolic state corresponding to the current stage and the ingredients of the formula. Individualized parameter adaptation module: used to adjust the parameters of the nutritional efficiency maximization model de1 according to the individual characteristics of the animal; Constraint processing module: Used to process nonlinear constraints using the penalty function method.

[0045] Specifically, the nutrient efficiency maximization model adopts a hierarchical optimization structure, consisting of the following main components: Nutrient Metabolism Pathway Module: Based on the metabolic pathway diagram, model the metabolic transformation process of each nutrient and calculate the utilization efficiency under a given metabolic state; Nutrient interaction adjustment module: Considers the synergistic and antagonistic relationships between nutrients and adjusts the initial efficiency; Multi-stage metabolic transformation prediction module: Combines the current metabolic state and formula ingredients to predict the transformation pathway and efficiency of nutrients in the body; Individualized parameter adaptation module: Adjusts model parameters according to individual animal characteristics (such as age, weight, genotype) to make them more consistent with the physiological characteristics of a specific individual; Constraint processing module: The penalty function method is used to handle nonlinear constraints to ensure that the generated formula meets various nutritional requirements and raw material restrictions.

[0046] In the solution process, the model first uses the non-dominated solution set generated by the NSGA-III algorithm, then performs a secondary screening using the objective function of maximizing nutrient efficiency, and finally outputs the optimal formula that comprehensively considers cost, nutrient balance, and growth efficiency. The model verifies the actual effect of the formula through metabolic network simulation, and adaptively adjusts the weight coefficients based on the verification results to continuously improve the practicality and bioavailability of the formula.

[0047] Based on the metabolic model parameters and key metabolic pathways obtained in the preceding steps, the parameters and constraints in the model are adjusted and optimized for the specific metabolic states of different individuals, generating individualized optimal feeding formulas. The optimization process considers characteristics such as the individual's growth stage, genetic background, and health status to ensure that the generated formulas can meet the nutritional needs of specific animals under their current metabolic state.

[0048] Step S500: Based on the metabolic state of the target individual in the current metabolic stage and the corresponding optimized formula, predict the time point when the target individual will transition from the current metabolic stage to the next metabolic stage, and generate the feeding plan corresponding to the next metabolic stage.

[0049] Step S500 also includes: Step S510: Based on historical growth data and the metabolic state corresponding to the current metabolic stage, predict the transition time point of the next generation of metabolic stage.

[0050] Based on historical growth data and current metabolic status, a metabolic pattern transition prediction model is constructed to predict the timing of future metabolic pattern transitions.

[0051] The metabolic pattern transition prediction model employs a time-series prediction network architecture, with past data as input. Physiological index sequence of the day ,in, , , They represent time. , , Collected physiological index vectors, Indicates the length of the time window for historical data. This represents the total time, and the output is the probability distribution of future metabolic transition points. ,in, , , They represent time. , , Events that cause a metabolic state transition, For the predicted time range.

[0052] Optimize the model parameters using the following objective function: ; in, Represents the set of parameters for the prediction model; This represents the total number of training samples; Represents the loss function; Indicates the first The actual label of each sample; Indicates the first Input features of each sample; The parameter is The prediction model function; Represents the regularization term; Represents the regularization coefficient. This represents the minimum value.

[0053] The temporal prediction network model employs an architecture combining a Long Short-Term Memory (LSTM) network and an attention mechanism. This network consists of the following key components: Input embedding layer: maps multidimensional physiological indicator sequences to a hidden representation space; Bidirectional LSTM layer: captures long-term dependencies in time series, where forward LSTM captures the influence of the past on the future, and backward LSTM captures the overall pattern of the sequence; Multi-head self-attention mechanism: Calculates the correlation between different time steps and uses attention weights. Indicates the first The time step for the first The degree of influence at each time step; Metabolic state representation layer: Combines LSTM output and attention weights to generate a contextual representation of the metabolic state; Prediction layer: Based on the state representation, outputs the probability of metabolic transformation occurring at each future time step.

[0054] In practical applications, the model is first trained on historical data using supervised learning, and its predictive performance is evaluated using the cross-entropy loss function. Then, Bayesian inference is used to generate probabilistic predictions, which, combined with confidence intervals, provide a reliable time window estimate for feeding decisions.

[0055] Step S520: Based on the metabolic state and nutrient intake corresponding to the current metabolic stage, predict the future growth trajectory.

[0056] A growth trajectory prediction algorithm is constructed to predict future growth trajectories based on current metabolic state and nutrient intake. This algorithm comprehensively considers historical growth data, current metabolic model parameters, and environmental factors to identify metabolic transformation trends in advance. The algorithm framework includes: Constructing a state-space model: ; in, Let be the state vector at time t. For feeding strategies, For environmental parameters, This is the state transition function. This is the predicted state vector at time t+1; The particle filter algorithm is used to estimate model parameters and predict future states. The uncertainty of the prediction results is analyzed using statistical methods to generate a confidence interval.

[0057] Step S530: Within a preset time period following the predicted transition time of the next generation of telogen effluvium, a transition feeding formula is determined based on a smooth transition strategy.

[0058] In the early stages of the predicted metabolic transition, a smooth transition adjustment to the feeding formula is implemented to avoid stress responses caused by the mutation. The smooth transition strategy is achieved through the following formula: ; in, Indicates time Transitional formula; This indicates the optimal formula for the current stage; This represents the optimal formula for the target stage; The smoothing coefficient is defined as follows: ; in, Represents the smoothing coefficient. Indicates the transition rate parameter; Indicates the time at the midpoint of the transition; The base of the natural logarithm; Indicates the current time.

[0059] This method uses the Sigmoid function to achieve smooth changes in feed composition over time, ensuring that the feed composition gradually transitions from the current stage formula to the target stage formula. This avoids metabolic disorders and stress responses caused by sudden changes, and improves the animal's adaptability to the new formula and the stability of its production performance.

[0060] Step S540: After a preset time period following the predicted transition time of the next generation metabolic stage, adjust to the feeding program corresponding to the next generation metabolic stage.

[0061] Specifically, the method for determining the feeding program corresponding to the next metabolic stage is the same as steps S100 to S400 above.

[0062] In one exemplary embodiment of this application, the method further includes: The parameters of the nutritional efficiency maximization model are adaptively updated based on the actual feeding effect corresponding to the feeding program.

[0063] Specifically, it includes the following sub-steps: First, feedback data collection and benchmark establishment; Feedback data was collected based on actual growth and health indicators, including weight gain rate, feed conversion ratio, and blood biochemical parameters. A benchmark was established based on this data to evaluate the accuracy of the current metabolic model and the effectiveness of the feeding strategy. The benchmark is represented by the following model: ; in, Indicates the benchmark set for comparison; Indicates the first The group's input index vector; Indicates the first The expected output index vector of the group; This indicates the total number of historical data records; , , These represent the complete input-output pairs of the first, second, and third groups, respectively, which are used for subsequent model parameter adjustment and performance evaluation.

[0064] Secondly, the metabolic model parameters are dynamically adjusted; Based on the collected feedback data, the parameters of the metabolic model are dynamically adjusted using Bayesian optimization methods to achieve continuous synchronization between the model and the actual situation.

[0065] Specifically, the model parameters are updated by minimizing the following objective function. : ; in, A set of parameters representing a metabolic model; This represents the updated model parameters; This indicates the number of samples used for updating model parameters; The parameter is Metabolic models; Indicates the first Input data for each sample; Indicates the first The actual observation results of each sample; This represents the squared error between the model's predicted values ​​and the actual observed values. Represents the regularization term; This represents the balance factor.

[0066] Through iterative optimization, the model parameters gradually converge to values ​​that better reflect actual metabolic conditions, thereby improving the model's accuracy in describing and predicting actual metabolic processes.

[0067] Then, reinforcement learning is used to optimize the formula parameters; This method employs a policy gradient-based reinforcement learning approach to continuously optimize the feed formulation parameters based on actual growth feedback. The feeding process is modeled as a Markov Decision Process (MDP), with a state space... Includes the animal's current physiological indicators and metabolic state, and range of motion. For feasible feeding formula adjustment schemes, reward functions Based on growth efficiency and health indicators, the optimization objective is to maximize cumulative reward. ; in, This represents the objective function of reinforcement learning; Indicating in strategy The expected value below; This represents a parameterized strategy function; Indicates the discount factor; Indicates the time step index; Indicates at time step The system status; Indicates at time step The feeding action performed; Indicates the state Next action The instant reward received.

[0068] Optimize policy parameters by updating policy gradient rules: ; in, Indicates time step The policy parameter vector at that time; This represents the updated policy parameter vector; Indicates the learning rate; Describe the objective function Regarding parameters The gradient.

[0069] This method can continuously optimize feeding strategies and improve the accuracy of metabolic models and feeding effects through constant interaction and learning during practical applications.

[0070] Reinforcement learning methods employ a multi-layered Actor-Critic architecture, which includes the following core components: State representation module: Transforms the animal's current physiological indicators, metabolic state, and environmental factors into low-dimensional vector representations, which serve as inputs to the policy network and value network; Actor Network: Generates feeding formula adjustment strategies based on the current state, adopts a multilayer perceptron structure, and uses the Softmax function in the output layer to map the output to the operation probability distribution; Critic Network: Evaluates the value of the current state and action, employs a double Q-learning structure to reduce overestimation, and outputs the Q-value of the state-action pair; Experience replay buffer: stores historical state transition sequences Random sampling reduces sample correlation and improves learning stability. Adaptive reward allocation module: Based on short-term growth indicators and long-term health indicators, it dynamically balances immediate rewards and delayed rewards to solve the problems of reward sparsity and delay.

[0071] This method employs a phased training strategy in practical applications. First, it pre-trains on a metabolic model in a simulation environment. Then, during actual feeding, it continuously adjusts the policy parameters through online learning. By introducing priority experience replay and target network update mechanisms, it reduces policy oscillations and improves learning efficiency and stability. Simultaneously, it utilizes representation learning techniques to extract state features, reducing the dimensionality of the state space and accelerating the learning convergence process.

[0072] In a preferred embodiment, a livestock management system 100 based on intelligent data analysis is used to execute a livestock management method based on intelligent data analysis, including: The data acquisition module 110 is used to collect basic data, metabolic characteristics and feed composition data of the target individual; The stage determination module 120 is used to determine the current metabolic stage of the target individual based on the target individual's basic data, metabolic characteristics, and feed composition data; wherein, the target individual has multiple corresponding metabolic stages; each metabolic stage has a corresponding metabolic state; The factor determination module 130 is used to determine the key nutritional bottlenecks and key limiting factors of the target individual at the current metabolic stage based on the key metabolic pathways and nutrient interaction matrix of the target individual at the current metabolic stage. The formula generation module 140 is used to generate an optimized formula for the target individual at the current metabolic stage based on key nutritional bottlenecks and key limiting factors. The transition time determination module 150 is used to predict the time point when the target individual transitions from the current metabolic stage to the next metabolic stage based on the metabolic state of the target individual in the current metabolic stage and the corresponding optimized formula, and to generate the feeding plan corresponding to the next metabolic stage.

[0073] Real-world application examples of this implementation method: This implementation method is applied to the precision feeding management of fattening pigs (30-120kg). 240 Duroc × Landrace × Large White three-way crossbred commercial pigs with similar genetic backgrounds were selected for the test and randomly divided into an experimental group and a control group, with 120 pigs in each group. The control group adopted a traditional three-stage feeding program, while the experimental group adopted the precision feeding method based on intelligent data analysis proposed in this implementation method. The experimental period was 12 weeks.

[0074] Implementation process example: Example of building a multi-stage metabolic model: First, blood and fecal samples were collected from 30 randomly selected pigs in the experimental group. Metabolite information was obtained using liquid chromatography-mass spectrometry (LC-MS / MS), identifying 217 key metabolites. Based on this metabolite information, an individual metabolic pathway map was constructed. Table 1 shows the node characteristics of some key metabolites in the pathway map.

[0075] Table 1: Key metabolite nodes in fattening pigs (partial);

[0076] By analyzing the differences in metabolite composition among pigs at different growth stages, the fattening period was divided into five metabolic stages (more refined than the traditional three-stage method), and a characteristic metabolic model library was established. Table 2 shows the division of the five metabolic stages and their key characteristics.

[0077] Table 2: Characteristics of multi-stage metabolic patterns in fattening pigs;

[0078] Key metabolic pathway analysis and formulation optimization examples: Response range analysis was performed on key metabolic pathways in the early fattening stage I, identifying 15 key metabolic pathways. Based on the key metabolic pathway analysis results, a multi-objective optimized formulation was generated for each stage using an improved NSGA-III algorithm and a nutrient efficiency maximization model. Taking the early fattening stage I as an example, the optimized formulation and its comparison with the traditional formulation are shown in Table 3.

[0079] Table 3: Comparison of optimized formula and traditional formula in early fattening stage I;

[0080] Examples of metabolic pattern switching prediction and feeding adjustment: Metabolic transition points were predicted using a time-series prediction network, and smooth transition adjustments were implemented. Table 4 shows a comparison between the predicted metabolic transition points and the actual observed transition points of an experimental pig during a 12-week feeding period.

[0081] Table 4: Evaluation of the predictive effectiveness of metabolic state transition points;

[0082] Technical effectiveness verification: Effects of improved feed utilization: The 12-week trial showed significant differences in feed utilization between the experimental and control groups. Table 5 shows the comparison of feed conversion ratio (FCR) between the two groups of pigs at different stages.

[0083] Table 5: Comparison of feed conversion rates between the experimental group and the control group;

[0084] Growth cycle shortening effect: At the end of the experiment, there was a significant difference in the time it took for the control group and the experimental group to reach the target slaughter weight (120kg). Table 6 shows the comparison of growth performance between the two groups of pigs.

[0085] Table 6: Comparison of growth performance between the experimental group and the control group;

[0086] The embodiments of this application have been described above, but these embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments based on the guidance of these embodiments, and all of them are within the protection scope of these embodiments.

Claims

1. A livestock management method based on intelligent data analysis, characterized by, The method comprises the following steps: Collecting basic data, metabolic characteristics and feed ingredient data of a target individual; Determining a metabolic stage currently occupied by the target individual according to the basic data, metabolic characteristics and feed ingredient data of the target individual; wherein the target individual has a plurality of corresponding metabolic stages; each metabolic stage has a corresponding metabolic state; Determining a key nutritional bottleneck and a key limiting factor corresponding to the target individual in the current metabolic stage according to a key metabolic pathway of the metabolic stage and a nutrient interaction matrix; Generating an optimized formula corresponding to the target individual in the current metabolic stage according to the key nutritional bottleneck and the key limiting factor; According to the metabolic state of the target individual in the current metabolic stage and the corresponding optimized formula, predicting a time point at which the target individual is converted from the current metabolic stage to a next metabolic stage, and generating a feeding scheme corresponding to the next metabolic stage.

2. The smart data analysis based livestock management method as claimed in claim 1, wherein, According to the basic data, metabolic characteristics and feed ingredient data of the target individual, the metabolic stage currently occupied by the target individual is determined, comprising: According to the basic data, metabolic characteristics and feed ingredient data of the target individual, a metabolic pathway graph of the target individual is constructed; wherein the metabolic pathway graph comprises nodes and edges; the nodes are metabolites; the edges are conversion relationships between metabolites; each edge has a corresponding reaction rate; According to the metabolic characteristics of each metabolic stage, a multi-stage metabolic model library is constructed; According to the feed ingredient data, a relationship model between spectral characteristics and the nutritional ingredient content of the feed is obtained; According to the hidden Markov model, the metabolic stage currently occupied by the target individual is determined.

3. The smart data analysis based livestock management method as claimed in claim 1, wherein, According to the key metabolic pathway of the metabolic stage currently occupied by the target individual and the nutrient interaction matrix, the key nutritional bottleneck and the key limiting factor corresponding to the target individual in the current metabolic stage are determined, comprising: According to the metabolic state of the metabolic stage currently occupied by the target individual, the key metabolic pathway is identified; The utilization efficiency of each nutrient in the current metabolic stage is analyzed; According to the influence degree between each nutrient, a nutrient interaction matrix is obtained; According to the characteristic values and eigenvectors of the key metabolic pathway, the utilization efficiency and the nutrient interaction matrix, the key nutritional bottleneck and the key limiting factor corresponding to the target individual in the current metabolic stage are determined.

4. The smart data analysis based livestock management method as claimed in claim 1, wherein, According to the key nutritional bottleneck and the key limiting factor, the optimized formula corresponding to the target individual in the current metabolic stage is generated, comprising: According to the key nutritional bottleneck and the key limiting factor, a multi-objective optimization model is obtained; According to the improved NSGA-III algorithm, a multi-objective optimization problem corresponding to the multi-objective optimization model is solved; wherein the improved NSGA-III algorithm introduces adaptive crossover and mutation operators, and a solution evaluation mechanism based on the metabolic model; According to the nutrient efficiency maximization model, the optimized formula corresponding to the target individual in the current metabolic stage is obtained.

5. The smart data analysis based livestock management method as claimed in claim 4, wherein, The nutrient efficiency maximization model comprises: A nutrient metabolic pathway module: used for modeling the metabolic conversion process of each nutrient based on the metabolic pathway graph, and calculating the utilization efficiency under a given metabolic state; A nutrient interaction adjustment module: used for considering the synergistic and antagonistic relationship between nutrients, and adjusting the initial efficiency; A multi-stage metabolic transformation prediction module is configured to predict the transformation path and efficiency of nutrients in vivo in combination with the metabolic state and formula ingredients corresponding to the current stage; An individualized parameter adaptation module is configured to adjust the model parameters of the nutrition efficiency maximization model de1 according to the individual characteristics of the animal; A constraint condition processing module is configured to process the nonlinear constraint conditions by using a penalty function method.

6. The smart data analysis based livestock management method as claimed in claim 1, wherein, The method further comprises: According to the historical growth data and the metabolic state corresponding to the current metabolic stage, the conversion time point of the next metabolic stage is predicted; According to the metabolic state corresponding to the current metabolic stage and the nutrition intake, the future growth trajectory is predicted; Within a preset time period after the predicted conversion time point of the next metabolic stage, a transition feeding formula is determined according to a smooth transition strategy; After the preset time period after the predicted conversion time point of the next metabolic stage, the feeding scheme corresponding to the next metabolic stage is adjusted. 7.The smart data analysis based livestock management method of claim 1, wherein, The method further comprises: According to the actual feeding effect corresponding to the feeding scheme, the model parameters of the nutrition efficiency maximization model are adaptively updated.

8. A livestock management system based on intelligent data analysis, characterized by, The method comprises: A collection module is configured to collect the basic data, metabolic characteristics and feed ingredient data of the target individual; A stage determination module is configured to determine the metabolic stage currently occupied by the target individual according to the basic data, metabolic characteristics and feed ingredient data of the target individual; wherein the target individual has a plurality of corresponding metabolic stages; each metabolic stage has a corresponding metabolic state; A factor determination module is configured to determine the key nutritional bottleneck and key limiting factor corresponding to the current metabolic stage of the target individual according to the key metabolic path of the current metabolic stage of the target individual and the nutrient interaction matrix; A formula generation module is configured to generate an optimized formula corresponding to the current metabolic stage of the target individual according to the key nutritional bottleneck and key limiting factor; A conversion time determination module is configured to predict the time point at which the target individual converts from the current metabolic stage to the next metabolic stage according to the metabolic state of the target individual in the current metabolic stage and the corresponding optimized formula, and generate a feeding scheme corresponding to the next metabolic stage.

Images