Method and system for predicting feed nutrient utilization rate of poultry

Abstract

The invention discloses a method and system for predicting the feed nutrient utilization rate of poultry, and relates to the technical field of feed nutrient utilization rate prediction.The method comprises the steps that feature monitoring data in the poultry feed feeding process are collected; performing correlation analysis on the feature monitoring data to obtain a weight distribution result; performing data conversion processing on the corresponding feature monitoring data according to a weight distribution result; taking the converted data as an input value, and inputting the input value into a trained deep neural network integration model to obtain a model output result; and performing data transformation on model output to obtain a final feed nutrient utilization rate. According to the method, a data set subjected to data conversion processing is simultaneously input into a BP neural network model, a double-delay depth deterministic strategy gradient neural network model and an LSTM neural network model of which initial weights and thresholds are optimized through a genetic algorithm, and the input data is analyzed through a plurality of neural network models; and determining a final analysis result according to a majority voting result.

Description

Technical Field

[0001] The invention relates to the technical field of feed nutrient utilization rate prediction, and more particularly to a method and system for predicting feed nutrient utilization rate of poultry. Background Art

[0002] With the continuous increase of population and the improvement of living standards, the demand for livestock and poultry products is gradually increasing. In order to meet the market demand, large-scale farming came into being. Large-scale farming refers to the use of modern scientific and technological means and management methods to raise livestock or poultry on a large scale in a centralized manner to achieve the purpose of improving production efficiency and economic benefits.

[0003] Large-scale poultry farming is one of the main forms of modern livestock and poultry farming. With the acceleration of agricultural modernization and the increase in market demand, many farms are gradually developing towards large-scale development, using modern equipment, technology and management methods for poultry farming. Large-scale poultry farms generally use various automated equipment, such as automatic feeders, automatic waterers, automatic cleaning systems, etc., which can reduce labor costs and improve farming efficiency.

[0004] Since the feed consumption of large-scale poultry farming is far greater than that of individual poultry farmers, existing large-scale poultry farming tends to adopt precise management methods, such as quantitative feeding and timed feeding, and realizes real-time monitoring and management of poultry feed consumption through data analysis and scientific and technological means. However, the existing technology does not have a technical solution to study the absorption and utilization rules of poultry for feed and nutrients, and to formulate more reasonable feed formulas, resulting in a waste of resources in large-scale poultry farming.

[0005] Therefore, how to propose a method and system for predicting the nutrient utilization rate of feed by poultry, achieve accurate prediction of the nutrient utilization rate of feed by poultry, improve breeding efficiency and reduce feed costs is an urgent problem to be solved by technical personnel in this field. Summary of the invention

[0006] In view of this, the present invention provides a method and system for predicting the nutrient utilization rate of poultry feed, so as to achieve accurate prediction of the nutrient utilization rate of poultry feed, improve breeding efficiency and reduce feed costs. In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A method for predicting feed nutrient utilization efficiency of poultry, comprising:

[0008] Collect characteristic monitoring data during poultry feeding;

[0009] Perform correlation analysis on the characteristic monitoring data to obtain the weight distribution result;

[0010] Perform data conversion processing on the corresponding feature monitoring data according to the weight distribution result;

[0011] The converted data is used as input value and input into the trained deep neural network integration model to obtain the model output result;

[0012] The model output is transformed to obtain the final feed nutrient utilization rate.

[0013] Optionally, the characteristic monitoring data include: poultry production performance, respiratory metabolism, energy metabolism and protein deposition characteristic data; poultry production performance characteristic data include dietary nutrients, feed intake, body weight gain and nutrient intake data; respiratory metabolism characteristic data include gas concentration, respiratory entropy, heat production and heat consumption; energy metabolism characteristic data include gross energy intake, metabolizable energy, net energy and precipitated energy; protein deposition characteristic data include edible nitrogen, nitrogen digestibility and deposited nitrogen.

[0014] Optionally, performing correlation analysis on the characteristic monitoring data to obtain a weight distribution result includes:

[0015] Based on grey correlation analysis, the correlation between characteristic monitoring data of poultry breeding process and feed nutrient utilization rate is calculated;

[0016] According to the correlation between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate, a correlation diagram is constructed;

[0017] Combined with the association graph, deep features are obtained by learning the deep dependency between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate through deep feature learning;

[0018] The deep features between the characteristic monitoring data of the poultry farming process and the feed nutrient utilization rate are fused through the attention mechanism, and the attention weights of each characteristic monitoring data are calculated and sorted.

[0019] Optionally, the deep dependency relationship between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate through deep feature learning includes:

[0020] Based on graph convolutional neural networks, the spatial dependency between characteristic monitoring data of poultry farming process and feed nutrient utilization is learned;

[0021] Based on LSTM neural network, the time dependency relationship between characteristic monitoring data of poultry breeding process and feed nutrient utilization rate is learned;

[0022] Based on the attention mechanism, spatial dependency and temporal dependency are fused to obtain the deep dependency between the characteristic monitoring data of the poultry farming process and the feed nutrient utilization rate.

[0023] Optionally, the data conversion process includes:

[0024] Filter the feature data set according to the ranking results of the attention weights of the feature monitoring data;

[0025] The Min-Max normalization algorithm is used to standardize the feature data set;

[0026] The PCA principal component analysis algorithm is used to reduce the dimension of the standardized feature data set;

[0027] Data segmentation is used to segment the feature data set into data with a window length of L.

[0028] Optionally, the Min-Max normalization algorithm formula is as follows:

[0029] Among them, X is the feature data, X min is the minimum value of the feature data, X max is the maximum value of the feature data, and X' is the standardized feature data.

[0030] Optionally, the step of inputting the converted data as input values ​​into a trained deep neural network integration model includes: inputting the data set after data conversion processing into a trained deep neural network integration model to obtain model output, analyzing the input data through multiple neural network models, and determining the final analysis result according to the result of majority voting.

[0031] Optionally, analyzing the input data through multiple network models includes:

[0032] The data sets after data conversion are simultaneously input into the BP neural network model, the double-delay deep deterministic policy gradient neural network model and the LSTM neural network model whose initial weights and thresholds are optimized by genetic algorithm. The input data are analyzed by multiple neural network models, and the final analysis result is determined according to the result of majority voting.

[0033] Optionally, the step of performing data transformation on the model output to obtain the final feed nutrient utilization rate includes:

[0034] Denormalize the model output:

[0035] X=X'×(X max -X min )+X max ; Restore the feature data X' after Min-Max standardization processing to the original feature data X.

[0036] Optionally, a system for predicting nutrient utilization of feed by poultry, comprising:

[0037] Collection module: used to collect characteristic monitoring data during the process of feeding poultry feed;

[0038] Correlation analysis module: used to perform correlation analysis on feature monitoring data to obtain weight distribution results;

[0039] Data conversion module: used to perform data conversion processing on the corresponding feature monitoring data according to the weight distribution results;

[0040] Model processing module: used to input the converted data as input value into the trained deep neural network integration model to obtain the model output result;

[0041] Utilization calculation module: used to perform data transformation on the model output to obtain the final feed nutrient utilization rate.

[0042] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method and system for predicting the nutrient utilization rate of feed by poultry, which has the following beneficial effects:

[0043] The present invention proposes a method for predicting the nutrient utilization rate of poultry feed, comprising: collecting characteristic monitoring data during the process of feeding poultry feed; performing correlation analysis on the characteristic monitoring data to obtain weight distribution results; performing data conversion processing on the corresponding characteristic monitoring data according to the weight distribution results; using the converted data as input values, inputting them into a trained deep neural network integration model to obtain model output results; performing data transformation on the model output to obtain the final nutrient utilization rate of feed. The present invention simultaneously inputs the data set after data conversion processing into a BP neural network model, a double-delayed deep deterministic policy gradient neural network model and an LSTM neural network model that optimize the initial weights and thresholds through a genetic algorithm, analyzes the input data through multiple neural network models, and determines the final analysis result according to the result of majority voting. By focusing the influence of various factors on the nutrient utilization rate of feed on the characteristic monitoring data, accurately predicting the nutrient utilization rate of feed in the next time period according to the characteristic monitoring data and the nutrient utilization rate of feed in the historical time period, and providing a prediction trend of the nutrient utilization rate of feed, more accurate energy-saving characteristic monitoring data can be obtained, thereby obtaining a more accurate setting of economic characteristic monitoring data, thereby achieving the purpose of saving feed. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0045] Figure 1 A schematic flow chart of a method for predicting feed nutrient utilization rate by poultry provided by the present invention.

[0046] Figure 2 A system structure framework diagram for predicting feed nutrient utilization rate of poultry provided by the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The embodiment of the present invention discloses a method for predicting the nutrient utilization rate of feed by poultry, such as Figure 1 As shown, including:

[0049] Collect characteristic monitoring data during poultry feeding;

[0050] Perform correlation analysis on the characteristic monitoring data to obtain the weight distribution result;

[0051] Perform data conversion processing on the corresponding feature monitoring data according to the weight distribution result;

[0052] The converted data is used as input value and input into the trained deep neural network integration model to obtain the model output result;

[0053] The model output is transformed to obtain the final feed nutrient utilization rate.

[0054] Furthermore, the characteristic monitoring data include: poultry production performance, respiratory metabolism, energy metabolism and protein deposition characteristic data; poultry production performance characteristic data include dietary nutrients, feed intake, body weight gain and nutrient intake data; respiratory metabolism characteristic data include gas concentration, respiratory entropy, heat production and heat consumption; energy metabolism characteristic data include gross energy intake, metabolizable energy, net energy and precipitated energy; protein deposition characteristic data include edible nitrogen, nitrogen digestibility and deposited nitrogen.

[0055] Furthermore, the weight distribution result obtained by performing correlation analysis on the characteristic monitoring data includes:

[0056] Based on grey correlation analysis, the correlation degree between characteristic monitoring data of poultry breeding process and feed nutrient utilization rate is calculated; the characteristic monitoring data of poultry breeding process and feed nutrient utilization rate are used as indicators of grey correlation analysis to form an input matrix and establish a grey correlation analysis model; after determining the reference sequence and comparison sequence, dimensionless variables and calculating the grey correlation coefficient, the grey correlation degree between the indicators and the grey correlation matrix they constitute are obtained;

[0057] According to the correlation between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate, a correlation diagram is constructed;

[0058] Combined with the association graph, deep features are obtained by learning the deep dependency between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate through deep feature learning;

[0059] The deep features between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate are integrated through the attention mechanism; the attention weight of each characteristic monitoring data in each time period is calculated; first, the information of each characteristic monitoring data and the feed nutrient utilization rate is aggregated and then concatenated with the information of each characteristic monitoring data in the t-th time period, and then passed through a fully connected layer to finally obtain the attention weight of all characteristic monitoring data;

[0060] The deep features between the characteristic monitoring data of the poultry farming process and the feed nutrient utilization rate are fused through the attention mechanism, and the attention weights of each characteristic monitoring data are calculated and sorted.

[0061] Specifically, the correlation analysis of the characteristic monitoring data includes: the gray correlation analysis adopted, the gray correlation analysis algorithm is based on the data samples of each factor, and describes the strength, size and order of the relationship between the factors. First, the reference sequence and the comparison sequence are determined, and the variables are dimensionless, so as to calculate the gray correlation coefficient and gray correlation degree.

[0062] Furthermore, the deep dependency relationship between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate through deep feature learning includes:

[0063] Based on graph convolutional neural networks, the spatial dependency relationship between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate is learned; the input of spatial dependency relationship learning includes time series data consisting of feed nutrient utilization rate and characteristic monitoring data in multiple time periods in the past, the association graph obtained through grey correlation analysis, and the adjacency matrix corresponding to the association graph;

[0064] Based on the LSTM neural network, the time dependency relationship between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate is learned; through the gating mechanism to remember as much effective historical information as possible, each unit of the LSTM can decide which information should be remembered and passed to the next layer, so as to obtain the time dependency relationship between the characteristic monitoring data and the feed nutrient utilization rate;

[0065] Based on the attention mechanism, the spatial dependency and temporal dependency are integrated to obtain the deep dependency between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate; after learning the spatial and temporal features of each time period, the attention mechanism module is added to further integrate the spatiotemporal information obtained from the previous learning. The feed nutrient utilization rate of the next time period is output in combination with the information obtained from the previous spatiotemporal feature learning; after learning the spatiotemporal features and fusing the spatiotemporal features of the mutual influence relationship between the characteristic monitoring data and the feed nutrient utilization rate of each time period, the learning results are connected alternately in time through the LSTM unit and the attention mechanism module corresponding to the time period, and the output of the last LSTM unit is decoded and output through a fully connected layer, and the output result is the final feed nutrient utilization rate.

[0066] Specifically, building an association graph includes:

[0067] The correlation matrix is ​​obtained based on the grey correlation analysis, and the characteristic monitoring data and feed nutrient utilization rate are used as the vertices of the graph. The correlation degree of each vertex is the weight of the edge, and a fully connected undirected graph is established;

[0068] Set a threshold, take fuel consumption as the center, retain the weights greater than the threshold, and delete the weights less than the threshold and above the fourth level edge;

[0069] Based on the vertices of the graph and the retained weights, a correlation graph of the relationship between the characteristic monitoring data and between the characteristic monitoring data and the feed nutrient utilization rate is obtained.

[0070] Furthermore, the data conversion process includes:

[0071] Filter the feature data set according to the ranking results of the attention weights of the feature monitoring data;

[0072] The Min-Max normalization algorithm is used to standardize the feature data set;

[0073] The PCA principal component analysis algorithm is used to reduce the dimension of the standardized feature data set;

[0074] Data segmentation is used to segment the feature data set into data with a window length of L.

[0075] Specifically, the PCA principal component analysis algorithm specifically includes the following steps:

[0076] Input the data set and reduce the dimension of the data set to 5 dimensions;

[0077] Centralize all input samples;

[0078] Calculate the covariance matrix of the data set: form a data matrix based on the standardized data set, and construct the covariance matrix of the sample based on the data matrix:

[0079] Perform eigenvalue decomposition on the covariance matrix of the sample and find the eigenvalues ​​and eigenvectors;

[0080] The eigenvectors corresponding to the five largest eigenvalues ​​are retained to form a projection matrix, and the PCA dimensionality reduction data is output.

[0081] Furthermore, the Min-Max normalization algorithm formula is as follows:

[0082] Among them, X is the feature data, X min is the minimum value of the feature data, X max is the maximum value of the feature data, and X' is the standardized feature data.

[0083] Furthermore, the analyzing the input data through multiple network models includes:

[0084] The data sets after data conversion are simultaneously input into the BP neural network model, the double-delay deep deterministic policy gradient neural network model and the LSTM neural network model whose initial weights and thresholds are optimized by genetic algorithm. The input data are analyzed by multiple neural network models, and the final analysis result is determined according to the result of majority voting.

[0085] Specifically, the data is input into the trained LSTM neural network black box model, and after being calculated by L isomorphic LSTM cells connected at previous and subsequent moments, its output is expressed as:

[0086] P={P1,P2,…,P L};

[0087] P p =LSTM forward (S p , C p-1 , H p-1 );

[0088] Among them, C p-1 and H p-1 are the state and output of the previous LSTM cell, respectively. forward For its forward calculation method.

[0089] Specifically, the dual-delay deep deterministic policy gradient neural network model includes:

[0090]

[0091] Where, G represents the state space, M represents the action space, and N represents the reward function; A(t) represents the real-time data at the current moment, b(t) represents the environmental impact of the real-time data, c(t) represents the relative difference between the real-time data and the predicted data at the current moment, d(t) represents the change rate of the predicted data at the current moment, e(t) represents the predicted data at the current moment, and f(t) represents the real-time data at the current moment; h(t) represents the change rate of the real-time data at the current moment; represents the weight factor of the real-time feed cost energy consumption reward function per unit time, τ represents the weight factor of the relative difference energy consumption reward function between real-time data and predicted data, ω represents the weight factor of the real-time data energy consumption reward function per unit time, θ represents the weight factor of the predicted data energy consumption reward function per unit time, j(t) is the real-time feed cost energy consumption reward function per unit time, k(t) is the relative difference energy consumption reward function between real-time data and predicted data per unit time, l(t) is the real-time data energy consumption reward function per unit time, and p(t) is the predicted data energy consumption reward function per unit time.

[0092] Specifically, the energy consumption reward function of the relative difference between real-time data and predicted data includes:

[0093]

[0094] Among them, d max d min They respectively represent the maximum relative difference and the minimum relative difference between real-time data and predicted data.

[0095] The real-time data energy consumption reward function per unit time includes:

[0096]

[0097] where v max and v min They respectively represent the maximum and minimum values ​​that real-time data can reach.

[0098] The reward function for predicting data energy consumption per unit time includes:

[0099]

[0100] where d max and d min Respectively represent the maximum and minimum values ​​that the predicted data can reach

[0101] Specifically, BP neural network has strong data fitting prediction ability and self-adaptation. The whole training process includes two processes: forward data transmission and error feedback correction. When training BP neural network, data is transmitted to the hidden layer through the input layer according to the weight, and finally acts on the output layer after nonlinear processing. Therefore, nonlinear BP neural network is selected to construct nonlinear feed nutrient utilization prediction problem.

[0102] In a specific implementation, the method of inputting the converted data as an input value into a trained deep neural network integration model includes: inputting the data set after data conversion processing into the trained deep neural network integration model to obtain a model output, analyzing the input data through multiple neural network models, and determining the final analysis result according to the result of majority voting.

[0103] Furthermore, the data transformation of the model output to obtain the final feed nutrient utilization rate includes:

[0104] Denormalize the model output:

[0105] X=X'×(X max -X min )+X max ; Restore the feature data X' after Min-Max standardization processing to the original feature data X.

[0106] In a specific embodiment, a system for predicting nutrient utilization of feed by poultry, such as Figure 2 As shown, including:

[0107] Collection module: used to collect characteristic monitoring data during the process of feeding poultry feed;

[0108] Correlation analysis module: used to perform correlation analysis on feature monitoring data to obtain weight distribution results;

[0109] Data conversion module: used to perform data conversion processing on the corresponding feature monitoring data according to the weight distribution results;

[0110] Model processing module: used to input the converted data as input value into the trained deep neural network integration model to obtain the model output result;

[0111] Utilization calculation module: used to perform data transformation on the model output to obtain the final feed nutrient utilization rate.

[0112] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0113] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one 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 present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the nutrient utilization rate of feed by poultry, characterized in that: include: Collect characteristic monitoring data during poultry feeding; Perform correlation analysis on the characteristic monitoring data to obtain the weight distribution result; Perform data conversion processing on the corresponding feature monitoring data according to the weight distribution result; The converted data is used as input value and input into the trained deep neural network integration model to obtain the model output result; The model output is transformed to obtain the final feed nutrient utilization rate.

2. A method for predicting nutrient utilization of feed by poultry according to claim 1, characterized in that: The characteristic monitoring data include: poultry production performance, respiratory metabolism, energy metabolism and protein deposition characteristic data; poultry production performance characteristic data include daily diet nutrients, feed intake, body weight gain and nutrient intake data; respiratory metabolism characteristic data include gas concentration, respiratory entropy, heat production and heat consumption; energy metabolism characteristic data include gross energy intake, metabolizable energy, net energy and precipitated energy; protein deposition characteristic data include edible nitrogen, nitrogen digestibility and deposited nitrogen.

3. A method for predicting feed nutrient utilization rate of poultry according to claim 1, characterized in that: The weight distribution result obtained by performing correlation analysis on the characteristic monitoring data includes: Based on grey correlation analysis, the correlation between characteristic monitoring data of poultry breeding process and feed nutrient utilization rate is calculated; According to the correlation between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate, a correlation diagram is constructed; Combined with the association graph, deep features are obtained by learning the deep dependency between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate through deep feature learning; The deep features between the characteristic monitoring data of the poultry farming process and the feed nutrient utilization rate are fused through the attention mechanism, and the attention weights of each characteristic monitoring data are calculated and sorted.

4. A method for predicting feed nutrient utilization rate of poultry according to claim 3, characterized in that: The deep dependency relationship between the characteristic monitoring data of the poultry breeding process and the feed nutrient utilization rate through deep feature learning includes: Based on graph convolutional neural networks, the spatial dependency between characteristic monitoring data of poultry farming process and feed nutrient utilization is learned; Based on LSTM neural network, the time dependency relationship between characteristic monitoring data of poultry breeding process and feed nutrient utilization rate is learned; Based on the attention mechanism, spatial dependency and temporal dependency are fused to obtain the deep dependency between the characteristic monitoring data of the poultry farming process and the feed nutrient utilization rate.

5. The method for predicting nutrient utilization of feed by poultry according to claim 1, characterized in that: The data conversion process includes: Filter the feature data set according to the ranking results of the attention weights of the feature monitoring data; The Min-Max normalization algorithm is used to standardize the feature data set; The PCA principal component analysis algorithm is used to reduce the dimension of the standardized feature data set; Data segmentation is used to segment the feature data set into data with a window length of L.

6. A method for predicting feed nutrient utilization rate of poultry according to claim 5, characterized in that: The formula of Min-Max normalization algorithm is as follows: Among them, X is the feature data, X min is the minimum value of the feature data, X max is the maximum value of the feature data, and X' is the standardized feature data.

7. A method for predicting feed nutrient utilization rate of poultry according to claim 1, characterized in that: The method of inputting the converted data as input value into the trained deep neural network integration model includes: inputting the data set after data conversion processing into the trained deep neural network integration model to obtain model output, analyzing the input data through multiple neural network models, and determining the final analysis result according to the result of majority voting.

8. A method for predicting feed nutrient utilization rate of poultry according to claim 7, characterized in that: The analyzing of the input data by using multiple network models includes: The data sets after data conversion are simultaneously input into the BP neural network model, the double-delay deep deterministic policy gradient neural network model and the LSTM neural network model whose initial weights and thresholds are optimized by genetic algorithm. The input data are analyzed by multiple neural network models, and the final analysis result is determined according to the result of majority voting.

9. A method for predicting feed nutrient utilization rate of poultry according to claim 1, characterized in that: The data transformation of the model output to obtain the final feed nutrient utilization rate includes: Denormalize the model output: X=X'×(X max -X min )+X max ; Restore the feature data X' after Min-Max standardization processing to the original feature data X.

10. A system for predicting the nutrient utilization rate of feed by poultry, characterized in that: include: Collection module: used to collect characteristic monitoring data during the process of feeding poultry feed; Correlation analysis module: used to perform correlation analysis on feature monitoring data to obtain weight distribution results; Data conversion module: used to perform data conversion processing on the corresponding feature monitoring data according to the weight distribution results; Model processing module: used to input the converted data as input value into the trained deep neural network integration model to obtain the model output result; Utilization calculation module: used to perform data transformation on the model output to obtain the final feed nutrient utilization rate.

Images