Search

CN-121981389-A - Cow net value multidimensional data modeling analysis method and system

CN121981389ACN 121981389 ACN121981389 ACN 121981389ACN-121981389-A

Abstract

The invention provides a method and a system for modeling and analyzing cow net value multidimensional data, which relate to the technical field of data processing, and the method comprises the following steps: step 1, acquiring time sequence data of individual identification, feeding event, lactation behavior and static lying posture of the dairy cows through a feeding channel inlet, a milking stall outlet and an acquisition point in the center of a bedridden area of the dairy cows, and forming a dairy cow behavior production original data set. According to the method, through full-dimensional processing and three-dimensional feature space modeling analysis of time sequence data of dairy cow production behaviors, accurate partitioning of dairy cow production modes and prediction of individual milk production efficiency and health level are achieved, production parameter adjustment requirements are quantified by means of net value compensation coefficients, data support is provided for pasture formulation targeted differentiated feeding, environment and health management measures, the fine management level of the pasture is effectively improved, milk production efficiency of dairy cows is optimized, and health states of the dairy cows are guaranteed.

Inventors

  • DONG FEI
  • ZHAO HUIQIU
  • LIU YUQUAN
  • XU WEI
  • MA XIUGUO
  • LI ZHIJIA

Assignees

  • 一牧科技（北京）有限公司

Dates

Publication Date
20260505
Application Date
20260123

Claims (10)

  1. 1. A method for modeling and analyzing net value multi-dimensional data of dairy cows, which is characterized by comprising the following steps: step 1, acquiring time sequence data of individual identification, feeding event, lactation behavior and static lying posture of a dairy cow through a dairy cow feeding channel inlet, a milking stall outlet and a collecting point in the center of a bedridden area to form a dairy cow behavior production original data set; Step 2, performing time alignment, abnormal data cleaning and feature extraction processing on the original data set, respectively calculating the net ingestion value, the net lactation value and the net behavior value of each acquisition point, and integrating the net ingestion value, the net lactation value and the net behavior value into a cow net feature set in a unified format; Step 3, based on the cow net value feature set, constructing a three-dimensional feature space by taking the feeding net value, the lactation net value and the behavior net value as coordinate axes, and based on the density distribution of data points in the cow net value feature set, constructing a space geometric partition plane capable of effectively separating data point groups with different distribution forms by calculating the discrete curvature and normal vector of each local data point group, Step 4, forming a plurality of characteristic partitions reflecting different dairy cow production modes on the space geometric partition surface through cluster analysis; Step 5, based on the characteristic partitions, a dairy cow net value comprehensive analysis model is established, the mapping relation between the dairy cow net value and the production performance index in each characteristic partition is determined, and a dairy cow net value efficiency prediction model is established based on the mapping relation; and 6, predicting the milk production efficiency and the health level of the dairy cows in each characteristic partition by using a dairy cow net value efficiency prediction model, calculating net value compensation coefficients and corresponding production parameter adjustment amounts of each characteristic partition, and making differentiated feeding, environment and health management measures.
  2. 2. The method for modeling analysis of net worth multidimensional data of dairy cows according to claim 1, wherein step2 comprises: The time sequence data of all individual cow identifications, feeding events, lactation behaviors and static and horizontal postures in the cow behavior production original data set are unified to the same standard time standard, and a time-aligned cow behavior production data set is formed; Screening a time-aligned dairy cow behavior production data set based on a preset dairy cow physiological threshold range, deleting abnormal data exceeding the physiological threshold range, finishing cleaning the abnormal data, and respectively performing feature extraction processing on the cleaned data set to obtain a ingestion net value, a lactation net value and a behavior net value; and aligning and combining the feeding net value, the lactation net value and the behavior net value according to the same dairy cow individual to form a dairy cow net value characteristic set in a unified format taking the dairy cow individual as a basic unit.
  3. 3. The method for modeling analysis of net worth multidimensional data of dairy cows of claim 2, wherein the feature extraction process comprises: the method comprises the steps of extracting feed intake variation values in a unit time period from data corresponding to acquisition points at the inlet of a milk cow feeding channel to serve as feeding net values, extracting lactation variation values in the unit time period from data corresponding to acquisition points at the outlet of a milking stall to serve as lactation net values, and extracting effective resting time duration variation values in the unit time period from data corresponding to acquisition points at the center of a bedridden area to serve as behavior net values.
  4. 4. The method for modeling analysis of net worth multidimensional data of dairy cows according to claim 3, wherein step 3 comprises: Mapping the feeding net value, the lactation net value and the behavior net value of each individual data point of the dairy cow in the dairy cow net value characteristic set to coordinate values on corresponding coordinate axes in a three-dimensional characteristic space respectively; in the three-dimensional feature space, based on the density distribution of data points in the cow net value feature set, identifying a plurality of local data point groups, calculating a spatial distribution covariance matrix of the local data point groups aiming at each local data point group, and determining the feature vector direction of the corresponding covariance matrix; based on the direction of the feature vector, constructing a plane which passes through the space distribution center of the corresponding point group and is orthogonal to the direction of the maximum variance, calculating the distribution discrete curvature of the space points of the corresponding point group to the plane, and taking the normal vector of the plane as the normal vector of the corresponding point group; Based on the discrete curvature and normal vector of all local data point groups, constructing a continuous scalar field function describing the data point distribution state in the three-dimensional feature space, forming an energy functional by the gradient square of the scalar field function and a potential energy item formed by the discrete curvature and normal vector space change constraint, driving the state distribution of the scalar field function to stably converge by iteratively updating the scalar field function to minimize the energy functional, and dynamically generating a space geometric division plane capable of effectively separating the data point groups with different distribution forms according to the equivalent surface of the converged scalar field function.
  5. 5. The method of modeling analysis of net worth multidimensional data of dairy cows of claim 4, wherein step 4 comprises: on the space geometric division plane, all the data points mapped to the space geometric division plane are calculated to obtain Euclidean distances between the data points according to the two-dimensional projection coordinates of the data points on the space geometric division plane; Based on Euclidean distance, data points with Euclidean distance smaller than a distance threshold value in a preset class are classified into the same group through an iterative process, the Euclidean distance of any data point between different groups is ensured to be larger than the distance threshold value between the preset classes until all the data points are distributed to one group and the group structure is not changed any more in continuous iteration, each finally formed data point group is defined as a characteristic partition, and each characteristic partition corresponds to a dairy cow production mode characterized by the combination of the net feeding value, the net lactation value and the net behavior value of the data points in the partition.
  6. 6. The method for modeling analysis of net worth multidimensional data of dairy cows according to claim 5, wherein step 5 comprises: for each characteristic partition, extracting feeding net value, lactation net value and behavior net value corresponding to all individual data of dairy cows in the characteristic partition to form an input variable set, simultaneously acquiring daily milk yield data corresponding to the individual dairy cows and recorded by a milking monitoring terminal as a milk yield efficiency index, and taking resting behavior data recorded by a lying monitoring terminal as a health level index to form an output variable set; based on the input variable set and the output variable set, a nonlinear regression method which fuses key parameters in a cow propagation physiological rule and a lactation curve standard equation is used for fitting to obtain a multi-element nonlinear mapping function, wherein the multi-element nonlinear mapping function is a cow net value comprehensive analysis model of a corresponding characteristic partition, and the multi-element nonlinear mapping function defines a mapping relation from a feeding net value, a lactation net value, a behavior net value to a milk production efficiency index and a health level index; Based on a multi-element nonlinear mapping function defined by a milk cow net value comprehensive analysis model, a calculation framework is constructed, the calculation framework takes the feeding net value, the lactation net value and the behavior net value as input data, and the predicted milk production efficiency index value and the health level index value are directly output through the calculation of the multi-element nonlinear mapping function, and the calculation framework is the milk cow net value efficiency prediction model corresponding to the characteristic partition.
  7. 7. The method of claim 6, wherein the fitting, based on the input variable set and the output variable set, by using a nonlinear regression method that fuses key parameters in a physiological law of cow reproduction with a standard equation of a lactation curve to obtain a multiple nonlinear mapping function, wherein the multiple nonlinear mapping function is a comprehensive analysis model of cow net values of corresponding feature partitions, and defines a mapping relationship from net ingestion values, net lactation values, net behavior values to milk production efficiency indexes and health level indexes, and the method comprises: Taking the net ingestion value, the net lactation value and the net behavior value in the input variable set as independent variables, taking the milk production efficiency index and the health level index in the output variable set as dependent variables, constructing an initial nonlinear function form which takes the fetal number and the pregnancy state parameter in the physiological rule of cow propagation as adjustment factors and takes the lactation curve standard equation as a basic function structure, optimizing the undetermined coefficient in the initial nonlinear function form by adopting a least square method by substituting the input variable set and the output variable set into the initial nonlinear function form, enabling the sum of squares of residual errors between the function calculated value and the actual output variable set data to be minimum, determining the final function coefficient, and completing fitting of the multi-element nonlinear mapping function.
  8. 8. The method of modeling analysis of net worth multidimensional data of dairy cows of claim 7, wherein step 6 comprises: inputting the current feeding net value, lactation net value and behavior net value of each dairy cow individual in each characteristic partition into a dairy cow net value efficacy prediction model corresponding to the characteristic partition, and obtaining a predicted milk production efficiency index value and a predicted health level index value corresponding to each dairy cow individual by executing defined multi-element nonlinear mapping function calculation; Calculating arithmetic average of predicted milk yield efficiency index values of all dairy cow individuals in each characteristic partition as a partition milk yield efficiency prediction average value, calculating arithmetic average of predicted health level index values of all dairy cow individuals as a partition health level prediction average value, dividing the partition milk yield efficiency prediction average value by a preset milk yield efficiency reference value, wherein the quotient value is defined as a milk yield efficiency net value compensation coefficient of the characteristic partition; According to the milk production efficiency net value compensation coefficient and the health level net value compensation coefficient, determining a daily feeding total amount adjustment parameter, a cowshed ventilation and temperature control parameter and a health inspection frequency adjustment parameter corresponding to the characteristic partition by referring to a preset mapping relation between the compensation coefficient and the production parameter, and generating a differential measure set comprising a specific feeding scheme, an environment control set value and a health management rule for each characteristic partition based on the daily feeding total amount adjustment parameter, the cowshed ventilation and temperature control parameter and the health inspection frequency adjustment parameter.
  9. 9. A dairy cow net worth multidimensional data modeling analysis system implementing the method of any of claims 1 to 8, comprising: The data acquisition module is used for acquiring time sequence data of individual identification, feeding event, lactation behavior and static lying posture of the dairy cows through a dairy cow feeding channel inlet, a dairy cow milking stall outlet and an acquisition point in the center of a bedridden area to form a dairy cow behavior production original data set; The feature extraction module is used for carrying out time alignment, abnormal data cleaning and feature extraction processing on the original data set, respectively calculating the feeding net value, the lactation net value and the behavior net value of each acquisition point, and integrating the feeding net value, the lactation net value and the behavior net value into a cow net value feature set in a unified format; A space construction module for constructing a three-dimensional characteristic space based on the cow net value characteristic set by taking the feeding net value, the lactation net value and the behavior net value as coordinate axes, constructing a space geometric division plane capable of effectively separating data point groups with different distribution forms by calculating the discrete curvature and normal vector of each local data point group based on the density distribution of data points in the cow net value characteristic set, The characteristic partition module is used for forming a plurality of characteristic partitions reflecting different dairy cow production modes on the space geometric partition surface through cluster analysis; The model construction module is used for establishing a dairy cow net value comprehensive analysis model based on the characteristic partitions, determining the mapping relation between the dairy cow net value and the production performance index in each characteristic partition, and establishing a dairy cow net value efficiency prediction model based on the mapping relation; And the regulation and control decision module is used for predicting the milk production efficiency and the health level of the dairy cows in each characteristic partition by utilizing the dairy cow net value efficiency prediction model, calculating net value compensation coefficients and corresponding production parameter adjustment amounts of the characteristic partitions, and making differentiated feeding, environment and health management measures.
  10. 10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a program which, when executed by a processor, implements the method according to any of claims 1 to 8.

Description

Cow net value multidimensional data modeling analysis method and system Technical Field The invention relates to the technical field of data processing, in particular to a method and a system for modeling and analyzing net value multidimensional data of dairy cows. Background In large-scale cow breeding, in order to realize fine management and increase production benefits, key behaviors such as feeding, lactation and rest of cow individuals are usually required to be monitored and analyzed, in the prior art, sensors are deployed at key points (such as feeding channels and milking parlors) of a farm, so that behavior time sequence data of the cow individuals can be continuously collected, and when analysis is carried out based on the data, one common method is to set independent thresholds or classification standards of each behavior index (such as daily feeding time length, daily milk yield and daily average bedridden time) respectively, or to carry out preliminary classification on cow groups by adopting a two-dimensional analysis method (such as a 'feeding amount-milk yield' scatter diagram) which is related in pairs. For example, the system may identify cows with a feeding duration lower than the threshold A and a milk yield lower than the threshold B as a group and uniformly recommend nutritional intervention, however, the final production performance of cows is the result of complex interweaving and dynamic balancing of multiple dimensional behaviors such as feeding, lactation and rest, and the like, such analysis methods based on independent indexes or simple two-dimensional correlations may not be capable of accurately distinguishing cow subgroups which are similar in performance in a single dimension or two dimensions but have substantial differences in overall behavior mode and production mechanism when three or more core behavior dimensions are processed simultaneously, specifically, the existing methods may face limitation when distinguishing the following two types of cows, supposing cows A and B, the two indexes of daily milk yield and daily bedridden time fall in similar numerical intervals, in two-dimensional analysis, the cows A may be classified into the same class, in fact, the cows A may have a behavior mode of 'high feeding efficiency and full active rest', and the second may be represented as a cow feeding efficiency, but the health condition (such as a passive time condition) is prolonged, the two-dimensional analysis may not be performed based on the same plane, and the two-dimensional analysis may not be similar to the prior art, and the three-dimensional analysis may not be performed on the two-dimensional analysis of the two-dimensional analysis may not be similar to the prior art, and the two-dimensional analysis may not be similar to the two-dimensional analysis. Disclosure of Invention The invention aims to solve the technical problem of providing a method and a system for modeling and analyzing net value multidimensional data of dairy cows, which clearly reflect group core production behavior characteristics corresponding to each characteristic partition, so that pastures can intuitively grasp the behavior pattern differences of different groups. In order to solve the technical problems, the technical scheme of the invention is as follows: In a first aspect, a method for modeling and analyzing net value multidimensional data of dairy cows, the method comprising: step 1, acquiring time sequence data of individual identification, feeding event, lactation behavior and static lying posture of a dairy cow through a dairy cow feeding channel inlet, a milking stall outlet and a collecting point in the center of a bedridden area to form a dairy cow behavior production original data set; Step 2, performing time alignment, abnormal data cleaning and feature extraction processing on the original data set, respectively calculating the net ingestion value, the net lactation value and the net behavior value of each acquisition point, and integrating the net ingestion value, the net lactation value and the net behavior value into a cow net feature set in a unified format; Step 3, based on the cow net value feature set, constructing a three-dimensional feature space by taking the feeding net value, the lactation net value and the behavior net value as coordinate axes, and based on the density distribution of data points in the cow net value feature set, constructing a space geometric partition plane capable of effectively separating data point groups with different distribution forms by calculating the discrete curvature and normal vector of each local data point group, Step 4, forming a plurality of characteristic partitions reflecting different dairy cow production modes on the space geometric partition surface through cluster analysis; Step 5, based on the characteristic partitions, a dairy cow net value comprehensive analysis model is established, the mapping relation between the dairy cow