CN-121970691-A - Meat sheep quantitative intelligent feeding management system based on Internet of things

CN121970691ACN 121970691 ACN121970691 ACN 121970691ACN-121970691-A

Abstract

The application relates to the technical field of intelligent livestock raising and discloses a quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which comprises a multi-mode sensing module, an edge computing gateway, a metabolic efficiency feedback factor computing unit, an individual nutrition digital twin modeling unit, an active disturbance response verification mechanism executing unit, an intelligent feeding terminal and a group knowledge evolution module. The system adopts an individual nutrition digital twin modeling unit which is combined with an online reinforcement learning algorithm to generate an accurate feeding decision according to a metabolic efficiency feedback factor and a nutrition elastic coefficient matrix, and an active disturbance response verification mechanism execution unit applies controllable micro disturbance when a model is uncertain to dynamically update the elastic coefficient matrix. The application utilizes the multi-mode data to sense the physiological state of the individual in real time, realizes the dynamic closed-loop control based on the metabolic efficiency, solves the cold start problem of the new individual through knowledge distillation, and remarkably improves the feed conversion efficiency and the culture benefit.

Inventors

ZHANG HUANYU
SUN LI
ZHAO XINJIAN
JIA LINYA
WANG LINJUAN
LU GANG
LI MINGXIA

Assignees

驻马店钲沅牧业有限公司

Dates

Publication Date: 20260505
Application Date: 20251128

Claims (10)

1. Mutton sheep quantification intelligent feeding management system based on thing networking, its characterized in that includes: The multi-mode sensing module (100) is used for collecting weight data, rumen region thermal imaging data, rumination acoustic signals, activity behavior data and environmental parameters of each mutton sheep individual; an edge computing gateway (200) for time synchronizing and binding the data collected by the multi-mode sensing module (100) according to the individual electronic identifier; A metabolic efficiency feedback factor calculation unit (300) for calculating a metabolic efficiency feedback factor of the individual based on the corrected weight gain and the total feed nutrient intake amount of the individual within a preset observation window; an individual nutrition digital twin modeling unit (400) for constructing a state vector containing static livestock parameters, dynamic physiological characteristics, environment covariates and nutrition elastic coefficient matrixes for each mutton sheep, and generating the total feeding amount and the feed component proportion of the next feeding period through online reinforcement learning based on metabolic efficiency feedback factors; The active disturbance response verification mechanism execution unit (500) is used for applying controllable micro disturbance to the feed component proportion when the uncertainty of the individual nutrition digital twin body output by the individual nutrition digital twin body modeling unit (400) exceeds a preset threshold value, and updating a nutrition elastic coefficient matrix according to the multimodal response acquired after the disturbance; An intelligent feeding terminal (600) for receiving the ratio of the total feeding amount to the feed components and performing quantitative mixing and feeding when the corresponding individual is identified as approaching the trough; and the group knowledge evolution module (700) is used for aggregating the nutrition digital twin states of all individuals, constructing a group metabolism phenotype library and supporting the initialization of cold start parameters of newly-entered individuals.
2. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things according to claim 1, wherein the metabolic efficiency feedback factor calculating unit (300) is characterized in that the corrected weight gain is obtained by performing inhibition correction on the original weight gain through an environment temperature-humidity index, the total feed nutrition intake amount is a weighted sum of each feed ingredient intake amount and a corresponding standardized nutrition equivalent, the metabolic efficiency feedback factor is defined as a ratio of the corrected weight gain to the total feed nutrition intake amount, and when the total feed nutrition intake amount is zero, the metabolic efficiency feedback factor of the last observation window is used.
3. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which is characterized in that in the individual nutrition digital twin modeling unit (400), dynamic physiological characteristics comprise metabolic efficiency feedback factors, ruminant strength indexes extracted from ruminant acoustic signals through a one-dimensional convolutional neural network and a transducer mixed model, temperature fluctuation variance of rumen region thermal imaging data, lying time length proportion and activity energy consumption estimated values, and a nutrition elasticity coefficient matrix is used for representing local sensitivity of the metabolic efficiency feedback factors to the proportion of each feed component.
4. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which is characterized in that a reward function adopted by on-line reinforcement learning in the individual nutrition digital twin modeling unit (400) is formed by weighting a metabolic efficiency feedback factor and the unit cost of a current feed formula, and the proportion of feed components output by on-line reinforcement learning meets the constraint condition that the sum is 1 and is not negative.
5. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which is disclosed by claim 1, is characterized in that the amplitude of controllable micro-disturbance does not exceed a preset disturbance upper limit in an execution unit (500) of the active disturbance response verification mechanism, the controllable micro-disturbance only acts on a single feed component, the proportions of the rest feed components are scaled and normalized, and the multi-modal response comprises the change rate of the metabolic efficiency feedback factor acquired in an observation window after disturbance and the ruminant strength index offset.
6. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things according to claim 5, wherein the active disturbance response verification mechanism execution unit (500) updates a nutrition elasticity coefficient matrix in an exponential smoothing mode, calculates an elasticity coefficient value at the next moment by carrying out weighted combination on the elasticity coefficient value at the previous moment and the ratio of the metabolic efficiency feedback factor variable caused by disturbance to the disturbance quantity, and the weighted combination is controlled by a preset learning rate.
7. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which is characterized in that the multi-mode sensing module (100) comprises a directional microphone array, a thermal infrared imager, an inertial measurement unit collar and an environmental sensor, wherein the directional microphone array is used for collecting ruminant acoustic signals, the thermal infrared imager is used for acquiring rumen area thermal imaging data, the inertial measurement unit collar is worn on the neck of an individual and used for collecting activity behavior data, and the environmental sensor is used for collecting environmental parameters, and the environmental parameters comprise environmental temperature, relative humidity and ammonia concentration.
8. The quantitative intelligent feeding management system for mutton sheep based on the internet of things according to claim 1, wherein the intelligent feeding terminal (600) confirms the electronic identifier of an individual approaching a feeding trough through a radio frequency identification or visual identification mode, and releases exclusive mixed feed which corresponds to the proportion of the total feeding amount to the feed components only for the identified individual.
9. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which is disclosed by claim 1, is characterized in that the group knowledge evolution module (700) is used for completing the initialization of cold start parameters based on initial ruminant strength index and early metabolic efficiency feedback factor trend of an individual when the new individual is in a column, matching similar historical individuals from a group metabolism phenotype library, and migrating partial prior parameters of a nutrition digital twin of the similar historical individuals through light knowledge distillation.
10. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which is disclosed in claim 1, is characterized in that the edge computing gateway (200) is communicated with the intelligent feeding terminal (600) and the multi-mode sensing module (100) through a low-power-consumption wide area network or a local wireless network, and all individual electronic identifiers are kept unique in the whole system and penetrate through the whole processes of data acquisition, modeling, decision making and execution.

Description

Meat sheep quantitative intelligent feeding management system based on Internet of things Technical Field The invention relates to the technical field of intelligent livestock raising, in particular to a quantitative intelligent feeding management system for mutton sheep based on the Internet of things. Background In the field of mutton sheep cultivation, the method realizes accurate feeding to improve feed conversion rate, optimize growth performance and ensure individual health, and is one of the core targets of modern animal husbandry. Existing mutton sheep feeding management practices rely to a large extent on feeding criteria based on population averages or growth stages. These methods typically group flocks according to static parameters such as age of day or average weight, and apply a uniform feed formulation and feed rate to the entire flock. This population-based management approach is technically difficult to take into account the significant physiological and metabolic differences between individual mutton sheep. In practical application, the unified standard can lead to insufficient feeding of partial individuals with high metabolic efficiency and limit the growth potential of the individuals, and the excessive feeding of the feed is caused for other individuals with low metabolic efficiency, so that the feed waste is caused, and the cultivation cost is increased. While some automated feeding systems have attempted to achieve individual identification and quantitative feeding through electronic identification, the strategy model upon which the feeding decision is based is typically static or open-loop. These systems lack an effective closed-loop feedback mechanism that cannot adjust the composition and total amount of feed in real time based on each individual's actual, dynamically changing energy conversion efficiency. In addition, the existing technical schemes are generally unable to effectively cope with the problem that the individual biological model drifts with time, namely, the prediction accuracy of the model can be reduced along with the change of the individual physiological state, and an online self-adaptive verification and calibration mechanism is lacked. Meanwhile, existing systems also present challenges in how to quickly and accurately set initial personalized feeding parameters when new individuals enter the rail due to lack of historical data. Disclosure of Invention Aiming at the defects of the prior art, the invention provides the quantitative intelligent feeding management system for mutton sheep based on the Internet of things, which solves the problems that in the prior art, the metabolic difference of individual mutton sheep breeding is difficult to quantify, the feeding strategy lacks dynamic adaptability and the cold start is difficult. The quantitative intelligent feeding management system for mutton sheep based on the Internet of things comprises a multi-mode sensing module, an edge computing gateway, a metabolic efficiency feedback factor computing unit, an individual nutrition digital twin modeling unit, an active disturbance response verification mechanism executing unit, an intelligent feeding terminal and a group knowledge evolution module. Preferably, the multi-modal sensing module is configured to collect weight data, rumen area thermal imaging data, ruminal acoustic signals, activity behavior data, and environmental parameters for each individual mutton sheep. The environmental parameters include ambient temperature, relative humidity, and ammonia concentration. The edge computing gateway is configured to receive the data collected by the multi-mode sensing module and perform time synchronization and binding processing on the multi-source data according to the individual electronic identifier. The electronic identifier remains unique throughout the system's data acquisition, modeling, decision making and execution. The communication between the edge computing gateway and the intelligent feeding terminal and between the edge computing gateway and the multi-mode sensing module can be a low-power consumption wide area network or a local wireless network. Preferably, the metabolic efficiency feedback factor calculating unit is configured to receive the synchronization data from the edge calculating gateway and periodically calculate the metabolic efficiency feedback factor of each individual mutton sheep. The calculation process utilizes the environmental temperature and humidity index to carry out inhibition correction on the original weight increment so as to obtain the corrected weight increment, and aims at eliminating the interference of environmental heat stress on the weight increment. The calculation formula of the environment temperature and humidity index is as follows: ; Wherein, the The ambient temperature in degrees celsius,Relative humidity in percent. The total feed nutrient intake is obtained by weighting and calculating the intake of each feed