Intelligent feeding algorithm-based accurate nutrition supply method for flock of sheep

Abstract

The invention discloses a sheep flock accurate nutrition supply method based on an intelligent feeding algorithm, and the method comprises the following steps: collecting the body type, physiology, movement and environment data of sheep through a multi-modal sensing device, and constructing individual three-dimensional representation based on a neural radiation field model, historical growth data and a real-time health state are analyzed by utilizing an individual nutritional requirement calculation model, an optimal feed distribution strategy is determined by combining an intelligent sheep flock optimization algorithm, a feeding scheme is dynamically adjusted by adopting deep reinforcement learning, and through a cloud-side collaborative calculation architecture, the calculation efficiency is improved and a feeding decision is optimized. A non-intrusive health monitoring technology is combined, a group health evaluation model is constructed, the feed ratio is adjusted in real time, accurate matching of individual and group nutritional requirements is achieved, and through prediction of the individual and group nutritional requirements, self-adaptive adjustment is conducted on feeding parameters, and a dynamic nutrition supply strategy is optimized. The system effectively improves the feeding precision, reduces the feed waste, and optimizes the health management of the sheep flock.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent feeding, and particularly to a precise nutrition supply method for sheep flocks based on an intelligent feeding algorithm. Background Art

[0002] With the development of intelligent agriculture, precise feeding technology has gradually become an important research direction in the livestock industry. During the process of raising sheep flocks, there are significant growth differences and nutritional requirement differences among individuals. Traditional feeding methods often adopt fixed feed ratios and unified feeding strategies, which are difficult to meet the nutritional needs of different individuals, resulting in feed waste and inefficient flock health management. Although existing intelligent feeding technologies have achieved feed optimization and feeding automation to a certain extent, there are still many technical bottlenecks, making it difficult to effectively improve the overall health level of the sheep flock and resource utilization rate.

[0003] In the prior art, common feeding methods mainly rely on empirical formulas, static feed feeding, or simple feeding models based on growth stages. These methods have the following deficiencies in practical applications: 1. Lack of individual precise modeling: Most of the prior art uses nutritional requirement prediction models based on static growth curves, which are difficult to accurately depict the real-time changes in individual body postures, bone structures, and fat distributions, resulting in a mismatch between nutritional supply and individual needs.

[0004] 2. Limitations in optimization strategies: Most existing feeding optimization algorithms are based on fixed rules or linear programming, and do not fully consider individual foraging behaviors, group interactions, and resource competition patterns, making it difficult to achieve a globally optimal feed allocation plan.

[0005] 3. Insufficient dynamic adjustment ability: Traditional feeding systems usually adopt preset feeding parameters and lack the ability of adaptive adjustment. They cannot perform dynamic nutritional adjustment according to the real-time health status, movement characteristics, and environmental factors of the sheep flock, resulting in low feeding accuracy.

[0006] 4. Low efficiency of the computing architecture: Some intelligent feeding systems rely on a centralized computing architecture, and there are high delays in data transmission and processing. It is difficult to complete real-time optimization calculations in a short time, affecting the response speed and accuracy of feeding decisions.

[0007] 5. Limited health monitoring means: Existing health assessment methods mainly rely on manual observation or invasive detection, and it is difficult to achieve non-contact and continuous individual health status assessment, resulting in a lag in the discovery of health problems and the inability to adjust feeding strategies in a timely manner.

[0008] 6. Insufficient nutritional supply prediction ability: The prior art mainly conducts nutritional requirement prediction based on historical data or simple statistical models, lacking long-term trend analysis of the health status of individuals and groups, and it is difficult to optimize feeding parameters to adapt to dynamic growth needs.

[0009] Therefore, how to provide a precise nutritional supplementation method for sheep based on intelligent feeding algorithms is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0010] An object of the present invention is to propose a precise nutritional supplementation method for sheep based on intelligent feeding algorithms. The present invention combines neural radiance field modeling, sheep flock intelligent optimization algorithms, deep reinforcement learning, and non-invasive health monitoring technologies, and details methods for individual precise modeling, dynamic nutritional requirement calculation, feed feeding optimization, health assessment, and adaptive nutritional adjustment, with the advantages of high precision, high feed utilization rate, strong personalized regulation ability, and optimized sheep flock health management.

[0011] According to an embodiment of the present invention, the precise nutritional supplementation method for sheep based on intelligent feeding algorithms includes the following steps: S1. Collect physiological data and behavioral characteristic data of individual sheep, use multi-modal sensing devices to obtain body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and movement behavior data, and construct an individual three-dimensional representation model based on the neural radiance field modeling method; S2. Based on the individual three-dimensional representation model, establish an individual nutritional requirement calculation model, use the multi-modal feature fusion method to extract key influencing factors, calculate the current nutritional requirements of the individual, and generate feeding parameters that change with the growth stage; S3. Use the sheep flock intelligent optimization algorithm to simulate individual foraging strategies, group interaction behaviors, and resource competition patterns, globally optimize the feeding parameters, and determine a feeding plan that meets the precise nutritional requirements of individuals while optimizing group resource allocation; S4. Combine deep reinforcement learning to construct a dynamic nutritional regulation model, use an intelligent agent to continuously monitor body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and movement behavior data, and optimize the feeding parameters based on long-term feedback learning to keep the feeding plan adaptively optimized in a dynamic environment; S5. Adopt a cloud-edge collaborative computing architecture, deploy the individual nutritional requirement calculation model and the sheep flock intelligent optimization algorithm in edge computing devices, and perform periodic model optimization and knowledge distillation in the cloud at the same time; S6. Combine non-invasive health monitoring technologies, construct a group health assessment model based on the dynamic monitoring results of body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and movement behavior data, and dynamically adjust the feed ratio based on the health assessment results; S7. Through the individual nutritional requirement calculation model and the group health assessment model, analyze the dynamic change trends of individual and group nutritional requirements, predictively adjust the feeding parameters, and optimize the dynamic nutritional supplementation strategies for individuals and groups.

[0012] Optionally, the specific steps of S2 include: S21. Extract the individual physiological parameter set based on the constructed individual three-dimensional representation model , where include body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and exercise behavior data. Perform normalization transformation on , use the min-max normalization method to process continuous variables, and use to normalize discrete variables and form a standardized physiological feature data set ; S22. Perform feature dimensionality reduction and fusion on the standardized physiological feature data set to construct a high-dimensional feature mapping , where represents the optimized feature space after deep feature learning. Adopt a multi-layer feature extraction method, calculate the feature contribution degree in combination with principal component analysis, and use a deep auto-encoding network for feature representation learning to screen out key nutrient influencing factors , where and ; S23. Based on the key nutrient influencing factors , construct an individual nutrient requirement calculation model, and use a time series prediction method to establish an individual nutrient requirement function , where represents the nutrient requirement value of the individual at time . Use a variational auto-regressive model to construct a time-dependent nutrition prediction model; S24. Based on the individual nutrient requirement function , combined with the feed composition matrix , the individual adaptability parameter matrix and the group resource allocation constraint, construct a dynamic nutrition supply optimization function: ; where, represents the absorption rate of the individual to the feed at time . Use an adaptive non-linear optimization method to solve the optimal feeding parameter set to make the individual nutrient intake meet the dynamic growth requirements; S25. Based on the optimal feeding parameter set , construct a personalized feeding regulation model, use a weighted moving average method to smooth the feeding parameter curve, and generate feeding parameters that change with the growth stage.

[0013] Optionally, the S24 specifically includes: S241. Based on the calculated individual nutrient requirement function , construct an optimized objective for dynamic nutritional supplementation, set an objective function for minimizing the individual nutritional intake error to optimize the deviation between the individual feed absorption amount and nutritional requirements; S242. Introduce an individual adaptability parameter matrix , adaptively adjust the individual's physiological state, metabolic capacity, and historical feeding records, and define an individual adaptability correction function : ; Among them, is the adaptability correction factor of individual , and the calculation formula is: ; Among them, , , are weight parameters, represents the long-term nutritional metabolism index of individual , is the length of the historical window; S243. Set the group resource allocation constraint , limit the total amount of individual feed supply to evenly distribute group resources, and the optimization constraint expression is as follows: ; Among them, represents the resource allocation constraint of the flock at time ; S244. Based on the constructed optimization constraint expression, use an adaptive non-linear optimization algorithm to solve, and set an iterative optimization function: ; Among them, is the feeding parameter matrix at time , is the learning rate, is the gradient of the objective function; S245. Smoothly regulate the obtained optimal feeding parameter set , use a parameter adjustment function to limit the parameter mutation amplitude, so as to ensure the stability and growth stage adaptability of the feeding plan. The adjusted feeding parameter set is calculated as: ; Among them, is the smoothing factor.

[0014] Optionally, the specific steps of S3 include: S31. Construct an individual foraging strategy model, and define individual at time Feeding plan vector : ; wherein, represents the intake ratio of individual selecting feed ingredient at time , and is the number of types of optional feed ingredients, satisfying the following constraint conditions: ; S32. Establish an individual interaction behavior model, and define the interaction influence factor of individual in the group: ; wherein, represents the neighbor set of individual in the group, is the interaction weight between individual and individual , is the feeding plan vector of individual at time ; S33. Establish a group resource competition model, and set the resource competition factor of individual at time : ; wherein, is the resource competition coefficient of individual , is the total amount of distributable feed of the group at time ; S34. Based on the individual foraging strategy, group interaction behavior, and resource competition models constructed in S31 - S33, use the sheep flock intelligent optimization algorithm to optimize the global feeding parameters, and set the optimization objective function: ; wherein, is the individual nutritional requirement function, is the absorption rate of individual for the feed ingredient, is the feed ingredient matrix in the th feed ingredient; S35. Based on the solved optimal feeding parameters, define a smoothing adjustment function to dynamically adjust the feeding plan vector and calculate the final feeding plan vector : ; Among them, is the smoothing factor.

[0015] Optionally, the specific steps of S4 are as follows: S41. Construct a dynamic nutritional regulation model driven by deep reinforcement learning, and set the state set of the intelligent agent at time : ; Among them, is the body shape parameter, bone structure information, and fat distribution characteristics, is the set of physiological indicators, is the exercise behavior data, is the environmental state variable; S42. Based on the constructed state set, set the action space of reinforcement learning, and define the feeding strategy of the intelligent agent : ; Among them, is the feeding plan vector adopted by the individual at time , which satisfies ; S43. Construct the reward function of reinforcement learning , set the evaluation index of the individual feeding effect, and define the reward value: ; S44. Based on the constructed reinforcement learning reward function, adopt the deep reinforcement learning optimization algorithm, and use the method to update the feeding strategy of the intelligent agent; S45. Based on the optimized feeding strategy parameters, update the individual feeding plan vector, and calculate the final feeding parameter set .

[0016] Optionally, the specific steps of S6 are as follows: S61. Based on non-invasive health monitoring technology, collect the health status data of the target sheep flock individuals, and construct the individual health feature vector set : ; Among them, represents the body shape parameter, bone structure characteristics, and fat distribution information of the individual at time , represents the set of physiological indicators, represents the exercise behavior data, Indicates an environmental variable; S62. Based on the obtained set of individual health feature vectors, calculate the individual health status evaluation value using a multimodal data fusion method, and define an individual health evaluation function : ; Wherein, , , , is the weighted coefficient of the health status factor; S63. Based on the calculated individual health evaluation value, construct a population health distribution model and calculate the population resource allocation constraint to measure the overall health status: ; Wherein, is the total number of population individuals, is the health balance adjustment coefficient, is the preset health target value; S64. Based on the calculated population resource allocation constraint , dynamically adjust the feed ratio and construct a feed composition correction function : ; Wherein, represents the supply amount of the optimized feed composition at time , is the original feed supply amount, controls the correction of the feed ratio by the health status deviation, controls the dynamic adjustment of the feed supply by the health status change rate; S65. Based on the dynamically adjusted feed ratio, update the individual feeding parameters and calculate the final feeding parameter matrix .

[0017] Optionally, the specific content of S7 includes: S71. Based on the individual nutritional requirement calculation model, construct an individual nutritional requirement time series , and define an individual nutritional requirement prediction function: ; Wherein, is the predicted nutritional requirement value of individual at time , is the historical nutritional requirement data, is the time window length, is the time series prediction model; S72. Based on the predicted value of individual nutritional requirements calculated, combined with the group resource allocation constraints calculated by the group health assessment model , construct an individual and group combined nutritional requirement prediction model, and define the combined nutritional requirement prediction function: ; Among them, is the predicted value of the combined nutritional requirements of individual at time , , is the weighted coefficient of individual requirements and the group equilibrium factor, satisfying ; S73. Based on the predicted value of the combined nutritional requirements calculated, adjust the individual feeding parameters, and define the feeding parameter prediction and adjustment function: ; Among them, is the corrected feeding parameter of individual at time , is the original feeding parameter, is the adaptive adjustment factor; S74. Based on the corrected feeding parameter calculated, optimize the individual feed intake strategy, define the feeding parameter smoothing adjustment function, and calculate the final feeding plan vector: ; Among them, is the smoothing factor; S75. Based on the optimized individual feeding plan obtained, calculate the final feeding parameter matrix , realize that the feed supply strategy adapts to the dynamic changes of individual and group nutritional requirements, optimize the feeding resource allocation, and combine the individual growth status to realize the adaptive adjustment of the feeding plan.

[0018] The beneficial effects of the present invention are: (1) By combining neural radiance field modeling, multi-modal feature fusion and individual nutritional requirement calculation models, the present invention realizes the accurate modeling and personalized nutritional assessment of sheep individuals, enabling the feeding system to real-time sense the individual body shape, bone structure, fat distribution and physiological state, dynamically calculate the individual nutritional requirements, so as to effectively match the growth requirements of different individuals, optimize the nutritional supply plan, and improve the feeding accuracy.

[0019] (2) By adopting the sheep flock intelligent optimization algorithm, simulating the individual foraging strategy, group interaction behavior and resource competition mode, globally optimize the feed feeding parameters, realize the dynamic balance of individual requirements and group resources, improve the rationality of feed distribution, reduce feed waste, and at the same time take into account the nutritional intake of weak individuals, and optimize the overall growth state of the group.

[0020] (3) The present invention constructs a dynamic nutrition regulation system by combining deep reinforcement learning. An intelligent agent is used to track the individual's health status, exercise behavior, and environmental changes in real time, and the feeding parameters are optimized through the Actor-Critic strategy, enabling the feeding plan to have the ability of adaptive optimization in a dynamic environment, thereby effectively improving the accuracy of feed feeding and reducing the sensitivity of the feeding strategy to environmental factors.

[0021] (4) The present invention adopts a cloud-edge collaborative computing architecture. NeRF modeling, SOA optimization algorithm, and real-time feeding decision-making are deployed on edge devices to reduce data transmission latency and computing energy consumption. At the same time, the cloud is used for periodic model optimization and knowledge distillation to improve computing efficiency and model generalization ability, ensuring the real-time performance and stability of the feeding strategy.

[0022] (5) The present invention combines non-invasive health monitoring technology. Based on the analysis of individual behavior patterns, body shape change trends, and physiological parameter monitoring, real-time health assessment is achieved, and the feed ratio is dynamically adjusted through the group health balance factor to ensure the balance of nutritional intake of individuals and groups and improve the overall health management ability of the flock.

[0023] (6) The present invention predicts the changing trends of the nutritional requirements of individuals and groups through multi-modal data fusion and intelligent prediction models, combined with historical growth data and health assessment results, and optimizes the feeding parameters based on time series analysis to achieve forward-looking adjustment of nutritional supply, ensuring that the feeding system has long-term stability and high adaptability. Description of the Drawings

[0024] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings: Figure 1 is a flowchart of the method for precise nutritional supplementation of sheep flock based on intelligent feeding algorithm proposed by the present invention. Detailed Embodiments

[0025] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic way, so they only show the components related to the present invention.

[0026] Reference Figure 1 , the method for precise nutritional supplementation of sheep flock based on intelligent feeding algorithm includes the following steps: S1. Collect the physiological data and behavioral characteristic data of individual sheep. Use multi-modal sensing devices to obtain body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and exercise behavior data, and construct an individual three-dimensional representation model based on the neural radiance field modeling method; S2. Based on the individual three-dimensional representation model, establish an individual nutritional requirement calculation model, extract key influencing factors using a multi-modal feature fusion method, calculate the current nutritional requirements of the individual, and generate feeding parameters that change with the growth stage; S3. Use the flock intelligence optimization algorithm to simulate the individual foraging strategy, group interaction behavior, and resource competition mode, globally optimize the feeding parameters, and determine a feeding plan that meets the precise nutritional requirements of the individual while optimizing the group resource allocation; S4. Combine deep reinforcement learning to construct a dynamic nutrition regulation model. Through intelligent agents, continuously monitor body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and exercise behavior data, and optimize the feeding parameters based on long-term feedback learning to keep the feeding plan adaptively optimized in a dynamic environment; S5. Adopt a cloud-edge collaborative computing architecture, deploy the individual nutritional requirement calculation model and the flock intelligence optimization algorithm in edge computing devices, and perform periodic model optimization and knowledge distillation in the cloud at the same time; S6. Combine non-invasive health monitoring technology, based on the dynamic monitoring results of body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and exercise behavior data, construct a group health assessment model, and dynamically adjust the feed ratio based on the health assessment results; S7. Through the individual nutritional requirement calculation model and the group health assessment model, analyze the dynamic change trends of the nutritional requirements of individuals and groups, predictively adjust the feeding parameters, and optimize the dynamic nutrition supply strategy for individuals and groups.

[0027] In this embodiment, the S2 specifically includes: S21. Based on the constructed individual three-dimensional representation model, extract the individual physiological parameter set , where includes body shape parameters, bone structure information, fat distribution characteristics, physiological indicators, and exercise behavior data. Perform normalization transformation on , use the min-max normalization method to process continuous variables, and use to normalize discrete variables, and form a standardized physiological feature data set ; S22. Perform feature dimensionality reduction and fusion on the standardized physiological feature data set to construct a high-dimensional feature mapping , where represents the optimized feature space after deep feature learning. Adopt a multi-layer feature extraction method, calculate the feature contribution degree in combination with principal component analysis, and use a deep autoencoder network for feature representation learning to screen out key nutritional influencing factors , where and ; S23. Based on key nutritional impact factors Construct an individual nutritional requirement calculation model and establish an individual nutritional requirement function using time series prediction methods , where represents the nutritional requirement value of the individual at time . Using a variational auto-regressive model, construct a time-dependent nutritional prediction model; S24. Based on the individual nutritional requirement function , combined with the feed composition matrix , the individual adaptability parameter matrix and the group resource allocation constraints, construct a dynamic nutritional replenishment optimization function: ; where represents the absorption rate of the individual for the feed at time . Use an adaptive non-linear optimization method to solve the optimal feeding parameter set to ensure that the individual's nutritional intake meets the dynamic growth requirements; S25. Based on the optimal feeding parameter set , construct a personalized feeding regulation model, use the weighted moving average method to smooth the feeding parameter curve, and generate feeding parameters that change with the growth stage.

[0028] In this embodiment, S24 specifically includes: S241. Based on the calculated individual nutritional requirement function , construct a dynamic nutritional replenishment optimization objective, set an individual nutritional intake error minimization objective function to optimize the deviation between the individual feed absorption amount and nutritional requirements; S242. Introduce the individual adaptability parameter matrix to adaptively adjust the individual's physiological state, metabolic capacity, and historical feeding records, and define an individual adaptability correction function : ; where is the adaptability correction factor of the individual , and the calculation formula is: ; where , , is the weight parameter, represents the long-term nutritional metabolism index of the individual , is the historical window length; S243. Set the group resource allocation constraint , limit the total individual feed supply, enable the balanced distribution of group resources, and the optimization constraint expression is as follows: ; Among them, represents the time the resource allocation constraint of the flock; S244. Based on the constructed optimization constraint expression, use the adaptive nonlinear optimization algorithm for solution, and set the iterative optimization function: ; Among them, is the feeding parameter matrix at time , is the learning rate, is the gradient of the objective function; S245. Smoothly regulate the obtained optimal feeding parameter set , use the parameter adjustment function to limit the parameter mutation amplitude, improve the stability and growth stage adaptability of the feeding plan, and the adjusted feeding parameter set The calculation formula is: ; Among them, is the smoothing factor.

[0029] In this embodiment, the specific content of S3 includes: S31. Build an individual foraging strategy model, and define the feeding plan vector of individual at time : ; Among them, represents the intake ratio of the feed component selected by individual at time , is the number of types of optional feed components, and satisfies the following constraint conditions: ; S32. Establish an individual interaction behavior model, and define the interaction influence factor of individual in the group: ; Among them, represents the neighbor set of individual in the group, is individual and individual The interaction weight between for an individual at time feeding plan vector; S33. Establish a group resource competition model, and set the resource competition factor of an individual at time : ; wherein, is the resource competition coefficient of the individual , is the total amount of distributable feed of the group at time ; S34. Based on the individual foraging strategy, group interaction behavior and resource competition model constructed in S31 - S33, use the flock intelligence optimization algorithm to optimize the global feeding parameters, and set the optimization objective function: ; wherein, is the individual nutritional requirement function, is the absorption rate of the individual to the feed components, is the feed component matrix in the th feed component; S35. Based on the obtained optimal feeding parameters, define a smoothing adjustment function to dynamically adjust the feeding plan vector , and calculate the final feeding plan vector : ; wherein, is the smoothing factor.

[0030] In this embodiment, the specific content of S4 includes: S41. Construct a dynamic nutrition regulation model driven by deep reinforcement learning, and set the state set of the intelligent agent at time : ; wherein, is the body size parameter, bone structure information, and fat distribution characteristics, is the set of physiological indicators, is the motion behavior data, is the environmental state variable; S42. Based on the constructed state set, set the action space of reinforcement learning, and define the feeding strategy of the intelligent agent: ; Among them, is the feeding plan vector adopted by the individual at time which satisfies ; S43. Construct the reward function of reinforcement learning , set the evaluation index of the individual feeding effect, and define the reward value: ; S44. Based on the constructed reinforcement learning reward function, adopt the deep reinforcement learning optimization algorithm, and use method to update the feeding strategy of the intelligent agent ; S45. Based on the optimized feeding strategy parameters, update the individual feeding plan vector and calculate the final feeding parameter set .

[0031] In this embodiment, the specific steps of S6 include: S61. Based on the non-invasive health monitoring technology, collect the health status data of the target flock individuals and construct the individual health feature vector set : ; Among them, represents the body shape parameters, bone structure characteristics and fat distribution information of the individual at time , represents the set of physiological indicators, represents the exercise behavior data, represents the environmental variables; S62. Based on the obtained individual health feature vector set, use the multi-modal data fusion method to calculate the individual health status evaluation value and define the individual health evaluation function : ; Among them, , , , are the weighted coefficients of the health status factors; S63. Based on the calculated individual health evaluation value, construct the group health distribution model and calculate the group resource allocation constraint to measure the overall health status: ; Among them, is the total number of group individuals, is the health balance adjustment coefficient, is the preset healthy target value; S64. Calculation-based group resource allocation constraint , dynamically adjust the feed ratio, and construct a feed composition correction function : ; Among them, represents the optimized feed composition at time supply volume, is the original feed supply volume, control the correction of the feed ratio by the healthy state deviation, control the dynamic adjustment of the feed supply by the change rate of the healthy state; S65. Based on the dynamically adjusted feed ratio, update the individual feeding parameters, and calculate the final feeding parameter matrix .

[0032] In this embodiment, the S7 specifically includes: S71. Based on the individual nutrition demand calculation model, construct an individual nutrition demand time series , and define an individual nutrition demand prediction function: ; Among them, is the individual at time predicted nutrition demand value, is the historical nutrition demand data, is the time window length, is the time series prediction model; S72. Based on the calculated individual nutrition demand prediction value, combined with the group resource allocation constraint calculated by the group health assessment model , construct an individual and group joint nutrition demand prediction model, and define a joint nutrition demand prediction function: ; Among them, is the individual at time joint nutrition demand prediction value, , is the weighted coefficient of the individual demand and the group equilibrium factor, satisfying ; S73. Based on the calculated joint nutrition demand prediction value, adjust the individual feeding parameters, and define a feeding parameter prediction adjustment function: ; Among them, is the individual at time The corrected feeding parameters are the original feeding parameters is the adaptive adjustment factor; S74. Based on the calculated corrected feeding parameters, optimize the individual feed intake strategy, define the smooth adjustment function of the feeding parameters, and calculate the final feeding plan vector: ; wherein is the smoothing factor; S75. Based on the optimized individual feeding plan, calculate the final feeding parameter matrix , realizing that the feed supply strategy adapts to the dynamic changes of individual and group nutritional requirements, optimizing the allocation of feeding resources, and realizing the adaptive adjustment of the feeding plan in combination with the individual growth state.

[0033] Example 1: To verify the application effect of the present invention in the precise nutritional supplementation of sheep flocks, the present invention was applied to a large ranch that raises more than 2,000 sheep and adopts a feeding mode combining captive breeding and grazing. The traditional feeding method mainly relies on a fixed feed ratio and fails to optimize according to individual differences, resulting in some sheep growing slowly due to insufficient nutrition, while some sheep have fat accumulation due to excessive intake, affecting the overall health level. In addition, the existing feeding system lacks the ability of real-time health monitoring and cannot dynamically adjust the feed supply, resulting in feed waste and imbalance between nutritional supply and demand. To solve these problems, the ranch introduced the intelligent feeding algorithm of the present invention and achieved precise nutritional supply and dynamic optimization adjustment through individual modeling, intelligent optimization, reinforcement learning control, and non-invasive health monitoring.

[0034] The implementation of the present invention is divided into four stages: data collection, modeling analysis, feeding optimization, and dynamic adjustment. In the data collection stage, the ranch installed multi-modal sensing devices in the feeding area, including infrared thermal imagers, 3D cameras, bioelectrical impedance sensors, and motion tracking devices, to collect sheep body size parameters, physiological indicators, motion behaviors, and environmental data. This data is transmitted to the cloud processing center in real time, and through the neural radiance field modeling technology, an individual three-dimensional representation is constructed, extracting bone structure, fat distribution, and body shape change information, and combining multi-modal data fusion to calculate the individual health state and establish an individual nutritional requirement calculation model.

[0035] In the modeling and analysis stage, the system calculates the nutritional requirements based on the individual health characteristics, and combines the long short-term memory network for time series prediction to obtain the future nutritional requirement trend. The system further combines the flock intelligence optimization algorithm to simulate the foraging behavior, resource competition and group interaction of the flock, and calculates the global feed optimization strategy to improve the rationality of feed allocation. To verify the effectiveness of this method, 100 sheep were randomly selected for a comparative experiment. One group used the traditional feeding method, and the other group used the method of the present invention for precise feeding.

[0036] In the feeding optimization stage, the system calculates the group health balance factor based on the individual health assessment results, and uses an adaptive optimization method to dynamically adjust the feed ratio. The system constructs an intelligent feeding control model through deep reinforcement learning, optimizes the feeding strategy based on the Actor-Critic algorithm, and updates the feeding parameters in real time to make the feed supply adapt to the individual growth requirements.

[0037] The experiment lasted for 90 days, and the system compared the effects of the traditional feeding scheme and the scheme of the present invention in terms of feed utilization rate, individual health improvement and feeding accuracy. The experimental data are as follows: Table 1 Comparison between the traditional feeding method and the method of the present invention

[0038] As can be seen from Table 1, compared with the traditional feeding method, the present invention reduces the feed consumption by 16.7%, reduces the feed waste rate by 52.9%, increases the individual daily weight gain by 23.1%, and significantly improves the overall health score of the flock.

[0039] In the dynamic adjustment stage, the present invention calculates the overall health status based on the group health balance factor and dynamically optimizes the feed ratio. During the experiment, the feed protein content and energy ratio were adjusted in real time according to the health status, so that the health score increased from 78.5 to 89.7, ensuring the balance between individual and group nutritional supply and demand and improving the feeding accuracy.

[0040] The present invention also adopts a cloud-edge collaborative computing architecture, and deploys NeRF modeling, SOA optimization algorithm and intelligent feeding decision on the edge device to reduce the data transmission delay and improve the system response speed. Compared with the traditional computing architecture, the data processing ability of the present invention is increased by 219%, and the calculation time of the feeding decision is shortened by 60.9%, significantly improving the real-time performance and calculation efficiency of the feeding system.

[0041] In summary, the present invention realizes precise nutritional supplementation for the flock through precise individual modeling, global feeding optimization, deep reinforcement learning control and non-invasive health monitoring, improves the feed utilization rate, optimizes the individual health status, and improves the calculation efficiency, solving the problems of low accuracy, serious resource waste and lagging health management of the traditional feeding method, and providing reliable technical support for large-scale intelligent livestock management.

[0042] As described above, it is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A method for accurate nutrition supply to sheep flocks based on intelligent feeding algorithm, characterized in that: The steps include: S1. Collect individual physiological data and behavioral characteristic data of sheep, use multimodal sensing equipment to obtain body parameters, bone structure information, fat distribution characteristics, physiological indicators and movement behavior data, and build an individual three-dimensional representation model based on the neural radiation field modeling method; S2. Based on the individual three-dimensional representation model, an individual nutritional requirement calculation model is established, and a multimodal feature fusion method is used to extract key influencing factors, calculate the individual's current nutritional requirements, and generate feeding parameters that change with the growth stage; S3, using flock intelligence optimization algorithm to simulate individual foraging strategies, group interaction behaviors and resource competition patterns, to globally optimize feeding parameters, and determine a feeding plan that meets the precise nutritional needs of individuals while optimizing group resource allocation; S4. Combine deep reinforcement learning to build a dynamic nutrition regulation model, monitor body shape parameters, bone structure information, fat distribution characteristics, physiological indicators and movement behavior data in real time through intelligent agents, optimize feeding parameters based on long-term feedback learning, and keep the feeding plan adaptively optimized in a dynamic environment; S5. Adopt cloud-edge collaborative computing architecture, deploy individual nutritional demand calculation model and flock intelligence optimization algorithm in edge computing devices, and perform periodic model optimization and knowledge distillation in the cloud; S6. Combined with non-invasive health monitoring technology, a group health assessment model is constructed based on the dynamic monitoring results of body shape parameters, bone structure information, fat distribution characteristics, physiological indicators and exercise behavior data, and the feed ratio is dynamically adjusted based on the health assessment results; S7. Through the individual nutritional demand calculation model and the group health assessment model, the dynamic changing trend of individual and group nutritional needs is analyzed, the feeding parameters are adjusted predictively, and the dynamic nutritional supply strategy of individuals and groups is optimized.

2. The method for precise nutrition supplementation of sheep flocks based on intelligent feeding algorithm according to claim 1, characterized in that: The S2 specifically includes: S21. Extract individual physiological parameter sets based on the constructed individual 3D representation model ,in Including body parameters, bone structure information, fat distribution characteristics, physiological indicators and exercise behavior data. Normalization transformation is performed, and the minimum-maximum normalization method is used to process continuous variables. Normalize discrete variables and form a standardized physiological characteristic data set ; S22. Standardized physiological characteristics dataset Perform feature dimensionality reduction and fusion to construct high-dimensional feature mapping ,in Represents the optimized feature space after deep feature learning, uses a multi-layer feature extraction method, combines principal component analysis to calculate feature contribution, and uses a deep automatic encoding network for feature representation learning to screen out key nutritional influencing factors ,in and ; S23. Based on key nutritional influencing factors Construct an individual nutritional requirement calculation model and use time series prediction methods to establish an individual nutritional requirement function ,in Indicates that an individual is at a certain time The nutritional requirement value is calculated and a time-dependent nutritional prediction model is constructed using a variational autoregressive model. S24. Based on individual nutritional requirement function , combined with feed ingredient matrix , individual adaptability parameter matrix And group resource allocation constraints, construct a dynamic nutrition supply optimization function: ； in, Represents an individual At the moment Feed The optimal feeding parameter set is solved by adaptive nonlinear optimization method. , so that individual nutritional intake meets dynamic growth needs; S25, based on the optimal feeding parameter set , a personalized feeding regulation model was constructed, the feeding parameter curve was smoothed by the weighted moving average method, and the feeding parameters that changed with the growth stage were generated.

3. The method for precise nutrition supplementation of sheep flocks based on intelligent feeding algorithm according to claim 2, characterized in that: The S24 specifically includes: S241. Calculation-based individual nutrition requirement function , construct dynamic nutrition supply optimization objectives, set the objective function of minimizing individual nutrition intake errors, and optimize the deviation between individual feed absorption and nutrition requirements; S242, introduce individual adaptability parameter matrix , adaptively adjust individual physiological status, metabolic capacity and historical feeding records, and define individual adaptability correction function : ； in, For individuals The adaptive correction factor is calculated as follows: ； in, , , is the weight parameter, Representative individual Long-term nutritional metabolic index, is the history window length; S243. Setting group resource allocation constraints , limit the total amount of individual feed supply to make the group resources evenly distributed, and the optimization constraint expression is as follows: ； in, Indicates time Resource allocation constraints for the flock; S244. Based on the constructed optimization constraint expression, an adaptive nonlinear optimization algorithm is used to solve it, and an iterative optimization function is set: ； in, For the moment The feeding parameter matrix, is the learning rate, is the objective function gradient; S245, the optimal feeding parameter set obtained by solving Perform smooth control and use parameter adjustment function Limit the amplitude of parameter mutation to ensure the stability of feeding scheme and adaptability to growth stages. Adjust the feeding parameter set The calculation formula is: ； in, is the smoothing factor.

4. The method for precise nutrition supplementation of sheep flocks based on intelligent feeding algorithm according to claim 1, characterized in that: The S3 specifically includes: S31. Construct an individual foraging strategy model and define individual At the moment Feeding plan vector : ； in, Represents an individual At the moment Selecting feed ingredients The intake ratio, is the number of optional feed ingredients, satisfying the following constraints: ； S32. Establish an interactive behavior model between individuals and define individuals Interaction factors in groups : ； in, Represents an individual The set of neighbors in a group, For individuals With individuals The interaction weight between For individuals At the moment The feeding plan vector of S33, establish a group resource competition model, set individual At the moment Resource competition factor : ； in, For individuals The resource competition coefficient, For the group at the moment Total amount of feed available for distribution; S34, based on the individual foraging strategy, group interaction behavior and resource competition model constructed in S31-S33, the flock intelligence optimization algorithm is used to optimize the global feeding parameters and set the optimization objective function: ； in, is the individual nutritional requirement function, For individuals The absorption rate of feed ingredients, For feed ingredient matrix The feed ingredients; S35. Based on the optimal feeding parameters solved, define a smooth adjustment function Feeding Scheme Vector Perform dynamic adjustments and calculate the final feeding plan vector : ； in, is the smoothing factor.

5. The method for precise nutrition supplementation of sheep flocks based on intelligent feeding algorithm according to claim 1, characterized in that: The S4 specifically includes: S41. Build a dynamic nutrition regulation model driven by deep reinforcement learning, and set the intelligent agent to The state collection: ； in, For body shape parameters, bone structure information, and fat distribution characteristics, is a set of physiological indicators, For sports behavior data, is the environment state variable; S42. Based on the constructed state set, set the action space of reinforcement learning and define the feeding strategy of the intelligent agent : ； in, For individuals At the moment The feeding scheme vector adopted satisfies ; S43. Constructing the reward function for reinforcement learning , set the evaluation index of individual feeding effect and define the reward value: ； S44, based on the constructed reinforcement learning reward function, adopts the deep reinforcement learning optimization algorithm and utilizes Method to update intelligent agent feeding strategy ; S45. Based on the optimized feeding strategy parameters, update the individual feeding plan vector and calculate the final feeding parameter set .

6. The method for precise nutrition supplementation of sheep flocks based on intelligent feeding algorithm according to claim 1, characterized in that: The S6 specifically includes: S61. Based on non-invasive health monitoring technology, collect health status data of target sheep individuals and construct a set of individual health feature vectors : ； in, Represents an individual At the moment body parameters, bone structure characteristics and fat distribution information, represents a set of physiological indicators, Represents movement behavior data, Represents environment variables; S62. Based on the acquired individual health feature vector set, a multimodal data fusion method is used to calculate the individual health status assessment value, and an individual health assessment function is defined. : ； in, , , , is the weighted coefficient of the health status factor; S63. Based on the calculated individual health assessment values, construct a group health distribution model and calculate group resource allocation constraints To measure overall health: ； in, is the total number of individuals in the group, is the health balance adjustment coefficient, To set the health target value; S64. Computation-based group resource allocation constraints , dynamically adjust feed ratio and construct feed composition correction function : ； in, Indicates the optimized feed composition At the moment The supply of is the original feed supply, Control the correction of feed ratio due to health status deviation, Control the rate of change of health status to dynamically adjust feed supply; S65: Based on the dynamically adjusted feed ratio, update the individual feeding parameters and calculate the final feeding parameter matrix .

7. The method for accurate nutrition supplementation of sheep flocks based on intelligent feeding algorithm according to claim 1, characterized in that: The S7 specifically includes: S71. Construct individual nutritional demand time series based on individual nutritional demand calculation model , define the individual nutritional demand prediction function: ； in, For individuals At the moment The predicted nutritional requirements of For historical nutritional requirements data, is the time window length, It is a time series forecasting model; S72. Based on the calculated individual nutritional demand prediction value, combined with the group resource allocation constraints calculated by the group health assessment model , build an individual and group joint nutritional demand prediction model, and define the joint nutritional demand prediction function: ； in, For individuals At the moment The predicted value of joint nutritional requirements, , is the weighted coefficient of individual demand and group equilibrium factor, satisfying ; S73. Based on the calculated predicted value of the combined nutritional requirement, adjust the individual feeding parameters and define a feeding parameter prediction adjustment function: ； in, For individuals At the moment Corrected feeding parameters, is the original feeding parameter, is the adaptive adjustment factor; S74, based on the calculated corrected feeding parameters, optimize the individual feed intake strategy, define the feeding parameter smoothing adjustment function, and calculate the final feeding plan vector: ； in, is the smoothing factor; S75. Calculate the final feeding parameter matrix based on the optimized individual feeding plan , so that the feed supply strategy can adapt to the dynamic changes in individual and group nutritional needs, optimize the allocation of feeding resources, and realize adaptive adjustment of feeding plans based on individual growth status.

Images