Intelligent optimization method and system for multi-area feeding decision of cattle and sheep

Abstract

The invention discloses an intelligent optimization method for multi-area feeding decision of cattle and sheep. The method comprises the following steps: obtaining individual parameters and environmental parameters of cattle and sheep; according to the individual parameters and the environmental parameters, an initial feeding strategy is constructed through a decision tree algorithm; dynamic physiological parameters and environmental parameters of cattle and sheep are obtained, the initial feeding strategy is dynamically optimized according to an incremental learning method, the dynamic physiological parameters and the environmental parameters, and a global optimal feeding strategy is obtained according to a Q-learning algorithm; and judging whether the global optimal feeding strategy meets a preset constraint condition or not, and when the constraint condition is met, converting the global optimal feeding strategy into an execution plan and monitoring execution. Individual and precise intelligent feeding of cattle and sheep is achieved, the feeding strategy can be adjusted in time according to the animal physiological state and the environment change, and the feeding efficiency and economic benefits are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of feeding control, and particularly relates to an intelligent optimization method and system for multi-region feeding decision-making of cattle and sheep. Background Art

[0002] In the intelligent optimization of multi-region feeding decision-making for cattle and sheep, there are technical problems of how to accurately formulate a feeding plan according to the individual needs, growth stages, health conditions of cattle and sheep, and environmental parameters. Due to the large individual differences among cattle and sheep, their feed requirements vary in different growth stages and health conditions, and the feeding environment parameters will also affect the feeding behavior of cattle and sheep. In addition, during the actual feeding process, the physiological states and environmental parameters of cattle and sheep are dynamically changing. How to obtain these data in real time and quickly adjust the feeding strategy also poses high requirements for the response speed and stability of the system. Therefore, constructing an efficient and reliable intelligent feeding system is also a technical problem to be solved urgently. Summary of the Invention

[0003] To solve the above technical problems, the present invention proposes an intelligent optimization method and system for multi-region feeding decision-making of cattle and sheep to solve the problems existing in the above prior art.

[0004] To achieve the above object, the present invention provides an intelligent optimization method for multi-region feeding decision-making of cattle and sheep, including the following steps:

[0005] Obtain the individual parameters and environmental parameters of cattle and sheep;

[0006] According to the individual parameters and environmental parameters, construct an initial feeding strategy through a decision tree algorithm;

[0007] Obtain the dynamic physiological parameters and environmental parameters of cattle and sheep, dynamically optimize the initial feeding strategy according to the incremental learning method and the dynamic physiological parameters and environmental parameters, and obtain the global optimal feeding strategy according to the Q-learning algorithm;

[0008] Judge whether the global optimal feeding strategy meets the preset constraint conditions. When the constraint conditions are met, convert the global optimal feeding strategy into an execution plan and monitor the execution.

[0009] Preferably, obtaining the individual parameters and environmental parameters of cattle and sheep includes;

[0010] The individual parameters include identity identification and physiological parameters;

[0011] Obtain the identity identification through an electronic ear tag, and the identity identification includes date of birth, breed, gender, parental information, and vaccination records;

[0012] Obtain the physiological parameters through an intelligent sensor, and the physiological parameters include body temperature, heart rate, and respiratory rate;

[0013] Obtain environmental parameters through an environmental sensor, where the environmental parameters include temperature, humidity, light intensity, and air quality.

[0014] Preferably, dynamically optimizing the initial feeding strategy includes:

[0015] Obtain real-time change data of the physiological parameters and environmental parameters of cattle and sheep as new training data for incremental learning; determine whether the new training data causes a decrease in the prediction accuracy of a certain branch of the decision tree model, and when it decreases, perform a pruning operation on that branch; according to the pruned decision tree branch, use the new data for re-growth training to obtain an optimized decision tree model.

[0016] Preferably, obtaining the global optimal feeding strategy according to the Q-learning algorithm includes:

[0017] Combine the physiological parameters and environmental parameters of cattle and sheep as states, and different feeding schemes as actions. Through the update of the Q-value table, learn the corresponding feeding schemes to be adopted in different states to obtain the global optimal feeding strategy.

[0018] Preferably, when the global optimal feeding strategy does not meet the constraint conditions, then through the selection, crossover, and mutation operations of the genetic algorithm, regenerate a new feeding decision scheme until the generated new feeding decision scheme meets the constraint conditions to obtain the global optimal feeding strategy that meets the constraint conditions.

[0019] The present invention also provides an intelligent optimization system for multi-region feeding decision of cattle and sheep, including:

[0020] A data acquisition module for acquiring the individual parameters and environmental parameters of cattle and sheep;

[0021] An initial feeding strategy generation module for constructing an initial feeding strategy according to the individual parameters and environmental parameters through a decision tree algorithm;

[0022] An optimization module for acquiring the dynamic physiological parameters and environmental parameters of cattle and sheep, dynamically optimizing the initial feeding strategy according to the incremental learning method and the dynamic physiological parameters and environmental parameters, and obtaining the global optimal feeding strategy according to the Q-learning algorithm;

[0023] An implementation module for determining whether the global optimal feeding strategy meets the preset constraint conditions, and when it meets the constraint conditions, converting the global optimal feeding strategy into an execution plan and monitoring the execution.

[0024] Preferably, the data acquisition module includes:

[0025] An identity identification acquisition unit, configured to acquire an identity identification through an electronic ear tag, where the identity identification includes a date of birth, a breed, a gender, parental information, and an immunization record;

[0026] A physiological parameter acquisition unit, configured to acquire physiological parameters through an intelligent sensor, where the physiological parameters include body temperature, heart rate, and respiratory rate;

[0027] An environmental parameter acquisition unit, configured to acquire environmental parameters through an environmental sensor, where the environmental parameters include temperature, humidity, light intensity, and air quality.

[0028] Preferably, the implementation module includes:

[0029] A judgment unit, configured to judge whether a globally optimal feeding strategy meets a preset constraint condition;

[0030] An adjustment unit, configured to, when the globally optimal feeding strategy does not meet the constraint condition, regenerate a new feeding decision plan through selection, crossover, and mutation operations of a genetic algorithm until the generated new feeding decision plan meets the constraint condition, so as to obtain a globally optimal feeding strategy that meets the constraint condition;

[0031] An execution unit, configured to convert the globally optimal feeding strategy into an execution plan and monitor the execution when the constraint condition is met.

[0032] The present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory, where the processor executes the computer program to implement the steps of the method.

[0033] The present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method are implemented.

[0034] Compared with the prior art, the present invention has the following advantages and technical effects:

[0035] The present invention discloses a method and system for intelligent optimization of feeding decisions for cattle and sheep in multiple regions. By collecting physiological parameters and environmental data of individual cattle and sheep in real time, an initial feeding strategy is constructed in combination with a decision tree algorithm, and the decision tree is dynamically optimized by using an incremental learning method. At the same time, a Q-learning reinforcement learning algorithm is introduced, and the feeding decision effect is used as feedback to learn the optimal feeding strategy. The present invention also sets constraint conditions to ensure the generation of a feasible optimal solution. The optimized feeding strategy is converted into a plan executable by an electronic feeding station and is issued and executed in real time. This method realizes individualized and precise intelligent feeding of cattle and sheep, can adjust the feeding strategy in a timely manner according to the physiological state of animals and environmental changes, and effectively improves the feeding efficiency and economic benefits. Description of the Drawings

[0036] The accompanying drawings, which form a part of this application, are used to provide a further understanding of this application. The schematic embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0037] Figure 1 It is a flowchart of the method according to an embodiment of the present invention. Detailed implementation manners

[0038] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will describe this application in detail with reference to the accompanying drawings and in combination with the embodiments.

[0039] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from that here.

[0040] Embodiment 1

[0041] As Figure 1 shown, in this embodiment, an intelligent optimization method for multi-region feeding decision-making of cattle and sheep is provided, including the following steps:

[0042] Obtain the individual parameters and environmental parameters of cattle and sheep;

[0043] According to the individual parameters and environmental parameters, construct an initial feeding strategy through a decision tree algorithm;

[0044] Obtain the dynamic physiological parameters and environmental parameters of cattle and sheep, dynamically optimize the initial feeding strategy according to the incremental learning method and the dynamic physiological parameters and environmental parameters, and obtain the global optimal feeding strategy according to the Q-learning algorithm;

[0045] Judge whether the global optimal feeding strategy meets the preset constraint conditions. When the constraint conditions are met, convert the global optimal feeding strategy into an execution plan and monitor the execution.

[0046] Specifically, it includes the following steps:

[0047] S101. Obtain the individual information of cattle and sheep, including the growth stage and health status, and collect the physiological parameters and environmental parameters of cattle and sheep in real time through intelligent sensors. When the physiological parameters or environmental parameters exceed the preset threshold, trigger the adjustment of the feeding strategy.

[0048] Specifically, according to the pre-established individual information database of cattle and sheep, obtain the identity identification information of each cattle and sheep, including age, breed, etc., so as to judge its growth stage. Through the intelligent sensors installed on the body surface or inside of cattle and sheep, collect real-time physiological parameter data such as body temperature, heart rate, and respiratory rate of cattle and sheep. Use environmental sensors to monitor parameters such as temperature, humidity, light intensity, and air quality of the environment where cattle and sheep are located in real time, and obtain detailed environmental parameter data.

[0049] In this embodiment, the establishment of the individual information database of cattle and sheep is the basis for precise feeding. Technologies such as electronic ear tags and RFID chips can be used to assign a unique identity identification to each cattle and sheep. The information database records the date of birth, breed, gender, parent information, vaccination records, etc. of each cattle and sheep. For example, a Simmental cow numbered "001" with a birth date of January 1, 2023 and a female gender, all this information will be recorded in the database. Through age and breed information, it can be judged that this cow is in different growth stages such as the fattening period, gestation period, or lactation period, providing a basis for subsequent precise feeding. Install intelligent sensors on the body surface or inside of cattle and sheep to collect real-time physiological parameters of cattle and sheep. For example, an implanted body temperature sensor can continuously monitor the body temperature change of cattle and sheep, and a sensor worn around the neck can record the heart rate and respiratory rate. Suppose the body temperature of cow "001" is 38.5°C, the heart rate is 70 beats per minute, and the respiratory rate is 20 times per minute. These data will be transmitted to the cloud server in real time through a wireless network. Environmental sensors also play an important role in the cowshed. Temperature and humidity sensors, light sensors, and air quality sensors are responsible for monitoring the temperature, humidity, light intensity, and concentrations of gases such as ammonia and carbon dioxide in the cowshed respectively. For example, the temperature in the cowshed is 25°C, the humidity is 60%, the light intensity is 500 lux, and the ammonia concentration is 10 ppm. These environmental parameter data will also be uploaded to the cloud server in real time.

[0050] S102. According to the obtained individual information and real-time parameters of cattle and sheep, use the decision tree algorithm to construct an initial feeding strategy. Take each parameter as a node of the decision tree, recursively split each node until a predetermined tree depth is reached or the node cannot be split any further. Obtain a complete decision tree as the initial feeding strategy.

[0051] Specifically, obtain the basic information and real-time physiological parameters of cattle and sheep individuals as the input data for the decision tree algorithm. According to the preset decision tree depth, use the recursive splitting method to take each parameter as a decision node and split the node. During the node splitting process, by calculating the information gain of each attribute, select the attribute with the largest information gain as the splitting node. Repeat the node splitting until the preset decision tree depth limit or the node sample number threshold is reached. Continue to use the split child nodes as new decision nodes until all nodes are split. According to the class labels of the leaf nodes of the decision tree, determine the feeding strategy under this path to obtain a complete initial feeding decision tree. Use the cross-validation method to evaluate the accuracy and generalization ability of the decision tree model, and perform model optimization and parameter adjustment if necessary.

[0052] In this embodiment, obtaining the basic information and real-time physiological parameters of cattle and sheep individuals is the basis for constructing an accurate feeding decision tree. For example, it is possible to obtain a Simmental cattle numbered 001, with an age of 1 year, a current weight of 300 kg, a body temperature of 38.5 °C, a heart rate of 70 beats per minute, and a respiratory rate of 20 breaths per minute. These information will be used as the input data for the decision tree algorithm for subsequent decision-making. The depth of the decision tree determines the complexity of the decision. The preset depth of the decision tree should not be too deep or too shallow. An overly deep decision tree is prone to overfitting, reducing the generalization ability of the model; an overly shallow decision tree may not be able to capture the complex relationships in the data, affecting the accuracy of the decision. For example, the depth of the decision tree can be set to 3, which means the decision tree will undergo three splits and finally form 8 leaf nodes. Node splitting is the core step in constructing the decision tree. During the node splitting process, it is necessary to select the attribute with the largest information gain as the splitting node. Information gain reflects the ability of an attribute to distinguish different categories of samples. For example, assume that the samples at the current node include 10 cattle, with 5 healthy and 5 diseased. If splitting according to body temperature, the samples can be divided into two groups: cattle with a body temperature higher than 39 °C and cattle with a body temperature lower than 39 °C. If 4 diseased and 1 healthy are in the high-temperature group, and 1 diseased and 4 healthy are in the low-temperature group, then the information gain of the body temperature attribute is relatively large and is suitable as the splitting node. Recursive splitting is the key technology for constructing the decision tree. Through recursive splitting, the decision rules can be continuously refined to improve the accuracy of the decision. For example, after the first split, the child nodes can continue to be split. Assume that the heart rate is selected as the second splitting node, and the cattle with a heart rate higher than 80 beats per minute and the cattle with a heart rate lower than 80 beats per minute are divided into two groups. And so on until the preset decision tree depth or the node sample quantity threshold is reached. The leaf nodes of the decision tree represent the final decision results. For example, a leaf node may represent "increase the feeding amount of concentrate feed", and another leaf node may represent "reduce the feeding amount of roughage feed". According to the various parameters of the cattle and sheep individuals, it will finally fall on a certain leaf node, thereby determining the corresponding feeding strategy. After the initial feeding decision tree is constructed, it needs to be evaluated and optimized. Cross-validation is a commonly used model evaluation method. For example, the data set can be divided into 10 parts, and each time 9 parts of the data are used to train the model, and the remaining 1 part of the data is used to test the performance of the model. Repeat 10 times and take the average value as the final performance index of the model. If the accuracy and generalization ability of the model are not ideal, then the model parameters need to be adjusted or the decision tree structure needs to be optimized. Through the above steps, an accurate feeding system based on the decision tree algorithm can be constructed.

[0053] S103. During the feeding process, continuously obtain data on the changes in the physiological states of cattle and sheep and the changes in environmental parameters, and dynamically optimize the existing decision tree through an incremental learning method. When the new data causes the prediction accuracy of a certain branch of the decision tree to decrease, prune and regrow this branch to enable the continuous evolution of the decision tree and the dynamic optimization of the feeding strategy.

[0054] Specifically, obtain the real-time change data of the physiological state parameters and environmental parameters of cattle and sheep, and use it as the new training data for incremental learning. Determine whether the new data causes the prediction accuracy of a certain branch of the decision tree model to decrease. If it decreases, trigger the pruning operation on this branch. According to the pruned decision tree branch, use the new data for regrowth training to obtain an optimized decision tree model. Use the optimized decision tree model to conduct real-time prediction and analysis on the physiological state parameters and environmental parameters of cattle and sheep. According to the prediction and analysis results of the decision tree model, dynamically adjust the feeding supply strategy to determine the optimal feeding plan. Continuously conduct incremental learning and decision tree optimization to enable the feeding strategy to adapt to the changes in the physiological states of cattle and sheep and environmental parameters. By continuously iteratively optimizing the decision tree model, achieve the continuous evolution and dynamic optimization of the feeding supply strategy, and improve the breeding efficiency of cattle and sheep.

[0055] In this embodiment, real-time change data of the physiological state parameters and environmental parameters of cattle and sheep are obtained and used as new training data for incremental learning. For example, physiological parameters such as the body temperature, heart rate, respiratory rate, and activity level of cattle and sheep, as well as environmental parameters such as environmental temperature, humidity, and light intensity, can be monitored in real time through sensors. These real-time data can be collected once per hour or every few hours to form a time series data stream, which serves as new training data for incremental learning. The advantage of this is that the model can continuously learn the latest data and adapt in a timely manner to changes in the physiological state of cattle and sheep and environmental parameters. Determine whether the new data causes a decrease in the prediction accuracy of a certain branch of the decision tree model. For example, assume that a branch of the initial decision tree model is: if the body temperature of a cow is higher than 39 degrees Celsius, then increase the supply of concentrate feed. Now, the new data shows that although the body temperature of some cows is higher than 39 degrees Celsius, their appetite has decreased. If the supply of concentrate feed is continued, it will instead cause indigestion. This indicates that the new data has led to a decrease in the prediction accuracy of this branch. At this time, pruning operation needs to be performed on this branch. If the new data causes a decrease in the prediction accuracy of a certain branch of the decision tree model, then trigger the pruning operation on this branch. The pruning operation can be implemented in various ways. For example, this branch can be directly deleted, or this branch can be merged into its parent node. In the above example, the branch "if the body temperature of a cow is higher than 39 degrees Celsius, then increase the supply of concentrate feed" can be modified to "if the body temperature of a cow is higher than 39 degrees Celsius and the appetite is normal, then increase the supply of concentrate feed", or this branch can be directly deleted. The pruning operation can prevent the model from overfitting and improve the generalization ability of the model. According to the pruned decision tree branch, use the new data for re-growth training. In the above example, the new data on the relationship between the body temperature, appetite, and supply of concentrate feed of cows can be used to retrain this branch. For example, by analyzing the new data, the optimal ratio of appetite to the supply of concentrate feed when the body temperature of a cow is higher than 39 degrees Celsius can be found, thereby adjusting the feeding strategy. Re-growth training can enable the model to better adapt to new data and improve the prediction accuracy. Adopt the optimized decision tree model to perform real-time prediction analysis on the physiological state parameters and environmental parameters of cattle and sheep. For example, using the optimized decision tree model, the changes in physiological parameters such as the body temperature and feed intake of cattle and sheep in the next few hours, as well as environmental parameters such as environmental temperature and humidity, can be predicted. These prediction results can provide a basis for dynamically adjusting the feed supply strategy. According to the prediction analysis results of the decision tree model, dynamically adjust the feed supply strategy to determine the optimal feeding plan. For example, if the model predicts that the environmental temperature will rise in the next few hours and the feed intake of cattle and sheep will decrease, then the supply of concentrate feed can be reduced accordingly, the supply of roughage can be increased, and sufficient drinking water can be provided. Dynamically adjusting the feed supply strategy can better meet the nutritional needs of cattle and sheep and improve the feed utilization rate. Continuously perform incremental learning and decision tree optimization so that the feeding strategy can adapt to changes in the physiological state of cattle and sheep and environmental parameters.For example, the physiological state parameters and environmental parameter data of cattle and sheep can be collected regularly (e.g., weekly or monthly) and used to update the decision tree model. Continuous incremental learning and decision tree optimization can enable the feeding strategy to evolve continuously and always maintain the best state. By continuously iteratively optimizing the decision tree model, the continuous evolution and dynamic optimization of the feed supply strategy are achieved, improving the breeding efficiency of cattle and sheep. By continuously learning new data and dynamically adjusting the feed supply strategy according to the prediction results, the nutritional needs of cattle and sheep can be better met, the feed utilization rate can be improved, thereby increasing the growth rate and health level of cattle and sheep, and ultimately improving the breeding efficiency of cattle and sheep.

[0056] S104. Use the Q-learning algorithm in reinforcement learning, and take the effect of each feeding decision as feedback. Combine the physiological state and environmental parameters of cattle and sheep as the state, and different feeding schemes as actions. Through the update of the Q-value table, learn which feeding scheme should be adopted in which state to obtain the globally optimal feeding strategy.

[0057] According to the physiological state parameters of cattle and sheep (such as body weight, health indicators, etc.) and environmental parameters (such as temperature, humidity, etc.), construct a state space and combine them as the state representation of the Q-learning algorithm. Take different feeding schemes (such as feed types, feeding times, feeding amounts, etc.) as the action space of the Q-learning algorithm. Initialize the Q-value table to store the Q-value estimates of each state-action pair. The rows of the Q-value table represent states, and the columns represent actions. At each decision time point, according to the current state, use the ε-greedy strategy to select a feeding action. That is, randomly select an action with probability ε, and select the action with the largest Q value with probability 1 - ε. Execute the selected feeding action and observe the changes in the state of cattle and sheep and the immediate rewards obtained (such as increased feed intake, accelerated growth rate, etc.). According to the observed state transition and immediate reward, use the update formula of Q-learning to update the Q-value estimate of the corresponding state-action pair in the Q-value table. Repeat steps 4 - 6 to continuously update the Q-value table until the Q value converges or reaches the preset number of training rounds. The finally obtained Q-value table is the learned globally optimal feeding strategy.

[0058] In this embodiment, first, by collecting the physiological state parameters of cattle and sheep such as weight and health indicators, as well as environmental parameters such as temperature and humidity, a detailed state space is constructed. The combination of these parameters defines each specific state. For example, a sheep weighing 50 kg in an environment with a temperature of 15 degrees and a humidity of 60% can be regarded as a unique state. Then, the action space is defined, including different types of feed, feeding time, and feeding amount. For example, an action can be to feed 1 kg of corn in the morning or 1.5 kg of forage in the afternoon. Each action corresponds to one or more states in the state space. During the implementation process, a Q-value table is initialized. This is a two-dimensional table where the rows correspond to states and the columns correspond to actions. The Q-value table is used to store the utility estimates for each state-action pair. Initially, these values can be random or set based on prior knowledge. At each decision time point, according to the current state, the ε-greedy strategy is used to select an action. This means that with a small probability ε (e.g., 0.1), the system will randomly select an action, which helps to explore the benefits that new actions may bring; while most of the time (probability 1 - ε), the system will select the currently known optimal action, that is, the action with the highest Q-value. After executing the selected action, observe the changes in the state of cattle and sheep and the immediate reward obtained. For example, if the selected feeding amount increases the feed intake of cattle and sheep, thereby accelerating the growth rate, this action will receive a positive immediate reward. According to the observed state transition and immediate reward, use the update formula of Q-learning to adjust the Q-value of the corresponding state-action pair in the Q-value table. This update formula takes into account the old Q-value, the learning rate, the immediate reward, and the maximum Q-value of the future state. By continuously repeating this process, the Q-value table gradually converges, reflecting the long-term value of each action in each state. Finally, this Q-value table can guide the decision maker to select the optimal feed supply strategy to maximize the overall health and growth efficiency of cattle and sheep. The advantage of this method is that it can dynamically adapt to changes in the environment and the state of cattle and sheep, continuously optimize the decision-making process, and ensure that the feed strategy is always in the best state. Through this intelligent feed management, the breeding efficiency and economic benefits can be significantly improved, while ensuring animal welfare.

[0059] S105. Set the constraint conditions for the feeding strategy, including nutritional balance and cost control factors. When generating a feeding decision plan, judge whether the plan meets the constraint conditions. If it does not meet, regenerate it until a feasible optimal plan is obtained.

[0060] Specifically, according to the constraint conditions of the feeding strategy, including nutritional balance and cost control factors, a feeding decision optimization model is established. A heuristic algorithm, such as a genetic algorithm, is used to generate an initial feeding decision plan. For the generated feeding decision plan, it is judged whether it meets the constraint conditions of nutritional balance and cost control. If the plan does not meet the constraint conditions, new feeding decision plans are regenerated through the selection, crossover, and mutation operations of the genetic algorithm. Repeat steps 3 and 4 until a feasible feeding decision plan that meets the constraint conditions is generated. Among the feasible plans that meet the constraint conditions, the nutritional balance and cost control of each plan are evaluated through the objective function, and the plan with the optimal comprehensive score is selected as the final feeding decision plan. The optimal feeding decision plan is output and applied to the actual feed formulation and supply process to achieve the goals of nutritional balance and cost control.

[0061] In this embodiment, the core of the feeding decision optimization model lies in balancing the nutritional requirements of cattle and sheep and the breeding costs. The inputs of the model include the nutritional components and prices of various feeds, as well as the physiological states and growth stages of cattle and sheep. The output of the model is the optimal feed ratio plan, which can not only meet the nutritional requirements of cattle and sheep but also control the costs within a reasonable range. For example, according to information such as the weight, age, and breed of cattle and sheep, the minimum and maximum values of nutritional indicators such as protein, energy, and minerals required by them can be set. At the same time, the prices of various feeds need to be considered, such as corn, soybean meal, alfalfa, etc., and their nutritional component contents. Heuristic algorithms, such as genetic algorithms, can be used to generate initial feeding decision plans. The genetic algorithm simulates the evolution process in nature and continuously optimizes the plan through operations such as selection, crossover, and mutation. For example, the dosage of each feed can be regarded as a gene, and a set of feed ratio plans can be regarded as a chromosome. Initially, multiple chromosomes can be randomly generated to represent different feeding plans. The generated feeding decision plans need to meet the constraint conditions of nutritional balance and cost control. For example, the protein content in a plan must reach the minimum value required by cattle and sheep, and at the same time, the cost should be controlled within the budget. Suppose an adult beef cattle needs 300 grams of protein per day, and the set budget is 10 yuan of feed cost per day. Then, a plan containing 5 kg of corn and 1 kg of soybean meal is not feasible if the total protein content is less than 300 grams or the cost is higher than 10 yuan. If the plan does not meet the constraint conditions, the genetic algorithm needs to be used for optimization. The selection operation will preferentially select better plans, such as plans with higher protein content and lower cost. The crossover operation will exchange some genes of two plans. For example, the corn dosage of plan A will be exchanged with the soybean meal dosage of plan B to generate a new plan. The mutation operation will randomly change a certain gene in the plan. For example, the corn dosage of plan A will be increased or decreased by a certain proportion. Through these operations, new plans can be continuously generated and gradually approach the optimal solution. This process will be repeated continuously until a feasible plan that meets the constraint conditions is generated. For example, after multiple rounds of iteration, a plan may be found, containing 4 kg of corn, 1.5 kg of soybean meal, and 0.5 kg of alfalfa, which not only meets the requirement of 300 grams of protein per day for cattle and sheep but also controls the cost within 10 yuan. Among the feasible plans that meet the constraint conditions, the plan with the optimal comprehensive score needs to be selected as the final feeding decision plan. Each plan can be scored according to nutritional balance and cost control. For example, the closer the protein content of a plan is to the ideal requirement value of cattle and sheep, the higher the score; the lower the cost, the higher the score. Finally, the plan with the highest score is selected as the final feeding decision plan. For example, among multiple feasible plans, a plan with a protein content of 310 grams and a cost of 9.5 yuan may have the highest score. Finally, the optimal feeding decision plan is output and applied to the actual feed ratio and supply process.For example, input the finally determined feed ratio plan into the automatic feeding system, and the system will automatically mix and dispense the feed according to the plan to achieve the goals of nutritional balance and cost control. In this way, not only can the healthy growth of cattle and sheep be ensured, but also the breeding efficiency can be improved.

[0062] S106. Convert the optimized feeding strategy into a feeding plan executable by the electronic feeding station, and send the plan to each electronic feeding station through the wireless communication module. The electronic feeding station feeds back the execution status to the central processing unit in real time. When a device failure occurs, the central processing unit senses it in time and adjusts the feeding plan.

[0063] Specifically, according to the optimized feeding strategy, the central processing unit generates an executable feeding plan. The feeding plan includes parameters such as the feeding time and feeding amount of each electronic feeding station. Through the wireless communication module, the generated feeding plan is sent to the corresponding electronic feeding station, and the electronic feeding station receives and stores the feeding plan. The electronic feeding station controls the feed dispensing device to dispense the specified amount of feed at the specified time according to the received feeding plan, and monitors the operation status of the equipment during the feeding process. During the feeding process, the electronic feeding station collects the equipment operation parameters in real time, such as the motor current and the signal of the level sensor, and uploads the collected data to the central processing unit through the wireless communication module. The central processing unit receives the equipment operation data uploaded by the electronic feeding station, analyzes the data using an anomaly detection algorithm, and determines whether there is a device failure or an abnormal situation. If a device failure is detected, the central processing unit immediately generates an adjusted feeding plan, sends the new feeding plan to the faulty electronic feeding station through the wireless communication module to guide it to take emergency measures, such as stopping feeding and sending an alarm signal. At the same time, the central processing unit also sends the adjusted feeding plan to other normally operating electronic feeding stations to dynamically optimize the feeding plan of the entire farm and ensure the normal progress of the breeding production.

[0064] In this embodiment, the central processing unit, similar to the "brain" of a farm, is responsible for formulating and issuing action plans for all feeding stations. It generates specific executable feeding plans according to the previously optimized feeding strategy. For example, Feed 50 kg of feed at Feeding Station No. 1 at 10:00, Feed 60 kg of feed at Feeding Station No. 2 at 10:30, Feed 55 kg of feed at Feeding Station No. 3 at 11:00, and so on. These plans contain key parameters such as the feeding time and feeding amount of each electronic feeding station, ensuring that each feeding station can execute according to the plan. The feeding plan is sent to each electronic feeding station through the wireless communication module. The wireless communication module can be understood as the "neural network" of the farm, responsible for transmitting the instructions of the central processing unit. For example, the central processing unit sends the instruction to feed 50 kg of feed at Feeding Station No. 1 at 10:00 to the receiver of Feeding Station No. 1 through a wireless network (such as WIFI, ZigBee, etc.). Each electronic feeding station has a receiver and a memory for receiving and storing the feeding plan, just like each feeding station has its own "task list". The electronic feeding station is like an "executor", and according to the received "task list", it controls the feed delivery device to deliver the specified amount of feed at the specified time. For example, after receiving the instruction, Feeding Station No. 1 will start the motor to drive the screw rod and deliver 50 kg of feed into the corresponding trough at 10:00. At the same time, the electronic feeding station will also monitor the operating status of the equipment during the feeding process, such as the motor current, the signal of the level sensor, etc., to ensure the smooth progress of the feeding process. The level sensor can monitor the remaining amount of feed to prevent feed shortage or overflow. The electronic feeding station collects the equipment operating parameters in real time and uploads them to the central processing unit through the wireless communication module. This is like an "executor" reporting the work progress and equipment status to the "brain". For example, Feeding Station No. 1 will upload data such as the motor current and the signal of the level sensor to the central processing unit in real time. After receiving the data uploaded by the electronic feeding station, the central processing unit will use an anomaly detection algorithm to analyze the data and judge whether there is a device failure or abnormal situation. The anomaly detection algorithm is like a "doctor" who can diagnose the "health" status of the equipment. For example, if the motor current of Feeding Station No. 1 suddenly increases and exceeds the preset threshold, the anomaly detection algorithm will judge that the motor may be faulty and immediately issue an alarm. If a device failure is detected, the central processing unit will immediately generate an adjusted feeding plan. For example, if the motor of Feeding Station No. 1 fails, the central processing unit will immediately stop the feeding task of Feeding Station No. 1 and allocate its unfinished feeding task to other normal feeding stations, such as allocating the remaining feed to Feeding Station No. 2 and Feeding Station No. 3. The central processing unit sends the new feeding plan to the faulty electronic feeding station to guide it to take emergency measures. For example, the central processing unit will send instructions such as stopping feeding and sending an alarm signal to Feeding Station No. 1.Meanwhile, the central processing unit will also send the adjusted feeding plan to other normally operating electronic feeding stations. For example, it will notify the 2nd and 3rd feeding stations to increase the feeding amount to make up for the absence of the 1st feeding station, ensuring that the feeding plan for the entire farm can be dynamically adjusted to maintain the normal progress of farming production. The advantage of this is that it can improve farming efficiency, reduce labor costs, and promptly detect and handle equipment failures to avoid greater losses.

[0065] This embodiment also provides an intelligent optimization system for multi-region feeding decision-making for cattle and sheep, including:

[0066] A data acquisition module for acquiring individual parameters and environmental parameters of cattle and sheep;

[0067] An initial feeding strategy generation module for constructing an initial feeding strategy according to individual parameters and environmental parameters through a decision tree algorithm;

[0068] An optimization module for acquiring dynamic physiological parameters and environmental parameters of cattle and sheep, dynamically optimizing the initial feeding strategy according to the incremental learning method and the dynamic physiological parameters and environmental parameters, and obtaining the globally optimal feeding strategy according to the Q-learning algorithm;

[0069] An implementation module for determining whether the globally optimal feeding strategy meets the preset constraint conditions, and when the constraint conditions are met, converting the globally optimal feeding strategy into an execution plan and monitoring the execution.

[0070] Furthermore, the data acquisition module includes:

[0071] An identity identification acquisition unit for acquiring identity identification through an electronic ear tag, where the identity identification includes date of birth, breed, gender, parental information, and immunization records;

[0072] A physiological parameter acquisition unit for acquiring physiological parameters through an intelligent sensor, where the physiological parameters include body temperature, heart rate, and respiratory rate;

[0073] An environmental parameter acquisition unit for acquiring environmental parameters through an environmental sensor, where the environmental parameters include temperature, humidity, light intensity, and air quality.

[0074] Furthermore, the implementation module includes:

[0075] A judgment unit for determining whether the globally optimal feeding strategy meets the preset constraint conditions;

[0076] An adjustment unit for, when the globally optimal feeding strategy does not meet the constraint conditions, regenerating a new feeding decision plan through selection, crossover, and mutation operations of the genetic algorithm until the generated new feeding decision plan meets the constraint conditions to obtain the globally optimal feeding strategy that meets the constraint conditions;

[0077] An execution unit, configured to convert a globally optimal feeding strategy into an execution plan and monitor the execution when a constraint condition is satisfied.

[0078] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory, where the processor executes the computer program to implement the steps of the method.

[0079] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method are implemented.

[0080] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent optimization method for multi-region feeding decision-making for cattle and sheep, characterized in that: The following steps are involved: Obtain individual parameters and environmental parameters of cattle and sheep; According to the individual parameters and environmental parameters, an initial feeding strategy is constructed by a decision tree algorithm; Acquire dynamic physiological parameters and environmental parameters of cattle and sheep, dynamically optimize the initial feeding strategy according to the incremental learning method and the dynamic physiological parameters and environmental parameters, and acquire the global optimal feeding strategy according to the Q-learning algorithm; It is determined whether the global optimal feeding strategy satisfies the preset constraints. When the constraints are satisfied, the global optimal feeding strategy is converted into an execution plan and the execution is monitored.

2. The method according to claim 1, characterized in that Obtaining individual and environmental parameters of cattle and sheep include; The individual parameters include identity identification and physiological parameters; Obtaining identification through electronic ear tags, said identification including date of birth, breed, sex, parent information and immunization records; Acquiring physiological parameters through intelligent sensors, wherein the physiological parameters include body temperature, heart rate and respiratory rate; Environmental parameters are acquired through environmental sensors, including temperature, humidity, light intensity and air quality.

3. The method according to claim 1, characterized in that Dynamic optimization of the initial feeding strategy includes: The real-time change data of the physiological parameters and environmental parameters of cattle and sheep are obtained as the new training data for incremental learning; it is determined whether the new training data leads to a decrease in the prediction accuracy of a branch of the decision tree model. If so, the branch is pruned; based on the pruned decision tree branches, the new data is used for regrowth training to obtain the optimized decision tree model.

4. The method according to claim 1, characterized in that: The global optimal feeding strategy obtained according to the Q-learning algorithm includes: The combination of the physiological parameters and environmental parameters of cattle and sheep is taken as the state, and different feeding plans are taken as actions. By updating the Q value table, the corresponding feeding plans that should be taken in different states are learned to obtain the global optimal feeding strategy.

5. The method according to claim 1, characterized in that When the global optimal feeding strategy does not meet the constraints, a new feeding decision plan is regenerated through the selection, crossover and mutation operations of the genetic algorithm until the generated new feeding decision plan meets the constraints and the global optimal feeding strategy that meets the constraints is obtained.

6. An intelligent optimization system for multi-region feeding decision-making for cattle and sheep, characterized in that: include: A data acquisition module is used to obtain individual parameters and environmental parameters of cattle and sheep; An initial feeding strategy generation module, used to construct an initial feeding strategy through a decision tree algorithm according to the individual parameters and environmental parameters; An optimization module is used to obtain dynamic physiological parameters and environmental parameters of cattle and sheep, dynamically optimize the initial feeding strategy according to the incremental learning method and the dynamic physiological parameters and environmental parameters, and obtain the global optimal feeding strategy according to the Q-learning algorithm; The implementation module is used to determine whether the global optimal feeding strategy meets the preset constraints. When the constraints are met, the global optimal feeding strategy is converted into an execution plan and the execution is monitored.

7. The system according to claim 6, characterized in that The data acquisition module comprises: An identity acquisition unit, used to acquire an identity through an electronic ear tag, wherein the identity includes date of birth, breed, sex, parent information and immunization record; A physiological parameter acquisition unit, used to acquire physiological parameters through intelligent sensors, wherein the physiological parameters include body temperature, heart rate and respiratory rate; The environmental parameter acquisition unit is used to acquire environmental parameters through environmental sensors, and the environmental parameters include temperature, humidity, light intensity and air quality.

8. The system according to claim 6, characterized in that The implementation module includes: A judgment unit, used to judge whether the global optimal feeding strategy meets the preset constraint conditions; The adjustment unit is used to regenerate a new feeding decision plan through the selection, crossover and mutation operations of the genetic algorithm when the global optimal feeding strategy does not meet the constraint conditions, until the generated new feeding decision plan meets the constraint conditions, thereby obtaining the global optimal feeding strategy that meets the constraint conditions; The execution unit is used to convert the global optimal feeding strategy into an execution plan and monitor the execution when the constraint conditions are met.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images