Feed formula optimization and regulation method and system and storage medium

Abstract

The invention provides a feed formula optimization, regulation and control method and system and a storage medium, and specifically, the method comprises the steps: building a nutritional ingredient probability distribution model to represent volatility through obtaining the cost, basic nutritional ingredients and historical batch data of to-be-selected feed raw materials; setting a standard nutritional requirement interval according to the physiological stage of the target cattle; constructing an interaction matrix to represent the synergistic and antagonistic effects of the nutritional ingredients, and calculating a formula nutritional ingredient vector; establishing an optimization model which takes the lowest total cost as a target and has the probability that components meet nutritional requirements greater than or equal to a confidence threshold; an improved particle swarm algorithm is adopted for solving, a learning factor is adjusted based on a comprehensive interaction effect index calculated by an interaction matrix, and the index reflects the interaction strength of the nutritional ingredients in the current formula.

Description

Technical Field

[0001] This application belongs to the field of feed formulation optimization and control, and in particular relates to a method, system and storage medium for feed formulation optimization and control. Background Technology

[0002] In animal husbandry, especially in intensive cattle farming, feed costs account for the majority of total production costs. Optimization of cattle feed formulations commonly employs mathematical programming methods, such as linear programming. These methods typically use the minimum total feed cost as the objective function, setting the nutritional requirements of cattle at specific physiological stages as a series of linear constraints to solve for the optimal combination ratio of various feed ingredients. However, due to factors such as origin, harvest season, processing technology, and storage conditions, the nutrient content of the same feed ingredient inevitably fluctuates randomly between different batches. Formulas based on average values ​​may, in actual production, deviate from the preset standard range due to the uncertainty of raw materials, failing to consistently meet the nutritional needs of cattle and thus affecting their growth, development, and production performance. Furthermore, existing feed formulation optimization technologies largely ignore the complex interactions of various nutrients during animal digestion and absorption when constructing models. Traditional models are generally based on a simplified assumption that various nutrients are independent, and the total effect is equal to the simple linear sum of the effects of each component. However, synergistic and antagonistic effects are widespread among nutrients, and linear superposition models cannot accurately reflect the amount of nutrients that cattle can absorb and utilize. Although some studies in recent years have attempted to apply intelligent optimization algorithms such as genetic algorithms and particle swarm optimization to solve feed formulation problems, the adjustment of internal parameters of these algorithms often relies on general strategies and is not combined with the characteristics of the formulation itself. This results in low efficiency of the algorithms when searching complex nutrient spaces, making it difficult to obtain a globally optimal formulation that balances cost and nutritional stability. Summary of the Invention

[0003] This invention proposes a method for optimizing and regulating feed formulation to address the problem that existing methods cannot reliably meet the nutritional needs of cattle and neglect the complex interactions of various nutrients during animal digestion and absorption. The method includes the following steps: Obtain the cost, basic nutrient composition, and historical batch data of the candidate feed ingredients, and establish a probability distribution model for each nutrient component of each ingredient; determine the standard nutrient requirement range that includes the minimum and maximum limits of each nutrient component based on the physiological stage of the target cattle. An interaction matrix representing the synergistic and antagonistic effects of different nutrients is constructed, and a nutrient vector provided by the formula is calculated based on the interaction matrix and the ratio of each ingredient in any formula. An optimization model is established with the objective function of minimizing the total cost of the feed formulation and the constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not lower than a preset confidence threshold. An improved particle swarm optimization algorithm is used to solve the optimization model and obtain the optimal ratio of each feed ingredient.

[0004] Optionally, the step of obtaining the cost, basic nutritional components, and historical batch data of the candidate feed ingredients, and establishing a probability distribution model for each nutrient component of each ingredient, includes: For each feed ingredient, collect the content test values ​​of each nutrient from at least 30 historical batches; The maximum likelihood estimation method is used to fit the content detection value to a normal distribution, and the expected value and standard deviation of the nutrient content are calculated, which are used as parameters of the probability distribution model.

[0005] Optionally, determining the standard nutritional requirement range, which includes the minimum and maximum limits of each nutrient component, based on the physiological stage of the target cattle includes: The physiological stages of the target cattle are divided into calf stage, growing stage, and fattening stage; For each stage, consult the beef cattle feeding standards to obtain the recommended intake ranges for 12 key nutrients, and set the lower limit of the range as the minimum limit and the upper limit as the maximum limit.

[0006] Optionally, the construction of the interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients includes: Construct an N×N square matrix M, where N is the number of different types of nutrients; The value of the element in the i-th row and j-th column of the matrix This represents the factor influencing the absorption rate of nutrient i on nutrient j, where, when i=j, =1; When nutrient i has a synergistic effect on j Set to greater than 1; When nutrient i has an antagonistic effect on j Set to less than 1; When there is no interaction between the two Set to 1.

[0007] Optionally, calculating the nutrient vector provided by the formula based on the interaction matrix and the proportions of each ingredient in any formula includes: Based on the proportions of each raw material and the probability distribution model of nutrient content, the nominal nutrient vector is calculated. Expected value vector Sum of variance vectors ; nutrient vector The components of the vector approximate a normal distribution, and the expected value vector is... and the variance of each component Calculated using the following formula: , ; Where M is the interaction matrix, Let be the transpose of M, where i and j are the indices of the nutrients.

[0008] Optionally, the constraint condition that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not lower than a preset confidence threshold includes: For each component in the nutrient vector, the probability that the content value falls within the standard nutrient requirement interval consisting of the minimum and maximum limits is calculated using the cumulative distribution function of the normal distribution model determined by the expected value and variance, and the probability value is required to be no less than the preset confidence threshold.

[0009] Optionally, the learning factor of the improved particle swarm optimization algorithm is adjusted according to a comprehensive interaction effect index calculated based on the interaction matrix, including: Calculate the arithmetic mean of all off-diagonal elements of the interaction matrix M, and use it as the comprehensive interaction effect index S; After each iteration, according to the formula and Update learning factors; among which, , This is the initial learning factor.

[0010] Furthermore, the present invention also relates to a feed formulation optimization and control system, comprising the following modules: The first module is used to obtain the cost, basic nutritional components and historical batch data of the candidate feed ingredients, and to establish a probability distribution model for each nutrient component of each ingredient; based on the physiological stage of the target cattle, it determines the standard nutritional requirement range that includes the minimum and maximum limits of each nutrient component. The calculation module is used to construct an interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients, and to calculate the nutrient vector provided by the formula based on the interaction matrix and the ratio of each raw material in any formula. The second module is used to establish an optimization model with the objective function of minimizing the total cost of the feed formulation and with the constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not lower than a preset confidence threshold. An adjustment module is used to solve the optimization model using an improved particle swarm optimization algorithm.

[0011] Preferably, the step of obtaining the cost, basic nutritional components, and historical batch data of the candidate feed ingredients, and establishing a probability distribution model for each nutrient component of each ingredient, includes: For each feed ingredient, collect the content test values ​​of each nutrient from at least 30 historical batches; The maximum likelihood estimation method is used to fit the content detection value to a normal distribution, and the expected value and standard deviation of the nutrient content are calculated, which are used as parameters of the probability distribution model.

[0012] Preferably, determining the standard nutritional requirement range, which includes the minimum and maximum limits of each nutrient component, based on the physiological stage of the target cattle includes: The physiological stages of the target cattle are divided into calf stage, growing stage, and fattening stage; For each stage, consult the beef cattle feeding standards to obtain the recommended intake ranges for 12 key nutrients, and set the lower limit of the range as the minimum limit and the upper limit as the maximum limit.

[0013] Preferably, the construction of the interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients includes: Construct an N×N square matrix M, where N is the number of different types of nutrients; The value of the element in the i-th row and j-th column of the matrix This represents the factor influencing the absorption rate of nutrient i on nutrient j, where, when i=j, =1; When nutrient i has a synergistic effect on j Set to greater than 1; When nutrient i has an antagonistic effect on j Set to less than 1; When there is no interaction between the two Set to 1.

[0014] Preferably, the step of calculating the nutrient vector provided by the formula based on the interaction matrix and the proportions of each ingredient in any formula includes: Based on the proportions of each raw material and the probability distribution model of nutrient content, the nominal nutrient vector is calculated. Expected value vector Sum of variance vectors ; nutrient vector The components of the vector approximate a normal distribution, and the expected value vector is... and the variance of each component Calculated using the following formula: , ; Where M is the interaction matrix, Let be the transpose of M, where i and j are the indices of the nutrients.

[0015] Preferably, the constraint condition that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not less than a preset confidence threshold includes: For each component in the nutrient vector, the probability that the content value falls within the standard nutrient requirement interval consisting of the minimum and maximum limits is calculated using the cumulative distribution function of the normal distribution model determined by the expected value and variance, and the probability value is required to be no less than the preset confidence threshold.

[0016] Preferably, the learning factor of the improved particle swarm optimization algorithm is adjusted according to a comprehensive interaction effect index calculated based on the interaction matrix, including: Calculate the arithmetic mean of all off-diagonal elements of the interaction matrix M, and use it as the comprehensive interaction effect index S; After each iteration, according to the formula and Update learning factors; among which, , This is the initial learning factor.

[0017] This invention establishes a probability distribution model for the nutrient components of feed ingredients, representing the content fluctuations caused by batch variations. Using the probability of meeting nutritional requirements within a given range as a core constraint, it ensures high reliability and quality stability of the formula in practical applications, reducing production risks caused by raw material fluctuations. An interaction matrix between nutrient components is constructed, enabling the calculation of the actual amount of nutrients that cattle can absorb, improving the biological accuracy of the formula. By integrating biological interactions into the optimization algorithm's solution process, the search process becomes more targeted, enabling a more accurate search for the global optimum. Therefore, this invention provides a nutritionally stable cattle feed formula that better meets the physiological needs of cattle, thereby improving farming efficiency. Attached Figure Description

[0018] Figure 1 A flowchart of the first embodiment; Figure 2 This is a schematic diagram of the probability distribution model of the nutrient content of raw materials; Figure 3 A schematic diagram showing the standard nutritional requirements range for target cattle. Figure 4 A schematic diagram illustrating the probability constraints for meeting nutritional requirement ranges; Figure 5 A schematic diagram of the learning factor adjustment strategy. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] In the first embodiment, the present invention proposes a method for optimizing and controlling feed formulation, such as... Figure 1 This includes the following steps: S1. Obtain the cost, basic nutrient composition, and historical batch data of the candidate feed ingredients, and establish a probability distribution model for each nutrient component of each ingredient; determine the standard nutrient requirement range that includes the minimum and maximum limits of each nutrient component based on the physiological stage of the target cattle. Specifically, the latest procurement costs of various candidate feed ingredients, such as corn, soybean meal, and alfalfa, are collected from suppliers. Laboratory test reports for at least 100 different batches of these ingredients are compiled, and data on the content of key basic nutrients such as crude protein, metabolizable energy, calcium, and phosphorus are extracted. For each nutrient component of each ingredient, such as the crude protein content of corn, statistical methods are used to fit the distribution of the crude protein content data. If the test results indicate that the data conforms to a normal distribution, a normal distribution model is established with the mean of historical data as the expected value and the standard deviation as the fluctuation parameter, thus mathematically representing the random uncertainty of the nutrient content.

[0021] Select a target cow, such as a Holstein dairy cow weighing 600 kg at its peak lactation. Consult authoritative feeding standards, such as the dairy cow feeding standards published by the National Research Council, to find the daily requirements of various nutrients for cows at this physiological stage. For example, the standard stipulates that the minimum daily metabolizable energy requirement for cows is 35 Mcal, and the maximum should not exceed 45 Mcal to prevent excessive obesity; the minimum crude protein requirement is 2800g, and the maximum should not exceed 3500g to avoid metabolic burden. Combine these upper and lower limits to obtain a multidimensional standard nutrient requirement interval vector.

[0022] In an optional embodiment, the step of acquiring the cost, basic nutritional components, and historical batch data of the candidate feed ingredients, and establishing a probability distribution model for each nutrient component of each ingredient, includes: For each feed ingredient, collect the content test values ​​of each nutrient from at least 30 historical batches; The maximum likelihood estimation method is used to fit the content detection value to a normal distribution, and the expected value and standard deviation of the nutrient content are calculated, which are used as parameters of the probability distribution model.

[0023] Taking corn as an example, crude protein content test data from the most recent 30 batches were collected, resulting in a set of 30 data points, such as 8.5%, 8.7%, and 8.4%. A normal distribution was chosen as the model to represent the fluctuations in the data because nutrient content typically varies randomly around a central value. Maximum likelihood estimation was used to determine the normal distribution parameters that best represent this set of data.

[0024] The 30 crude protein content data points were processed using maximum likelihood estimation to calculate the mean (expected value) and standard deviation of the data set. Assuming the expected value is 8.55% and the standard deviation is 0.15%, these two values ​​constitute the probability distribution model of the crude protein content in corn. This process will be repeated for all other nutrients in corn, such as fat and fiber, and further, independent probability distribution models will be established for each nutrient in all used feed ingredients, such as soybean meal and wheat bran. Figure 2 .

[0025] In an optional embodiment, determining the standard nutritional requirement range, which includes minimum and maximum limits for each nutrient component, based on the physiological stage of the target cattle includes: The physiological stages of the target cattle are divided into calf stage, growing stage, and fattening stage; For each stage, consult the beef cattle feeding standards to obtain the recommended intake ranges for 12 key nutrients, and set the lower limit of the range as the minimum limit and the upper limit as the maximum limit.

[0026] For example, when formulating feed for beef cattle entering the weight gain phase, corresponding to the fattening period, refer to the beef cattle feeding standards and find the nutritional recommendations for fattening cattle with a target weight of 500 kg. Find the recommended daily intake ranges for 12 key nutrients from the standards, such as... Figure 3 For example, the standard might specify that the daily crude protein intake of beef cattle at this stage should be between 3.0 kg and 3.5 kg. Therefore, 3.0 kg would be set as the minimum limit for crude protein, and 3.5 kg as the maximum limit. Similarly, for metabolizable energy, the recommended range might be 30 MPa to 33 MPa, thus establishing the minimum and maximum limits for metabolizable energy. This process will complete the assessment of all 12 key nutrients, thereby establishing a complete set of standard nutritional requirement ranges for fattening beef cattle.

[0027] S2, construct an interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients, and calculate the nutrient vector provided by the formula based on the interaction matrix and the ratio of each raw material in any formula. Specifically, based on animal nutrition literature and experimental data, an N×N interaction matrix M is established, where N represents the quantity of nutrients. Matrix elements This represents the effect of nutrient j on the absorption efficiency of nutrient i: a positive value indicates synergistic promotion, such as the promoting effect of vitamin D on calcium absorption; a negative value indicates antagonistic inhibition, such as the inhibitory effect of high calcium on zinc absorption; and zero indicates no significant interaction. For a given feed formulation, the nominal nutrient vector X provided by the formulation is calculated using the proportions of each ingredient and the mean nutrient composition. The interaction matrix A is used to correct the vector X, and the nutrient composition is then calculated. For example, calcium content. equal to nominal calcium content Multiply by a correction factor, which is calculated using the corresponding coefficients in the interaction matrix based on the vitamin D and zinc content in the formula. The content of all nutrients together constitutes the nutrient vector Y.

[0028] In an optional embodiment, constructing the interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients includes: Construct an N×N square matrix M, where N is the number of different types of nutrients; The value of the element in the i-th row and j-th column of the matrix This represents the factor influencing the absorption rate of nutrient i on nutrient j, where, when i=j, =1; When nutrient i has a synergistic effect on j Set to greater than 1; When nutrient i has an antagonistic effect on j Set to less than 1; When there is no interaction between the two Set to 1.

[0029] Assuming we focus on three nutrients: calcium, phosphorus, and zinc, N=3, requiring a 3x3 matrix M. The rows and columns of the matrix represent calcium, phosphorus, and zinc in sequence. According to the rules, the diagonal elements of the matrix... , , All values ​​are set to 1, indicating that the nutrient has no effect on its absorption. For example, it is known that excess calcium inhibits zinc absorption, which is an antagonistic effect; therefore, the element representing the effect of calcium on zinc... Set to less than 1. Phosphorus promotes calcium absorption in a synergistic way; therefore, the element representing phosphorus's effect on calcium is... Set to greater than 1. If there is no known interaction between phosphorus and zinc, then the value representing the effect of phosphorus on zinc is... and representing the effect of zinc on phosphorus All values ​​are set to 1. By filling the entire matrix in the above manner, an interaction matrix M representing the interaction relationships between nutrients is obtained. The interaction matrix reflects the linearized absorption effect.

[0030] To simplify calculations, it is assumed that the fluctuations in the content of various nutrients in different raw materials are independent of each other, and that the nutrient content approximately follows a normal distribution. In an optional embodiment, the calculation of the nutrient vector provided by the formula based on the interaction matrix and the proportion of each raw material in any formula includes: Based on the proportions of each raw material and the probability distribution model of nutrient content, the nominal nutrient vector is calculated. Expected value vector Sum of variance vectors ; nutrient vector The components of the vector approximate a normal distribution, and the expected value vector is... and the variance of each component Calculated using the following formula: , ; Where M is the interaction matrix, Let be the transpose of M, where i and j are the indices of the nutrients.

[0031] Suppose a feed formulation contains 60% corn and 40% soybean meal. First, calculate the expected value and variance of each nutrient component in the mixed feed; these are called nominal values. For example, the nominal expected value of crude protein is 60% of the expected value of corn plus 40% of the expected value of soybean meal, and the nominal variance is the variance of corn. Double the variance of soybean meal This calculation is performed on all nutrients to obtain a vector of nominal expected values. and nominal variance vector .

[0032] The interaction matrix M is used to adjust nominal values, reflecting the absorption effects between nutrients. Nutrient expectation vector. By transposing the interaction matrix With the nominal expected value vector The results are obtained by multiplication. The calculation adjusts the expected value of each nutrient component upwards or downwards based on synergistic or antagonistic effects. The variance of each nutrient component is calculated. Nutritional composition vector Its components follow a normal distribution, and are determined by the corrected expected value. and variance Full description. The above formula considers how fluctuations in other nutrients affect the current nutrient volatility through interactions. This yields a new set of expected values ​​and variances, representing the probability distribution of nutrients that cattle can actually absorb and utilize.

[0033] Since the contents of each nutrient are interrelated, in another embodiment, calculating the nutrient vector provided by the formula based on the interaction matrix and the proportions of each ingredient in any formula includes: Based on the proportions of each raw material and the probability distribution model of nutrient content, the nominal nutrient vector is calculated. Expected value vector Sum of variance-covariance matrix ; nutrient vector The components of are approximately normally distributed, and their expected value vector is... Sum of variance vectors Calculated using the following approximate model: ; Where i and j are indices of the nutrients, and Let be the expected value and variance of the i-th effective nutrient component, respectively. for The variance of the formula is as follows. The nutrient vector provided by the formula is not a fixed numerical vector, but a random vector, in which the content of each component, that is, each nutrient, follows a probability distribution, which is approximated as a normal distribution in this invention.

[0034] S3. An optimization model is established with the objective function of minimizing the total cost of the feed formulation and the constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not lower than a preset confidence threshold. The optimization model takes the raw material ratio of feed formulation as the core decision variable and is constructed in a hierarchical manner according to the definition of objective function, the quantification of constraints, and the qualification judgment, forming a closed-loop optimization framework that takes into account both economic efficiency and nutritional reliability. The model takes minimizing the sum of the products of raw material unit price and ratio as the core objective function and takes the probability of the effective nutrient component vector meeting the standard demand range as the hard constraint, and realizes the formulation screening through quantitative calculation and logical judgment.

[0035] Specifically, let the proportions of various raw materials in the feed formulation be the decision variables. The objective function is set as the sum of the products of the unit price of all raw materials and their proportions, i.e., minimizing the total cost. The constraints stipulate that for the calculated nutrient composition vector Y, each component... Falling within the standard nutritional requirement range determined in the second step [ , The probability of [the result] must be no less than a pre-set confidence threshold, such as 95%. This indicates that the formula obtained by the model, after considering the fluctuation of raw material content, has a greater than 95% confidence that each nutrient, such as crude protein and calcium, can meet the physiological needs of cattle.

[0036] In an optional embodiment, the constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not less than a preset confidence threshold includes: For each component in the nutrient vector, the probability that the content value falls within the standard nutrient requirement interval consisting of the minimum and maximum limits is calculated using the cumulative distribution function of the normal distribution model determined by the expected value and variance, and the probability value is required to be no less than the preset confidence threshold.

[0037] A reliability standard is established, with a pre-set reliability threshold of 95%. This means that a qualified feed formulation provides each key nutrient with at least a 95% probability of meeting the nutritional requirements of cattle. This threshold is a hard constraint that the optimization algorithm must adhere to when selecting formulations.

[0038] For each nutrient, such as crude protein, the calculated expected value is used. and variance Let's define a normal distribution. Assume the standard nutrient requirement range is from a minimum of 3.0 kg to a maximum of 3.5 kg. Calculate the probability that the content value falls within this range using the cumulative distribution function of this normal distribution. If the calculated probability is 97%, since 97% > 95%, the crude protein meets the constraint. This probability calculation and comparison process will be performed on all 12 key nutrients one by one. Only when the probability of meeting the standard for all nutrients is greater than or equal to 95% is the feed formulation considered a candidate solution. Figure 4 .

[0039] S4. The improved particle swarm optimization algorithm is used to solve the optimization model to obtain the optimal ratio of each feed ingredient.

[0040] Specifically, each particle is considered a potential feed formulation, and the particle's position vector represents the proportion of each ingredient. In each iteration of the algorithm, for each formulation represented by a particle, a comprehensive interaction effect index is calculated using the nutrient vector and the interaction matrix. In one embodiment, the index is obtained by calculating the arithmetic mean of the synergistic and antagonistic effects among all nutrients; a higher index value indicates better overall nutritional synergy in the formulation. The self-learning factor of the particle is then adjusted based on this index value. and group learning factors When the indicator value is high, the learning factor should be increased appropriately. ,reduce This encourages particles to conduct deeper local searches and discover better solutions near the current formulation; when the index value is low, it is reduced. Increase This encourages particles to learn from the globally optimal particle and escape regions with poor nutrient interactions. Through continuous iteration, the algorithm outputs the position of the globally optimal particle that satisfies all constraints and has the lowest cost, which represents the optimal ratio of each feed ingredient.

[0041] In an optional embodiment, the learning factor of the improved particle swarm optimization algorithm is adjusted according to a comprehensive interaction effect index calculated based on the interaction matrix, including: Calculate the arithmetic mean of all off-diagonal elements of the interaction matrix M, and use it as the comprehensive interaction effect index S; After each iteration, according to the formula and Update learning factors; among which, , This is the initial learning factor.

[0042] The comprehensive interaction effect index S is calculated by extracting all non-diagonal elements from the constructed N×N nutrient interaction matrix M, summing them all, and then dividing by the total number of elements to obtain the arithmetic mean S. The S value reflects the overall trend of interactions among all nutrients in the formula. If S is greater than 1, synergistic effects are dominant; if S is less than 1, antagonistic effects are dominant.

[0043] Update the learning factor by setting the initial self-learning factor at the start of the particle swarm optimization algorithm. and group learning factors Both are 2.0. In each iteration of the algorithm's optimization process, the calculated S value is used to adjust the two factors. For example, when the synergistic effect is strong, resulting in S of 1.1, It will increase. This will decrease, causing particles to rely more on their historical optimal positions for deep local exploitation; when the antagonistic effect is strong enough to result in S = 0.9, It will decrease. This increases the likelihood of particles learning more towards the optimal position within the group, allowing them to escape regions with poor nutrient interactions. The adjustment strategy enables the algorithm to intelligently balance global search and local exploitation capabilities based on the characteristics of nutrient interactions. Figure 5 .

[0044] In another embodiment, a dynamic adjustment strategy that varies with the number of iterations is employed to balance the algorithm's global exploration and local exploitation capabilities. For example, a linearly decreasing inertia weight w is used: Where t is the current iteration number, The maximum number of iterations, and These are the initial and final values ​​of the inertia weights. The learning factors c1 and c2 can be adjusted using a non-linear dynamic adjustment strategy, increasing the learning factors in the early stages of iteration. and reduce In the later stages of iteration, the opposite is true, in order to promote the particles to converge quickly to the global optimum.

[0045] In a second embodiment, the present invention also provides a feed formulation optimization and control system, comprising the following modules: The first module is used to obtain the cost, basic nutritional components and historical batch data of the candidate feed ingredients, and to establish a probability distribution model for each nutrient component of each ingredient; based on the physiological stage of the target cattle, it determines the standard nutritional requirement range that includes the minimum and maximum limits of each nutrient component. The calculation module is used to construct an interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients, and to calculate the nutrient vector provided by the formula based on the interaction matrix and the ratio of each raw material in any formula. The second module is used to establish an optimization model with the objective function of minimizing the total cost of the feed formulation and with the constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not lower than a preset confidence threshold. The adjustment module is used to solve the optimization model using an improved particle swarm optimization algorithm to obtain the optimal ratio of each feed ingredient.

[0046] In this specification, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise limited, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. In this document, "a," "an," "the," "the," and "its" may also include plural forms unless the context clearly indicates otherwise. "Multiple" refers to at least two, such as 2, 3, 5, or 8, etc. "And / or" includes any and all combinations of the associated listed items.

[0047] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0048] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for optimizing and controlling feed formulation, characterized in that, Includes the following steps: Obtain the cost, basic nutrient composition, and historical batch data of the candidate feed ingredients, and establish a probability distribution model for each nutrient component of each ingredient; determine the standard nutrient requirement range that includes the minimum and maximum limits of each nutrient component based on the physiological stage of the target cattle. An interaction matrix representing the synergistic and antagonistic effects of different nutrients is constructed, and a nutrient vector provided by the formula is calculated based on the interaction matrix and the ratio of each ingredient in any formula. An optimization model is established with the objective function of minimizing the total cost of the feed formulation and the constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not lower than a preset confidence threshold. An improved particle swarm optimization algorithm is used to solve the optimization model and obtain the optimal ratio of each feed ingredient.

2. The method according to claim 1, characterized in that, The process of acquiring the cost, basic nutritional components, and historical batch data of the candidate feed ingredients, and establishing a probability distribution model for each nutrient component of each ingredient, includes: For each feed ingredient, collect the content test values ​​of each nutrient from at least 30 historical batches; The maximum likelihood estimation method is used to fit the content detection value to a normal distribution, and the expected value and standard deviation of the nutrient content are calculated, which are used as parameters of the probability distribution model.

3. The method according to claim 1, characterized in that, The determination of standard nutritional requirement ranges, including minimum and maximum limits for each nutrient component, based on the physiological stage of the target cattle includes: The physiological stages of the target cattle are divided into calf stage, growing stage, and fattening stage; For each stage, consult the beef cattle feeding standards to obtain the recommended intake ranges for 12 key nutrients, and set the lower limit of the range as the minimum limit and the upper limit as the maximum limit.

4. The method according to claim 1, characterized in that, The construction of the interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients includes: Construct an N×N square matrix M, where N is the number of different types of nutrients; The value of the element in the i-th row and j-th column of the matrix This represents the factor influencing the absorption rate of nutrient i on nutrient j, where, when i=j, =1; When nutrient i has a synergistic effect on j Set to greater than 1; When nutrient i has an antagonistic effect on j Set to less than 1; When there is no interaction between the two Set to 1.

5. The method according to claim 4, characterized in that, The calculation of the nutrient vector provided by the formula based on the interaction matrix and the proportions of each ingredient in any formula includes: Based on the proportions of each raw material and the probability distribution model of nutrient content, the nominal nutrient vector is calculated. Expected value vector Sum of variance vectors ; nutrient vector The components of the vector approximate a normal distribution, and the expected value vector is... and the variance of each component Calculated using the following formula: , ; Where M is the interaction matrix, Let be the transpose of M, where i and j are the indices of the nutrients.

6. The method according to claim 1, characterized in that, The constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not less than a preset confidence threshold includes: For each component in the nutrient vector, the probability that the content value falls within the standard nutrient requirement interval consisting of the minimum and maximum limits is calculated using the cumulative distribution function of the normal distribution model determined by the expected value and variance, and the probability value is required to be no less than the preset confidence threshold.

7. The method according to claim 1, characterized in that, The improved particle swarm optimization algorithm dynamically adjusts its learning factor, including: Calculate the arithmetic mean of all off-diagonal elements of the interaction matrix M, and use it as the comprehensive interaction effect index S; The learning factor of the improved particle swarm optimization algorithm is determined according to the formula after each iteration. and Update learning factors; among which, , This is the initial learning factor.

8. A feed formulation optimization and control system, characterized in that, Includes the following modules: The first module is used to obtain the cost, basic nutritional components and historical batch data of the candidate feed ingredients, and to establish a probability distribution model for each nutrient component of each ingredient. Based on the physiological stage of the target cattle, determine the standard nutritional requirement range that includes the minimum and maximum limits of each nutrient component; The calculation module is used to construct an interaction matrix representing the synergistic and antagonistic effects of absorption among different nutrients, and to calculate the nutrient vector provided by the formula based on the interaction matrix and the ratio of each raw material in any formula. The second module is used to establish an optimization model with the objective function of minimizing the total cost of the feed formulation and with the constraint that the probability of the nutrient component vector satisfying the standard nutrient requirement range is not lower than a preset confidence threshold. The adjustment module is used to solve the optimization model using an improved particle swarm optimization algorithm to obtain the optimal ratio of each feed ingredient.

9. The system according to claim 8, characterized in that, The process of acquiring the cost, basic nutritional components, and historical batch data of the candidate feed ingredients, and establishing a probability distribution model for each nutrient component of each ingredient, includes: For each feed ingredient, collect the content test values ​​of each nutrient from at least 30 historical batches; The maximum likelihood estimation method is used to fit the content detection value to a normal distribution, and the expected value and standard deviation of the nutrient content are calculated, which are used as parameters of the probability distribution model.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program, when executed by a processor, implements the method as described in any one of claims 1-7.

Images