Method for increasing proportion of unsaturated fatty acid in sturgeon compound feed

Abstract

The invention discloses a method for increasing the proportion of unsaturated fatty acid in sturgeon compound feed, and relates to the technical field of production control, a raw material information set and a nutritional requirement information set are collected in advance, a formula optimization model is constructed based on the raw material information set and the nutritional requirement information set, and an optimization solver is used for solving the formula optimization model, so that the proportion of the unsaturated fatty acid in the sturgeon compound feed is increased. And iteratively solving and outputting an optimized formula result, inputting the optimized formula result into a production process system, monitoring production process monitoring data in a production process in real time, constructing a production stability fitness function based on the production process monitoring data, and calculating a production stability fitness function based on the production stability fitness function and the production process monitoring data. A reinforcement learning model is used for adjusting production parameters; the proportion of unsaturated fatty acid in the feed is remarkably increased, and the stability and consistency of the quality of the feed are ensured.

Description

Technical Field

[0001] The invention relates to the technical field of production control, in particular to a method for increasing the proportion of unsaturated fatty acids in sturgeon compound feed. Background Art

[0002] In the production process of sturgeon compound feed, unsaturated fatty acids (UFAs) are one of the most important nutrients, especially ω-3 and ω-6 fatty acids, which have a significant effect on the growth, development, immune function and reproductive capacity of sturgeons. However, the traditional sturgeon compound feed production process relies too much on subjective experience and manual adjustment. Producers select feed formulas and set production parameters based on experience, and often cannot scientifically control the proportion of unsaturated fatty acids in the feed. Parameter fluctuations and instability in the production process will also lead to inconsistent feed quality, further affecting the healthy growth of sturgeons.

[0003] At present, the selection of formula and adjustment of production parameters in the feed production process mainly rely on the subjective experience of production personnel. Experienced technicians can guarantee the quality of feed to a certain extent through long-term practical accumulation, but this method has significant limitations. First, subjective experience is difficult to quantify, and there are large differences in experience levels among different producers, resulting in unstable feed quality; second, empirical methods are difficult to adapt to the rapidly changing market demand and production environment, and it is impossible to adjust and optimize the feed formula in time; finally, the manual adjustment method is inefficient and it is difficult to achieve automation and intelligence of the production process. In the face of these challenges, it is necessary to introduce scientific formula optimization and production parameter optimization methods to improve the stability and efficiency of feed production.

[0004] Chinese patent application with publication number CN101209089A provides a method for increasing n-3 polyunsaturated fatty acids in grass carp meat, wherein hybrid Pennisetum is combined with commercially available freshwater fish compound feed to feed grass carp, and the weight ratio of dry matter of the compound feed and hybrid Pennisetum is 1.1:1. However, this method fails to regulate production parameters from the perspective of data processing.

[0005] To this end, the present invention provides a method for increasing the proportion of unsaturated fatty acids in sturgeon compound feed. Summary of the invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention provides a method for increasing the proportion of unsaturated fatty acids in sturgeon compound feed, which significantly increases the proportion of unsaturated fatty acids in the feed and ensures the stability and consistency of the feed quality.

[0007] To achieve the above object, a method for increasing the proportion of unsaturated fatty acids in sturgeon feed is proposed, comprising the following steps:

[0008] Step 1: Collect raw material information and nutritional requirement information in advance;

[0009] Step 2: Building a recipe optimization model based on the raw material information set and the nutritional requirement information set;

[0010] Step 3: Use the optimization solver to solve the recipe optimization model, iteratively solve and output the optimized recipe results;

[0011] Step 4: Input the optimized formula results into the production process system and monitor the production process monitoring data in real time;

[0012] Step 5: Construct a production stability fitness function based on production process monitoring data;

[0013] Step 6: Based on the production stability fitness function and production process monitoring data, use the reinforcement learning model to adjust the production parameters.

[0014] The raw material information set is collected in the following manner:

[0015] By using analytical instruments in the laboratory to analyze and test the nutritional data and physical and chemical properties of various raw materials for making sturgeon feed;

[0016] Collect economic parameters of various raw materials based on market pricing;

[0017] The nutritional component data, the physicochemical properties and the economic parameters constitute a raw material information set.

[0018] The nutritional requirement information set is collected in the following manner:

[0019] Divide the growth process of sturgeons into stages and obtain a list of growth stages;

[0020] For each growth stage, the nutritional requirement ratio of various nutritional factors is calculated.

[0021] The method of constructing the recipe optimization model based on the raw material information set and the nutritional requirement information set is:

[0022] Label each growth stage of the sturgeon as k;

[0023] Mark the numbers of various raw materials in the production formula as i, and the quantities of all raw material types as I;

[0024] The unit content of unsaturated fatty acids in the i-th raw material is marked as UNi;

[0025] The numbers of other nutrients except unsaturated fatty acids in the nutritional composition data are marked as j; the quantities of all nutrients except unsaturated fatty acids are marked as J;

[0026] The unit content of the jth nutrient element in the i-th raw material is marked as Nij;

[0027] Setting a preset target ratio UTk of unsaturated fatty acids for the kth growth stage;

[0028] For the kth growth stage, the jth nutrient element sets the preset target ratio Tkj of the corresponding nutrient element;

[0029] The unit cost of the i-type raw material is marked as Ci, and the cost budget of the unit quantity of compound feed is marked as Cmax;

[0030] Set the recipe ratio variable xki for the i-th raw material in the k-th growth stage;

[0031] Set the minimum required contents UDk and Dkj for unsaturated fatty acids and the jth nutrient element in the kth growth stage, respectively;

[0032] For the kth growth stage:

[0033] Constructing a formula objective function based on the unit content of unsaturated fatty acids, the unit content of nutrient elements, the preset target ratio of unsaturated fatty acids, the preset target ratio of nutrient elements, and formula ratio variables;

[0034] Construct a set of constraints based on the formula ratio variables, the minimum required content of unsaturated fatty acids and the jth nutrient element, the unit cost of each raw material and the cost budget;

[0035] Taking maximizing the recipe objective function as the optimization goal and the constraint condition set as the constraint, the constructed linear programming model is used as the recipe optimization model.

[0036] The method of using the optimization solver to solve the recipe optimization model and iteratively solve and output the optimized recipe result is:

[0037] For each growth stage of sturgeon:

[0038] A linear programming solver is used to solve the formula optimization model corresponding to the growth stage to obtain the variable values ​​of the formula proportion variables of each raw material; the variable values ​​of the formula proportion variables of each raw material constitute the optimized formula result.

[0039] The result of inputting the optimized formula into the production process system is:

[0040] According to the specific growth stage of the sturgeons to be raised, the raw materials are proportioned according to the variable values ​​of the formula proportion variables of each raw material in the optimization formula result of the corresponding growth stage to obtain the production raw materials of the mixed compound feed.

[0041] The method of real-time monitoring of production process monitoring data in the production process is:

[0042] Use spectral analysis technology to monitor the unsaturated fatty acid content of mixed raw materials in real time during the production of compound feed;

[0043] Through sensors of various physical parameters, real-time parameter values ​​of various physical parameters in the process of compound feed production are collected in real time;

[0044] The real-time unsaturated fatty acid content and the real-time parameter values ​​of various physical parameters constitute the production process monitoring data.

[0045] The method of constructing the production stability fitness function based on the production process monitoring data is as follows:

[0046] The initial moment of compound feed production is marked as 0:

[0047] Let any time of compound feed production be marked as t;

[0048] The number of each physical parameter is marked as p;

[0049] The standard deviation of the pth physical parameter from time 0 to time t is marked as ;

[0050] Based on the standard deviation of various physical parameters, a production stability fitness function is constructed.

[0051] The method of adjusting the production parameters using the reinforcement learning model based on the production stability fitness function and the production process monitoring data is as follows:

[0052] The reinforcement learning model is set to an Actor-Critic model;

[0053] Define the state space, action space, and reward function of the reinforcement learning model;

[0054] The state space includes the content of unsaturated fatty acids at any time t in the production process and the real-time parameter values ​​of various physical parameters;

[0055] The action space includes the adjustment amounts of all regulated production parameters;

[0056] The reward function is a weighted sum of the production stability fitness function and the unsaturated fatty acid consumption value, wherein the weight of the unsaturated fatty acid consumption value is a negative value, and the weights are all preset values;

[0057] Initialize the model parameters of the Actor network and the Critic network, and set the optimizer to the Adam optimizer;

[0058] At each time t:

[0059] Get the current state from the environment, use the Actor network to select an action, and execute the selected action to observe the state at the next moment, calculate the reward value, and store the experience in the experience playback buffer. The experience includes the current state, the selected action, the state at the next moment, and the reward value;

[0060] Randomly sample a batch of experiences from the experience replay buffer and calculate the target Q value;

[0061] Use the mean square error function as the loss function of the Critic network model; use the soft update strategy to update the parameters of the Critic network;

[0062] Through multiple iterative updates, the production parameters are optimized to maximize the reward function.

[0063] A computer-readable storage medium is provided, on which a rewritable computer program is stored;

[0064] When the computer program is run on a computer device, the computer device is caused to execute the above-mentioned method for increasing the proportion of unsaturated fatty acids in a sturgeon feed.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] The present invention collects raw material information set and nutritional requirement information set in advance, constructs a formula optimization model based on the raw material information set and the nutritional requirement information set, uses an optimization solver to solve the formula optimization model, iteratively solves and outputs the optimized formula result, inputs the optimized formula result into the production process system, monitors the production process monitoring data in real time during the production process, constructs a production stability fitness function based on the production process monitoring data, and uses a reinforcement learning model to adjust production parameters based on the production stability fitness function and the production process monitoring data; collects and analyzes raw material information and nutritional requirement information, constructs a fitness function, and uses a linear programming solver to optimize the formula combination to ensure that the unsaturated fatty acid content in the feed reaches the target value. In the process of production parameter optimization, the state space, action space and reward function are defined, the production stability fitness function is constructed, the Actor-Critic model is trained, and the automation and intelligence of sturgeon feed production are realized through parameter adjustment and model optimization, the proportion of unsaturated fatty acids in the feed is significantly improved, and the stability and consistency of the feed quality are ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a flow chart of a method for increasing the proportion of unsaturated fatty acids in sturgeon feed in Example 1 of the present invention;

[0068] Figure 2This is an example diagram of the unsaturated fatty acid requirements of sturgeon at various growth stages in Example 1 of the present invention. DETAILED DESCRIPTION

[0069] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] Example 1

[0071] like Figure 1 As shown, a method for increasing the proportion of unsaturated fatty acids in a sturgeon feed comprises the following steps:

[0072] Step 1: Collect raw material information and nutritional requirement information in advance;

[0073] Step 2: Building a recipe optimization model based on the raw material information set and the nutritional requirement information set;

[0074] Step 3: Use the optimization solver to solve the recipe optimization model, iteratively solve and output the optimized recipe results;

[0075] Step 4: Input the optimized formula results into the production process system and monitor the production process monitoring data in real time;

[0076] Step 5: Construct a production stability fitness function based on production process monitoring data;

[0077] Step 6: Based on the production stability fitness function and production process monitoring data, use the reinforcement learning model to adjust the production parameters.

[0078] Wherein, the collection method of the raw material information set is:

[0079] By using analytical instruments in the laboratory to analyze and test the nutritional data and physical and chemical properties of various raw materials for making sturgeon feed;

[0080] Collect economic parameters of various raw materials based on market pricing;

[0081] The nutritional component data, the physicochemical properties and the economic parameters constitute a raw material information set.

[0082] Specifically, the analytical instrument includes a gas chromatograph or a liquid chromatograph, which uses gas chromatography or liquid chromatography to analyze the nutritional component data and physical and chemical properties of the raw materials respectively;

[0083] The nutritional component data includes protein content, fat content, content of various types of unsaturated fatty acids, such as the specific content of ω-3 fatty acids and ω-6 fatty acids, and other key nutritional components, such as the content of vitamins, minerals, cellulose, etc.;

[0084] The physicochemical properties include moisture content, particle size distribution, and antioxidant properties;

[0085] The economic parameters include the cost of various raw materials.

[0086] It should be noted that the collection of nutritional data is not only necessary to ensure the proportion of unsaturated fatty acids in the subsequent formula optimization process, but also to ensure that the designed feed formula can meet the nutritional needs of sturgeons at all growth stages and promote their healthy growth and reproduction;

[0087] The collection of physical and chemical properties is needed to ensure the stability, palatability and processing performance of the compound feed, thereby improving the overall quality of the feed;

[0088] Economic parameters can help to reasonably control feed production costs and ensure the sustainability of production while meeting nutritional needs.

[0089] The nutritional requirement information set is collected in the following manner:

[0090] Divide the growth process of sturgeons into stages and obtain a list of growth stages;

[0091] Specifically, Figure 2 The requirements of unsaturated fatty acids for sturgeons at various growth stages are shown in the figure. The list of growth stages includes juvenile stage, adult stage, breeding stage, etc. At different growth stages, the requirements of sturgeons for unsaturated fatty acids vary. For example, at the juvenile stage, the requirements for ω-3 unsaturated fatty acids are between 2% and 3%, and the requirements for ω-6 unsaturated fatty acids are between 1% and 2%. At the adult stage, the requirements for ω-3 unsaturated fatty acids are between 1% and 2%, and the requirements for ω-6 unsaturated fatty acids are between 0.5 and 1.5%. At the breeding stage, the requirements for ω-3 unsaturated fatty acids are between 3% and 4%, and the requirements for ω-6 unsaturated fatty acids are between 1% and 3%. Therefore, it is necessary to optimize the formula of compound feed for different growth stages.

[0092] For each growth stage, the nutritional requirement ratio of various nutritional factors is calculated through scientific literature consultation or feeding experience.

[0093] Specifically, the nutritional factors include protein requirements, fat requirements, unsaturated fatty acid requirements, carbohydrate requirements, vitamin requirements, and mineral requirements, etc.

[0094] Furthermore, the method of constructing the recipe optimization model based on the raw material information set and the nutritional requirement information set is:

[0095] Label each growth stage of sturgeon as k;

[0096] Mark the numbers of various raw materials in the production formula as i, and the quantities of all raw material types as I;

[0097] The unit content of unsaturated fatty acids in the i-th raw material is marked as UNi;

[0098] The numbers of other nutrients except unsaturated fatty acids in the nutritional composition data are marked as j; the quantities of all nutrients except unsaturated fatty acids are marked as J;

[0099] The unit content of the jth nutrient element in the i-th raw material is marked as Nij;

[0100] A preset target ratio UTk of unsaturated fatty acids is set for the kth growth stage; it is understood that since unsaturated fatty acids are consumed during the production of compound feeds, in order to increase the ratio of unsaturated fatty acids in compound feeds, the value of UT needs to be set greater than the upper limit of the required ratio of unsaturated fatty acids in the production stage, for example, the value of UT in the juvenile stage should be greater than 5% (the sum of the upper limits of ω-3 unsaturated fatty acids and ω-6 unsaturated fatty acids);

[0101] For the kth growth stage, the jth nutrient element sets the preset target ratio Tkj of the corresponding nutrient element;

[0102] The unit cost of the i-type raw material is marked as Ci, and the cost budget of the unit quantity of compound feed is marked as Cmax;

[0103] Set the recipe ratio variable xki for the i-th raw material in the k-th growth stage;

[0104] Set the minimum required contents UDk and Dkj for unsaturated fatty acids and the jth nutrient element in the kth growth stage, respectively;

[0105] For the kth growth stage:

[0106] Constructing a formula objective function based on the unit content of unsaturated fatty acids, the unit content of nutrient elements, the preset target ratio of unsaturated fatty acids, the preset target ratio of nutrient elements, and formula ratio variables;

[0107] Construct a set of constraints based on the formula ratio variables, the minimum required content of unsaturated fatty acids and the jth nutrient element, the unit cost of each raw material and the cost budget;

[0108] Taking maximizing the recipe objective function as the optimization goal and the constraint condition set as the constraint, the constructed linear programming model is used as the recipe optimization model.

[0109] Preferably, the expression of the formulation objective function can be constructed as follows:

[0110] The formulation objective function of the kth growth stage is labeled as Fk;

[0111] Then the formula of the recipe objective function Fk is set as:

[0112] ; Wherein, w0 and wj are both preset proportional coefficients; It can be understood that the first term in the expression of the formula objective function Fk measures the difference between the unsaturated fatty acid ratio and the target ratio, and the second term measures the difference between various other nutritional elements and the corresponding target ratios. Therefore, maximizing Fk takes into account the goal of increasing the ratio of unsaturated fatty acids and the nutritional balance of balancing other nutritional elements;

[0113] The set of constraints can be constructed as follows:

[0114] The set of constraints in the kth growth stage is labeled Uk;

[0115] Then the constraints contained in Uk include:

[0116] ; The constraint is that the unit content of unsaturated fatty acids must not be lower than the minimum required content;

[0117] ; The constraint is that the total content of any j-th nutrient element shall not be lower than the minimum required content corresponding to the j-th nutrient element;

[0118] ; The constraint is that the sum of the proportions of all raw materials should be 1;

[0119] ; Cost constraints per unit quantity of compound feed.

[0120] Furthermore, the method of using the optimization solver to solve the recipe optimization model and iteratively solve and output the optimized recipe result is:

[0121] For each growth stage of sturgeon:

[0122] A linear programming solver is used to solve the formula optimization model corresponding to the growth stage to obtain the variable values ​​of the formula ratio variables of each raw material; the variable values ​​of the formula ratio variables of each raw material constitute the optimized formula result; the linear programming solver includes but is not limited to existing tools such as CPlex and Gurobi;

[0123] Generally, since linear programming problems are non-NP-hard problems and can be solved in polynomial time, the linear programming solver can obtain the optimal solution of the recipe optimization model, that is, the optimized recipe result is essentially the optimal recipe of the recipe optimization model.

[0124] Furthermore, the optimized formula result input into the production process system is:

[0125] According to the specific growth stage of the sturgeons to be raised, the raw materials are proportioned according to the variable values ​​of the formula proportion variables of each raw material in the optimization formula results of the corresponding growth stage to obtain the production raw materials of the mixed compound feed; providing a basis for the subsequent production of compound feed.

[0126] Furthermore, the method of real-time monitoring of the production process monitoring data in the production process is:

[0127] Using spectral analysis technology to monitor the unsaturated fatty acid content of mixed raw materials in real time during the production of compound feed; specifically, the spectral analysis technology can be near infrared spectroscopy technology or Raman spectroscopy technology;

[0128] Through sensors of various physical parameters, real-time parameter values ​​of various physical parameters in the process of compound feed production are collected in real time;

[0129] The real-time unsaturated fatty acid content and the real-time parameter values ​​of various physical parameters constitute the production process monitoring data.

[0130] Specifically, the physical parameters include temperature, humidity, pressure, oil coating amount, raw material flow, power consumption, etc. Among these physical parameters, temperature, pressure and oil coating amount are controllable parameters, and all these physical parameters need to be kept stable during the production process.

[0131] Furthermore, the method of constructing the production stability fitness function based on the production process monitoring data is as follows:

[0132] The initial moment of compound feed production is marked as 0:

[0133] Let any time of compound feed production be marked as t;

[0134] The number of each physical parameter is marked as p;

[0135] The standard deviation of the pth physical parameter from time 0 to time t is marked as ;

[0136] Based on the standard deviation of each physical parameter, a production stability fitness function is constructed;

[0137] Preferably, the production stability fitness function may be constructed in the following manner:

[0138] The production stability fitness function is labeled G;

[0139] Then the expression of the production stability fitness function G can be set as: ; It can be understood that by maximizing the production stability fitness function G, the stability of the production process can be controlled to improve the production quality of compound feed.

[0140] Furthermore, the method of adjusting the production parameters using the reinforcement learning model based on the production stability fitness function and the production process monitoring data is as follows:

[0141] The reinforcement learning model is set to an Actor-Critic model;

[0142] Define the state space, action space, and reward function of the reinforcement learning model;

[0143] Among them, the state space includes the content of unsaturated fatty acids at any time t in the production process and the real-time parameter values ​​of various physical parameters;

[0144] The action space includes adjustable values ​​of all adjustable production parameters; for example, the temperature can be increased by 0.5 degrees Celsius, decreased by 0.5 degrees Celsius, increased by 1 degree Celsius, etc., to describe how to change the current production status;

[0145] The reward function is the weighted sum of the production stability fitness function and the unsaturated fatty acid consumption value, wherein the weight of the unsaturated fatty acid consumption value is a negative value, and the weights are all preset values; the unsaturated fatty acid consumption value is the difference between the unsaturated fatty acids at the current moment and the unsaturated fatty acids at the previous moment, and the larger the difference, the greater the loss of unsaturated fatty acids.

[0146] Initialize the model parameters of the Actor network and the Critic network, and set the optimizer to the Adam optimizer;

[0147] At each time t:

[0148] Get the current state from the environment, use the Actor network to select an action, and execute the selected action to observe the state at the next moment, calculate the reward value, and store the experience (current state, selected action, state at the next moment, reward value) in the experience replay buffer;

[0149] Randomly sample a batch of experiences from the experience replay buffer and calculate the target Q value;

[0150] The target Q value is ;in, is the target Q value, γ is the discount factor, is the estimated value output by the Critic network, and Rt is the reward value;

[0151] Use the mean square error function as the loss function of the Critic network model; use the soft update strategy to update the parameters of the Critic network;

[0152] Through multiple iterative updates, the Actor-Critic model gradually learns to optimize production parameters to maximize the reward function.

[0153] The following is the experimental setup and experimental conclusions during the implementation of the present invention to illustrate that the solution of the present invention achieves the expected technical effect:

[0154] 1. Experimental Objectives

[0155] 1. Core verification:

[0156] Whether the invented method (formula optimization + parameter reinforcement learning) can significantly improve the ratio of ω-3 / ω-6 fatty acids in sturgeon feed, whether the intelligent regulation of production parameters can reduce the fluctuation of physical parameters and improve production stability, and the comprehensive advantages of this method in terms of economy (cost) and sturgeon growth performance.

[0157] 2. Segment verification:

[0158] The impact of formula optimization on nutritional balance (such as the synergistic effect of protein, fat and unsaturated fatty acids) and the adaptability of reinforcement learning models to different production stages (juvenile stage → adult stage);

[0159] 2. Experimental Group Design

[0160] The specific experimental settings and comparative experimental table are shown in Table 1:

[0161] Table 1 Experimental settings and comparative experiment table

[0162] Experimental groups Description of experimental method Core technical parameters Control group / Experimental group basis Control group (Group A) Traditional manual formula + manual parameter adjustment Formula: ω-3=2.5%, ω-6=1.2% Parameters: Temperature 18-22℃, Pressure 0.3-0.5MPa Industry standard operation Experimental group 1 (group B) Recipe optimization model (linear programming) + traditional manual parameter adjustment Formula: Optimized based on the invention model (target ω-3=3.5%, ω-6=2.0%) Parameters: Same as Group A Verify independent effects of recipe optimization Experimental group 2 (Group C) Traditional formula + production parameter reinforcement learning adjustment Recipe: Same as Group A Parameters: Real-time control by reinforcement learning model (temperature, pressure, flow) Verify the independent effect of parameter optimization Experimental group 3 (group D) Recipe optimization model + production parameter reinforcement learning adjustment (complete invention method) Recipe: Same as group B Parameters: Real-time control of reinforcement learning model Systematic advantages of verification methods Additional control group (Group E) Commercial formulation (positive control) Formula: Commercially available high-quality sturgeon feed (ω-3=2.8%, ω-6=1.5%) Parameters: Manufacturer recommended parameters Compare the performance of commercially available products

[0163] 3. Experimental Setup

[0164] 1. Experimental subjects

[0165] Species: Siberian sturgeon (Acipenser baerii);

[0166] Source: juveniles from the same batch in the same hatchery, initial weight (50 ± 5 g);

[0167] Farming system:

[0168] Circulating aquaculture system (flow rate 100L / min, water temperature 18±1℃);

[0169] Breeding pond specifications: 3m×2m×1m (effective water volume 5m³);

[0170] Group management:

[0171] Each group had three replicate pools, randomly assigned (a total of 15 pools, 600 fish);

[0172] Feed 4 times a day (08:00, 12:00, 16:00, 20:00), and the feeding amount is 3% of body weight.

[0173] 2. Raw materials and production process

[0174] The raw material list is shown in Table 2:

[0175] Table 2 Raw materials list

[0176] Raw material name Omega-3 (%) Omega-6 (%) protein(%) Fat(%) Moisture (%) Cost (yuan / kg) Deep sea fish oil 22.5 18.0 0.5 99.0 ≤0.5 15.0 Soybean meal 0.8 5.2 45.0 2.0 ≤12.0 3.5 Corn Gluten Meal 0.3 1.2 60.0 0.5 ≤10.0 4.0 Soybean lecithin 3.0 12.0 1.0 95.0 ≤2.0 8.0 wheat flour 0.2 0.5 12.0 1.5 ≤14.0 2.5

[0177] Production process:

[0178] Raw material pretreatment: crush to 80 mesh (pass rate ≥ 95%);

[0179] Mixing: horizontal mixer (mixing time 10 minutes, CV ≤ 5%);

[0180] Puffing: twin-screw extruder (die diameter 4mm, screw speed 300rpm);

[0181] Drying: fluidized bed dryer (temperature 80°C, moisture ≤10%);

[0182] 3. Evaluation indicators

[0183] Nutritional indicators:

[0184] ω-3 / ω-6 fatty acid content (GB / T 22223-2008 gas chromatography method);

[0185] Crude protein (Kjeldahl method, GB / T 6432-2018);

[0186] Crude fat (Soxhlet extraction method, GB / T 6433-2006);

[0187] Moisture content (105℃ constant weight method, GB / T 6435-2014);

[0188] Production indicators:

[0189] Physical parameter stability: temperature / pressure / flow rate standard deviation (σ) and coefficient of variation (CV);

[0190] Production efficiency: energy consumption per ton of material (kW·h / t), output (kg / h);

[0191] Product consistency: feed particle hardness (N), powder rate (%);

[0192] Economic indicators:

[0193] Raw material cost (yuan / kg), energy cost (yuan / t), comprehensive cost (yuan / kg);

[0194] Sturgeon growth indicators:

[0195] Weight gain rate (WG) = [(final weight - initial weight) / initial weight] × 100%;

[0196] Specific growth rate (SGR) = (ln final weight - ln initial weight) / days × 100%;

[0197] Feed conversion ratio (FCR) = total feed amount / total weight gain;

[0198] Survival rate (SR) = (final number of fish / initial number of fish) × 100%;

[0199] IV. Experimental Procedure

[0200] 1. Control group (Group A)

[0201] formula:

[0202] Fish oil 5%, soybean meal 40%, corn gluten meal 30%, lecithin 3%, wheat flour 22%;

[0203] Theoretical nutritional value: ω-3 = 2.6%, ω-6 = 1.3%, protein = 36.2%;

[0204] parameter:

[0205] Extruder temperature: 18-22℃ (manual adjustment);

[0206] Screw pressure: 0.3-0.5MPa (manually adjusted according to the discharge status);

[0207] 2. Experimental Group 1 (Group B)

[0208] Recipe Optimization:

[0209] Using the invented linear programming model (larval stage parameters):

[0210] Optimization results: fish oil 8%, soybean meal 35%, corn protein meal 25%, lecithin 5%, wheat flour 27%;

[0211] Theoretical nutritional value: ω-3 = 3.4%, ω-6 = 1.9%, protein = 38.5%;

[0212] Parameters: Same as group A;

[0213] 3. Experimental Group 2 (Group C)

[0214] Formula: Same as group A;

[0215] Parameter adjustment:

[0216] Reinforcement learning model (Actor-Critic architecture);

[0217] 4. Experimental Group 3 (Group D)

[0218] Formula: Same as the optimized formula of group B;

[0219] Parameter adjustment: Same as the reinforcement learning model in group C;

[0220] 5. Additional control group (Group E)

[0221] Formula: a certain brand of sturgeon feed on the market (labeled ω-3=2.8%, ω-6=1.5%);

[0222] Parameters: as recommended by the manufacturer (temperature 20°C, pressure 0.4MPa).

[0223] V. Data Collection and Analysis

[0224] 1. Data Collection

[0225] Feed analysis:

[0226] Take samples every Monday, Wednesday and Friday (take 5 particles each time, mix them and test them);

[0227] Testing agency: third-party feed testing center (CMA certification);

[0228] Production monitoring:

[0229] Sensor data: automatically recorded every 5 minutes (PLC system storage);

[0230] Energy consumption data: real-time monitoring of electric meter (accuracy 0.1kW·h).

[0231] Growth performance:

[0232] The weight was measured after fasting for 24 hours every 10 days (electronic balance, accuracy ±0.1 g);

[0233] Food intake record: the difference between daily feeding amount and remaining amount (accuracy ±10g).

[0234] 2. Data Analysis

[0235] Statistical tools: SPSS 26.0, GraphPad Prism 9;

[0236] method:

[0237] Normality test: Shapiro-Wilk test;

[0238] Variance homogeneity test: Levene's test;

[0239] Comparison between groups: one-way analysis of variance (ANOVA) + Tukey post hoc test (p < 0.05);

[0240] Correlation analysis: Pearson correlation coefficient (study the relationship between production parameters and nutritional indicators).

[0241] 3. Quality Control

[0242] Keep samples of each batch of feed (stored at 4°C, shelf life 6 months);

[0243] The sensor is calibrated every 2 weeks (standard thermometer, pressure gauge);

[0244] Experimenter training: unify operating procedures and reduce human errors.

[0245] VI. Experimental Data and Conclusions

[0246] The experimental data are shown in Table 3:

[0247] Table 3 Experimental data table

[0248] index Control group (A) Experimental group 1 (B) Experimental group 2 (C) Experimental group 3 (D) Additional control group (E) Omega-3 fatty acids (%) 2.6±0.3 3.4±0.2 2.7±0.2 3.7±0.1 2.8±0.1 Omega-6 fatty acids (%) 1.3±0.1 1.9±0.1 1.4±0.1 2.1±0.1 1.5±0.1 Temperature standard deviation (℃) 1.2 1.1 0.5 0.4 0.8 Unit cost (yuan / kg) 8.5 8.2 8.4 7.9 9.2 Weight gain rate (%) 120 135 125 140 128 Feed conversion ratio (FCR) 1.8 1.6 1.7 1.5 1.7

[0249] in conclusion:

[0250] 1. Formulation optimization (Group B):

[0251] The ω-3 / ω-6 ratio increased by 30%, but the production stability (temperature σ=1.1℃) was not significantly different from the control group;

[0252] Costs fell 3.5% (due to increased use of fish oil but less soybean meal).

[0253] 2. Parameter adjustment (Group C):

[0254] The temperature standard deviation decreased by 58% (σ = 0.5°C), but there was no significant improvement in nutritional indicators;

[0255] Energy consumption was reduced by 4% (due to parameter optimization and reduced equipment idling).

[0256] 3. Complete Invention Method (Group D):

[0257] The synergistic effect was significant: the ω-3 / ω-6 ratio reached 3.7% / 2.1%, which was 42% / 61% higher than the control group;

[0258] The production stability is the best (σ temperature = 0.4℃) and the cost is the lowest (7.9 yuan / kg);

[0259] Sturgeon had the best growth performance (weight gain rate 140%, FCR=1.5).

[0260] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained by simulating a large amount of data.

[0261] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for increasing the proportion of unsaturated fatty acids in sturgeon feed, characterized in that: The following steps are involved: Step 1: Collect raw material information and nutritional requirement information in advance; Step 2: Building a recipe optimization model based on the raw material information set and the nutritional requirement information set; Step 3: Use the optimization solver to solve the recipe optimization model, iteratively solve and output the optimized recipe results; Step 4: Input the optimized formula results into the production process system and monitor the production process monitoring data in real time; Step 5: Construct a production stability fitness function based on production process monitoring data; Step 6: Based on the production stability fitness function and production process monitoring data, use the reinforcement learning model to adjust the production parameters.

2. A method for increasing the proportion of unsaturated fatty acids in sturgeon feed according to claim 1, characterized in that: The raw material information set is collected in the following manner: By using analytical instruments in the laboratory to analyze and test the nutritional data and physical and chemical properties of various raw materials for making sturgeon feed; Collect economic parameters of various raw materials based on market pricing; The nutritional component data, the physicochemical properties and the economic parameters constitute a raw material information set.

3. A method for increasing the proportion of unsaturated fatty acids in sturgeon feed according to claim 2, characterized in that: The nutritional requirement information set is collected in the following manner: Divide the growth process of sturgeons into stages and obtain a list of growth stages; For each growth stage, the nutritional requirement ratio of various nutritional factors is calculated.

4. A method for increasing the proportion of unsaturated fatty acids in a sturgeon feed according to claim 3, characterized in that: The method of constructing the recipe optimization model based on the raw material information set and the nutritional requirement information set is: Label each growth stage of the sturgeon as k; Mark the numbers of various raw materials in the production formula as i, and the quantities of all raw material types as I; The unit content of unsaturated fatty acids in the i-th raw material is marked as UNi; The numbers of other nutrients except unsaturated fatty acids in the nutritional composition data are marked as j; the quantities of all nutrients except unsaturated fatty acids are marked as J; The unit content of the jth nutrient element in the i-th raw material is marked as Nij; Setting a preset target ratio UTk of unsaturated fatty acids for the kth growth stage; For the kth growth stage, the jth nutrient element sets the preset target ratio Tkj of the corresponding nutrient element; The unit cost of the i-type raw material is marked as Ci, and the cost budget of the unit quantity of compound feed is marked as Cmax; Set the recipe ratio variable xki for the i-th raw material in the k-th growth stage; Set the minimum required contents UDk and Dkj for unsaturated fatty acids and the jth nutrient element in the kth growth stage, respectively; For the kth growth stage: Constructing a formula objective function based on the unit content of unsaturated fatty acids, the unit content of nutrient elements, the preset target ratio of unsaturated fatty acids, the preset target ratio of nutrient elements, and formula ratio variables; The formula of the recipe objective function Fk is set as: ;Wherein, w0 and wj are both preset proportional coefficients; Construct a set of constraints based on the formula ratio variables, the minimum required content of unsaturated fatty acids and the jth nutrient element, the unit cost of each raw material and the cost budget; The set of constraints is constructed as follows: The set of constraints in the kth growth stage is labeled Uk; Then the constraints contained in Uk include: ; The constraint is that the unit content of unsaturated fatty acids must not be lower than the minimum required content; ; The constraint is that the total content of any j-th nutrient element must not be lower than the minimum required content corresponding to the j-th nutrient element; ; The constraint is that the sum of the proportions of all raw materials should be 1; ; Cost constraints per unit quantity of compound feed; Taking maximizing the recipe objective function as the optimization goal and the constraint condition set as the constraint, the constructed linear programming model is used as the recipe optimization model.

5. A method for increasing the proportion of unsaturated fatty acids in a sturgeon feed according to claim 4, characterized in that: The method of using the optimization solver to solve the recipe optimization model and iteratively solve and output the optimized recipe result is: For each growth stage of sturgeon: A linear programming solver is used to solve the formula optimization model corresponding to the growth stage to obtain the variable values ​​of the formula proportion variables of each raw material; the variable values ​​of the formula proportion variables of each raw material constitute the optimized formula result.

6. A method for increasing the proportion of unsaturated fatty acids in a sturgeon feed according to claim 5, characterized in that: The result of inputting the optimized formula into the production process system is: According to the growth stage of the sturgeons to be raised, the raw materials are proportioned according to the variable values ​​of the formula proportion variables of each raw material in the optimization formula result of the corresponding growth stage to obtain the production raw materials of the mixed compound feed.

7. A method for increasing the proportion of unsaturated fatty acids in sturgeon feed according to claim 6, characterized in that: The method of real-time monitoring of production process monitoring data in the production process is: Use spectral analysis technology to monitor the unsaturated fatty acid content of mixed raw materials in real time during the production of compound feed; Through sensors of various physical parameters, real-time parameter values ​​of various physical parameters in the process of compound feed production are collected in real time; The real-time unsaturated fatty acid content and the real-time parameter values ​​of various physical parameters constitute the production process monitoring data.

8. A method for increasing the proportion of unsaturated fatty acids in sturgeon feed according to claim 7, characterized in that: The method of constructing the production stability fitness function based on the production process monitoring data is as follows: The initial moment of compound feed production is marked as 0: Let any time of compound feed production be marked as t; The number of each physical parameter is marked as p; The standard deviation of the pth physical parameter from time 0 to time t is marked as ; Based on the standard deviation of each physical parameter, a production stability fitness function is constructed; The expression of the production stability fitness function G is set as: .

9. A method for increasing the proportion of unsaturated fatty acids in a sturgeon feed according to claim 8, characterized in that: The method of adjusting the production parameters using the reinforcement learning model based on the production stability fitness function and the production process monitoring data is as follows: The reinforcement learning model is set to an Actor-Critic model; Define the state space, action space, and reward function of the reinforcement learning model; The state space includes the content of unsaturated fatty acids at any time t in the production process and the real-time parameter values ​​of various physical parameters; The action space includes the adjustment amounts of all regulated production parameters; The reward function is a weighted sum of the production stability fitness function and the unsaturated fatty acid consumption value, wherein the weight of the unsaturated fatty acid consumption value is a negative value; Initialize the model parameters of the Actor network and the Critic network, and set the optimizer to the Adam optimizer; At each time t: Get the current state from the environment, select an action using the Actor network, and execute the selected action to observe the state at the next moment, calculate the reward value, and store the experience in the experience playback buffer, which includes the current state, the selected action, the state at the next moment, and the reward value; Randomly sample a batch of experiences from the experience replay buffer and calculate the target Q value; Use the mean square error function as the loss function of the Critic network model; use the soft update strategy to update the parameters of the Critic network; Through multiple iterative updates, the production parameters are optimized to maximize the reward function.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon; When the computer program is executed on a computer device, the computer device executes the method for increasing the proportion of unsaturated fatty acids in a sturgeon feed as described in any one of claims 1 to 9 in the background.

Images