Method for culturing seawater rotifers by using prawn shell fermentation product

Abstract

The invention discloses a method for culturing seawater rotifers by using prawn shell leavening, which comprises the following steps: S1, cleaning prawn shells to obtain a fermentation substrate; s2, dynamically optimizing fermentation conditions by adopting a multi-agent collaborative reinforcement learning algorithm; s3, predicting the optimal completion time of fermentation by adopting a mixed graph convolutional neural network and an adaptive genetic algorithm, and providing dilution suggestions through a spectral analysis technology; s4, dynamically calculating the dilution ratio, and mixing the fermentation liquor with seawater; s5, pouring the diluent into a rotifer culture pond, predicting the health state of the rotifer by adopting a Vision Transform model, and automatically adjusting a nutrition formula or environmental conditions; s6, monitoring the density of the rotifers in real time through a YOLOv5 target detection model and a camera, and keeping the rotifers within a preset density range; and S7, performing intelligent management on waste treatment based on a multi-objective evolutionary algorithm of non-dominated sorting. According to the method, a multi-agent collaborative algorithm and a deep learning technology are adopted, and intelligent culture of seawater rotifers and optimized treatment of wastes are realized.

Description

Technical Field

[0001] The invention relates to the technical field of aquaculture, and in particular to a method for culturing seawater rotifers by utilizing shrimp shell fermentation products. Background Art

[0002] Marine rotifers are an important aquaculture organism, which is widely used in feeds for aquatic animals such as shrimps and fish. Especially in the seedling stage, rotifers provide rich nutrients and easy digestibility. The cultivation of marine rotifers has important economic and production significance. However, traditional rotifer cultivation methods have many challenges in terms of efficiency, cost and environmental impact. Especially in large-scale aquaculture, how to provide a stable nutrient supply, optimize culture conditions, and deal with the large amount of waste generated during the aquaculture process has become an urgent problem to be solved in the aquaculture industry.

[0003] In the prior art, the cultivation of rotifers mainly relies on artificial feeding and control of environmental conditions. Traditional breeding systems usually need to rely on experience and judgment, and manually adjust the nutritional ratio and environmental parameters. They are easily affected by external factors and cannot achieve stable production efficiency. At the same time, in traditional fermentation systems, fermentation conditions (such as temperature, humidity, pH and oxygen concentration) rely on simple fixed settings and lack real-time adjustment capabilities, resulting in resource waste and instability during the fermentation process. In addition, the judgment of the fermentation completion time usually relies on human experience, which is not only time-consuming and labor-intensive, but also has low accuracy, which may lead to uneven nutritional components of the fermented product and affect the growth of rotifers.

[0004] In the process of rotifer cultivation, feeding strategy and nutrient supply are also one of the difficulties in the existing technology. Since the nutritional requirements of rotifers change significantly with the changes in their growth stages, the traditional timed and quantitative feeding method is difficult to effectively meet the actual needs of rotifers, which can easily lead to nutritional imbalance and affect the reproductive efficiency and health status of rotifers. In addition, the existing technology also has shortcomings in the monitoring and management of rotifer density. It usually relies on manual sampling and observation, resulting in inaccurate density control, which is prone to problems such as over-reproduction of rotifers or low density.

[0005] Waste disposal is another urgent problem to be solved in aquaculture. In the prior art, the waste generated during the rotifer culture process is usually directly treated or simply utilized, but this method not only wastes recyclable resources, but also increases the burden on the environment. How to achieve efficient resource recovery and minimize environmental impact in the waste treatment process is a major challenge in modern aquaculture. Traditional waste treatment methods often lack intelligent management methods, and it is difficult to find an effective balance between environmental impact and resource utilization, resulting in high treatment costs and low recycling efficiency.

[0006] With the rapid development of artificial intelligence technology, intelligent aquaculture solutions have gradually attracted widespread attention and application. Some existing technologies attempt to introduce intelligent means into fermentation control, nutrient supply and waste treatment, but most technical solutions have problems such as single intelligent algorithm, limited optimization ability and poor system adaptability. For example, some intelligent control systems can only make feedback adjustments based on single sensor data during the fermentation process and cannot cope with the dynamic changes of complex multi-dimensional environmental parameters. In addition, most of these systems lack collaborative optimization capabilities and cannot effectively integrate the parameters of fermentation, cultivation and waste treatment, resulting in low overall system efficiency.

[0007] In the existing intelligent fermentation control schemes, the application of reinforcement learning algorithms is relatively rare, and most of them are simple single-agent reinforcement learning. Due to the complexity of the rotifer culture environment, single-agent systems are difficult to cope with multi-dimensional environmental control needs. At the same time, there are limitations in the prediction of fermentation completion time in existing technologies. Most of them rely on fixed experience values ​​and lack dynamic adjustment and prediction algorithms based on real-time data, resulting in insufficient flexibility and precision in the fermentation process.

[0008] In addition, most existing nutrition supply strategies use pre-set nutrition ratios, while ignoring the real-time health status and changes in nutrition needs of rotifers. This fixed nutrition supply method is not only prone to excessive or insufficient nutrition, but may also affect the reproductive efficiency and health of rotifers due to incompatibility of nutrients. Even if some systems have introduced nutrient analysis technology, its analysis speed and accuracy are still difficult to meet the needs of practical applications.

[0009] In addition, in terms of waste treatment, traditional technical means usually rely on simple physical or chemical methods to treat waste, but lack systematic and intelligent treatment processes. Traditional waste treatment solutions are mostly focused on reducing treatment costs, and fail to fully consider environmental burdens and resource recycling issues, resulting in a large number of potential resources not being effectively recycled, and the environmental protection and economic benefits of waste treatment cannot be optimally balanced.

[0010] Therefore, how to provide a method for culturing marine rotifers using shrimp shell fermentation products is an urgent problem that technicians in this field need to solve. Summary of the invention

[0011] One purpose of the present invention is to propose a method for cultivating marine rotifers using fermented shrimp shells. The present invention adopts a multi-agent collaborative reinforcement learning algorithm, a hybrid graph convolutional neural network and a non-dominated sorting multi-objective evolutionary algorithm, combined with deep learning technologies such as the Vision Transformer model, to achieve intelligent optimization management of the entire process from fermentation process, nutrient supply, rotifer health monitoring to waste treatment. By dynamically controlling fermentation conditions, accurately predicting fermentation completion time, real-time monitoring of rotifer health status and automatically adjusting the culture environment and nutrient formula, the purpose of efficiently cultivating rotifers, improving resource utilization and reducing environmental burden is finally achieved, with the advantages of high efficiency, precision, intelligence and environmental protection.

[0012] A method for culturing marine rotifers using shrimp shell fermentation products according to an embodiment of the present invention comprises the following steps:

[0013] S1, washing the collected shrimp shells, removing impurities and crushing them into particles of 3-5 mm in size to obtain a fermentation matrix;

[0014] S2, mixing the fermentation substrate with water in a ratio of 1:3 to 1:5, using a multi-agent collaborative reinforcement learning algorithm to control the fermentation system, and dynamically optimizing the fermentation conditions through interaction between agents based on temperature, humidity, pH value and oxygen concentration;

[0015] S3, using a hybrid graph convolutional neural network and adaptive genetic algorithm to predict the optimal completion time of fermentation based on real-time data, and analyze the nutrients in the fermentation liquid through spectral analysis technology before the end of fermentation to provide dilution suggestions;

[0016] S4, dynamically calculating the dilution ratio according to the dilution suggestion and the growth requirements of rotifers, and mixing the fermentation liquid with seawater through an intelligent dilution device;

[0017] S5, pouring the diluted solution into the rotifer culture tank, using the Vision Transformer model to analyze the behavior of the rotifers in real time, predicting the health status of the rotifers and automatically adjusting the nutritional formula or environmental conditions;

[0018] S6, monitor the rotifer density in real time through the YOLOv5 target detection model and camera, automatically adjust the rotifer density and dynamically control the harvesting and feeding strategies to keep the rotifer within the preset density range;

[0019] S7. A multi-objective evolutionary algorithm based on non-dominated sorting is used to intelligently manage waste treatment. By optimizing the environmental burden and resource utilization efficiency, the Pareto optimal solution between waste treatment and resource recycling is found, and the optimal waste treatment solution is automatically selected.

[0020] Optionally, the S2 specifically includes:

[0021] S21, preliminarily mixing the fermentation substrate and water in a ratio of 1:3 to 1:5, setting multiple intelligent agents, the intelligent agents monitor the environmental state in real time through multi-dimensional environmental sensors, and constructing an initial state vector ,in represents the initial temperature, represents the initial humidity, represents the initial pH, represents the initial oxygen concentration;

[0022] S22, the agent based on the current state of the parameters , select the optimal action through reinforcement learning strategy , the action is determined by the policy network The policy network is continuously optimized based on global environmental feedback, and agents share state information and coordinate behaviors;

[0023] S23. Global Reward Function Designed as a nonlinear function combination of agents, taking into account the interdependence between fermentation parameters:

[0024] ;

[0025] in, represents the total number of agents, Indicates Agents at time Partial rewards, Indicates The oscillation factor of each agent, Indicates The sensitivity of an agent to state deviations, Indicates The difference between the current state and the target state of an agent, Indicates Agent and The state difference of each agent, Indicates Agent and The dependency coefficient between agents, Indicates The state fluctuations of the agents, Indicates The global influence weight of each agent;

[0026] S24, update the agent's strategy:

[0027] ;

[0028] in, represents the strategy parameters of the agent, represents the mathematical expectation, represents the policy parameter gradient, Indicates Agents in state Next Select Behavior The probability of Indicates The oscillation factor of each agent, Indicates Agent and The coefficient of state dependence between agents, Indicates The agent is The influence coefficient of each agent, Indicates The current state of the agent, Indicates The target state of an agent, Indicates The adjustment sensitivity of each agent, Indicates The contribution weight of each agent;

[0029] S25, the agent continuously adjusts the fermentation conditions through adaptive algorithms and dynamically optimizes the fermentation conditions based on historical data and real-time status feedback;

[0030] S26. When the temperature, humidity, pH value and oxygen concentration of the fermentation environment gradually approach the preset target state values, the intelligent agent slows down the frequency of parameter adjustment through the automatic adjustment mechanism, and automatically triggers the stop of the fermentation operation when the fermentation environment reaches the optimal state.

[0031] Optionally, the S3 specifically includes:

[0032] S31, collecting multi-dimensional data through the sensor network, inputting the hybrid graph convolutional neural network to extract spatiotemporal features, and generating a node state matrix;

[0033] S32, hybrid graph convolutional neural network converts the collected multi-dimensional data into a spatiotemporal graph ,in represents the set of sensor nodes, Represents the spatiotemporal correlation edges between nodes, each node The state is represented by , the state of the corresponding node is the monitored fermentation parameter value, and the hybrid graph convolutional neural network extracts the spatiotemporal features of each node and its neighborhood through the convolution layer and outputs the spatiotemporal feature matrix , where the features of each node are shared with neighboring nodes through a complex topological structure, generating a spatiotemporal relationship matrix;

[0034] S33, spatiotemporal feature matrix extracted by hybrid graph convolutional neural network , combined with adaptive genetic algorithm to dynamically predict the fermentation completion time:

[0035] ;

[0036] in, represents the total number of agents, represents the set of internal parameters used to control the behavior of the agent, represents the time variable in the fermentation process, represents the fitness function, represents the hyperbolic sine function, and Indicates The oscillation factor of each agent, represents the control factor, Indicates The deviation between the current nutrient composition of each agent and the target concentration, Indicates Agents at time The humidity at Indicates Agent current time The difference between the pH and the target pH, Indicates the current time The oxygen concentration, Indicates the current time The fermentation temperature, Indicates The sensitivity control parameters of each agent to the interaction between oxygen and temperature, Indicates The internal parameter set of the agent The weight coefficient of

[0037] S34, through the fitness function The calculated results determine the optimal end time of fermentation. At this time, the fitness function value reaches the maximum value, the system determines that the fermentation is completed, and outputs an end signal;

[0038] S35. Before the end of fermentation, the nutrients in the fermentation liquid are analyzed in real time by spectral analysis technology, and the nutrient concentration matrix is ​​generated by spectral analysis algorithm using the spectral feature matrix. , the nutrient concentration matrix Target nutrient concentration Conduct comparative analysis to determine whether the quality and composition of the fermentation broth meet the requirements and provide dilution recommendations.

[0039] Optionally, the S4 specifically includes:

[0040] S41. Collecting real-time nutrient concentration vectors in fermentation broth ,in Indicates time The fermentation liquid The concentration of the nutrients in the seawater is combined with the initial concentration vector of the nutrients in the seawater. ,in Indicates time Time in the sea The concentration of various nutrients is determined, and the target nutrient concentration is set according to the growth stage of the rotifer. ,in Indicates time Moment Target concentrations of nutrients, Indicates the number of nutrient types;

[0041] S42, based on the current real-time nutrient concentration vector of the fermentation liquid and target nutrient concentration and the initial nutrient concentration vector in seawater Calculate the dilution ratio , to meet the nutritional needs of rotifers;

[0042] S43. Calculate the dilution ratio , adjust the dilution amount based on the current real-time nutrient concentration of the fermentation broth and the growth requirements of rotifers:

[0043] ;

[0044] in, Indicates The factors that affect the periodic changes of nutrients over time, represents the smoothing term, Indicates the difference between the actual nutrient concentration and the target concentration, Indicates the temperature regulation sensitivity, Indicates the difference between the current fermentation liquid temperature and the target temperature. Indicates pH adjustment sensitivity, Indicates the difference between the current pH of the fermentation liquid and the target pH. represents the dilution dynamic adjustment coefficient, represents a time function;

[0045] S44, based on the calculated dilution ratio , the intelligent dilution device adjusts the volume of fermentation liquid and seawater so that the ratio of the volume of fermentation liquid to the volume of seawater is equal to the dilution ratio ;

[0046] S45, intelligent dilution device monitors the concentration of nutrients after dilution through real-time feedback mechanism , with the target concentration For comparison, if the concentration deviation exceeds the tolerance range, the dilution system recalculates the dilution ratio and automatically adjusts the mixing ratio of fermentation liquid and seawater:

[0047] ;

[0048] in, Indicates the new dilution ratio, Indicates Parameters related to fluctuations in nutrients, represents the exponential function, It indicates the difference between the actual concentration and the target concentration after dilution. Represents the dilution dynamic adjustment coefficient.

[0049] Optionally, the S5 specifically includes:

[0050] S51, collecting behavioral image data of rotifers in real time through a camera, generating a multi-dimensional behavioral data matrix, and capturing the movement trajectory, morphology and swimming speed characteristics of the rotifers;

[0051] S52: Input the collected image data into the Vision Transformer model, which divides each image into multiple small blocks to form an image block matrix. ,in Indicates image blocks, Indicates the total number of image blocks;

[0052] S53. Through the multi-layer self-attention mechanism of the Vision Transformer model, the dependencies between image blocks are calculated to generate a behavior feature matrix :

[0053] ;

[0054] in, Indicates time Moment The self-attention weight of each image patch, Indicates The feature value of the image block, Indicates The feature value of the image block, represents the exponential function, and represents the time function of adjusting the weights in response to different feature blocks, Indicates image blocks and The dependencies between image patches, Indicates image blocks and The parameters of the temporal dependence between image patches, and represents the nonlinear term that controls the feature dependency between image blocks, represents the hyperbolic cosine function;

[0055] S54, based on the behavioral characteristics matrix , calculate the predicted value of the health status of rotifers :

[0056] ;

[0057] in, Indicates The behavior feature value extracted from the image block, represents the time fluctuation parameter that captures the behavior characteristics, and represents the control factor, Indicates The difference between the behavioral characteristics of each image block and the normal healthy state, represents the coefficient of adjusting the influence of health status on characteristic deviation, represents the fluctuation parameter;

[0058] S55. Predicted value based on health status and target health status The difference in the nutrient composition of the rotifer culture medium is automatically adjusted;

[0059] S56. While adjusting the nutritional components according to the predicted health status value, the system monitors and adjusts the environmental parameters of rotifer culture in real time.

[0060] Optionally, the S7 specifically includes:

[0061] S61, collecting multi-dimensional data in the waste treatment process, wherein the multi-dimensional data includes environmental burden parameters, treatment costs, and resource recycling rates, to form a multi-dimensional data set;

[0062] S62, initializing waste treatment plans, generating a set of random initial plans, each plan including specific waste treatment steps, resource utilization methods and environmental indicators, and the initial plan set is used as input of the multi-objective evolutionary algorithm;

[0063] S63, using a non-dominated sorting genetic algorithm to evaluate each waste treatment plan, the evaluation objectives include minimizing environmental burden, minimizing treatment costs and maximizing resource recycling rate, and the system preliminarily sorts each plan according to the weight of each goal;

[0064] S64, performing non-dominated sorting on the evaluated treatment plans to determine a Pareto optimal solution, where the Pareto optimal solution represents a solution that cannot further optimize other goals without affecting a certain goal, thereby forming a set of optimal solutions that balance environmental benefits and economic benefits;

[0065] S65, generating a new processing scheme through a crossover operation and a mutation operation, wherein the crossover operation exchanges the processing strategies and parameters in the existing scheme, and the mutation operation introduces random changes;

[0066] S66. After several generations of evolution, the treatment plan with the highest fitness is selected as the final waste treatment strategy, and the waste treatment process is optimized.

[0067] The beneficial effects of the present invention are:

[0068] First, the present invention uses a multi-agent collaborative reinforcement learning algorithm to perform global dynamic optimization of the fermentation process. Based on multi-dimensional environmental parameters such as temperature, humidity, pH value and oxygen concentration, the agents interact collaboratively to jointly adjust the fermentation conditions to ensure that all parameters in the fermentation process are always in the best state. This innovative fermentation control method overcomes the shortcomings of the prior art that the fermentation conditions are fixed and cannot be dynamically adjusted, avoids resource waste and instability, greatly improves the quality and stability of the fermentation liquid, and provides a high-quality nutrient matrix for subsequent rotifer culture.

[0069] In addition, the present invention adopts a method combining a hybrid graph convolutional neural network and an adaptive genetic algorithm to dynamically predict the fermentation completion time based on real-time sensor data, avoiding the uncertainty and low precision of the traditional reliance on experience to judge the end time of fermentation. Through the spatiotemporal feature extraction capability of the graph convolutional network, the present invention can effectively capture complex environmental changes during the fermentation process, ensuring that the nutrients in the fermentation liquid reach the ideal state at the optimal time point. At the same time, spectral analysis technology is introduced into the real-time monitoring of the fermentation liquid to accurately analyze the changes in nutrients, making the dilution recommendations more targeted and scientific, and meeting the nutritional needs of rotifers at different growth stages.

[0070] In the core link of rotifer cultivation, the present invention uses the Vision Transformer model to monitor the behavior, morphology and swimming state of rotifers in real time and predict their health status, ensuring that the growth environment and nutritional supply of rotifers are always in the best state. Compared with the traditional method that relies on manual observation and simple sensor monitoring, the Vision Transformer model uses the self-attention mechanism of deep learning to more accurately capture subtle changes in rotifer behavior and automatically adjust nutritional formulas and environmental conditions according to fluctuations in health status. This intelligent and refined health management not only improves the survival rate and reproductive efficiency of rotifers, but also reduces the risk of abnormal growth due to environmental discomfort or malnutrition.

[0071] In the nutrient supply link, the present invention realizes the dynamic mixing of fermentation liquid and seawater based on the intelligent regulation of the dilution device, and accurately calculates the dilution ratio according to the real-time monitored nutrient concentration and the actual needs of rotifers, avoiding the nutrient waste or insufficient problems caused by the traditional fixed feeding method. The dilution device can automatically adjust the nutrient ratio through the feedback control mechanism to ensure that each round of dilution can meet the growth needs of rotifers, greatly improving the culture efficiency and nutrient utilization rate.

[0072] At the same time, the present invention introduces a multi-objective evolutionary algorithm of non-dominated sorting in waste treatment, and systematically optimizes the environmental burden, treatment cost and resource utilization efficiency in the waste treatment process. Through the collection and analysis of multi-dimensional data, the present invention can find the best balance between environmental protection and economic benefits, effectively reduce the impact on the environment during waste treatment, and improve the recycling rate of resources. Compared with the prior art, traditional waste treatment methods often cannot effectively balance environmental protection and economic benefits, while the optimization method of the present invention can intelligently adjust the waste treatment process and achieve efficient recycling and reuse of waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0074] Figure 1 A flow chart of a method for culturing marine rotifers using shrimp shell fermentation products proposed by the present invention;

[0075] Figure 2 The present invention provides a structural schematic diagram of a method for culturing marine rotifers using shrimp shell fermentation products. DETAILED DESCRIPTION

[0076] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0077] refer to Figure 1 and Figure 2 A method for culturing marine rotifers using shrimp shell fermentation products comprises the following steps:

[0078] S1, washing the collected shrimp shells, removing impurities and crushing them into particles of 3-5 mm in size to obtain a fermentation matrix;

[0079] S2, mixing the fermentation substrate with water in a ratio of 1:3 to 1:5, using a multi-agent collaborative reinforcement learning algorithm to control the fermentation system, and dynamically optimizing the fermentation conditions through interaction between agents based on temperature, humidity, pH value and oxygen concentration;

[0080] S3, using a hybrid graph convolutional neural network and adaptive genetic algorithm to predict the optimal completion time of fermentation based on real-time data, and analyze the nutrients in the fermentation liquid through spectral analysis technology before the end of fermentation to provide dilution suggestions;

[0081] S4, dynamically calculating the dilution ratio according to the dilution suggestion and the growth requirements of rotifers, and mixing the fermentation liquid with seawater through an intelligent dilution device;

[0082] S5, pouring the diluted solution into the rotifer culture tank, using the Vision Transformer model to analyze the behavior of the rotifers in real time, predicting the health status of the rotifers and automatically adjusting the nutritional formula or environmental conditions;

[0083] S6, monitor the rotifer density in real time through the YOLOv5 target detection model and camera, automatically adjust the rotifer density and dynamically control the harvesting and feeding strategies to keep the rotifer within the preset density range;

[0084] S7. A multi-objective evolutionary algorithm based on non-dominated sorting is used to intelligently manage waste treatment. By optimizing the environmental burden and resource utilization efficiency, the Pareto optimal solution between waste treatment and resource recycling is found, and the optimal waste treatment solution is automatically selected.

[0085] In this implementation, S2 specifically includes:

[0086] S21, preliminarily mixing the fermentation substrate and water in a ratio of 1:3 to 1:5, setting multiple intelligent agents, the intelligent agents monitor the environmental state in real time through multi-dimensional environmental sensors, and constructing an initial state vector ,in represents the initial temperature, represents the initial humidity, represents the initial pH, represents the initial oxygen concentration;

[0087] S22, the agent based on the current state of the parameters , select the optimal action through reinforcement learning strategy , the action is determined by the policy network The policy network is continuously optimized based on global environmental feedback, and agents share state information and coordinate behaviors;

[0088] S23. Global Reward Function Designed as a nonlinear function combination of agents, taking into account the interdependence between fermentation parameters:

[0089] ;

[0090] in, represents the total number of agents, Indicates Agents at time Partial rewards, Indicates The oscillation factor of each agent, Indicates The sensitivity of an agent to state deviations, Indicates The difference between the current state and the target state of an agent, Indicates Agent and The state difference of each agent, Indicates Agent and The dependency coefficient between agents, Indicates The state fluctuations of the agents, Indicates The global influence weight of each agent;

[0091] S24, update the agent's strategy:

[0092] ;

[0093] in, represents the strategy parameters of the agent, represents the mathematical expectation, represents the policy parameter gradient, Indicates Agents in state Next Select Behavior The probability of Indicates The oscillation factor of each agent, Indicates Agent and The coefficient of state dependence between agents, Indicates The agent is The influence coefficient of each agent, Indicates The current state of the agent, Indicates The target state of an agent, Indicates The adjustment sensitivity of each agent, Indicates The contribution weight of each agent;

[0094] S25, the agent continuously adjusts the fermentation conditions through adaptive algorithms and dynamically optimizes the fermentation conditions based on historical data and real-time status feedback;

[0095] S26. When the temperature, humidity, pH value and oxygen concentration of the fermentation environment gradually approach the preset target state values, the intelligent agent slows down the frequency of parameter adjustment through the automatic adjustment mechanism, and automatically triggers the stop of the fermentation operation when the fermentation environment reaches the optimal state.

[0096] In this implementation, S3 specifically includes:

[0097] S31, collecting multi-dimensional data through the sensor network, inputting the hybrid graph convolutional neural network to extract spatiotemporal features, and generating a node state matrix;

[0098] S32, hybrid graph convolutional neural network converts the collected multi-dimensional data into a spatiotemporal graph ,in represents the set of sensor nodes, Represents the spatiotemporal correlation edges between nodes, each node The state is represented by , the state of the corresponding node is the monitored fermentation parameter value, and the hybrid graph convolutional neural network extracts the spatiotemporal features of each node and its neighborhood through the convolution layer and outputs the spatiotemporal feature matrix , where the features of each node are shared with neighboring nodes through a complex topological structure, generating a spatiotemporal relationship matrix;

[0099] S33, spatiotemporal feature matrix extracted by hybrid graph convolutional neural network , combined with adaptive genetic algorithm to dynamically predict the fermentation completion time:

[0100] ;

[0101] in, represents the total number of agents, represents the set of internal parameters used to control the behavior of the agent, represents the time variable in the fermentation process, represents the fitness function, represents the hyperbolic sine function, and Indicates The oscillation factor of each agent, represents the control factor, Indicates The deviation between the current nutrient composition of each agent and the target concentration, Indicates Agents at time The humidity at Indicates Agent current time The difference between the pH and the target pH, Indicates the current time The oxygen concentration, Indicates the current time The fermentation temperature, Indicates The sensitivity control parameters of each agent to the interaction between oxygen and temperature, Indicates The internal parameter set of the agent The weight coefficient of

[0102] S34, through the fitness function The calculated results determine the optimal end time of fermentation. At this time, the fitness function value reaches the maximum value, the system determines that the fermentation is completed, and outputs an end signal;

[0103] S35. Before the end of fermentation, the nutrients in the fermentation liquid are analyzed in real time by spectral analysis technology, and the nutrient concentration matrix is ​​generated by spectral analysis algorithm using the spectral feature matrix. , the nutrient concentration matrix Target nutrient concentration Conduct comparative analysis to determine whether the quality and composition of the fermentation broth meet the requirements and provide dilution recommendations.

[0104] In this implementation manner, the S4 specifically includes:

[0105] S41. Collecting real-time nutrient concentration vectors in fermentation broth ,in Indicates time The fermentation liquid The concentration of the nutrients in the seawater is combined with the initial concentration vector of the nutrients in the seawater. ,in Indicates time Time in the sea The concentration of various nutrients is determined, and the target nutrient concentration is set according to the growth stage of the rotifer. ,in Indicates time Moment Target concentrations of nutrients, Indicates the number of nutrient types;

[0106] S42, based on the current real-time nutrient concentration vector of the fermentation liquid and target nutrient concentration and the initial nutrient concentration vector in seawater Calculate the dilution ratio , to meet the nutritional needs of rotifers;

[0107] S43. Calculate the dilution ratio , adjust the dilution amount based on the current real-time nutrient concentration of the fermentation broth and the growth requirements of rotifers:

[0108] ;

[0109] in, Indicates The factors that affect the cyclical changes of nutrients over time, represents the smoothing term, Indicates the difference between the actual nutrient concentration and the target concentration, Indicates the temperature regulation sensitivity, Indicates the difference between the current fermentation liquid temperature and the target temperature. Indicates pH adjustment sensitivity, Indicates the difference between the current pH of the fermentation liquid and the target pH. represents the dilution dynamic adjustment coefficient, represents a time function;

[0110] S44, based on the calculated dilution ratio , the intelligent dilution device adjusts the volume of fermentation liquid and seawater so that the ratio of the volume of fermentation liquid to the volume of seawater is equal to the dilution ratio ;

[0111] S45, intelligent dilution device monitors the concentration of nutrients after dilution through real-time feedback mechanism , with the target concentration For comparison, if the concentration deviation exceeds the tolerance range, the dilution system recalculates the dilution ratio and automatically adjusts the mixing ratio of fermentation liquid and seawater:

[0112] ;

[0113] in, Indicates the new dilution ratio, Indicates Parameters related to fluctuations in nutrients, represents the exponential function, It indicates the difference between the actual concentration and the target concentration after dilution. Represents the dilution dynamic adjustment coefficient.

[0114] In this implementation, S5 specifically includes:

[0115] S51, collecting behavioral image data of rotifers in real time through a camera, generating a multi-dimensional behavioral data matrix, and capturing the movement trajectory, morphology and swimming speed characteristics of the rotifers;

[0116] S52: Input the collected image data into the Vision Transformer model, which divides each image into multiple small blocks to form an image block matrix. ,in Indicates image blocks, Indicates the total number of image blocks;

[0117] S53. Through the multi-layer self-attention mechanism of the Vision Transformer model, the dependencies between image blocks are calculated to generate a behavior feature matrix :

[0118] ;

[0119] in, Indicates time Moment The self-attention weight of each image patch, Indicates The feature value of the image block, Indicates The feature value of the image block, represents the exponential function, and represents the time function of adjusting the weights in response to different feature blocks, Indicates image blocks and The dependencies between image patches, Indicates image blocks and The parameters of the temporal dependence between image patches, and represents the nonlinear term that controls the feature dependency between image blocks, represents the hyperbolic cosine function;

[0120] S54, based on the behavioral characteristics matrix , calculate the predicted value of the health status of rotifers :

[0121] ;

[0122] in, Indicates The behavior feature value extracted from the image block, represents the time fluctuation parameter that captures the behavior characteristics, and represents the control factor, Indicates The difference between the behavioral characteristics of each image block and the normal healthy state, represents the coefficient of adjusting the influence of health status on characteristic deviation, represents the fluctuation parameter;

[0123] S55. Predicted value based on health status and target health status The difference in the nutrient composition of the rotifer culture medium is automatically adjusted;

[0124] S56. While adjusting the nutritional components according to the predicted health status value, the system monitors and adjusts the environmental parameters of rotifer culture in real time.

[0125] In this implementation manner, the S7 specifically includes:

[0126] S61, collecting multi-dimensional data in the waste treatment process, wherein the multi-dimensional data includes environmental burden parameters, treatment costs, and resource recycling rates, to form a multi-dimensional data set;

[0127] S62, initializing waste treatment plans, generating a set of random initial plans, each plan including specific waste treatment steps, resource utilization methods and environmental indicators, and the initial plan set is used as input of the multi-objective evolutionary algorithm;

[0128] S63, using a non-dominated sorting genetic algorithm to evaluate each waste treatment plan, the evaluation objectives include minimizing environmental burden, minimizing treatment costs and maximizing resource recycling rate, and the system preliminarily sorts each plan according to the weight of each goal;

[0129] S64, performing non-dominated sorting on the evaluated treatment plans to determine a Pareto optimal solution, where the Pareto optimal solution represents a solution that cannot further optimize other goals without affecting a certain goal, thereby forming a set of optimal solutions that balance environmental benefits and economic benefits;

[0130] S65, generating a new processing scheme through a crossover operation and a mutation operation, wherein the crossover operation exchanges the processing strategies and parameters in the existing scheme, and the mutation operation introduces random changes;

[0131] S66. After several generations of evolution, the treatment plan with the highest fitness is selected as the final waste treatment strategy, and the waste treatment process is optimized.

[0132] Embodiment 1:

[0133] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a large shrimp farm. The farm has as many as 100 breeding ponds, and the demand for rotifers in each pond is huge. However, the traditional rotifer cultivation method is overly dependent on manual regulation and the fermentation conditions are difficult to accurately control, resulting in unstable rotifer production and quality, low nutrient utilization, and waste treatment methods that are not environmentally friendly, resulting in serious waste of resources.

[0134] In this embodiment, we installed a complete set of intelligent rotifer cultivation system for the farm, which includes an intelligent fermentation system, an intelligent dilution device, a rotifer health and density monitoring system, and an intelligent waste treatment system. First, the pretreatment of shrimp shells is carried out through special cleaning and crushing equipment. About 5 tons of shrimp shells are collected each month. After pretreatment, they are crushed into 3-5 mm particles, which will be put into the intelligent fermentation system as a fermentation matrix.

[0135] The intelligent fermentation system runs based on a multi-agent collaborative reinforcement learning algorithm, and monitors the temperature, humidity, pH value and oxygen concentration during the fermentation process in real time. Each fermentation takes about 5 days, and the system automatically adjusts the fermentation conditions based on the data collected by the sensors. Unlike the farm's previous practice of manually setting fermentation conditions, the intelligent fermentation system can dynamically adjust parameters according to environmental changes during the fermentation process to ensure that the fermentation liquid is always in the best condition.

[0136] After fermentation is completed, the system uses spectral analysis technology to analyze the nutrients in the fermentation liquid in real time and determine the dilution ratio. The demand for rotifers varies according to their growth stages. The traditional fixed feeding method is difficult to meet the actual needs of rotifers, often resulting in excess or insufficient nutrition. In the present invention, the intelligent dilution device can dynamically adjust the dilution ratio of the fermentation liquid and seawater according to the spectral analysis results and the growth status of the rotifers, ensuring that each batch of dilution liquid can accurately meet the nutritional needs of the rotifers.

[0137] During the rotifer cultivation process, the health monitoring system uses the Vision Transformer model to analyze the behavior of rotifers in real time. The camera collects image data of rotifers 24 hours a day. By analyzing this data, the system can monitor the health status of rotifers in real time. For example, when the swimming speed of rotifers slows down significantly, the system will determine that there may be problems with insufficient nutrition or environmental discomfort, and then adjust the nutrient supply or the dilution ratio of the fermentation liquid. At the same time, the density monitoring system uses the YOLOv5 target detection model to monitor the density of rotifers in real time and automatically adjust the feeding and harvesting strategies.

[0138] In addition, the intelligent system also solves the problem of waste disposal. The waste generated during the rotifer cultivation process is treated by the waste treatment system and the resources are recycled. The waste is first treated by biodegradation and converted into organic fertilizer. At the same time, the system optimizes the cost and environmental impact of waste treatment through a non-dominated sorting multi-objective evolutionary algorithm to ensure a balance between resource recycling and environmental protection.

[0139] During the three-month implementation, the intelligent rotifer cultivation system significantly improved the yield and quality of rotifers. The experimental data are shown in the following table:

[0140] Table 1 Comparison of key indicators of the traditional method and the method of the present invention in rotifer culture

[0141] project Traditional methods Method of the present invention Fermentation liquid temperature fluctuation range (℃) 28-34 29.5-30 Humidity Level 80%-95% Stable at 90%-92% pH Range 6.0-7.5 6.5-6.8 Oxygen concentration (%) 8.5%-12% 10%-12% Fluctuation range of protein content in fermentation broth (%) 5.8%-8.0% 6.7%-7.1% Rotifer density fluctuation range (pcs / ml) 15,000-50,000 20,000-40,000 Average daily production of rotifers (liters / day) 950-1050 1500-1700 Average survival rate of rotifers (%) 82%-85% 94%-96% Labor requirements (number of operators) 6 3 Monthly fermentation cycle 4 times 4 times Total production of lunar rotifer (liters) 30,000-32,000 47,000-51,000 Dilution ratio per batch (fermentation liquid: seawater) Fixed to 1:4 Dynamic adjustment to 1:3-1:5 Waste recycling volume (organic fertilizer, kg / month) 100-150 800-850 Carbon emissions (month, kg) 2,000-2,200 1,300-1,500 Error rate of rotifer nutrition supply ±10% ±2% Reduction ratio of breeding costs (%) - 15%-17%

[0142] According to the comparative data in Table 1 above, we can clearly see that after adopting the intelligent system of the present invention, there are significant improvements and optimizations in multiple key indicators.

[0143] First, in terms of controlling the temperature of the fermentation liquid, the temperature fluctuation range of the traditional method is between 28°C and 34°C, which means that the temperature control is unstable and can easily lead to fluctuations in the quality of the fermentation liquid. However, through the intelligent fermentation system of the present invention, the temperature fluctuation is effectively controlled between 29.5°C and 30°C, which greatly reduces the negative impact of temperature on the fermentation process and makes the fermentation environment more stable. Similar improvements are also reflected in humidity control. The humidity fluctuations of the traditional method are large, while the method of the present invention can maintain the humidity between 90%-92%. This stability ensures that the fermentation environment conditions required by rotifers are more ideal.

[0144] The fluctuation range of pH value is also significantly reduced. The fluctuation range of pH value in traditional fermentation system is 6.0 to 7.5, which may lead to imbalance of nutrients in fermentation liquid. However, through the real-time monitoring and regulation of the present invention, the pH value is strictly controlled between 6.5 and 6.8, which effectively improves the stability and quality of fermentation liquid. In addition, the oxygen concentration sometimes fluctuates greatly under traditional methods, but after adopting the intelligent fermentation system, the oxygen concentration is maintained between 10% and 12%, which further ensures the smooth progress of the fermentation process.

[0145] In terms of nutrient supply, the protein content in the fermentation broth is an important indicator for measuring the fermentation effect. Under the traditional method, the protein content fluctuates between 5.8% and 8.0%, indicating that the supply of nutrients is uneven, and it is easy for rotifers to have insufficient or excessive nutrient intake. However, through the intelligent dilution device of the present invention, the protein content is accurately controlled between 6.7% and 7.1%, and the error rate is reduced to ±2%, which greatly improves the accuracy of rotifer nutrition supply and ensures that the needs of rotifers at different growth stages can be met.

[0146] The fluctuation range of rotifer density is also significantly improved. The density of rotifers under traditional methods often fluctuates between 15,000 and 50,000 per milliliter, but after adopting the intelligent system, the density is stabilized between 20,000 and 40,000 per milliliter. This not only improves the growth efficiency of rotifers, but also makes the feeding and harvesting of rotifers more accurate and efficient. Furthermore, the average daily output of rotifers has been greatly increased from 950 to 1050 liters in traditional methods to 1500 to 1700 liters, showing that the system of the present invention has significantly improved the cultivation efficiency of rotifers.

[0147] The survival rate of rotifers is another important measurement indicator. In traditional methods, the average survival rate of rotifers is 82%-85%, while after adopting the method of the present invention, the survival rate is increased to 94%-96%, reflecting the positive role of intelligent monitoring and environmental regulation on the health of rotifers. The high survival rate not only means an increase in the yield of rotifers, but also greatly reduces the mortality rate of rotifers caused by adverse environments or malnutrition.

[0148] The invention also significantly reduces labor requirements. The traditional method requires six operators to manage 100 breeding ponds, but with the help of the intelligent system, the labor requirements are reduced by half, and only three operators are needed to complete the same workload. This not only greatly reduces labor costs, but also improves the automation level and efficiency of operations.

[0149] In terms of waste treatment and resource recycling, the present invention demonstrates outstanding environmental benefits. The traditional method only recycles 100 to 150 kilograms of waste per month, but through the intelligent waste treatment system, the recycling volume is increased to 800 to 850 kilograms, and the carbon emissions are also reduced from 2000-2200 kilograms / month to 1300-1500 kilograms / month. This achievement not only significantly reduces the environmental burden, but also improves the economic benefits of the farm through the reuse of waste.

[0150] In terms of cost, the invention also shows obvious economic advantages. The overall breeding cost of the farm was reduced by 15%-17% after the introduction of the intelligent system. Through intelligent control of nutrient supply, fermentation process optimization, waste treatment and reduction of personnel costs, the farm significantly improved its economic benefits while taking into account environmental protection and resource sustainability.

[0151] In general, the intelligent rotifer cultivation system of the present invention significantly improves the core indicators of rotifer cultivation, not only increasing the yield and survival rate, but also reducing labor input and breeding costs. The introduction of the intelligent monitoring system ensures the precision of nutrient supply and the stability of the fermentation process, greatly improving the production efficiency of the farm and significantly reducing the environmental impact.

[0152] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for culturing marine rotifers using shrimp shell fermentation products, characterized in that: The steps include: S1, washing the collected shrimp shells, removing impurities and crushing them into particles of 3-5 mm in size to obtain a fermentation matrix; S2, mixing the fermentation substrate with water in a ratio of 1:3 to 1:5, using a multi-agent collaborative reinforcement learning algorithm to control the fermentation system, and dynamically optimizing the fermentation conditions through interaction between agents based on temperature, humidity, pH value and oxygen concentration; S3, using a hybrid graph convolutional neural network and adaptive genetic algorithm to predict the optimal completion time of fermentation based on real-time data, and analyze the nutrients in the fermentation liquid through spectral analysis technology before the end of fermentation to provide dilution suggestions; S4, dynamically calculating the dilution ratio according to the dilution suggestion and the growth requirements of rotifers, and mixing the fermentation liquid with seawater through an intelligent dilution device; S5, pouring the diluted solution into the rotifer culture tank, using the Vision Transformer model to analyze the behavior of the rotifers in real time, predicting the health status of the rotifers and automatically adjusting the nutritional formula or environmental conditions; S6, monitor the rotifer density in real time through the YOLOv5 target detection model and camera, automatically adjust the rotifer density and dynamically control the harvesting and feeding strategies to keep the rotifer within the preset density range; S7. A multi-objective evolutionary algorithm based on non-dominated sorting is used to intelligently manage waste treatment. By optimizing the environmental burden and resource utilization efficiency, the Pareto optimal solution between waste treatment and resource recycling is found, and the optimal waste treatment solution is automatically selected.

2. The method for culturing marine rotifers using shrimp shell fermentation products according to claim 1, characterized in that: The S2 specifically includes: S21, preliminarily mixing the fermentation substrate and water in a ratio of 1:3 to 1:5, setting multiple intelligent agents, the intelligent agents monitor the environmental state in real time through multi-dimensional environmental sensors, and constructing an initial state vector ,in represents the initial temperature, represents the initial humidity, represents the initial pH, represents the initial oxygen concentration; S22, the agent based on the current state of the parameters , select the optimal action through reinforcement learning strategy , the action is determined by the policy network The policy network is continuously optimized based on global environmental feedback, and agents share state information and coordinate behaviors; S23. Global Reward Function Designed as a nonlinear function combination of agents, taking into account the interdependence between fermentation parameters: ； in, represents the total number of agents, Indicates Agents at time Partial rewards, Indicates The oscillation factor of each agent, Indicates The sensitivity of an agent to state deviations, Indicates The difference between the current state and the target state of an agent, Indicates Agent and The state difference of each agent, Indicates Agent and The dependency coefficient between agents, Indicates The state fluctuations of the agents, Indicates The global influence weight of each agent; S24, update the agent's strategy: ； in, represents the strategy parameters of the agent, represents the mathematical expectation, represents the policy parameter gradient, Indicates Agents in state Next Select Behavior The probability of Indicates The oscillation factor of each agent, Indicates Agent and The coefficient of state dependence between agents, Indicates The agent is The influence coefficient of each agent, Indicates The current state of the agent, Indicates The target state of an agent, Indicates The adjustment sensitivity of each agent, Indicates The contribution weight of each agent; S25, the agent continuously adjusts the fermentation conditions through adaptive algorithms and dynamically optimizes the fermentation conditions based on historical data and real-time status feedback; S26. When the temperature, humidity, pH value and oxygen concentration of the fermentation environment gradually approach the preset target state values, the intelligent agent slows down the frequency of parameter adjustment through the automatic adjustment mechanism, and automatically triggers the stop of the fermentation operation when the fermentation environment reaches the optimal state.

3. The method for culturing marine rotifers using shrimp shell fermentation products according to claim 1, characterized in that: The S3 specifically includes: S31, collecting multi-dimensional data through the sensor network, inputting the hybrid graph convolutional neural network to extract spatiotemporal features, and generating a node state matrix; S32, hybrid graph convolutional neural network converts the collected multi-dimensional data into a spatiotemporal graph ,in represents the set of sensor nodes, Represents the spatiotemporal correlation edges between nodes, each node The state is represented by , the state of the corresponding node is the monitored fermentation parameter value, and the hybrid graph convolutional neural network extracts the spatiotemporal features of each node and its neighborhood through the convolution layer and outputs the spatiotemporal feature matrix , where the features of each node are shared with neighboring nodes through a complex topological structure, generating a spatiotemporal relationship matrix; S33, spatiotemporal feature matrix extracted by hybrid graph convolutional neural network , combined with adaptive genetic algorithm to dynamically predict the fermentation completion time: ； in, represents the total number of agents, represents the set of internal parameters used to control the behavior of the agent, represents the time variable in the fermentation process, represents the fitness function, represents the hyperbolic sine function, and Indicates The oscillation factor of each agent, represents the control factor, Indicates The deviation between the current nutrient composition of each agent and the target concentration, Indicates Agents at time The humidity at Indicates Agent current time The difference between the pH and the target pH, Indicates the current time The oxygen concentration, Indicates the current time The fermentation temperature, Indicates The sensitivity control parameters of each agent to the interaction between oxygen and temperature, Indicates The internal parameter set of the agent The weight coefficient of S34, through the fitness function The calculated results determine the optimal end time of fermentation. At this time, the fitness function value reaches the maximum value, the system determines that the fermentation is completed, and outputs an end signal; S35. Before the end of fermentation, the nutrients in the fermentation liquid are analyzed in real time by spectral analysis technology, and the nutrient concentration matrix is ​​generated by spectral analysis algorithm using the spectral feature matrix. , the nutrient concentration matrix Target nutrient concentration Conduct comparative analysis to determine whether the quality and composition of the fermentation broth meet the requirements and provide dilution recommendations.

4. The method for culturing marine rotifers using shrimp shell fermentation products according to claim 1, characterized in that: The S4 specifically includes: S41. Collecting real-time nutrient concentration vectors in fermentation broth ,in Indicates time The fermentation liquid The concentration of the nutrients in the seawater is combined with the initial concentration vector of the nutrients in the seawater. ,in Indicates time Time in the sea The concentration of various nutrients is determined, and the target nutrient concentration is set according to the growth stage of the rotifer. ,in Indicates time Moment Target concentrations of nutrients, Indicates the number of nutrient types; S42, based on the current real-time nutrient concentration vector of the fermentation liquid and target nutrient concentration and the initial nutrient concentration vector in seawater Calculate the dilution ratio , to meet the nutritional needs of rotifers; S43. Calculate the dilution ratio , adjust the dilution amount based on the current real-time nutrient concentration of the fermentation broth and the growth requirements of rotifers: ； in, Indicates The factors that affect the periodic changes of nutrients over time, represents the smoothing term, Indicates the difference between the actual nutrient concentration and the target concentration, Indicates the temperature regulation sensitivity, Indicates the difference between the current fermentation liquid temperature and the target temperature. Indicates pH adjustment sensitivity, Indicates the difference between the current pH of the fermentation liquid and the target pH. represents the dilution dynamic adjustment coefficient, represents a time function; S44, based on the calculated dilution ratio , the intelligent dilution device adjusts the volume of fermentation liquid and seawater so that the ratio of the volume of fermentation liquid to the volume of seawater is equal to the dilution ratio ; S45, intelligent dilution device monitors the concentration of nutrients after dilution through real-time feedback mechanism , with the target concentration For comparison, if the concentration deviation exceeds the tolerance range, the dilution system recalculates the dilution ratio and automatically adjusts the mixing ratio of fermentation liquid and seawater: ； in, Indicates the new dilution ratio, Indicates Parameters related to fluctuations in nutrients, represents the exponential function, It indicates the difference between the actual concentration and the target concentration after dilution. Represents the dilution dynamic adjustment coefficient.

5. The method for culturing marine rotifers using shrimp shell fermentation products according to claim 1, characterized in that: The S5 specifically includes: S51, collecting behavioral image data of rotifers in real time through a camera, generating a multi-dimensional behavioral data matrix, and capturing the movement trajectory, morphology and swimming speed characteristics of the rotifers; S52: Input the collected image data into the Vision Transformer model, which divides each image into multiple small blocks to form an image block matrix. ,in Indicates image blocks, Indicates the total number of image blocks; S53. Through the multi-layer self-attention mechanism of the Vision Transformer model, the dependencies between image blocks are calculated to generate a behavior feature matrix : ； in, Indicates time Moment The self-attention weight of each image patch, Indicates The feature value of the image block, Indicates The feature value of the image block, represents the exponential function, and represents the time function of adjusting the weights in response to different feature blocks, Indicates image blocks and The dependencies between image patches, Indicates image blocks and The parameters of the temporal dependence between image patches, and represents the nonlinear term that controls the feature dependency between image blocks, represents the hyperbolic cosine function; S54, based on the behavioral characteristics matrix , calculate the predicted value of the health status of rotifers : ； in, Indicates The behavior feature value extracted from the image block is represents the time fluctuation parameter that captures the behavior characteristics, and represents the control factor, Indicates The difference between the behavioral characteristics of each image block and the normal healthy state, represents the coefficient of adjusting the influence of health status on characteristic deviation, represents the fluctuation parameter; S55. Predicted value based on health status and target health status The difference in the nutrient composition of the rotifer culture medium is automatically adjusted; S56. While adjusting the nutritional components according to the predicted health status value, the system monitors and adjusts the environmental parameters of rotifer culture in real time.

6. The method for culturing marine rotifers using shrimp shell fermentation products according to claim 1, characterized in that: The S7 specifically includes: S61, collecting multi-dimensional data in the waste treatment process, wherein the multi-dimensional data includes environmental burden parameters, treatment costs, and resource recycling rates, to form a multi-dimensional data set; S62, initializing waste treatment plans, generating a set of random initial plans, each plan including specific waste treatment steps, resource utilization methods and environmental indicators, and the initial plan set is used as input of the multi-objective evolutionary algorithm; S63, using a non-dominated sorting genetic algorithm to evaluate each waste treatment plan, the evaluation objectives include minimizing environmental burden, minimizing treatment costs and maximizing resource recycling rate, and the system preliminarily sorts each plan according to the weight of each goal; S64, performing non-dominated sorting on the evaluated treatment plans to determine a Pareto optimal solution, where the Pareto optimal solution represents a solution that cannot further optimize other goals without affecting a certain goal, thereby forming a set of optimal solutions that balance environmental benefits and economic benefits; S65, generating a new processing scheme through a crossover operation and a mutation operation, wherein the crossover operation exchanges the processing strategies and parameters in the existing scheme, and the mutation operation introduces random changes; S66. After several generations of evolution, the treatment plan with the highest fitness is selected as the final waste treatment strategy, and the waste treatment process is optimized.

Images