Ecological aquaculture system

Abstract

The invention relates to the technical field of aquaculture, and discloses an ecological aquaculture system, which comprises a culture pond as a basic component of the whole device and used for bearing and assembling a central data processing system and a subordinate lower structural component thereof; the central data processing system is used for receiving, processing and storing data from each monitoring module; the multi-parameter water quality monitoring module monitors key water quality indexes including dissolved oxygen, temperature, pH value, ammonia nitrogen content and nitrite content in real time; the aquatic animal growth monitoring module is used for measuring growth parameters including body length and body weight of aquatic animals through an image recognition technology; through the multi-parameter water quality monitoring device, the culture environment and the state of the aquatic animals can be mastered in real time, so that accurate management is performed, and the disease early warning device can find disease risks in an early stage, take prevention and treatment measures in time and reduce the influence of diseases on the growth of the aquatic animals.

Description

Technical Field

[0001] The invention relates to the technical field of aquaculture, in particular to an ecological aquaculture system. Background Art

[0002] With the continuous development of aquaculture, traditional aquaculture methods are facing many challenges. On the one hand, the feeding of feed in traditional aquaculture is often based on experience, which is difficult to achieve precise control, easily leading to feed waste and water quality deterioration. Overfeeding not only increases the cost of breeding, but may also cause eutrophication of water bodies and affect the living environment of aquatic animals. On the other hand, the difficulty of disease prevention and control is also a prominent problem in traditional breeding. Due to the lack of effective monitoring methods and early warning mechanisms, diseases are often already serious when they are discovered, causing huge economic losses to farmers.

[0003] At the same time, traditional aquaculture's monitoring and management of water quality are relatively extensive. It is unable to grasp the changes in water quality in a timely manner and cannot take effective measures to maintain the stability of water quality. Moreover, traditional aquaculture methods are inefficient in resource utilization and it is difficult to achieve sustainable development. Under the pressure of environmental protection and resource shortages, the transformation and upgrading of traditional aquaculture methods is imminent.

[0004] In order to solve these problems, people began to explore ecological aquaculture systems. By introducing advanced technical means, such as sensor technology, data analysis technology and machine learning algorithms, they can achieve precise management of the breeding process, improve breeding efficiency, reduce breeding costs, and reduce pollution to the environment, thereby achieving sustainable development of aquaculture. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an ecological aquaculture system, which solves the problems of low feed utilization and great difficulty in disease prevention and control in the existing ecological aquaculture system.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an ecological aquaculture system, comprising:

[0007] The breeding pond, as the basic component of the whole device, is used to carry and assemble the central data processing system and its subordinate structural parts;

[0008] A central data processing system for receiving, processing and storing data from each monitoring module;

[0009] Multi-parameter water quality monitoring module, real-time monitoring of key water quality indicators including dissolved oxygen, temperature, pH value, ammonia nitrogen content and nitrite content;

[0010] Aquatic animal growth monitoring module, which measures the growth parameters of aquatic animals including body length and weight through image recognition technology;

[0011] Intelligent feeding module, which adjusts feeding according to the optimal feed amount calculated by the deep learning model;

[0012] The disease analysis module uses an integrated learning model to provide early warning and prevention of aquatic animal diseases.

[0013] Preferably, the multi-parameter water quality monitoring module includes:

[0014] The water quality index collection unit uses multi-parameter water quality sensors to monitor key water quality indicators including dissolved oxygen, temperature, pH value, ammonia nitrogen content, and nitrite content in real time;

[0015] The water quality feature extraction unit calculates the changing trends of water quality parameters in different time periods, including the correlation between different water quality parameters including the 24-hour change rate of dissolved oxygen and the 7-day standard deviation of pH, and extracts relevant features;

[0016] The change capture unit uses the long short-term memory network LSTM in deep learning to capture the dynamic changes of water quality parameter data.

[0017] Preferably, the aquatic organism growth monitoring module comprises:

[0018] The growth monitoring unit uses underwater high-definition cameras and image analysis software to measure growth parameters of aquatic animals, including body length and weight, and combines behavioral analysis algorithms to monitor the activity status and feeding behavior of aquatic animals;

[0019] The growth feature extraction unit calculates features including growth rate and weight growth rate based on the body length and weight data of aquatic animals, and extracts shape features by analyzing the changes in the body shape of aquatic animals.

[0020] Preferably, the intelligent feeding module comprises:

[0021] The remaining amount monitoring unit is equipped with a weight sensor on the feeding device to record the amount of feed put in each time and the amount of remaining feed;

[0022] Feeding feature extraction unit: Analyze the feeding time, feeding frequency and feeding amount of aquatic animals from the images recorded by the camera, calculate the relationship characteristics between feeding frequency and growth rate, and the ratio characteristics of the remaining feed amount to the input feed amount;

[0023] Model building unit, the LSTM layer is used to process time series data to capture the dynamic changes of water quality parameters and aquatic animal growth data, and the MLP layer is used to nonlinearly combine the output of LSTM and other features to predict the optimal feed feeding amount;

[0024] The feeding decision unit makes feeding decisions based on the multi-parameter water quality monitoring module and the aquatic organism growth monitoring module.

[0025] Preferably, the steps of the feeding decision unit include:

[0026] When feeding decisions need to be made, factors including current water quality parameters, aquatic animal growth data, and feeding behavior data are input into the trained model to calculate the optimal feed amount Ft in real time;

[0027] The specific formula is as follows:

[0028] Ft=f(Wt,Gt,Ht,Tt,St)

[0029] Among them, Ft represents the optimal feeding amount at time t, Wt is the water quality parameter vector at time t, including dissolved oxygen, temperature, pH value, ammonia nitrogen content, and nitrite content, Tt represents the temperature at time t, Gt represents the growth data vector of aquatic animals at time t, including body weight, body length, and growth rate, Ht represents the historical feeding data vector, including the feeding amount and feeding time in the past period of time, St represents the type and breeding density of aquatic animals, and f represents the complex functional relationship established by the deep learning model. The nonlinear mapping of input features is realized through the combination of LSTM and MLP to predict the optimal feeding amount;

[0030] The feeding amount and feeding time are automatically adjusted according to the calculated optimal feeding amount. If the optimal feeding amount decreases, the feeding device will correspondingly reduce the feeding amount each time and extend the feeding time interval; if the optimal feeding amount increases, the feeding device will increase the feeding amount. At the same time, water quality parameters and feeding behavior of aquatic animals are closely monitored to ensure that they are not overfed.

[0031] Preferably, the feeding decision unit also includes an optimization unit, which uses a stochastic gradient descent optimization algorithm to train the model, defines a loss function as the mean square error between the predicted feeding amount and the actual feeding amount, and adjusts the parameters of the model through multiple iterative training to minimize the loss function.

[0032] Preferably, the disease analysis module includes:

[0033] Physiological index sensing unit, real-time monitoring of physiological indicators of aquatic animals including body surface color, activity status and respiratory rate;

[0034] Microbial community monitoring unit, which analyzes the microbial community structure in water, including bacteria, fungi, and viruses, through metagenomic sequencing;

[0035] The environmental feature extraction unit collects environmental factors including light intensity, water flow speed and water level changes through sensors.

[0036] Preferably, the steps of the environmental feature extraction unit include:

[0037] Extract the color change rate of the computer table, the relative abundance changes of different microbial species, and the characteristics of light intensity changes;

[0038] Based on these feature vectors Xt, the integrated learning module is used to combine the feature vectors Yt of historical disease data and the factor vectors Zt of aquatic animal species and breeding density;

[0039] The probability of disease occurrence Pd is calculated by combining random forest and gradient boosting tree. The specific formula is as follows:

[0040] Pd=g(Xt,Yt,Zt)

[0041] Among them, Xt represents the feature vector of multi-source data at time t, including physiological indicator data, microbial community data, and environmental factor data; Yt represents the feature vector of historical disease data at time t, including the time, frequency, and severity of past diseases; Zt represents the type and breeding density of aquatic animals; g represents the complex functional relationship established by the integrated learning model, which realizes nonlinear mapping of input features through the combination of random forest and gradient boosting tree to predict the probability of disease occurrence.

[0042] Preferably, the disease analysis module further includes an early warning unit, and when the probability of disease occurrence exceeds 50%, the system issues a disease early warning signal.

[0043] Preferably, the steps of the early warning unit include:

[0044] When new data is input, the multi-source data features are input into the trained disease warning model to calculate the probability of disease occurrence. When the probability of disease occurrence exceeds 50%, the system issues a disease warning signal.

[0045] Analyze possible causes based on the model's prediction results and the input multi-source data characteristics;

[0046] Based on the results of the cause analysis, the central data processing system makes decisions on prevention and control measures, including adjusting water quality parameters and adding water quality regulators.

[0047] Working principle: The sensors in the multi-parameter water quality monitoring device, aquatic animal growth monitoring device, and disease early warning device continuously collect data and transmit the data to the data receiving module of the central data processing system. After the data receiving module converts and verifies the data format, it sends the data to the data processing module. The data processing module cleans, preprocesses, and extracts features from the data, and then inputs the extracted features into the deep learning model and the integrated learning model for calculation. The deep learning model calculates the optimal feed amount, and the intelligent feeding device adjusts the feeding amount according to the amount. The integrated learning model calculates the probability of disease occurrence. When the probability exceeds a certain threshold, the disease early warning device sends a warning signal, analyzes the cause of the disease and recommends Prevention and control measures. Throughout the entire workflow, each device collaborates with each other through data transmission and algorithm calculation to achieve precise management of the aquaculture process. For example, the multi-parameter water quality monitoring device provides water quality parameter data, which is not only used to calculate the optimal feed dosage, but also serves as an important basis for disease warning. The aquatic animal growth monitoring device provides growth data and behavioral characteristics of aquatic animals, which can help adjust feed dosage and judge the health status of aquatic animals. The disease warning device detects disease risks in advance through the fusion analysis of multi-source data, providing support for timely prevention and control measures. The central data processing system is responsible for coordinating data transmission and algorithm calculation between devices to ensure the efficient operation of the entire system.

[0048] The present invention provides an ecological aquaculture system, which has the following beneficial effects:

[0049] 1. The present invention can grasp the breeding environment and the status of aquatic animals in real time through the coordinated work of a multi-parameter water quality monitoring device, an aquatic animal growth monitoring device, an intelligent feeding device and a disease early warning device, so as to carry out precise management. The intelligent feeding device adjusts the feeding amount according to the optimal feed feeding amount calculated by the deep learning model, thereby avoiding feed waste and meeting the nutritional needs of aquatic animals and promoting their healthy growth. The disease early warning device can detect disease risks at an early stage, take preventive measures in time, and reduce the impact of diseases on the growth of aquatic animals.

[0050] 2. The present invention can timely detect water quality problems through the multi-parameter water quality monitoring device, take corresponding measures to adjust, and avoid aquatic animal diseases and deaths caused by water quality deterioration. The precise feeding of the intelligent feeding device avoids feed waste and reduces feed costs. The early warning and prevention measures of the disease warning device can reduce the incidence of diseases, reduce the use of drugs, and reduce breeding costs.

[0051] 3. The present invention reduces environmental pollution through real-time monitoring and precise management of the breeding environment. The precise feeding of the intelligent feeding device and the early prevention and control measures of the disease early warning device reduce the incidence of diseases, reduce the use of drugs, and avoid drug pollution to the environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a schematic diagram of the aquaculture system of the present invention;

[0053] Figure 2 This is a schematic diagram of the multi-parameter water quality monitoring module of the present invention;

[0054] Figure 3 This is a schematic diagram of the aquatic animal growth monitoring module of the present invention;

[0055] Figure 4 This is a schematic diagram of the intelligent feeding module of the present invention;

[0056] Figure 5 This is the architecture diagram of the disease analysis module in the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not 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.

[0058] Example:

[0059] Please refer to the attached Figure 1 - Attachment Figure 5 The embodiment of the present invention provides an ecological aquaculture system, comprising:

[0060] The breeding pond, as the basic component of the whole device, is used to carry and assemble the central data processing system and its subordinate structural parts;

[0061] A central data processing system for receiving, processing and storing data from each monitoring module;

[0062] Multi-parameter water quality monitoring module, real-time monitoring of key water quality indicators including dissolved oxygen, temperature, pH value, ammonia nitrogen content and nitrite content;

[0063] Aquatic animal growth monitoring module, which measures the growth parameters of aquatic animals including body length and weight through image recognition technology;

[0064] Intelligent feeding module, which adjusts feeding according to the optimal feed amount calculated by the deep learning model;

[0065] The disease analysis module uses an integrated learning model to provide early warning and prevention of aquatic animal diseases.

[0066] The multi-parameter water quality monitoring module includes:

[0067] The water quality index collection unit uses multi-parameter water quality sensors to monitor key water quality indicators including dissolved oxygen, temperature, pH value, ammonia nitrogen content, and nitrite content in real time;

[0068] The water quality feature extraction unit calculates the changing trends of water quality parameters in different time periods, including the correlation between different water quality parameters including the 24-hour change rate of dissolved oxygen and the 7-day standard deviation of pH, and extracts relevant features;

[0069] The change capture unit uses the long short-term memory network LSTM in deep learning to capture the dynamic changes of water quality parameter data.

[0070] Specifically, multi-parameter water quality sensors including dissolved oxygen sensors, temperature sensors, pH sensors, ammonia nitrogen content sensors and nitrite content sensors are installed in the breeding pond to monitor key water quality indicators including dissolved oxygen, temperature, pH value, ammonia nitrogen content and nitrite content in real time. These sensors collect data every 15 minutes and send the data to the central data processing system through wireless transmission technology. The dissolved oxygen sensor uses electrochemical principles to determine the dissolved oxygen content by measuring the potential change of the electrode in the water. The temperature sensor usually uses thermistor or thermocouple technology to measure the water temperature according to the change in resistance or thermoelectric potential caused by temperature change. The pH sensor determines the pH of the water by measuring the hydrogen ion concentration, generally using glass electrodes or composite electrodes. The ammonia nitrogen content sensor and the nitrite content sensor use chemical analysis. Methods, such as spectrophotometry or ion-selective electrode method, are used to detect the concentrations of ammonia nitrogen and nitrite in water. These sensors can monitor the key parameters of water quality in real time, provide an important basis for aquaculture management, and calculate the changing trends of water quality parameters in different time periods, including the average change rate of dissolved oxygen in the past 24 hours and the standard deviation of pH value in a week. At the same time, the correlation between different water quality parameters is analyzed, and relevant features are extracted to timely grasp the water quality status, so as to take appropriate measures to adjust the water quality. For example, if the dissolved oxygen content is too low, the central data processing system can automatically control the oxygenation equipment to increase the dissolved oxygen content in the water; if the pH value deviates from the appropriate range, the system can recommend the introduction of appropriate regulators to adjust the water quality. This can ensure that aquatic animals live in a suitable environment, improve their growth rate and immunity, and reduce the incidence of diseases.

[0071] The aquatic biological growth monitoring module includes:

[0072] Growth monitoring unit, which uses underwater high-definition cameras and image analysis software to measure growth parameters of aquatic animals, combined with behavioral analysis algorithms to monitor the activity status and feeding behavior of aquatic animals;

[0073] The growth feature extraction unit calculates features including growth rate and weight growth rate based on the body length and weight data of aquatic animals, and extracts shape features by analyzing the changes in the body shape of aquatic animals.

[0074] Specifically, underwater high-definition cameras and image analysis software are used to regularly take images of aquatic animals once a day, and computer vision technology is used to analyze and process the images of aquatic animals. The edge detection algorithm can identify the outline of aquatic animals, and the feature extraction algorithm can extract the shape and color characteristics of aquatic animals, thereby realizing the measurement of their body length and weight growth parameters. The behavior analysis algorithm analyzes the changes in continuous images, calculates the movement speed and activity range indicators of aquatic animals, and determines their activity status. According to the behavioral characteristics of aquatic animals in the feeding area, their feeding situation can be understood, providing a basis for adjusting the feed feeding amount. The growth parameters of aquatic animals including body length and weight are measured by image recognition technology. At the same time, combined with the behavior analysis algorithm, the activity status and feeding behavior of aquatic animals are monitored, and the growth rate and weight growth rate characteristics are calculated according to the body length and weight data of aquatic animals. In addition, the body shape changes of aquatic animals can be analyzed and shape characteristics can be extracted.

[0075] The intelligent feeding module comprises:

[0076] The remaining amount monitoring unit is equipped with a weight sensor on the feeding device to record the amount of feed put in each time and the amount of remaining feed;

[0077] Feeding feature extraction unit: Analyze the feeding time, feeding frequency and feeding amount of aquatic animals from the images recorded by the camera, calculate the relationship characteristics between feeding frequency and growth rate, and the ratio characteristics of the remaining feed amount to the input feed amount;

[0078] Model building unit, the LSTM layer is used to process time series data to capture the dynamic changes of water quality parameters and aquatic animal growth data, and the MLP layer is used to nonlinearly combine the output of LSTM and other features to predict the optimal feed feeding amount;

[0079] The feeding decision unit makes feeding decisions based on the multi-parameter water quality monitoring module and the aquatic organism growth monitoring module.

[0080] The steps of the feeding decision unit include:

[0081] When feeding decisions need to be made, factors including current water quality parameters, aquatic animal growth data, and feeding behavior data are input into the trained model to calculate the optimal feed amount Ft in real time;

[0082] The specific formula is as follows:

[0083] Ft=f(Wt,Gt,Ht,Tt,St)

[0084] Among them, Ft represents the optimal feeding amount at time t, Wt is the water quality parameter vector at time t, including dissolved oxygen, temperature, pH value, ammonia nitrogen content, and nitrite content, Tt represents the temperature at time t, Gt represents the growth data vector of aquatic animals at time t, including body weight, body length, and growth rate, Ht represents the historical feeding data vector, including the feeding amount and feeding time in the past period of time, St represents the type and breeding density of aquatic animals, and f represents the complex functional relationship established by the deep learning model. The nonlinear mapping of input features is realized through the combination of LSTM and MLP to predict the optimal feeding amount;

[0085] The feeding amount and feeding time are automatically adjusted according to the calculated optimal feeding amount. If the optimal feeding amount decreases, the feeding device will correspondingly reduce the feeding amount each time and extend the feeding time interval; if the optimal feeding amount increases, the feeding device will increase the feeding amount. At the same time, water quality parameters and feeding behavior of aquatic animals are closely monitored to ensure that they are not overfed.

[0086] Specifically, a weight sensor is installed on the feeding device to record the amount of feed put in each time and the amount of remaining feed. From the images recorded by the camera, the feeding time, feeding frequency and feeding amount of aquatic animals are analyzed. The relationship characteristics between feeding frequency and growth rate, as well as the ratio characteristics of the remaining feed amount to the input feed amount are calculated, and the extracted feature data and the corresponding actual feeding amount data are organized into a training data set. When calculating the optimal feed feeding amount Ft, the deep learning model first analyzes the input water quality parameter vector Wt, calculates the average change rate of dissolved oxygen in the past period of time, and the standard deviation statistical characteristics of pH value; then processes the aquatic animal growth data vector Gt, such as calculating the growth rate and weight growth rate characteristics; then combines the historical feeding data vector Ht, the external factor vector Tt; and the aquatic animal species and breeding density factor vector St, and captures the dynamic changes of time series data through the long short-term memory network LSTM, and then uses the multi-layer perceptron MLP for nonlinear combination, and finally obtains the optimal feeding amount Ft.

[0087] The feeding decision unit also includes an optimization unit, which uses a stochastic gradient descent optimization algorithm to train the model, defines a loss function as the mean square error between the predicted feeding amount and the actual feeding amount, and adjusts the parameters of the model through multiple iterative training to minimize the loss function.

[0088] The disease analysis module comprises:

[0089] Physiological index sensing unit, real-time monitoring of physiological indicators of aquatic animals including body surface color, activity status and respiratory rate;

[0090] Microbial community monitoring unit, which analyzes the microbial community structure in water, including bacteria, fungi, and viruses, through metagenomic sequencing;

[0091] The environmental feature extraction unit collects environmental factors including light intensity, water flow speed and water level changes through sensors.

[0092] The steps of the environmental feature extraction unit include:

[0093] Extract the color change rate of the computer table, the relative abundance changes of different microbial species, and the characteristics of light intensity changes;

[0094] Based on these feature vectors Xt, the integrated learning module is used to combine the feature vectors Yt of historical disease data and the factor vectors Zt of aquatic animal species and breeding density;

[0095] The probability of disease occurrence Pd is calculated by combining random forest and gradient boosting tree. The specific formula is as follows:

[0096] Pd=g(Xt,Yt,Zt)

[0097] Among them, Xt represents the feature vector of multi-source data at time t, including physiological indicator data, microbial community data, and environmental factor data; Yt represents the feature vector of historical disease data at time t, including the time, frequency, and severity of past diseases; Zt represents the type and breeding density of aquatic animals; g represents the complex functional relationship established by the integrated learning model, which realizes nonlinear mapping of input features through the combination of random forest and gradient boosting tree to predict the probability of disease occurrence.

[0098] Specifically, in the multi-source data fusion stage: physiological index monitoring, using biosensors to monitor the surface color, activity status, and respiratory rate of aquatic animals. These sensors can collect data in real time and send it to the central data processing system through wireless transmission technology. Microbial community monitoring: using microbial detection technology, water samples are collected from the breeding pond once a week, and the structure of the microbial community in the water is analyzed through metagenomic sequencing methods, including the types and quantities of bacteria, fungi, and viruses. Environmental factor monitoring: installing environmental sensors to monitor environmental factors such as light intensity, water flow rate, and water level changes. These sensors collect data every 30 seconds and send the data to the central data processing system.

[0099] Feature engineering stage: physiological indicator characteristics, calculate the rate of change of body surface color, calculate the degree of color change by comparing the current body surface color with the average body surface color in the past period of time, analyze the abnormality of the activity state, and determine whether there is abnormal behavior by monitoring the swimming speed and activity range parameters of aquatic animals. Calculate changes in respiratory rate, such as fluctuations in respiratory rate in the past 24 hours, microbial community characteristics, calculate the relative abundance changes of different microbial species, the increase or decrease ratio of the relative abundance of a certain harmful bacteria within a week, and count the number of harmful microorganisms, such as the concentration of virus particles, environmental factor characteristics, and take light intensity change characteristics, such as the maximum, minimum and average change rate of light intensity in a day. Analyze changes in water flow velocity and calculate the standard deviation and change trend of water flow velocity. Calculate the water level change rate, that is, the rate at which the water level rises or falls within a certain period of time;

[0100] Model building phase: data preparation, organizing the extracted multi-source data features and the corresponding historical disease data into a training data set, encoding the data, and converting categorical variables into numerical variables, for example, encoding disease types into different numbers and encoding seasons into different seasonal indices;

[0101] The model architecture adopts a model that combines random forest and gradient boosting tree in ensemble learning. The random forest consists of multiple decision trees. By voting on the prediction results of multiple decision trees, the accuracy and stability of the prediction are improved. The gradient boosting tree gradually builds multiple weak learners by continuously optimizing the loss function and combines them into a strong learner. In the process of architecture, a large amount of data related to aquaculture is first collected and sorted, including water quality parameters, aquatic animal growth data, feeding behavior data, disease occurrence records and other relevant environmental factors, and meaningful features are extracted from the original data, such as calculating the change rate of water quality parameters, statistically analyzing the mean and variance of aquatic animal growth data, and constructing characteristic indicators reflecting feeding behavior. When constructing each decision tree, all features are extracted. A part of the features is randomly selected as the split features. For each node split, a part of the samples is randomly selected from the training data as the basis for splitting. A decision tree is constructed recursively until the stopping condition is met (the number of node samples is less than the threshold). The above steps are repeated to construct multiple decision trees to form a random forest. The prediction result of the random forest is obtained by voting on the prediction results of all decision trees. Assuming that there are N decision trees in the random forest, for a sample prediction category C, the number nc predicted by each decision tree as category C is calculated, and the final prediction category is the category with the largest nc. For example, assuming that the prediction result of the random forest is R, the prediction result of the gradient boosting tree is G, and the weights are wR and wG (wR+wG=1), then the final fusion prediction result F is:

[0102] F=wR×R+wG×G

[0103] Model training: Use the cross-validation method to divide the training data set into multiple subsets for training and validating the model. Improve the performance of the model by adjusting the model's hyperparameters, such as the number of trees in the random forest and the learning rate in the gradient boosting tree. Use the binary cross entropy loss function, which is defined as the difference between the predicted probability of disease occurrence and the actual disease occurrence.

[0104] The central data processing system integrates these multi-source data and extracts features related to disease warning, such as calculating the rate of change of body surface color, changes in the relative abundance of different microbial species, and changes in light intensity. Then, the integrated learning model calculates the probability of disease occurrence Pd based on these feature vectors Xt, combined with the feature vectors Yt of historical disease data and the factor vectors Zt of aquatic animal species and breeding density, by combining random forests and gradient boosting trees.

[0105] The disease analysis module also includes an early warning unit. When the probability of disease occurrence exceeds 50%, the system sends out a disease early warning signal.

[0106] The steps of the early warning unit include:

[0107] When new data is input, the multi-source data features are input into the trained disease warning model to calculate the probability of disease occurrence. When the probability of disease occurrence exceeds 50%, the system issues a disease warning signal.

[0108] Analyze possible causes based on the model's prediction results and the input multi-source data characteristics;

[0109] Based on the results of the cause analysis, the central data processing system makes decisions on prevention and control measures, including adjusting water quality parameters and adding water quality regulators.

[0110] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An ecological aquaculture system, characterized in that: include: The breeding pond, as the basic component of the whole device, is used to carry and assemble the central data processing system and its subordinate structural parts; A central data processing system for receiving, processing and storing data from each monitoring module; Multi-parameter water quality monitoring module, real-time monitoring of key water quality indicators including dissolved oxygen, temperature, pH value, ammonia nitrogen content and nitrite content; Aquatic animal growth monitoring module, which measures the growth parameters of aquatic animals including body length and weight through image recognition technology; Intelligent feeding module, which adjusts feeding according to the optimal feed amount calculated by the deep learning model; The disease analysis module uses an integrated learning model to provide early warning and prevention of aquatic animal diseases.

2. The ecological aquaculture system according to claim 1, characterized in that: The multi-parameter water quality monitoring module includes: The water quality index collection unit uses multi-parameter water quality sensors to monitor key water quality indicators including dissolved oxygen, temperature, pH value, ammonia nitrogen content, and nitrite content in real time; The water quality feature extraction unit calculates the changing trends of water quality parameters in different time periods, including the correlation between different water quality parameters including the 24-hour change rate of dissolved oxygen and the 7-day standard deviation of pH, and extracts relevant features; The change capture unit uses the long short-term memory network LSTM in deep learning to capture the dynamic changes of water quality parameter data.

3. The ecological aquaculture system according to claim 1, characterized in that: The aquatic biological growth monitoring module includes: Growth monitoring unit, which uses underwater high-definition cameras and image analysis software to measure growth parameters of aquatic animals, combined with behavioral analysis algorithms to monitor the activity status and feeding behavior of aquatic animals; The growth feature extraction unit calculates features including growth rate and weight growth rate based on the body length and weight data of aquatic animals, and extracts shape features by analyzing the changes in the body shape of aquatic animals.

4. The ecological aquaculture system according to claim 1, characterized in that: The intelligent feeding module comprises: The remaining amount monitoring unit is equipped with a weight sensor on the feeding device to record the amount of feed put in each time and the amount of remaining feed; Feeding feature extraction unit: Analyze the feeding time, feeding frequency and feeding amount of aquatic animals from the images recorded by the camera, calculate the relationship characteristics between feeding frequency and growth rate, and the ratio characteristics of the remaining feed amount to the input feed amount; Model building unit, the LSTM layer is used to process time series data to capture the dynamic changes of water quality parameters and aquatic animal growth data, and the MLP layer is used to nonlinearly combine the output of LSTM and other features to predict the optimal feed feeding amount; The feeding decision unit makes feeding decisions based on the multi-parameter water quality monitoring module and the aquatic organism growth monitoring module.

5. The ecological aquaculture system according to claim 4, characterized in that: The steps of the feeding decision unit include: When feeding decisions need to be made, factors including current water quality parameters, aquatic animal growth data, and feeding behavior data are input into the trained model to calculate the optimal feed amount Ft in real time; The specific formula is as follows: Ft=f(Wt,Gt,Ht,Tt,St) Among them, Ft represents the optimal feeding amount at time t, Wt is the water quality parameter vector at time t, including dissolved oxygen, temperature, pH value, ammonia nitrogen content, and nitrite content, Tt represents the temperature at time t, Gt represents the growth data vector of aquatic animals at time t, including body weight, body length, and growth rate, Ht represents the historical feeding data vector, including the feeding amount and feeding time in the past period of time, St represents the type and breeding density of aquatic animals, and f represents the complex functional relationship established by the deep learning model. The nonlinear mapping of input features is realized through the combination of LSTM and MLP to predict the optimal feeding amount; The feeding amount and feeding time are automatically adjusted according to the calculated optimal feeding amount. If the optimal feeding amount decreases, the feeding device will correspondingly reduce the feeding amount each time and extend the feeding time interval; if the optimal feeding amount increases, the feeding device will increase the feeding amount. At the same time, water quality parameters and feeding behavior of aquatic animals are closely monitored to ensure that they are not overfed.

6. The ecological aquaculture system according to claim 1, characterized in that: The feeding decision unit also includes an optimization unit, which uses a stochastic gradient descent optimization algorithm to train the model, defines a loss function as the mean square error between the predicted feeding amount and the actual feeding amount, and adjusts the parameters of the model through multiple iterative training to minimize the loss function.

7. The ecological aquaculture system according to claim 1, characterized in that: The disease analysis module includes: Physiological index sensing unit, real-time monitoring of physiological indicators of aquatic animals including body surface color, activity status and respiratory rate; Microbial community monitoring unit, which analyzes the microbial community structure in water, including bacteria, fungi, and viruses, through metagenomic sequencing; The environmental feature extraction unit collects environmental factors including light intensity, water flow speed and water level changes through sensors.

8. The ecological aquaculture system according to claim 7, characterized in that: The steps of the environmental feature extraction unit include: Extract the color change rate of the computer table, the relative abundance changes of different microbial species, and the characteristics of light intensity changes; Based on these feature vectors Xt, the integrated learning module is used to combine the feature vectors Yt of historical disease data and the factor vectors Zt of aquatic animal species and breeding density; The probability of disease occurrence Pd is calculated by combining random forest and gradient boosting tree. The specific formula is as follows: Pd=g(Xt,Yt,Zt) Among them, Xt represents the feature vector of multi-source data at time t, including physiological indicator data, microbial community data, and environmental factor data; Yt represents the feature vector of historical disease data at time t, including the time, frequency, and severity of past diseases; Zt represents the type and breeding density of aquatic animals; g represents the complex functional relationship established by the integrated learning model, which realizes nonlinear mapping of input features through the combination of random forest and gradient boosting tree to predict the probability of disease occurrence.

9. The ecological aquaculture system according to claim 7, characterized in that: The disease analysis module also includes an early warning unit. When the probability of disease occurrence exceeds 50%, the system sends out a disease early warning signal.

10. The ecological aquaculture system according to claim 9, characterized in that: The steps of the early warning unit include: When new data is input, the multi-source data features are input into the trained disease warning model to calculate the probability of disease occurrence. When the probability of disease occurrence exceeds 50%, the system issues a disease warning signal. Analyze possible causes based on the model's prediction results and the input multi-source data characteristics; Based on the results of the cause analysis, the central data processing system makes decisions on prevention and control measures, including adjusting water quality parameters and adding water quality regulators.

Images