Aquaculture intelligent collaborative management system integrating deep learning and Internet of Things

Abstract

The invention discloses an aquaculture intelligent collaborative management system integrating deep learning and the Internet of Things, and particularly relates to the technical field of intelligent agricultural aquaculture management. The system is formed by sequentially connecting an Internet of Things data acquisition unit, a deep learning processing unit and an intelligent collaborative management unit. The Internet of Things data acquisition unit acquires multi-dimensional environmental parameters such as water temperature, dissolved oxygen, pH value and ammonia nitrogen concentration in real time through an aquaculture water area sensor network; the deep learning processing unit adopts a pre-trained long-short-term memory network model to dynamically analyze time series data, and environment trend prediction and abnormal state recognition are achieved; the intelligent collaborative management unit generates a collaborative control instruction according to the analysis result in combination with a rule base and an optimization algorithm, and drives an aerator, a bait casting machine, a water pump and other execution devices to operate in a linkage mode. According to the system, intelligent sensing, accurate prediction and cooperative control of the aquaculture environment are realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent agricultural aquaculture management technology, and more specifically, to an intelligent collaborative management system for aquaculture that integrates deep learning and the Internet of Things. Background Technology

[0002] Aquaculture is one of my country's important agricultural industries, and its production management methods are gradually shifting from traditional experience-based to modern intelligent models. Early aquaculture processes relied primarily on manual observation and experience-based judgment, resulting in low efficiency and high risk. With the development of information technology, the Internet of Things (IoT) has been introduced into the field of aquaculture environmental monitoring, enabling the automated collection of key parameters such as water temperature and dissolved oxygen. In recent years, artificial intelligence (AI) technology, especially deep learning, has made significant progress in pattern recognition and predictive analysis, providing a new technological path for the intelligent upgrading of aquaculture management. Currently, integrating IoT and deep learning technologies to build intelligent management systems with predictive warning and optimization decision-making capabilities has become an important development direction in the aquaculture industry.

[0003] However, existing aquaculture monitoring and management systems still have significant limitations. Most systems only have basic data acquisition and display functions, relying on fixed thresholds for out-of-limit alarms, making it difficult to cope with the complex characteristics of multi-parameter coupling and dynamic changes in the aquaculture environment. Some systems that use traditional data analysis methods lack the ability to process nonlinear and time-series data, resulting in limited prediction accuracy and adaptability. Furthermore, the data perception, analysis, decision-making, and control execution stages of existing systems are often disconnected, lacking effective coordination mechanisms, leading to simplistic and lagging management strategies and making it difficult to achieve precise and efficient closed-loop control.

[0004] Therefore, this paper proposes an intelligent collaborative management system for aquaculture that integrates deep learning and the Internet of Things to address the aforementioned problems. The main technical challenges to be solved are: how to overcome the technical bottlenecks of existing systems, such as lagging environmental condition assessment, insufficient predictive capabilities, and poor coordination in management processes, thereby improving the accuracy, foresight, and overall efficiency of aquaculture management. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an intelligent collaborative management system for aquaculture that integrates deep learning and the Internet of Things, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent collaborative management system for aquaculture that integrates deep learning and the Internet of Things, the system comprising an Internet of Things data acquisition unit, a deep learning processing unit, and an intelligent collaborative management unit; The IoT data acquisition unit is configured to collect environmental parameter data related to aquaculture in real time through multiple IoT sensor nodes deployed in the aquaculture environment, and transmit the environmental parameter data to the deep learning processing unit. The IoT sensor nodes include, but are not limited to, temperature sensors, pH sensors, dissolved oxygen sensors, ammonia nitrogen sensors and turbidity sensors. These sensors form a network in a wireless or wired manner, and data aggregation and protocol conversion are performed through an IoT gateway. The deep learning processing unit is configured to receive the environmental parameter data and process and analyze the environmental parameter data using at least one pre-trained deep learning model to generate predictive information or anomaly detection results about the aquaculture status. The deep learning model includes, but is not limited to, recurrent neural networks or long short-term memory networks based on time series analysis, used to model the changing trends of environmental parameters. The intelligent collaborative management unit is configured to generate corresponding management operation instructions based on the predicted information or anomaly detection results, and send the management operation instructions to the execution equipment at the aquaculture site through a communication network to realize intelligent collaborative management of the aquaculture process. The execution equipment includes, but is not limited to, aerators, feeders, water pumps and alarms.

[0007] Preferably, the IoT data acquisition unit further includes a data preprocessing subunit, which is used to clean, normalize, and extract features from the acquired raw environmental parameter data. The cleaning operation includes removing noisy data and filling in missing values, the normalization operation scales the data to a uniform range, and the feature extraction operation derives feature vectors from the raw data for use as input to the deep learning model, so as to generate a standardized and enhanced dataset for use by the deep learning processing unit.

[0008] Preferably, the deep learning model in the deep learning processing unit is trained using historical aquaculture data and is equipped with a model adaptive update mechanism. This mechanism periodically adjusts the model parameters based on newly collected environmental parameter data to adapt to dynamic changes in the aquaculture environment. The model adaptive update includes, but is not limited to, online learning or incremental learning methods to ensure that the prediction accuracy is maintained over time.

[0009] Preferably, the intelligent collaborative management unit further includes a decision support subunit, which integrates a rule base and optimization algorithms to evaluate various possible management strategies based on prediction information or anomaly detection results, and select the optimal strategy through multi-objective optimization to generate management operation instructions. The rule base contains logical rules based on the knowledge of aquaculture experts, and the optimization algorithms include, but are not limited to, genetic algorithms or particle swarm optimization, to balance cost, efficiency and risk factors.

[0010] Preferably, the system further includes a user interaction interface module, which provides a graphical interface and remote access function for real-time display of aquaculture environment data, prediction results of deep learning processing unit and management instructions generated by intelligent collaborative management unit, and allows users to manually modify system parameters or override automatic decision-making through input devices to achieve human-machine collaborative operation.

[0011] Preferably, the system is deployed in a distributed manner through a cloud computing platform, wherein some or all of the functions of the IoT data acquisition unit, deep learning processing unit and intelligent collaborative management unit run on the cloud server in the form of software services. The cloud platform provides data storage, computing resources and network communication services, and supports simultaneous access by multiple users and system expansion.

[0012] Preferably, the environmental parameter data includes, but is not limited to, various data such as water temperature, pH, dissolved oxygen concentration, turbidity, ammonia nitrogen concentration, nitrite concentration, and microbial indicators. These data are collected continuously in time series form, and the sampling frequency can be adjusted from one minute to one hour according to the aquaculture needs.

[0013] Preferably, the management operation instructions include commands to control the start and stop and speed of the aeration equipment, adjust the feeding amount and time of the feeder, control the flow rate and cycle of the water pump, or trigger the audible and visual alarm device. These instructions are generated based on preset thresholds or dynamic model outputs and are sent to the execution device via wireless communication protocols.

[0014] The technical effects and advantages of this invention are as follows: Compared to existing technologies, this invention achieves a shift from passive response to proactive intervention in aquaculture environment monitoring by constructing a seamless integration mechanism between IoT data acquisition and deep learning analysis. The system continuously collects multi-dimensional environmental parameters using a sensor network deployed at the aquaculture site and inputs the real-time data stream into a prediction model based on a long short-term memory network for time-series analysis. By mining the coupling relationships and changing patterns between parameters, it enables early warning of key risks such as abnormal dissolved oxygen and water quality deterioration. This approach overcomes the lag of traditional threshold alarms, allowing managers to take preventative measures in advance, effectively reducing aquaculture risks and improving the scientific rigor and foresight of management decisions.

[0015] Compared to the prediction accuracy decay problem caused by fixed models in existing technologies, this invention introduces an online update mechanism based on incremental learning and elastic weight consolidation algorithms, enabling the analysis model to continuously optimize. This mechanism periodically fine-tunes model parameters using newly collected aquaculture environment data and constrains the variation of key parameters during optimization, thus retaining existing experience while incorporating new knowledge. This process allows the model to dynamically adapt to environmental dynamics such as seasonal changes and variations in stocking density, maintaining long-term prediction accuracy and solving the technical problem of static models failing due to environmental drift, thereby improving the system's reliability and lifespan.

[0016] Compared to the limitations of independent control of single devices in existing technologies, this invention achieves collaborative intelligent regulation of aquaculture equipment by integrating a rule base and a decision support subunit based on multi-objective optimization algorithms. After receiving the prediction results from the deep learning module, the system, guided by the expert rule base, uses optimization algorithms to comprehensively evaluate multiple objectives such as equipment energy consumption, adjustment efficiency, and operating costs, generating a globally optimal sequence of control commands. This method can coordinate the timing and intensity of operations such as aeration, feeding, and water changes, avoiding equipment action conflicts and resource waste, and achieving a balance between energy conservation, consumption reduction, and refined management while stabilizing the aquaculture environment. Attached Figure Description

[0017] Figure 1 This is a system overall framework diagram of the present invention.

[0018] Figure 2 This is a system workflow diagram of the present invention.

[0019] Figure 3 This is a flowchart illustrating the adaptive update process of the model in this invention.

[0020] Figure 4 This is a flowchart of the intelligent decision generation process of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1: As attached Figures 1 to 4The invention presents an intelligent collaborative management system for aquaculture that integrates deep learning and the Internet of Things. The complete implementation process of this system is a closed-loop process from data perception of the physical world to intelligent decision-making and feedback control within the physical environment. Its implementation encompasses multiple stages, including hardware deployment, software configuration, data processing, model computation, and decision execution. The following will provide a detailed description of the complete implementation process, including startup preparation, daily operation, and continuous optimization.

[0023] Phase 1: System Initialization and Deployment The implementation of the system begins with the hardware deployment and software environment setup at the aquaculture site. In the aquaculture ponds or water areas, various IoT sensor nodes are scientifically deployed based on the aquaculture species and water area. These nodes include, but are not limited to, temperature sensors, pH sensors, dissolved oxygen sensors, ammonia nitrogen sensors, and turbidity sensors. Each sensor node has a built-in microcontroller and communication module (such as a LoRa or NB-IoT module), which self-organize into a network wirelessly. The IoT gateway is placed in the center of the aquaculture site or in an area with good signal coverage to collect data uploaded by all sensor nodes. The gateway needs to be configured with network parameters to establish a stable communication connection with a remote cloud platform or local server (e.g., via 4G / 5G or Ethernet). Simultaneously, actuators such as aerators, feeders, and water pumps need to be automated or equipped with intelligent models to ensure they can receive and execute control commands from the system. These actuators also need to have communication capabilities and communicate with the system server either through the gateway or directly.

[0024] On the server side (cloud platform or local server), the system's core software needs to be deployed. First, a database needs to be established to store historical aquaculture data, real-time collected data, model parameters, and operation logs. Then, the deep learning processing unit is initialized. This step is crucial, requiring the collection and organization of a large amount of historical aquaculture environmental parameter data and its corresponding aquaculture outcome records (such as fish growth status, disease occurrence records, etc.). This historical data undergoes rigorous preprocessing, including: Data cleaning: Remove obviously erroneous outliers (such as data that exceeds the physical reasonable range), and fill in missing values ​​caused by equipment failure or communication interruption using time series interpolation methods (such as linear interpolation or spline interpolation).

[0025] Data normalization: To eliminate the impact of differences in the dimensions of environmental parameters on model training, a minimum-maximum scaling method is used to uniformly transform the values ​​of each parameter to the interval [0, 1]. The formula is: X_norm = (X - X_min) / (X_max - X_min), where X_norm represents the normalized data value, X represents the original measurement value, and X_min and X_max represent the minimum and maximum values ​​of the parameter in the historical dataset, respectively.

[0026] Constructing training samples: The processed time series data is cut into sample segments of fixed length. Each sample segment contains a multidimensional environmental parameter sequence for a past period of time (e.g., the past 6 hours, with a sampling interval of 10 minutes, for a total of 36 time points). The actual environmental state or whether an anomaly occurs in the period following the sequence (e.g., the next 1 hour) is used as the label.

[0027] Using these prepared training samples, the deep learning model is trained offline. This invention preferably uses a Long Short-Term Memory (LSTM) network model. The calculation process for the LSTM model at each time step t involves the following formula: (The Forgotten Gate: determines which information to discard) (Input gate, determines which information to update) (Candidate cell status) (Update cell status) (Output gate, determines which information is output) (Current hidden state, including information at the current moment) Where t is the time step, It is the input vector of the current time step (i.e., the normalized environment parameters). It is the hidden state of the previous time step. and Here, σ and tanh are the weight matrix and bias vector for the corresponding gate, respectively; σ is the sigmoid activation function; and tanh is the hyperbolic tangent activation function. The symbol indicates vector concatenation, and * indicates element-wise multiplication. It represents the cellular state, carrying long-term memory. The model learns the complex nonlinear relationships and temporal dynamics between environmental parameters through backpropagation and time-series training. The training objective is to improve the model's predicted output. (or by) The derived predicted values ​​have the smallest error compared to the true labels. After training, a set of optimized model parameters is obtained. and This forms the initial prediction and anomaly detection model.

[0028] Simultaneously, the intelligent collaborative management unit needs to initialize its rule base. The rule base is built upon the knowledge of aquaculture experts and contains a series of "IF-THEN" rules, such as "IF predicts dissolved oxygen will be below 4 mg / L in the next 2 hours; THEN triggers an oxygenation plan." The parameters of the optimization algorithm (such as a genetic algorithm) also need to be initially set, including population size, number of iterations, crossover and mutation probabilities, and weighting coefficients in multi-objective optimization (such as cost weights). Efficiency weight Risk weights The user interaction interface module also needs to be configured, including setting the data refresh frequency, warning threshold, and alert method.

[0029] Phase Two: System Online Operation and Closed-Loop Control After the system initialization is completed, it enters a continuous online operation phase, forming a closed loop of "perception-analysis-decision-execution-feedback".

[0030] Step 1: Real-time Data Acquisition and Transmission. Various sensors deployed in the aquaculture area continuously collect environmental parameters at a preset frequency (e.g., every 5 minutes). The collected raw data is timestamped and assigned a device ID, then transmitted to the IoT gateway via the sensor network. The gateway performs preliminary verification and protocol conversion on the data, packages it into a standard format (e.g., JSON), and transmits it to the system's data receiving interface via the internet.

[0031] Step Two: Data Preprocessing and Feature Extraction. The data preprocessing subunit (located in the gateway or server) processes the incoming real-time data stream in real time. This includes: Real-time cleaning: Re-inspect and remove extreme values ​​or invalid data generated during transmission.

[0032] Real-time normalization: Using X_min and X_max determined during the initialization phase, normalization calculations are performed on the new original data X_new. X_norm_new=(X_new-X_min) / (X_max-X_min) to make it the same dimension as the training data.

[0033] Feature construction: The latest normalized data points are combined with historical data from a previous period to form a fixed-length real-time feature vector (time window sequence) with the same structure as the training samples. This feature vector reflects the overall state of environmental parameters over a recent period.

[0034] Step 3: Deep learning model inference and prediction. The preprocessed real-time feature vectors are fed into the trained LSTM model for forward propagation (inference).

[0035] The model is based on the current input sequence and the past state of memory and The hidden state at the current time step is calculated according to the LSTM formula described above. . This can then be mapped to a specific predicted value, such as the dissolved oxygen concentration prediction X_predicted for the next hour, through a fully connected layer. Simultaneously, the model calculates an anomaly probability score based on the learned normal patterns. A simplified anomaly detection method is to calculate the Z-score of the prediction error: Z=|X_actual-X_predicted| / σ_error Where X_actual is the actual sensor measurement at the current moment (after normalization), X_predicted is the model's prediction for that moment, and σ_error is the standard deviation of the model's prediction error on the validation set. If the Z value exceeds a set threshold (e.g., 3), the data at that moment is considered abnormal.

[0036] Step Four: Intelligent Collaborative Decision Generation. The prediction results (e.g., "Dissolved oxygen will drop to 3.8 mg / L in the next hour") and / or anomaly indicators (e.g., "Abnormal ammonia nitrogen concentration") output by the deep learning processing unit are sent to the intelligent collaborative management unit. The decision support subunit activates the rule engine, matching the received information against conditions in the rule base. For example, matching the rule "Dissolved oxygen too low" triggers an oxygenation strategy evaluation. The optimization algorithm (using a fitness function, for example, the weighted sum method) begins to work, evaluating multiple possible strategies (e.g., operating only aerator A, operating both A and B simultaneously, operating A while also performing a water change): Where x represents a policy code. Estimate the energy cost of this strategy. Estimate the time required for the environment to return to normal. Assess equipment wear and tear risks. , , The corresponding weights are used. The optimization algorithm finds the strategy x that optimizes (e.g., minimizes) the fitness. Finally, the optimal strategy is decoded into a specific, time-sequential sequence of equipment control instructions, such as "Immediately start aerator A at 80% power and run it continuously for 40 minutes".

[0037] Step 5: Command Issuance and Equipment Execution. The generated control commands are sent to the corresponding execution equipment controllers at the aquaculture site via a communication network (such as the MQTT protocol). The controllers parse the commands and execute the operations, such as starting the aerator, adjusting the feeder's feeding amount, and starting or stopping the water pump. The execution status (e.g., "Aerator A has started") is fed back to the system.

[0038] Step Six: Result Monitoring and Continuous Feedback. After the equipment actions are executed, the aquaculture environment changes accordingly. The sensor network continues to collect new environmental data, and a new cycle of "sensing-analysis-decision-making" begins. The system evaluates the effectiveness of the previous decision by comparing predicted values ​​with actual values ​​and observing whether anomalies have been eliminated. This new data and behavioral records are stored in the database.

[0039] Phase 3: Model Adaptive Update and System Optimization To address the dynamic nature of the aquaculture environment, which varies with seasons and stocking densities, the system incorporates a built-in adaptive model update mechanism. This process is typically executed at a low frequency (e.g., weekly) when the system is relatively idle.

[0040] New data accumulation: The system continuously stores new environmental data and corresponding real results.

[0041] Triggered update: When new data accumulates to a certain scale (such as a week's worth of data), or when the model's predictive performance is monitored to continuously decline over a certain period of time, the model update process is triggered.

[0042] Incremental learning: An elastic weight consolidation algorithm is used for incremental learning. This algorithm utilizes new data... When updating the model parameters θ, a regularization term is added to the loss function to prevent the application of old knowledge. Forgetting. The modified loss function is: Where L(θ) is the loss on the new data, It is the value of the i-th parameter before the update. This parameter represents the importance of the old task (usually approximated by the diagonal elements of the Fisher information matrix), and λ is a hyperparameter controlling the strength of regularization. By minimizing... While the model learns new knowledge, the important parameters will not deviate too far from the old model.

[0043] Model Validation and Deployment: The updated model will be evaluated on a reserved validation set. If the performance is better than or equal to the old model, the new model will replace the old model; otherwise, the old model will be retained and the reasons will be analyzed. Simultaneously, users can adjust the weight parameters of rules in the rule base or optimization algorithms based on actual farming experience through an interactive interface, making the system's decisions more suitable for specific scenarios.

[0044] In summary, the implementation process of this invention is a dynamic, self-improving intelligent management cycle. Through continuous sensing via the Internet of Things, deep learning for in-depth insights, and intelligent algorithms for collaborative decision-making, it operates on the aquaculture environment in a closed-loop control manner. Simultaneously, through adaptive learning, it continuously optimizes its own performance, ultimately achieving refined, intelligent, and automated management of the entire aquaculture process.

[0045] Finally, the following points should be noted: First, in the description of this application, it should be noted that, unless otherwise specified and limited, the terms "installation", "connection", and "linkage" should be interpreted broadly, and can be mechanical or electrical connections, or internal connections between two components, or direct connections. "Up", "down", "left", "right", etc. are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may change. Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart collaborative management system for aquaculture integrating deep learning and the Internet of Things, characterized in that, The system includes an IoT data acquisition unit, a deep learning processing unit, and an intelligent collaborative management unit. The IoT data acquisition unit is configured to collect environmental parameter data related to aquaculture in real time through multiple IoT sensor nodes deployed in the aquaculture environment, and transmit the environmental parameter data to the deep learning processing unit. The IoT sensor nodes include, but are not limited to, temperature sensors, pH sensors, dissolved oxygen sensors, ammonia nitrogen sensors and turbidity sensors. These sensors form a network in a wireless or wired manner, and data aggregation and protocol conversion are performed through an IoT gateway. The deep learning processing unit is configured to receive the environmental parameter data and process and analyze the environmental parameter data using at least one pre-trained deep learning model to generate predictive information or anomaly detection results about the aquaculture status. The deep learning model includes, but is not limited to, recurrent neural networks or long short-term memory networks based on time series analysis, used to model the changing trends of environmental parameters. The intelligent collaborative management unit is configured to generate corresponding management operation instructions based on the predicted information or anomaly detection results, and send the management operation instructions to the execution equipment at the aquaculture site through a communication network to realize intelligent collaborative management of the aquaculture process. The execution equipment includes, but is not limited to, aerators, feeders, water pumps and alarms.

2. The intelligent collaborative management system for aquaculture integrating deep learning and the Internet of Things as described in claim 1, characterized in that, The IoT data acquisition unit also includes a data preprocessing subunit, which is used to clean, normalize, and extract features from the acquired raw environmental parameter data. The cleaning operation includes removing noisy data and filling in missing values. The normalization operation scales the data to a uniform range. The feature extraction operation derives feature vectors from the raw data for use as input to the deep learning model, so as to generate a standardized and enhanced dataset for use by the deep learning processing unit.

3. The intelligent collaborative management system for aquaculture integrating deep learning and the Internet of Things as described in claim 1, characterized in that, The deep learning model in the deep learning processing unit is trained using historical aquaculture data and is equipped with a model adaptive update mechanism. This mechanism periodically adjusts the model parameters based on newly collected environmental parameter data to adapt to dynamic changes in the aquaculture environment. The model adaptive update includes, but is not limited to, online learning or incremental learning methods to ensure that the prediction accuracy is maintained over time.

4. The intelligent collaborative management system for aquaculture integrating deep learning and the Internet of Things as described in claim 1, characterized in that, The intelligent collaborative management unit also includes a decision support subunit, which integrates a rule base and optimization algorithms. This subunit is used to evaluate various possible management strategies based on prediction information or anomaly detection results, and select the optimal strategy through multi-objective optimization to generate management operation instructions. The rule base contains logical rules based on the knowledge of aquaculture experts, and the optimization algorithms include, but are not limited to, genetic algorithms or particle swarm optimization, used to balance cost, efficiency and risk factors.

5. The intelligent collaborative management system for aquaculture integrating deep learning and the Internet of Things as described in claim 1, characterized in that, The system also includes a user interaction interface module, which provides a graphical interface and remote access function to display aquaculture environment data, prediction results of deep learning processing unit and management instructions generated by intelligent collaborative management unit in real time. It also allows users to manually modify system parameters or override automatic decision-making through input devices to achieve human-machine collaborative operation.

6. The intelligent collaborative management system for aquaculture integrating deep learning and the Internet of Things as described in claim 1, characterized in that, The system is deployed in a distributed manner through a cloud computing platform. Some or all of the functions of the IoT data acquisition unit, deep learning processing unit, and intelligent collaborative management unit run on the cloud server as software services. The cloud platform provides data storage, computing resources, and network communication services, and supports simultaneous access by multiple users and system expansion.

7. The intelligent collaborative management system for aquaculture integrating deep learning and the Internet of Things as described in claim 1, characterized in that, The environmental parameter data includes, but is not limited to, various data such as water temperature, pH, dissolved oxygen concentration, turbidity, ammonia nitrogen concentration, nitrite concentration, and microbial indicators. These data are collected continuously in time series form, and the sampling frequency can be adjusted from one minute to one hour according to the aquaculture needs.

8. The intelligent collaborative management system for aquaculture integrating deep learning and the Internet of Things as described in claim 1, characterized in that, The management operation instructions include commands to control the start and stop and speed of the aeration equipment, adjust the feeding amount and time of the feeder, control the flow rate and cycle of the water pump, or trigger the audible and visual alarm device. These instructions are generated based on preset thresholds or dynamic model outputs and are sent to the execution device through wireless communication protocols.

Images