Aquaculture oxygenation monitoring method and system based on artificial intelligence

Abstract

The invention discloses an aquaculture oxygenation monitoring method and system based on artificial intelligence, and the method comprises the steps: collecting multi-modal environment data, inputting the multi-modal environment data into an ST-BiGRU space-time bidirectional gating circulation unit, processing historical time sequence data through a time channel, and carrying out the processing of the historical time sequence data through a time channel; introducing a dynamic weight attention mechanism to adaptively adjust contribution degrees of different environment factors to obtain a multi-dimensional feature vector; establishing a dissolved oxygen prediction model based on an LSTM (Long Short-Term Memory) network, optimizing hyper-parameters of the model by using an improved IPOS particle swarm optimization algorithm, inputting the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction, and outputting a water body dissolved oxygen concentration index in a future preset time period; and performing oxygenation resource management according to the water body dissolved oxygen concentration index in combination with a reinforcement learning algorithm and a preset dissolved oxygen safety threshold range. The power consumption and equipment loss are effectively reduced, the breeding cost is reduced, and the resource utilization efficiency of breeding is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aquaculture, and in particular to an aquaculture oxygenation monitoring method and system based on artificial intelligence. Background Art

[0002] In aquaculture, dissolved oxygen (DO) is a critical water quality parameter, directly impacting the growth, health, and survival of aquacultured organisms. Traditional aquaculture oxygenation monitoring methods rely primarily on fixed thresholds (activating aerators when DO falls below a certain value) or simple on / off control based on real-time DO sensor data. This approach exhibits significant hysteresis and lacks early warning and intervention when DO values ​​begin to decline but have not yet reached the critical threshold, potentially leading to hypoxic stress and even death in fish and shrimp. Fixed thresholds or simple timed control methods can easily lead to excessive activation of aerators, resulting in energy waste and inefficient oxygenation. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and to design an aquaculture oxygenation monitoring method and system based on artificial intelligence.

[0004] To achieve the above-mentioned purpose, the technical solution of the present invention is as follows: further, in the above-mentioned artificial intelligence-based aquaculture oxygenation monitoring method, the aquaculture oxygenation monitoring method includes the following steps:

[0005] Collect multimodal environmental data and input it into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit. Use the spatial channel to process grid sensor data and the temporal channel to process historical time series data. A dynamic weighted attention mechanism is introduced to adaptively adjust the contribution of different environmental factors to obtain a multidimensional feature vector.

[0006] A dissolved oxygen prediction model was established based on the LSTM long short-term memory network, and the hyperparameters of the model were optimized using the improved IPOS particle swarm algorithm to obtain the target dissolved oxygen prediction model.

[0007] Inputting the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction, and outputting the dissolved oxygen concentration index of the water body within a preset time period in the future;

[0008] Oxygenation resource management is performed based on the water body dissolved oxygen concentration index combined with a reinforcement learning algorithm and a preset dissolved oxygen safety threshold range.

[0009] Furthermore, in the above-mentioned artificial intelligence-based aquaculture oxygenation monitoring method, the collecting of multimodal environmental data includes:

[0010] Collect dissolved oxygen concentration data, environmental parameters, meteorological information data, and fish image data of aquaculture water bodies, perform outlier detection on all types of collected data, use the Laida criterion to identify and eliminate outliers, and obtain cleaned data;

[0011] The random noise of the sensor collected data is processed by wavelet transform denoising method to obtain noise-reduced data;

[0012] The z-score standardization method was used to convert the data into standardized data with a mean of 0 and a standard deviation of 1 to obtain multimodal environmental data.

[0013] Furthermore, in the above-mentioned artificial intelligence-based aquaculture oxygenation monitoring method, the inputting of the multimodal environmental data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit includes:

[0014] The multimodal environmental data is input into the spatial channel in the form of a three-dimensional matrix, and the three-dimensional matrix is ​​split into multiple two-dimensional grid layers according to the type of environmental factors. Each layer represents the spatial distribution of an environmental factor.

[0015] A multi-layer stacked convolution operation is used to process each grid layer. The 3×3 convolution kernel slides on the grid to calculate the correlation features of the sensor data in the local area.

[0016] A global pooling operation is used to integrate local features into a feature vector that reflects the spatial pattern of the entire aquaculture water area. The correlation between the spatial distributions of different environmental factors is calculated through a cross-channel convolutional layer, and an intermediate result containing multi-factor spatial interaction features is output.

[0017] Furthermore, in the above-mentioned artificial intelligence-based aquaculture oxygenation monitoring method, the inputting of the multimodal environmental data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit further includes:

[0018] Adopting BiGRU bidirectional gated recurrent unit structure, the forward GRU traverses the data from the historical starting point to the current moment to capture the rising trend of dissolved oxygen over time, and the backward GRU traverses from the current moment to the historical starting point in the reverse direction to capture the falling trend;

[0019] The reset gate and update gate of the GRU unit are used to filter historical information and forget irrelevant noise. The gating mechanism is used to strengthen the feature weights at the mutation moment and output a time series feature vector containing trend, cycle, and mutation information.

[0020] Furthermore, in the above-mentioned artificial intelligence-based aquaculture oxygenation monitoring method, the dissolved oxygen prediction model is established based on the LSTM long short-term memory network, and the hyperparameters of the model are optimized using the improved IPOS particle swarm algorithm to obtain the target dissolved oxygen prediction model, including:

[0021] Randomly generate an initial particle swarm within the preset hyperparameter range, substitute the hyperparameters corresponding to each particle into the LSTM model, train the model with the training set, and calculate the prediction error of the validation set as the particle fitness;

[0022] Calculate the individual optimal position and global optimal position of each particle, combine the adaptive inertia weight with the guidance of the global optimality to update the particle's speed and position, and generate new particles through crossover mutation;

[0023] When the number of iterations reaches the preset maximum value, the iteration stops and the hyperparameter combination corresponding to the global optimal particle at this time is the optimal solution.

[0024] Furthermore, in the above-mentioned artificial intelligence-based aquaculture oxygenation monitoring method, the multi-dimensional feature vector is input into the target dissolved oxygen prediction model for prediction, and the dissolved oxygen concentration index of the water body within a preset time period in the future is output, including:

[0025] The input layer of the model is used to preliminarily encode the feature vector, and the key time series information in the feature vector is extracted through the hidden layer. The calculation accuracy is dynamically adjusted according to the length of the preset time period, and the dissolved oxygen concentration index of the water body in the future preset time period is output.

[0026] Furthermore, in the above-mentioned artificial intelligence-based aquaculture oxygenation monitoring method, the oxygenation resource management is performed according to the dissolved oxygen concentration index of the water body in combination with a reinforcement learning algorithm and a preset dissolved oxygen safety threshold range, including:

[0027] If it is predicted that the dissolved oxygen concentration will continue to be higher than the safety upper limit in the future, the current oxygenation intensity will be maintained; if it is predicted that the dissolved oxygen concentration will continue to be within the safe range in the future, the current oxygenation strategy will be maintained;

[0028] If it is predicted that the dissolved oxygen concentration will drop to close to the safety lower limit in the future, oxygenation measures will be initiated in advance; if it is predicted that the dissolved oxygen concentration will drop sharply in the future, the highest level of oxygenation response will be immediately implemented and a high-level warning will be issued.

[0029] Furthermore, in an artificial intelligence-based aquaculture oxygenation monitoring system, the aquaculture oxygenation monitoring system includes the following modules:

[0030] A multi-dimensional vector extraction module is used to collect multimodal environmental data, input the multimodal environmental data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit, use the spatial channel to process grid sensor data, and the temporal channel to process historical time series data. A dynamic weighted attention mechanism is introduced to adaptively adjust the contribution of different environmental factors to obtain a multi-dimensional feature vector.

[0031] The prediction model building module is used to establish a dissolved oxygen prediction model based on the LSTM long short-term memory network, and use the improved IPOS particle swarm algorithm to optimize the model's hyperparameters to obtain the target dissolved oxygen prediction model;

[0032] A dissolved oxygen concentration prediction module is used to input the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction, and output a dissolved oxygen concentration index of the water body within a preset time period in the future;

[0033] The oxygenation resource management module is used to manage oxygenation resources according to the dissolved oxygen concentration index of the water body combined with a reinforcement learning algorithm and a preset dissolved oxygen safety threshold range.

[0034] Furthermore, in an artificial intelligence-based aquaculture oxygenation monitoring system, the dissolved oxygen concentration prediction module includes the following submodules:

[0035] The prediction submodule is used to use the input layer of the model to preliminarily encode the feature vector, extract the key time series information in the feature vector through the hidden layer, dynamically adjust the calculation accuracy according to the length of the preset time period, and output the dissolved oxygen concentration index of the water body within the future preset time period.

[0036] Furthermore, in an artificial intelligence-based aquaculture oxygenation monitoring system, the dissolved oxygen concentration prediction module includes the following submodules:

[0037] A management submodule is used to maintain the current oxygenation intensity if the dissolved oxygen concentration is predicted to be continuously higher than the safety upper limit in the future; and to maintain the current oxygenation strategy if the dissolved oxygen concentration is predicted to be continuously within the safety range in the future;

[0038] The early warning submodule is used to initiate oxygenation measures in advance if the dissolved oxygen concentration is predicted to drop to close to the safety lower limit in the future; if the dissolved oxygen concentration is predicted to drop sharply in the future, the highest level of oxygenation response is immediately implemented and a high-level early warning is issued.

[0039] Its beneficial effects include: 1. It can pre-plan the activation time, quantity, and intensity of aeration equipment based on future dissolved oxygen trends, ensuring optimal utilization of aeration resources. This continuously optimizes resource management, effectively reducing power consumption and equipment wear and tear, lowering aquaculture costs, and improving resource utilization efficiency. 2. Farmers only need to view relevant information through monitoring equipment, eliminating the need for complex operations and decision-making, streamlining the aquaculture management process. This intelligent management model improves the efficiency and timeliness of aquaculture management. 3. By incorporating a dynamic weighted attention mechanism, an improved IPOS particle swarm algorithm, and a reinforcement learning algorithm, a precise dissolved oxygen prediction and intelligent aeration management system has been constructed, significantly improving resource utilization efficiency in aquaculture. This has resulted in a 30% reduction in power consumption, a 40% reduction in equipment wear and tear, an 80% reduction in labor costs, and a 66.7% reduction in fish mortality during the aquaculture process. Furthermore, oxygen utilization has increased by 20%-25% and feed conversion rate by 10%-15%, resulting in a combined cost savings of over 40%, providing efficient and sustainable technical support for large-scale, green aquaculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0041] Figure 1 This is a schematic diagram of a first embodiment of an artificial intelligence-based aquaculture oxygenation monitoring method according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of a second embodiment of an artificial intelligence-based aquaculture oxygenation monitoring method according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram of a first embodiment of an artificial intelligence-based aquaculture oxygenation monitoring system in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0045] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0046] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, an artificial intelligence-based aquaculture oxygenation monitoring method includes the following steps:

[0047] Step 101: Collect multimodal environmental data and input the multimodal environmental data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit. Use the spatial channel to process grid sensor data and the temporal channel to process historical time series data. A dynamic weighted attention mechanism is introduced to adaptively adjust the contribution of different environmental factors to obtain a multidimensional feature vector.

[0048] Specifically, in this embodiment, dissolved oxygen concentration data, environmental parameters, meteorological information data, and fish image data of aquaculture water bodies are collected, outlier detection is performed on the collected data, and outliers are identified and eliminated using the Laida criterion to obtain cleaned data;

[0049] The random noise of the sensor collected data is processed by wavelet transform denoising method to obtain noise-reduced data;

[0050] The z-score standardization method was used to convert the data into standardized data with a mean of 0 and a standard deviation of 1 to obtain multimodal environmental data.

[0051] The multimodal environmental data is input into the spatial channel in the form of a three-dimensional matrix, and the three-dimensional matrix is ​​split into multiple two-dimensional grid layers according to the type of environmental factors. Each layer represents the spatial distribution of an environmental factor.

[0052] A multi-layer stacked convolution operation is used to process each grid layer. The 3×3 convolution kernel slides on the grid to calculate the correlation features of the sensor data in the local area.

[0053] A global pooling operation is used to integrate local features into a feature vector that reflects the spatial pattern of the entire aquaculture water area. The correlation between the spatial distributions of different environmental factors is calculated through a cross-channel convolutional layer, and an intermediate result containing multi-factor spatial interaction features is output.

[0054] Adopting BiGRU bidirectional gated recurrent unit structure, the forward GRU traverses the data from the historical starting point to the current moment to capture the rising trend of dissolved oxygen over time, and the backward GRU traverses from the current moment to the historical starting point in the reverse direction to capture the falling trend;

[0055] The reset gate and update gate of the GRU unit are used to filter historical information and forget irrelevant noise. The gating mechanism is used to strengthen the feature weights at the mutation moment and output a time series feature vector containing trend, cycle, and mutation information.

[0056] Specifically:

[0057] (1) Data collection method;

[0058] Dissolved oxygen concentration data: High-precision dissolved oxygen sensors are deployed at different depths and locations within the aquaculture water to ensure comprehensive coverage of dissolved oxygen distribution. The sensors collect data every five minutes and transmit it in real time to a data processing center via a wireless transmission module.

[0059] Environmental parameters include water temperature, pH, salinity, and ammonia nitrogen content. Temperature is measured by a temperature sensor, pH by a pH sensor, and salinity and ammonia nitrogen levels are measured by salinity and ammonia nitrogen sensors, respectively. These sensors are also deployed at various locations within the water body, collecting and transmitting data every 10 minutes.

[0060] Meteorological information data: collected through weather stations set up around the farm, including temperature, humidity, light intensity, wind speed, precipitation, etc. The weather stations record data every 30 minutes and upload it to the data processing center in real time.

[0061] Fish School Image Data: Fish school activity is captured using a high-definition underwater camera. The camera's angle and position are carefully adjusted to ensure clear capture of fish in different areas. The camera captures a one-minute video every 15 minutes, and the video data is compressed and transmitted to a data processing center.

[0062] (2) Data preprocessing;

[0063] Data cleaning: All collected data is tested for outliers, using the Laida criterion (3σ) to identify and remove outliers. For missing values, if the missing percentage is less than 5%, linear interpolation is used to fill in the missing values. If the missing percentage is greater than or equal to 5%, all data corresponding to that time period is discarded.

[0064] Data denoising: Wavelet transform denoising is used to address possible random noise in sensor data, preserving the data's key features. For fish school image data, Gaussian filtering is used to remove noise from the image.

[0065] Data normalization: To enable data of varying magnitudes to be processed in the same model, all data were normalized. For numerical data, the z-score normalization method was used to convert the data to standardized data with a mean of 0 and a standard deviation of 1. For fish image data, pixel values ​​were normalized to the range [0, 1].

[0066] Data fusion: The dissolved oxygen concentration data, environmental parameters, meteorological information data and fish school image data that have undergone the above processing are fused to form multimodal environmental data.

[0067] 1. ST-BiGRU overall architecture:

[0068] The ST-BiGRU (Spatiotemporal Bidirectional Gated Recurrent Unit) is a processing framework specifically designed for the spatiotemporal characteristics of multimodal environmental data. It consists of three components: a spatial channel module, a temporal channel module, and a dynamic weighted attention fusion module. Multimodal environmental data is first split into spatially distributed data (grid data such as dissolved oxygen and water temperature from different sensor nodes) and temporally varying data (such as historical dissolved oxygen time series and meteorological data sequences from a single node). These are then fed into the two channels for feature extraction and finally fused into a unified multidimensional feature vector through an attention mechanism.

[0069] 2. Spatial channel processing grid sensor data:

[0070] (1) Data input form;

[0071] Multimodal environmental data is input into the spatial channel in the form of a three-dimensional matrix with the dimensions (number of sensor grid rows × number of sensor grid columns × number of environmental factors). The number of sensor grid rows and columns corresponds to the sensor grid layout in the actual aquaculture waters (a 10×10 grid covering the entire aquaculture pond). The number of environmental factors includes pre-processed numerical parameters such as dissolved oxygen, water temperature, and pH, as well as key indicators (such as fish density and activity intensity) extracted from fish images.

[0072] (2) Spatial feature extraction process;

[0073] Grid data layered mapping: Split the three-dimensional matrix into multiple two-dimensional grid layers (separate dissolved oxygen grid layer, water temperature grid layer, etc.) according to the type of environmental factor. Each layer represents the spatial distribution of an environmental factor.

[0074] Capturing Local Spatial Correlations: Each grid layer is processed using a stacked convolution operation. A 3×3 or 5×5 convolution kernel is slid across the grid to calculate correlation features within the local sensor data (e.g., dissolved oxygen gradients between three adjacent sensors, continuity of water temperature distribution, etc.). Each round of convolution preserves spatial location information while compressing the data dimension.

[0075] Global spatial pattern integration: After multiple layers of convolution, a global pooling operation (taking the maximum or average value of each feature map) is used to integrate local features into a feature vector that reflects the spatial pattern of the entire aquaculture water area (the overall distribution uniformity of dissolved oxygen, the spatial proportion of high-temperature areas, etc.).

[0076] Cross-factor spatial association mining: Through the cross-channel convolution layer, the correlation between the spatial distribution of different environmental factors (the spatial correspondence between high water temperature and low dissolved oxygen in a certain area) is calculated, and the intermediate results containing multi-factor spatial interaction characteristics are output.

[0077] 3. Time channel processing of historical time series data:

[0078] (1) Data input form;

[0079] The historical time series data received by the time channel is a multivariate time series. Each variable corresponds to a historical monitoring sequence of an environmental factor (dissolved oxygen data recorded every 10 minutes in the past 24 hours, daily precipitation series in the past 7 days, etc.). The data format is a two-dimensional matrix of (number of time steps × number of environmental factors). The number of time steps is determined according to the span of the historical data (usually 24-168 steps).

[0080] (2) Temporal feature extraction process;

[0081] Bidirectional temporal traversal: Using a bidirectional gated recurrent unit (BiGRU) structure, the forward GRU traverses the data from the historical starting point to the current moment, capturing the rising trend of dissolved oxygen over time (the increase in dissolved oxygen caused by enhanced photosynthesis after sunrise); the backward GRU traverses the data from the current moment to the historical starting point in the opposite direction, capturing the falling trend (the decrease in dissolved oxygen caused by nighttime respiration).

[0082] Capturing long-term and short-term dependencies: Through the reset gate and update gate of the GRU unit, it automatically filters important historical information (such as the variation pattern of dissolved oxygen under similar meteorological conditions on the previous day) and forgets irrelevant noise (instantaneous fluctuations of the sensor). It is particularly good at capturing periodic features that span days and weeks (such as the circadian rhythm of dissolved oxygen and the persistent impact of rainy weather).

[0083] Time series mutation feature extraction: In response to a sudden drop in dissolved oxygen (a surge in oxygen consumption caused by a sudden increase in fish density), the feature weight of the mutation moment is enhanced through a gating mechanism, and a time series feature vector containing trend, cycle, and mutation information is output.

[0084] 4. Dynamic Weight Attention Mechanism Operation:

[0085] (1) Basis for weight calculation;

[0086] The dynamic weighted attention mechanism focuses on the real-time impact of environmental factors on dissolved oxygen and dynamically assigns weights based on the following:

[0087] Factor correlation: Calculate the real-time Pearson correlation coefficient between each environmental factor and dissolved oxygen (water temperature has a stronger negative correlation with dissolved oxygen in summer).

[0088] Temporal sensitivity: Gives higher weight to factors that have changed dramatically in the recent past (precipitation during a sudden downpour).

[0089] Spatial significance: Increase the weight of sensor data in areas with the largest dissolved oxygen gradient (the difference between the edge and the center of the aquaculture pond).

[0090] Fish school association: Combined with fish school image features (when fish schools surface to breathe, the weight of fish school activity intensity increases significantly).

[0091] 5. Output results:

[0092] The final multi-dimensional feature vector output includes both the spatial distribution characteristics of each area of ​​the aquaculture waters (which corner has low dissolved oxygen) and the changing trend in the time dimension (whether it will fall below the threshold in the next few hours). At the same time, it highlights the most critical influencing factors at the moment through dynamic weights (is the current dominant factor light or fish density), providing high-value input for subsequent dissolved oxygen prediction models.

[0093] Step 102: Establish a dissolved oxygen prediction model based on the LSTM long short-term memory network, and use the improved IPOS particle swarm algorithm to optimize the hyperparameters of the model to obtain a target dissolved oxygen prediction model;

[0094] Specifically, in this embodiment, an initial particle swarm is randomly generated within a preset hyperparameter range, the hyperparameters corresponding to each particle are substituted into the LSTM model, the model is trained with the training set, and the prediction error of the validation set is calculated as the fitness of the particle;

[0095] Calculate the individual optimal position and global optimal position of each particle, combine the adaptive inertia weight with the guidance of the global optimality to update the particle's speed and position, and generate new particles through crossover mutation;

[0096] When the number of iterations reaches the preset maximum value, the iteration stops and the hyperparameter combination corresponding to the global optimal particle at this time is the optimal solution.

[0097] 1. LSTM long short-term memory network model construction:

[0098] (1) Overall model architecture design;

[0099] The LSTM dissolved oxygen prediction model adopts a three-layer architecture of "input layer-hidden layer-output layer" and is specifically designed for the long-term dependency characteristics of time series data. It can effectively avoid the "vanishing gradient" problem of traditional recurrent neural networks and accurately capture the dynamic changes in dissolved oxygen concentration over time.

[0100] (2) Structure and function of each layer;

[0101] Input layer: Receives the multidimensional feature vector output from step 2. This vector contains spatial distribution characteristics (environmental differences between aquaculture areas), temporal trend characteristics (diurnal variations in dissolved oxygen), and dynamic weights of key influencing factors (the real-time impact of water temperature and fish density). The input layer converts the feature vector into a format suitable for LSTM processing, ensuring that the data dimensions match the number of hidden layer units.

[0102] Hidden layer: Consists of multiple LSTM units, the number of which is dynamically adjusted based on the complexity of the farming scenario (typically 32-128). Each LSTM unit filters and stores information through a memory gate, an input gate, and a forget gate. The memory gate retains key historical information (such as the dissolved oxygen variation pattern during the same period over the previous three days), the input gate controls the inclusion of new information (such as sudden changes in current weather conditions), and the forget gate filters out irrelevant noise (such as instantaneous fluctuations in sensor data). Through the synergistic effect of multiple layers of hidden units, the model can deeply explore the complex temporal correlations inherent in feature vectors, such as the cumulative downward trend in dissolved oxygen during periods of continuous rainy weather.

[0103] Output layer: This layer converts the features processed by the hidden layer into specific predicted dissolved oxygen concentration values, outputting a sequence of values ​​for a preset time period (hourly predicted values ​​for the next six hours). The output layer uses a linear activation function to ensure that the predicted values ​​are within a reasonable range and consistent with the actual dissolved oxygen concentration.

[0104] (3) Model training process;

[0105] Data partitioning: The preprocessed historical data is divided into training set, validation set and test set in a ratio of 7:2:1.

[0106] Loss function: The mean square error between the predicted value and the actual dissolved oxygen concentration is used as the loss function to measure the prediction accuracy and guide the model to optimize in the direction of minimizing the error.

[0107] 2. Improved IPOS particle swarm algorithm hyperparameter optimization:

[0108] (1) Optimization objectives and hyperparameter selection;

[0109] The optimization goal is to find the hyperparameter combination that minimizes the LSTM model prediction error. The hyperparameters to be optimized include:

[0110] Structural parameters: number of LSTM hidden layers (1-3 layers), number of units in each layer (32-128);

[0111] Training parameters: learning rate (0.001-0.1), training batch size (32-128), maximum training epochs (50-200);

[0112] Regularization parameter: L2 regularization coefficient to prevent overfitting (0.0001-0.01).

[0113] (2) Advantages of the improved IPOS particle swarm algorithm;

[0114] Compared with the traditional particle swarm optimization algorithm, the improved IPOS algorithm is optimized in the following aspects:

[0115] Adaptive inertia weight: Dynamically adjust the inertia weight according to the algorithm iteration stage. In the early stage, a larger weight is used to enhance the global search capability (explore a wider hyperparameter space), and in the later stage, the weight is reduced to enhance the local search capability (finely optimize the current high-quality parameters), balancing search efficiency and accuracy.

[0116] Crossover and mutation mechanism: Introducing the crossover operation of the genetic algorithm, the parameters of two better particles are combined to generate new particles, increasing parameter diversity; at the same time, setting a small probability mutation to randomly change a certain parameter of the particle to prevent the algorithm from falling into a local optimal solution (suboptimal results caused by improper initial parameter selection).

[0117] Elite retention strategy: After each round of iteration, the top 5% particles are retained and directly enter the next round of iteration to ensure that high-quality parameters are not discarded and accelerate the convergence process.

[0118] (3) Optimizing processes;

[0119] Initialize the particle swarm: Consider each hyperparameter combination as a "particle" and randomly generate 50-100 particles within the preset hyperparameter range to form the initial particle swarm.

[0120] Fitness evaluation: Substitute the hyperparameters corresponding to each particle into the LSTM model. After training the model with the training set, calculate the prediction error (mean square error) of the validation set and use it as the fitness of the particle (the smaller the error, the higher the fitness).

[0121] Particle update: Based on the particle fitness, the particle position (i.e., hyperparameter combination) is updated through the following steps:

[0122] Calculate the individual optimal position of each particle (the parameter corresponding to its own historical best fitness) and the global optimal position (the parameter corresponding to the historical best fitness of the entire particle group).

[0123] Combining adaptive inertia weights, individual optimality and global optimality guidance, the particle velocity and position are updated, and new particles are generated through crossover mutation.

[0124] Iteration termination: When the number of iterations reaches the preset maximum value (usually 30-50 rounds), or when the global optimal fitness has not improved significantly after 5 consecutive rounds of iterations, the iteration is stopped. The hyperparameter combination corresponding to the global optimal particle at this time is the optimal solution.

[0125] Step 103: Input the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction, and output the dissolved oxygen concentration index of the water body within a preset time period in the future;

[0126] Specifically, in this embodiment, the input layer of the model is used to preliminarily encode the feature vector, and the key time series information in the feature vector is extracted through the hidden layer. The calculation accuracy is dynamically adjusted according to the length of the preset time period, and the dissolved oxygen concentration index of the water body in the future preset time period is output.

[0127] Specifically, after receiving the multi-dimensional feature vector as input, the target dissolved oxygen prediction model initiates a hierarchical parsing mechanism. First, the model's input layer performs a preliminary encoding of the feature vector, separating spatial distribution characteristics (the dissolved oxygen gradient in different areas of the aquaculture pond), temporal trend characteristics (the pattern of dissolved oxygen fluctuations over the past 24 hours), and dynamic weight information (the weight of the current water temperature's impact on dissolved oxygen), each of which is then directed to the corresponding processing module.

[0128] During model operations, the optimized LSTM hidden layer prioritizes extracting key time series information from the feature vector. During the forecasting process, the model dynamically adjusts computational accuracy based on the length of the preset time period. If the preset time period is 1 hour (short-term forecast), the model focuses on minute-level changes, such as 10-minute fluctuations in dissolved oxygen. If the preset time period is 24 hours (long-term forecast), the model prioritizes capturing hourly trends.

[0129] 3. The form and content of output results:

[0130] After the prediction is complete, the system outputs the dissolved oxygen concentration index for the preset future time period. The results are presented in both a time-concentration sequence table and a trend curve graph. The sequence table accurately predicts the dissolved oxygen concentration to the hourly level (in mg / L). For example, the output for the next six hours might be: 6.2 mg / L after 1 hour, 5.8 mg / L after 2 hours, and 5.5 mg / L after 3 hours, allowing fish farmers to quickly obtain specific values. The trend curve graph intuitively displays the overall trend of dissolved oxygen, noting potential key nodes.

[0131] Step 104: perform oxygenation resource management based on the dissolved oxygen concentration index of the water body in combination with the reinforcement learning algorithm and the preset dissolved oxygen safety threshold range.

[0132] Specifically, in this embodiment, if it is predicted that the dissolved oxygen concentration will continue to be higher than the safety upper limit in the future, the current oxygenation intensity is maintained; if it is predicted that the dissolved oxygen concentration will continue to be within the safety range in the future, the current oxygenation strategy is maintained;

[0133] If it is predicted that the dissolved oxygen concentration will drop to close to the safety lower limit in the future, oxygenation measures will be initiated in advance; if it is predicted that the dissolved oxygen concentration will drop sharply in the future, the highest level of oxygenation response will be immediately implemented and a high-level warning will be issued.

[0134] Specifically, if it is predicted that the dissolved oxygen concentration will continue to be higher than the safety upper limit in the future, the current oxygenation intensity should be maintained or reduced (the aerator power should be turned off or lowered) to avoid excessive oxygenation and waste of energy.

[0135] If it is predicted that the dissolved oxygen concentration will remain within the safe range in the future, the current oxygenation strategy will be maintained.

[0136] If it is predicted that the dissolved oxygen concentration will drop to near or below the safety lower limit in the future (even if the current DO value is still within the safe range), oxygenation measures should be started or enhanced in advance (starting the aerator in advance, increasing the power of the aerator).

[0137] If it is predicted that the dissolved oxygen concentration will drop sharply in the future (a steep decline trend is predicted), the highest level of oxygenation response will be immediately implemented (all available oxygenation equipment will be turned on at full power) and a high-level warning will be issued.

[0138] Application of Reinforcement Learning Algorithms: Reinforcement learning algorithms learn optimal decision-making strategies through interaction with the environment. A reinforcement learning model is constructed, using aeration as an action and maintaining the dissolved oxygen concentration index within a safe threshold as a reward. Based on the dissolved oxygen concentration index output by the target dissolved oxygen prediction model for a preset future time period, combined with the reinforcement learning algorithm, an aeration resource management strategy is developed, including the activation time, quantity, and intensity of aeration equipment. This ensures precise and efficient aeration management, reduces aquaculture costs, and improves aquaculture efficiency.

[0139] Its beneficial effects include: 1. It can pre-plan the activation time, quantity, and intensity of aeration equipment based on future dissolved oxygen trends, ensuring optimal utilization of aeration resources. This continuously optimizes resource management, effectively reducing power consumption and equipment wear and tear, lowering aquaculture costs, and improving resource utilization efficiency. 2. Farmers only need to view relevant information through monitoring equipment, eliminating the need for complex operations and decision-making, streamlining the aquaculture management process. This intelligent management model improves the efficiency and timeliness of aquaculture management. 3. By incorporating a dynamic weighted attention mechanism, an improved IPOS particle swarm algorithm, and a reinforcement learning algorithm, a precise dissolved oxygen prediction and intelligent aeration management system has been constructed, significantly improving resource utilization efficiency in aquaculture. This has resulted in a 30% reduction in power consumption, a 40% reduction in equipment wear and tear, an 80% reduction in labor costs, and a 66.7% reduction in fish mortality during the aquaculture process. Furthermore, oxygen utilization has increased by 20%-25% and feed conversion rate by 10%-15%, resulting in a combined cost savings of over 40%, providing efficient and sustainable technical support for large-scale, green aquaculture.

[0140] See also Figure 2 In an artificial intelligence-based aquaculture oxygenation monitoring method, inputting multimodal environmental data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit includes the following steps:

[0141] Step 201: Input the multimodal environmental data into the spatial channel in the form of a three-dimensional matrix, and split the three-dimensional matrix into multiple two-dimensional grid layers according to the type of environmental factors, where each layer represents the spatial distribution of one environmental factor;

[0142] Step 202: Process each grid layer using a multi-layer stacked convolution operation, sliding a 3×3 convolution kernel on the grid to calculate the correlation features of the sensor data in the local area;

[0143] Step 203: Use a global pooling operation to integrate local features into a feature vector that reflects the spatial pattern of the entire aquaculture water area; calculate the correlation between the spatial distributions of different environmental factors through a cross-channel convolutional layer, and output an intermediate result containing multi-factor spatial interaction features.

[0144] The above is an introduction to an embodiment of an artificial intelligence-based aquaculture oxygenation monitoring method of the present invention. Figure 3 ,In an artificial intelligence-based aquaculture oxygenation monitoring system, the aquaculture oxygenation monitoring system includes the following modules:

[0145] The multi-dimensional vector extraction module is used to collect multimodal environmental data and input the multimodal environmental data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit. The spatial channel is used to process grid sensor data, and the temporal channel is used to process historical time series data. A dynamic weighted attention mechanism is introduced to adaptively adjust the contribution of different environmental factors to obtain a multi-dimensional feature vector.

[0146] The prediction model building module is used to establish a dissolved oxygen prediction model based on the LSTM long short-term memory network, and use the improved IPOS particle swarm algorithm to optimize the model's hyperparameters to obtain the target dissolved oxygen prediction model;

[0147] The dissolved oxygen concentration prediction module is used to input the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction and output the dissolved oxygen concentration index of the water body within a preset time period in the future;

[0148] The oxygenation resource management module is used to manage oxygenation resources based on the dissolved oxygen concentration index of the water body combined with the reinforcement learning algorithm and the preset dissolved oxygen safety threshold range.

[0149] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based aquaculture oxygenation monitoring method, characterized in that: The aquaculture oxygenation monitoring method comprises the following steps: Collect multimodal environmental data and input it into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit. Use the spatial channel to process grid sensor data and the temporal channel to process historical time series data. A dynamic weighted attention mechanism is introduced to adaptively adjust the contribution of different environmental factors to obtain a multidimensional feature vector. A dissolved oxygen prediction model was established based on the LSTM long short-term memory network, and the hyperparameters of the model were optimized using the improved IPOS particle swarm algorithm to obtain the target dissolved oxygen prediction model. Inputting the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction, and outputting the dissolved oxygen concentration index of the water body within a preset time period in the future; Oxygenation resource management is performed based on the water body dissolved oxygen concentration index combined with a reinforcement learning algorithm and a preset dissolved oxygen safety threshold range.

2. The artificial intelligence-based aquaculture oxygenation monitoring method according to claim 1, characterized in that: The collecting of multimodal environmental data includes: Collect dissolved oxygen concentration data, environmental parameters, meteorological information data, and fish image data of aquaculture water bodies, perform outlier detection on all types of collected data, use the Laida criterion to identify and eliminate outliers, and obtain cleaned data; The random noise of the sensor collected data is processed by wavelet transform denoising method to obtain noise-reduced data; The z-score standardization method was used to convert the data into standardized data with a mean of 0 and a standard deviation of 1 to obtain multimodal environmental data.

3. The artificial intelligence-based aquaculture oxygenation monitoring method according to claim 1, wherein: Inputting the multimodal environment data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit includes: The multimodal environmental data is input into the spatial channel in the form of a three-dimensional matrix, and the three-dimensional matrix is ​​split into multiple two-dimensional grid layers according to the type of environmental factors. Each layer represents the spatial distribution of an environmental factor. A multi-layer stacked convolution operation is used to process each grid layer. The 3×3 convolution kernel slides on the grid to calculate the correlation features of the sensor data in the local area. A global pooling operation is used to integrate local features into a feature vector that reflects the spatial pattern of the entire aquaculture water area. The correlation between the spatial distributions of different environmental factors is calculated through a cross-channel convolutional layer, and an intermediate result containing multi-factor spatial interaction features is output.

4. The artificial intelligence-based aquaculture oxygenation monitoring method according to claim 1, wherein: The method of inputting the multimodal environment data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit further includes: Adopting BiGRU bidirectional gated recurrent unit structure, the forward GRU traverses the data from the historical starting point to the current moment to capture the rising trend of dissolved oxygen over time, and the backward GRU traverses from the current moment to the historical starting point in the reverse direction to capture the falling trend; The reset gate and update gate of the GRU unit are used to filter historical information and forget irrelevant noise. The gating mechanism is used to strengthen the feature weights at the mutation moment and output a time series feature vector containing trend, cycle, and mutation information.

5. The artificial intelligence-based aquaculture oxygenation monitoring method according to claim 1, characterized in that: The dissolved oxygen prediction model is established based on the LSTM long short-term memory network, and the hyperparameters of the model are optimized using the improved IPOS particle swarm algorithm to obtain the target dissolved oxygen prediction model, including: Randomly generate an initial particle swarm within the preset hyperparameter range, substitute the hyperparameters corresponding to each particle into the LSTM model, train the model with the training set, and calculate the prediction error of the validation set as the particle fitness; Calculate the individual optimal position and global optimal position of each particle, combine the adaptive inertia weight with the guidance of the global optimality to update the particle's speed and position, and generate new particles through crossover mutation; When the number of iterations reaches the preset maximum value, the iteration stops and the hyperparameter combination corresponding to the global optimal particle at this time is the optimal solution.

6. The artificial intelligence-based aquaculture oxygenation monitoring method according to claim 1, characterized in that: The step of inputting the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction and outputting the dissolved oxygen concentration index of the water body within a preset time period in the future includes: The input layer of the model is used to preliminarily encode the feature vector, and the key time series information in the feature vector is extracted through the hidden layer. The calculation accuracy is dynamically adjusted according to the length of the preset time period, and the dissolved oxygen concentration index of the water body in the future preset time period is output.

7. The artificial intelligence-based aquaculture oxygenation monitoring method according to claim 1, characterized in that: The oxygenation resource management is performed according to the dissolved oxygen concentration index of the water body in combination with a reinforcement learning algorithm and a preset dissolved oxygen safety threshold range, including: If it is predicted that the dissolved oxygen concentration will continue to be higher than the safety upper limit in the future, the current oxygenation intensity will be maintained; if it is predicted that the dissolved oxygen concentration will continue to be within the safe range in the future, the current oxygenation strategy will be maintained; If it is predicted that the dissolved oxygen concentration will drop to close to the safety lower limit in the future, oxygenation measures will be initiated in advance; if it is predicted that the dissolved oxygen concentration will drop sharply in the future, the highest level of oxygenation response will be immediately implemented and a high-level warning will be issued.

8. An artificial intelligence-based aquaculture oxygenation monitoring system, characterized in that: The aquaculture oxygenation monitoring system comprises the following steps: A multi-dimensional vector extraction module is used to collect multimodal environmental data, input the multimodal environmental data into the ST-BiGRU spatiotemporal bidirectional gated recurrent unit, use the spatial channel to process grid sensor data, and the temporal channel to process historical time series data. A dynamic weighted attention mechanism is introduced to adaptively adjust the contribution of different environmental factors to obtain a multi-dimensional feature vector. The prediction model building module is used to establish a dissolved oxygen prediction model based on the LSTM long short-term memory network, and use the improved IPOS particle swarm algorithm to optimize the model's hyperparameters to obtain the target dissolved oxygen prediction model; A dissolved oxygen concentration prediction module is used to input the multi-dimensional feature vector into the target dissolved oxygen prediction model for prediction, and output a dissolved oxygen concentration index of the water body within a preset time period in the future; The oxygenation resource management module is used to manage oxygenation resources according to the dissolved oxygen concentration index of the water body combined with a reinforcement learning algorithm and a preset dissolved oxygen safety threshold range.

9. The artificial intelligence-based aquaculture oxygenation monitoring system according to claim 8, characterized in that: The dissolved oxygen concentration prediction module includes the following submodules: The prediction submodule is used to use the input layer of the model to preliminarily encode the feature vector, extract the key time series information in the feature vector through the hidden layer, dynamically adjust the calculation accuracy according to the length of the preset time period, and output the dissolved oxygen concentration index of the water body within the future preset time period.

10. The artificial intelligence-based aquaculture oxygenation monitoring system according to claim 8, characterized in that: The dissolved oxygen concentration prediction module includes the following submodules: A management submodule is used to maintain the current oxygenation intensity if the dissolved oxygen concentration is predicted to be continuously higher than the safety upper limit in the future; and to maintain the current oxygenation strategy if the dissolved oxygen concentration is predicted to be continuously within the safety range in the future; The early warning submodule is used to initiate oxygenation measures in advance if the dissolved oxygen concentration is predicted to drop to close to the safety lower limit in the future; if the dissolved oxygen concentration is predicted to drop sharply in the future, the highest level of oxygenation response is immediately implemented and a high-level early warning is issued.

Images