Self-adaptive adjustment method and system for prawn culture environment
By extracting control coefficients through a collaborative analysis model, the control conflict caused by the coupling of water temperature and dissolved oxygen was resolved, enabling precise and intelligent adjustment of the shrimp farming environment and improving farming efficiency and survival rate.
Patent Information
- Application Number
- CN202511925778.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-03-13
AI Technical Summary
In existing shrimp farming environmental control technologies, the coupling relationship between water temperature and dissolved oxygen parameters is neglected, leading to control conflicts, energy waste, and frequent equipment start-ups and shutdowns, making it difficult to stabilize within the optimal range.
By using a collaborative analysis model to extract control coefficients based on historical datasets and combining them with current environmental parameters, the collaborative control quantities of temperature and oxygenation are accurately analyzed, driving the device to adjust and achieve synchronous regulation of water temperature and dissolved oxygen.
It improves the accuracy and flexibility of environmental control, reduces energy consumption and equipment conflicts, ensures that the breeding environment is stable in the optimal state, reduces the need for human intervention, and improves breeding efficiency and survival rate.
Smart Images
Figure CN121647205A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automated control technology in aquaculture, specifically to a method and system for adaptive adjustment of the shrimp farming environment. Background Technology
[0002] Intensive shrimp farming is an important model in my country's aquaculture industry, and its output and efficiency are closely related to the stability of environmental conditions. Water temperature and dissolved oxygen concentration are core parameters in the farming process, directly affecting shrimp metabolism, feeding, growth, and immune status. Maintaining these two parameters within a suitable and stable range is crucial for ensuring successful farming and improving yield and quality.
[0003] Currently, large-scale shrimp farms are generally equipped with temperature sensors, dissolved oxygen sensors, and corresponding actuators, such as heating devices, refrigeration units, and aerators, to monitor and regulate environmental factors.
[0004] The limitations of existing technologies include at least the following problems: Although existing aquaculture environment control technologies can monitor parameters such as water temperature and dissolved oxygen through sensors and drive corresponding equipment, such as heaters and aerators, to make adjustments when they exceed the threshold, this control strategy treats water temperature and dissolved oxygen concentration as two independent variables and sets up an independent control loop for each parameter.
[0005] However, in the complex biochemical system of actual aquaculture water, there is a strong physical and biochemical coupling between these parameters. For example, starting an aerator increases dissolved oxygen, but the water temperature drops due to water agitation and gas exchange. Conversely, adjusting the water temperature also affects microbial activity and gas solubility, thereby changing the dynamics of dissolved oxygen. When the system independently increases aeration based on low dissolved oxygen and simultaneously starts heating due to low water temperature, the two control actions will inadvertently conflict with each other. That is, increasing aeration and lowering temperature, and heating may indirectly affect dissolved oxygen, thus creating a control conflict. This not only causes additional energy consumption, but may also cause the system to oscillate continuously around the target value, making it difficult to stabilize in the optimal range, and may even shorten the lifespan of the equipment due to frequent start-ups and shutdowns. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides an adaptive adjustment method and system for shrimp farming environments, which solves the problem of control conflicts caused by neglecting the parameter coupling of water temperature and dissolved oxygen in existing technologies.
[0007] To achieve the above objectives, the present invention provides the following technical solution: an adaptive adjustment method for shrimp farming environment, comprising the following steps: based on a set time period, synchronously acquiring a preset historical collaborative control time series dataset of a temperature control device and an oxygenation device in the shrimp farming water, including historical temperature control time series data, historical oxygenation control time series data, historical water temperature time series data, and historical dissolved oxygen concentration time series data; inputting the historical collaborative control time series dataset into a pre-trained collaborative analysis model to extract a set of control coefficients; acquiring the current values of water temperature and dissolved oxygen concentration in the shrimp farming water and preset target values, and combining the control coefficient set to analyze the temperature collaborative control quantity and oxygenation collaborative control quantity of the shrimp farming water; and driving the preset temperature control device and oxygenation device to adjust the shrimp farming water according to the temperature collaborative control quantity and oxygenation collaborative control quantity.
[0008] Furthermore, the historical temperature control time series data includes the temperature control power values of the temperature control devices at several historical time points, the historical oxygenation control time series data includes the oxygenation control power values of the oxygenation devices at several historical time points, the historical water temperature time series data includes the temperature measurement values of the shrimp farming water body at several historical time points, and the historical dissolved oxygen concentration time series data includes the dissolved oxygen concentration measurement values of the shrimp farming water body at several historical time points.
[0009] Furthermore, the set of control coefficients includes temperature-controlled water temperature coefficient, temperature-controlled dissolved oxygen coefficient, oxygen-enriched water temperature coefficient, and oxygen-enriched dissolved oxygen coefficient.
[0010] Furthermore, the collaborative analysis model includes a data alignment and splicing layer, a feature attention weighting layer, a coefficient regression analysis layer, and a control coefficient output layer.
[0011] Further, the specific steps for extracting the control coefficient set are as follows: In the data alignment and stitching layer of the collaborative analysis model, the historical collaborative control time series dataset is time-aligned and feature-stitched to obtain a fused time series tensor; in the feature attention weighting layer of the collaborative analysis model, the fused time series tensor is subjected to attention weight calculation and weighting to obtain a weighted fused time series tensor; in the coefficient regression analysis layer of the collaborative analysis model, the weighted fused time series tensor is forward-propagated to obtain the control coefficient set; in the control coefficient output layer of the collaborative analysis model, the control coefficient set is output.
[0012] Further, the specific steps for obtaining the fused time series tensor are as follows: The historical temperature control time series data, historical oxygenation control time series data, historical water temperature time series data, and historical dissolved oxygen concentration time series data are aligned using a unified timestamp as a reference; at each aligned timestamp, the corresponding temperature control power value, oxygenation control power value, temperature measurement value, and dissolved oxygen concentration measurement value are concatenated to obtain a combined feature vector for the corresponding aligned timestamp; the combined feature vectors of each aligned timestamp are arranged in chronological order, and the fused time series tensor is constructed.
[0013] Furthermore, the specific steps to obtain the weighted fused temporal tensor are as follows: calculate the attention score for each aligned timestamp combination feature vector in the fused temporal tensor; weight the corresponding aligned timestamp combination feature vectors according to the attention score; and reassemble all the weighted combination feature vectors in their original temporal order to obtain the weighted fused temporal tensor.
[0014] Furthermore, the specific steps to obtain the control coefficient set are as follows: perform forward propagation calculation on the weighted fusion time series tensor to obtain initial coefficient data; perform structuring processing on the initial coefficient data to obtain a structured data set; and perform output encapsulation processing on the structured data set to obtain the control coefficient set.
[0015] Further, the specific steps for analyzing the temperature-coordinated control and oxygenation-coordinated control quantities of shrimp farming water are as follows: Perform difference analysis on the current values of water temperature and dissolved oxygen concentration in the shrimp farming water and the preset target values to obtain the target deviation values of water temperature and dissolved oxygen concentration; combine the target deviation values of water temperature and dissolved oxygen concentration with the control coefficient set for comprehensive analysis to obtain the temperature-coordinated control and oxygenation-coordinated control quantities of the shrimp farming water.
[0016] An adaptive adjustment system for shrimp farming environment includes: a historical data acquisition unit, used to synchronously acquire a preset time-series dataset of historical coordinated control of a temperature control device and an oxygenation device in the shrimp farming water body based on a set time period, including historical temperature control time-series data, historical oxygenation control time-series data, historical water temperature time-series data, and historical dissolved oxygen concentration time-series data; a coordinated analysis unit, used to input the historical coordinated control time-series dataset into a pre-trained coordinated analysis model to extract a set of control coefficients; a coordinated control analysis unit, used to acquire the current values of water temperature and dissolved oxygen concentration in the shrimp farming water body and preset target values, and analyze the temperature coordinated control quantity and oxygenation coordinated control quantity of the shrimp farming water body in combination with the control coefficient set; and a control execution unit, used to drive the preset temperature control device and oxygenation device to adjust the shrimp farming water body according to the temperature coordinated control quantity and oxygenation coordinated control quantity.
[0017] The present invention has the following beneficial effects:
[0018] (1) The adaptive adjustment method for shrimp farming environment acquires historical collaborative control time sequence data of temperature control device and oxygenation device by setting a time period, covering temperature and oxygenation control power, water temperature and dissolved oxygen concentration. Then, the pre-trained collaborative analysis model extracts temperature-controlled water temperature, temperature-controlled dissolved oxygen, oxygenation water temperature and oxygenation dissolved oxygen coefficient. Combined with the current water temperature and dissolved oxygen concentration and the preset target value, the temperature and oxygenation collaborative control quantities are accurately analyzed, thereby avoiding the blindness of traditional farming based solely on experience. It can accurately capture the change pattern of water temperature and dissolved oxygen concentration and the control logic of the device, making the adjustment instructions more in line with the actual farming needs, effectively reducing the stress response of shrimp caused by environmental parameter fluctuations, and improving the survival rate of farming. Compared with the traditional extensive control, this method can stabilize the water temperature and dissolved oxygen concentration within the target range, providing a suitable environment for shrimp growth and reducing the risk of disease caused by parameter imbalance.
[0019] (2) The adaptive adjustment method for shrimp farming environment improves the efficiency of regulation through the multi-level processing of the collaborative analysis model. The collaborative analysis model first realizes the time alignment and feature fusion of multi-dimensional time series data through the data alignment and splicing layer, then highlights the feature weights of key time nodes through the feature attention weighting layer, and finally outputs the precise control coefficients by the coefficient regression analysis layer. Subsequently, combined with the deviation analysis between the current value and the target value, the collaborative control quantity is quickly derived and directly drives the device to perform adjustment. The whole process does not require manual intervention, which greatly shortens the response time of parameter monitoring and device regulation. At the same time, it avoids the conflict that may occur when temperature and oxygenation devices are regulated separately. It can reduce the labor cost of manual inspection and operation, improve the timeliness of regulation, ensure that the farming environment parameters quickly return to the appropriate range, and ensure that the growth rhythm of shrimp is not affected.
[0020] (3) The adaptive adjustment method for shrimp farming environment improves environmental adaptability by dynamically extracting control coefficients and analyzing real-time deviations. It continuously optimizes the control coefficient set based on historical collaborative data, which can adapt to the different needs of water temperature and dissolved oxygen concentration at different farming stages. It can also cope with the parameter adjustment needs brought about by external environmental changes such as seasonal changes and water quality fluctuations. When analyzing control quantities, it comprehensively considers the dual deviations of water temperature and dissolved oxygen concentration, and performs collaborative regulation in combination with the control coefficient set to avoid the adverse effects of adjusting one parameter on another parameter. Whether it is the fine parameter control during the seedling cultivation period or the stable environment maintenance during the adult shrimp farming period, it can adapt to different scenarios through adaptive adjustment, improve the flexibility and applicability of farming environment regulation, and provide stable guarantee for shrimp growth under different farming conditions.
[0021] (4) The shrimp farming environment adaptive adjustment system integrates a historical data acquisition unit, a collaborative analysis unit, a collaborative control analysis unit, and a control execution unit to form an automated and intelligent environmental control system. The system can acquire key environmental parameters such as water temperature and dissolved oxygen concentration in the shrimp farming water in real time. At the same time, it learns and analyzes the historical operation of temperature control and oxygenation devices through a preset historical collaborative control dataset. Through this collaborative analysis method, the system can extract the best control coefficient, intelligently optimize the control strategy, and automatically calculate the most suitable temperature and oxygenation adjustment amount based on the difference between the current water state and the target value, thereby accurately adjusting the farming environment. This adaptive adjustment method not only improves the accuracy and flexibility of environmental control, but also significantly reduces the need for manual intervention, enabling the farm to achieve more efficient and refined management and ensuring that the water environment is always in the optimal state.
[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description
[0023] Figure 1 This is a flowchart of a method for adaptive adjustment of the shrimp farming environment according to the present invention.
[0024] Figure 2 This is a flowchart illustrating the specific steps involved in extracting the control coefficient set in a method for adaptive adjustment of the shrimp farming environment according to the present invention.
[0025] Figure 3 This is a block diagram of an adaptive adjustment system for shrimp farming environment according to the present invention. Detailed Implementation
[0026] Please see Figure 1 This invention provides a technical solution: an adaptive adjustment method for shrimp farming environment, comprising the following steps: based on a set time period, synchronously acquiring historical collaborative control time series datasets generated by a preset temperature control device and an oxygenation device in step response or impulse response control experiments in shrimp farming water, including historical temperature control time series data, historical oxygenation control time series data, historical water temperature time series data, and historical dissolved oxygen concentration time series data; inputting the historical collaborative control time series dataset into a pre-trained collaborative analysis model to extract a set of control coefficients; acquiring the current values of water temperature and dissolved oxygen concentration in the shrimp farming water and preset target values, and combining the control coefficient set to analyze the temperature collaborative control quantity and oxygenation collaborative control quantity of the shrimp farming water; and driving the preset temperature control device and oxygenation device to adjust the shrimp farming water according to the temperature collaborative control quantity and oxygenation collaborative control quantity.
[0027] The time period setting includes the data collection period and the model update period. The data collection period is 1-5 minutes, and the model update period is 12-48 hours or triggered when the amount of new data reaches 500-2000 sets.
[0028] The preset temperature control device includes, but is not limited to, immersion heating rods or circulating water chiller units, and the control quantity is a power percentage.
[0029] The preset oxygenation devices include, but are not limited to, nano-aeration discs or impeller aerators, and their control parameters are power percentage or gas flow rate.
[0030] The step response control experiment specifically includes: In the first experimental stage, the power of the aeration device is kept constant, and the power of the temperature control device is increased stepwise from the first reference value to the first target value, while the first steady-state changes in water temperature and dissolved oxygen concentration are recorded simultaneously; In the second experimental stage, the power of the temperature control device is kept constant, and the power of the aeration device is increased stepwise from the second reference value to the second target value, while the second steady-state changes in water temperature and dissolved oxygen concentration are recorded simultaneously.
[0031] Its adaptive adjustment specifically involves adding new time-series data obtained during continuous operation to the historical collaborative control time-series dataset in a rolling update manner, and retraining the collaborative analysis model and updating the control coefficient set according to the model update cycle.
[0032] Specifically, the historical temperature control time series data includes the temperature control power values of the temperature control devices at several historical time points, the historical oxygenation control time series data includes the oxygenation control power values of the oxygenation devices at several historical time points, the historical water temperature time series data includes the temperature measurement values of the shrimp farming water body at several historical time points, and the historical dissolved oxygen concentration time series data includes the dissolved oxygen concentration measurement values of the shrimp farming water body at several historical time points.
[0033] The control coefficient set includes the temperature control water temperature coefficient characterizing the influence of the temperature control device on water temperature, the temperature control dissolved oxygen coefficient characterizing the influence of the temperature control device on dissolved oxygen concentration, the oxygenation water temperature coefficient characterizing the influence of the aeration device on water temperature, and the oxygenation dissolved oxygen coefficient characterizing the influence of the aeration device on dissolved oxygen concentration. The collaborative analysis model includes a data alignment and stitching layer, a feature attention weighting layer, a coefficient regression analysis layer, and a control coefficient output layer.
[0034] The pre-training steps for the collaborative analysis model are as follows:
[0035] From the historical database, select several time periods that fully record the "step response" or "impulse response" control experiments. Each time period must contain four sets of strictly synchronized raw data: the power percentage time series of the temperature control device, the power percentage (or gas flow rate) time series of the aeration device, the water temperature measurement time series, and the dissolved oxygen concentration measurement time series.
[0036] For each selected experimental period, the operation is strictly carried out in accordance with the method flow defined in claim 6: First, based on the highest acquisition frequency, the four sets of data are timestamped and missing values are filled by linear interpolation; then, at each aligned time point, the four scalar values are concatenated into a four-dimensional feature vector; finally, all feature vectors are stacked in chronological order to form the standard fused temporal tensor corresponding to the experimental period.
[0037] For each fused time series tensor generated in the previous step, offline analysis is performed using classical control system identification theory methods. Specifically, the least squares parameter estimation method is employed, using the equipment power data in the tensor as the system input sequence and the water temperature and dissolved oxygen data as the system output sequence, to fit and calculate a mathematical model that best explains the input-output relationship. From this model, four steady-state gain parameters characterizing the intensity of the input's influence on the output are extracted: the true temperature control coefficient, the temperature control dissolved oxygen coefficient, the oxygenation temperature coefficient, and the oxygenation dissolved oxygen coefficient. These parameters are then organized into a two-dimensional matrix, serving as the unique true label for this sample.
[0038] The standard fused temporal tensor for each experimental period is paired with the corresponding true coefficient matrix to form a training sample for supervised learning. All such samples are then collected to form the dataset used for model pre-training.
[0039] Build a complete network model that includes a data alignment and splicing layer, a feature attention weighting layer, a coefficient regression analysis layer, and a control coefficient output layer.
[0040] The parameters that need to be learned in the model (mainly concentrated in the feature attention weighting layer and the coefficient regression analysis layer) are scientifically initialized. For example, the "Xavier initialization" method is used to automatically set the initial value range of the weight matrix according to the input and output dimensions of each layer to ensure gradient stability in the early stage of training.
[0041] Configure the key hyperparameters for model training. This includes: setting the optimization algorithm to an adaptive moment estimator and giving it an initial learning rate; determining the batch size of data used for each parameter update; and defining the total number of training iterations that will traverse the entire dataset.
[0042] A batch of fused temporal tensors is extracted from the training dataset and input into the model. The data flows sequentially through each layer: the structure is confirmed by the data alignment and splicing layer; in the feature attention weighting layer, the model automatically learns and assigns higher weights to data from key experimental periods; in the coefficient regression analysis layer, the weighted data undergoes multiple nonlinear transformations, ultimately outputting a predicted two-dimensional coefficient matrix.
[0043] The difference between the coefficient matrix predicted by the calculation model and the corresponding true coefficient matrix in the dataset is used as the loss value for this prediction. The loss function is mainly based on the mean squared error, and an additional physical constraint penalty term can be added. For example, a penalty is imposed for cases where the predicted temperature control coefficient and dissolved oxygen coefficient are negative (which does not conform to physical common sense).
[0044] Automatic differentiation is used to calculate the gradient of the loss value with respect to all trainable parameters in the model. The gradient indicates the magnitude of each parameter's "responsibility" for the prediction error and the direction of adjustment. Subsequently, the optimizer updates the model parameters based on this gradient information, according to the set learning strategy, with the goal of reducing the loss value.
[0045] The process of forward prediction, loss assessment, and backpropagation is repeated until multiple iterations of all training data are completed. Through numerous iterations, the model continuously adjusts its internal parameters, making its output coefficient matrix increasingly approximate the true coefficient matrix obtained by the system identification method in the dataset.
[0046] The trained model is periodically tested using a separate set of historical experimental data (validation set) that was not used in the training. The evaluation metric is the average error between its predicted and true coefficients. When this error no longer decreases significantly in multiple consecutive validations and meets the preset accuracy requirements, the model is considered sufficiently trained, and iteration stops.
[0047] Specifically, such as Figure 2 As shown, the specific steps for extracting the control coefficient set are as follows: In the data alignment and stitching layer of the collaborative analysis model, the historical collaborative control time series dataset is time-aligned and feature-stitched to obtain a fused time series tensor; in the feature attention weighting layer of the collaborative analysis model, the fused time series tensor is subjected to attention weight calculation and weighting to obtain a weighted fused time series tensor; in the coefficient regression analysis layer of the collaborative analysis model, the weighted fused time series tensor is subjected to forward propagation calculation to obtain the control coefficient set; in the control coefficient output layer of the collaborative analysis model, the control coefficient set is output.
[0048] In this implementation scheme, the control coefficient set is accurately extracted through multiple processing layers of the collaborative analysis model, significantly improving the intelligence level of shrimp farming environment adjustment. First, in the data alignment and stitching layer, the historical collaborative control time-series dataset undergoes time alignment and feature stitching processing, ensuring that all data are compared and analyzed within the same time frame, thus avoiding control errors caused by time deviations. Next, through the feature attention weighting layer, the model can automatically identify which data features have a greater impact on the control results, assigning higher weights to key features and improving the accuracy of the analysis. Finally, after forward propagation calculations in the coefficient regression analysis layer, the system... An optimized set of control coefficients was generated, which accurately reflected the regulatory effects of temperature control and aeration devices on the aquatic environment. Finally, the control coefficient output layer outputs this result for use by subsequent execution units, ensuring that the regulation of temperature and dissolved oxygen is more scientific and efficient. It not only improves the automation of data processing, but also ensures the accuracy and real-time performance of each link in the regulation process. Through intelligent analysis, the system can automatically optimize the regulation strategy based on historical data and current environmental conditions, thereby providing more stable and efficient environmental control for the aquaculture farm, reducing the complexity of manual operation, improving the aquaculture effect, and ensuring that shrimp grow in the optimal environment.
[0049] Specifically, the steps to obtain the fused temporal tensor are as follows:
[0050] Historical time-series data on temperature control, oxygenation control, water temperature, and dissolved oxygen concentration were aligned using a unified timestamp. Specifically, based on the highest data acquisition frequency, the four sets of historical time-series data were resampled to the same time point sequence [t1, t2, ..., t]. n For any target time point t i If a set of data has no direct sample value at a given time point, linear interpolation of the sample values from adjacent time points is used to fill the gap, ensuring that all four sets of data have corresponding values at time point tᵢ. For example, water temperature data at time point tᵢ... i No time was recorded, but in t i-1 The temperature was 28.5℃ at time t i+1 If the temperature is 28.7℃, then t is calculated using linear interpolation. i The water temperature at that moment was 28.6℃;
[0051] At each aligned timestamp, the corresponding temperature control power value, oxygenation control power value, temperature measurement value, and dissolved oxygen concentration measurement value are concatenated to obtain the combined feature vector of the corresponding aligned timestamp. Specifically, for each aligned time point t... i Extract the four-dimensional value at that moment: temperature control power value Pt(t) i(Unit: %), Oxygenation control power value Po(t) i (Unit: %), Temperature measurement value T(t) i (Unit: °C), Dissolved oxygen concentration measurement value Do(t) i (Unit: mg / L). These four scalars are concatenated in a fixed order to form a four-dimensional row vector V(t). i )=[Pt(t i ),Po(t i ),T(t i Do(t) i This vector is the combined feature vector at time tᵢ;
[0052] Arrange the combined feature vectors of each aligned timestamp in chronological order and construct a fused temporal tensor, specifically by: arranging the time point sequence [t1, t2, ..., t...] into a combined feature vector. n All corresponding combined feature vectors [V(t1),V(t2),...,V(t)] n The time points are stacked sequentially to form a two-dimensional tensor TS with dimensions n×4, where n represents the total number of time points and 4 represents the dimension of each feature vector. This two-dimensional tensor TS is the fused temporal tensor.
[0053] In this implementation scheme, the four sets of time-series data are first resampled based on the highest data acquisition frequency to ensure that all data have corresponding values at the same time point, avoiding data inconsistencies caused by different data acquisition frequencies. During processing, linear interpolation is used to fill in missing data points, ensuring that each time point reflects the real environmental changes and guaranteeing data continuity and integrity. Then, the temperature control power, oxygenation control power, water temperature, and dissolved oxygen concentration at each time point are concatenated into a feature vector in a fixed order. In this way, the system can comprehensively capture the control effect and water state at each moment. Finally, these feature vectors are stacked in chronological order to form a fused time-series tensor, which facilitates subsequent analysis and processing. This time-series tensor not only provides complete historical environmental control data but also accurately reflects the dynamic changes of each control parameter over time, effectively improving the system's response speed and accuracy to environmental changes, reducing human intervention, and enhancing the adaptive adjustment capability of the aquaculture environment, thus providing a more intelligent and precise control solution for shrimp farming.
[0054] Specifically, the steps to obtain the weighted fusion temporal tensor are as follows:
[0055] For each aligned timestamp combined feature vector in the fused temporal tensor, an attention score is calculated. Specifically, based on the records of step response or impulse response control experiments, the continuous time interval [t] during which the experiment occurs is determined. start ,tend ]; Traverse the time sequence [t1, t2, ..., t n For each time point t i and its corresponding combined feature vector V(t) i ), determine t i Whether it is within the experimental interval; if t i ∈[t start ,t end If α is assigned a high attention score, then it is assigned a high attention score. high (e.g., 1.0), otherwise assign a low attention score α. low (e.g., 0.1); thus, an attention score sequence [α(t1), α(t2), ..., α(t)] corresponding to the time point sequence is obtained. n For example, if the experiment is at t 50 To t 100 If it proceeds, then α(t) 50 ) to α(t 100 All are 1.0, the rest are 0.1;
[0056] Based on the attention score, the combined feature vectors of the corresponding aligned timestamps are weighted, specifically by traversing the time sequence [t1, t2, ..., t...]. n For each time point t i Combine them into feature vector V(t) i )=[Pt(t i ),Po(t i ),T(t i Do(t) i )] and the corresponding attention score α(t) i Perform element-wise scalar multiplication to obtain the weighted combined eigenvector V'(ti)=[α(t)]. i )*Pt(t i ),α(t i )*Po(t i ),α(t i )*T(t i ),α(t i )*Do(t i This operation amplifies the values of data within the experimental period while reducing the influence of data from outside the experimental period.
[0057] Reassemble all the weighted combined feature vectors in their original temporal order to obtain a weighted fused temporal tensor, specifically: [V'(t1), V'(t2), ..., V'(t...]] n According to its original time sequence [t1, t2, ..., t] nThe data are stacked to form a new n×4 dimensional tensor TS with the exact same dimensions as the original fused temporal tensor TS. This TS is the weighted fused temporal tensor, which retains the complete temporal structure, but the data values have been recalibrated.
[0058] In this implementation plan, firstly, high and low attention scores are determined based on the experimental time intervals to ensure that data from key experimental periods are given sufficient attention, while data from non-experimental periods are appropriately downplayed. For example, during the experiment, attention scores at specific time points are set to high values (e.g., 1.0), while other time points are given lower attention scores (e.g., 0.1). This approach ensures that experimental data dominates the analysis, effectively enhancing its representativeness and influence. Subsequently, element-wise scalar multiplication is used to weight the combined feature vectors, ensuring that data from the experimental periods are amplified, thereby strengthening the impact of these data on the final decision. Data from non-experimental periods... The data from the previous segment was weakened due to its low attention score, reducing interference with the results. Finally, all the feature vectors after weighting were stacked in chronological order to form a weighted fusion temporal tensor. This tensor not only preserved the original temporal structure but also calibrated the influence of the data through weighting, ensuring that the system can accurately reflect environmental changes during key periods. It can efficiently identify and focus on experimental data that are crucial to the regulation of the aquaculture environment, thereby providing more accurate and effective data support for subsequent control strategies. Through this weighted analysis, the system can make more precise adaptive adjustments, improve the intelligence level of aquaculture environment regulation, and thus improve aquaculture results and economic benefits.
[0059] Specifically, the steps to obtain the set of control coefficients are as follows:
[0060] The weighted fusion temporal tensor is processed by forward propagation to obtain initial coefficient data. Specifically, the weighted fusion temporal tensor TS is used as the overall input and fed into the coefficient regression analysis layer of the pre-trained co-analysis model. This layer is a feedforward neural network with fixed parameters, which performs standard, feedback-free forward propagation calculations: First, global feature aggregation (such as calculating the mean along the time dimension) is performed on TS to obtain a 1×4 comprehensive feature vector; then, this vector is nonlinearly transformed by several fully connected layers and activation functions; finally, a vector [k1,k2,k3,k4] containing four floating-point numbers is generated in the network output layer, which is the initial coefficient data.
[0061] The initial coefficient data is structured to obtain a structured data set. Specifically, the four elements k1, k2, k3, and k4 in the initial coefficient data are assigned explicit physical definitions and encapsulated: Let... =k1, used to characterize the intensity of the effect of the temperature control device on the water temperature; let =k2, used to characterize the intensity of the effect of the temperature control device on the dissolved oxygen concentration; let =k3, used to characterize the intensity of the aeration device's influence on water temperature; let =k4, used to characterize the intensity of the effect of the aeration device on the dissolved oxygen concentration. Subsequently, these four coefficients are organized into a preset 2×2 matrix structure as [[ , ],[ , This matrix is a structured data set;
[0062] The structured dataset is encapsulated to obtain a set of control coefficients, specifically by: encapsulating the output of the data. , , , The four coefficients form a 2×2 matrix, encapsulated in a standard data format agreed upon internally by the system. For example, create a dictionary object with key-value pairs like {"temperature control water temperature coefficient":} "Temperature-controlled dissolved oxygen coefficient": "Oxygenation water temperature coefficient": "Oxygenation and dissolved oxygen coefficient": Alternatively, it can be serialized into a JSON object containing these four fields. The resulting data packet, containing four coefficients with clearly defined physical meanings, is the final output set of control coefficients.
[0063] In this implementation scheme, a weighted fusion time series tensor is first used as input. After passing through the coefficient regression analysis layer of the collaborative analysis model, standard forward propagation calculations are performed. This process generates initial coefficient data through global feature aggregation and multi-layer nonlinear transformation. This initial coefficient data accurately reflects the influence of temperature control and aeration devices on water temperature and dissolved oxygen concentration in the aquaculture water environment. Through structured processing of the initial coefficient data, each coefficient is assigned a clear physical meaning, representing the influence of temperature control on water temperature and dissolved oxygen, and the intensity of the influence of aeration devices on water temperature and dissolved oxygen, respectively. These coefficients are organized in a preset matrix format, making the data more... With clear physical and structural representations, it facilitates subsequent analysis and processing. Finally, by encapsulating the structured data, a standard data format (such as a dictionary object or JSON) is generated, ensuring that the system can easily use these control coefficients for subsequent adjustments and optimizations. This method not only extracts accurate control coefficients from historical data, but also enables these coefficients to be efficiently and accurately transmitted and applied within the system through structuring and encapsulation. In this way, the system can automatically adjust its control strategy according to environmental changes, improving the precision and response speed of the aquaculture environment regulation, thereby ensuring the healthy growth of shrimp farming and improving farming efficiency and management efficiency.
[0064] Specifically, the steps for analyzing the temperature and oxygenation co-control quantities of shrimp farming water are as follows: Perform a difference analysis on the current values of water temperature and dissolved oxygen concentration in the shrimp farming water and the preset target values to obtain the target deviation values for water temperature (preset target value minus the current value) and dissolved oxygen concentration (preset target value minus the current value). Combine these target deviation values with the control coefficient set for comprehensive analysis to obtain the temperature and oxygenation co-control quantities of the shrimp farming water, which are obtained by solving the following co-control matrix equation: ;in, , , , The values, in order, are: temperature-coordinated control parameters, oxygenation-coordinated control parameters, target deviation values for water temperature, and target deviation values for dissolved oxygen concentration in shrimp farming water. , , , In order, they are: temperature control water temperature coefficient, temperature control dissolved oxygen coefficient, oxygenation water temperature coefficient, and oxygenation dissolved oxygen coefficient;
[0065] The calculation formulas for temperature-coordinated control and oxygenation-coordinated control are as follows: .
[0066] In this implementation plan, the difference between the actual values of the current water temperature and dissolved oxygen concentration and the preset target values is first calculated to obtain the target deviation value. These deviation values directly reflect the difference between the current environment and the target environment and are the basic data for adjusting the environment. Then, by combining the deviation value with the control coefficient set, a comprehensive analysis is performed using a collaborative control matrix equation to finally obtain the collaborative control quantity of temperature and oxygenation. This control quantity provides a scientific basis for subsequent equipment adjustment, ensuring that the temperature control and oxygenation devices can be accurately adjusted according to the real-time needs of the environment during execution. The calculation formulas for the temperature collaborative control quantity and the oxygenation collaborative control quantity clearly link each control coefficient with the environmental deviation, enabling each control step to be dynamically optimized based on actual needs. This method not only ensures that each adjustment is accurately calculated based on the difference between the current water state and the target value, but also optimizes the intelligence and accuracy of the control process through a systematic formula. Through this adaptive adjustment strategy, the system can efficiently respond to changes in the aquatic environment, ensuring that shrimp farming grows in the optimal environment, improving farming efficiency and reducing risks.
[0067] Please see Figure 3This invention provides a technical solution: an adaptive adjustment system for shrimp farming environment, comprising: a historical data acquisition unit, used to synchronously acquire historical collaborative control time-series datasets generated by a preset temperature control device and an oxygenation device in a shrimp farming water body through step response or impulse response control experiments based on a set time period, including historical temperature control time-series data, historical oxygenation control time-series data, historical water temperature time-series data, and historical dissolved oxygen concentration time-series data; a collaborative analysis unit, used to input the historical collaborative control time-series datasets into a pre-trained collaborative analysis model to extract a set of control coefficients; a collaborative control analysis unit, used to acquire the current values of water temperature and dissolved oxygen concentration in the shrimp farming water body and preset target values, and analyze the temperature collaborative control quantity and oxygenation collaborative control quantity of the shrimp farming water body in combination with the control coefficient set; and a control execution unit, used to drive the preset temperature control device and oxygenation device to adjust the shrimp farming water body according to the temperature collaborative control quantity and oxygenation collaborative control quantity.
[0068] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.
[0069] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A method for adaptive adjustment of the shrimp farming environment, characterized in that, Includes the following steps: Based on a set time period, the historical collaborative control time series dataset of the preset temperature control device and oxygenation device in the shrimp farming water body is synchronously acquired, including historical temperature control time series data, historical oxygenation control time series data, historical water temperature time series data, and historical dissolved oxygen concentration time series data. The historical collaborative control time series dataset is input into a pre-trained collaborative analysis model to extract the control coefficient set; The current values of water temperature and dissolved oxygen concentration in shrimp farming water and the preset target values are obtained. Combined with the control coefficient set, the temperature-coordinated control quantity and oxygenation-coordinated control quantity of shrimp farming water are analyzed. Based on the temperature and oxygenation control parameters, the preset temperature control and oxygenation devices are driven to adjust and treat the shrimp farming water.
2. The method for adaptive adjustment of shrimp farming environment according to claim 1, characterized in that, Historical temperature control time series data includes temperature control power values of temperature control devices at several historical time points; historical oxygenation control time series data includes oxygenation control power values of oxygenation devices at several historical time points; historical water temperature time series data includes temperature measurement values of shrimp farming water bodies at several historical time points; and historical dissolved oxygen concentration time series data includes dissolved oxygen concentration measurement values of shrimp farming water bodies at several historical time points.
3. The method for adaptive adjustment of shrimp farming environment according to claim 1, characterized in that, The set of control coefficients includes temperature-controlled water temperature coefficient, temperature-controlled dissolved oxygen coefficient, oxygen-enriched water temperature coefficient, and oxygen-enriched dissolved oxygen coefficient.
4. The method for adaptive adjustment of shrimp farming environment according to claim 1, characterized in that, The collaborative analysis model includes a data alignment and splicing layer, a feature attention weighting layer, a coefficient regression analysis layer, and a control coefficient output layer.
5. The method for adaptive adjustment of shrimp farming environment according to claim 4, characterized in that, The specific steps for extracting the control coefficient set are as follows: In the data alignment and stitching layer of the collaborative analysis model, the historical collaborative control time series dataset is processed by time alignment and feature stitching to obtain a fused time series tensor. In the feature attention weighting layer of the collaborative analysis model, attention weights are calculated and weighted on the fusion temporal tensor to obtain the weighted fusion temporal tensor. In the coefficient regression analysis layer of the collaborative analysis model, forward propagation calculation is performed on the weighted fusion time series tensor to obtain the control coefficient set; In the control coefficient output layer of the collaborative analysis model, the control coefficient set is processed for output.
6. The method for adaptive adjustment of shrimp farming environment according to claim 5, characterized in that, The specific steps to obtain the fused temporal tensor are as follows: The historical temperature control time series data, historical oxygenation control time series data, historical water temperature time series data, and historical dissolved oxygen concentration time series data are aligned based on a unified timestamp. At each aligned timestamp, the corresponding temperature control power value, oxygenation control power value, temperature measurement value, and dissolved oxygen concentration measurement value are concatenated to obtain the combined feature vector of the corresponding aligned timestamp; Arrange the combined feature vectors of each aligned timestamp in chronological order and construct a fused temporal tensor.
7. The method for adaptive adjustment of shrimp farming environment according to claim 6, characterized in that, The specific steps to obtain the weighted fusion temporal tensor are as follows: Attention scores are calculated for the combined feature vectors of each aligned timestamp in the fused temporal tensor. Based on the attention score, the combined feature vectors of the corresponding aligned timestamps are weighted. All the combined feature vectors after weighting are recombined in their original time order to obtain the weighted fused temporal tensor.
8. The method for adaptive adjustment of shrimp farming environment according to claim 7, characterized in that, The specific steps to obtain the control coefficient set are as follows: The weighted fusion temporal tensor is processed by forward propagation to obtain initial coefficient data; The initial coefficient data is processed into a structured dataset. The structured data set is encapsulated to obtain a set of control coefficients.
9. The method for adaptive adjustment of shrimp farming environment according to claim 1, characterized in that, The specific steps for analyzing the synergistic control parameters of temperature and oxygenation in shrimp farming water are as follows: The difference between the current values of water temperature and dissolved oxygen concentration in the shrimp farming water and the preset target values is analyzed to obtain the target deviation values of water temperature and dissolved oxygen concentration in the shrimp farming water. By combining the target deviation values of water temperature and dissolved oxygen concentration in shrimp farming water with the control coefficient set for comprehensive analysis, the temperature-coordinated control quantity and oxygenation-coordinated control quantity of shrimp farming water are obtained.
10. A shrimp farming environment adaptive adjustment system, employing the shrimp farming environment adaptive adjustment method according to any one of claims 1-9, characterized in that, include: The historical data acquisition unit is used to synchronously acquire a set historical collaborative control time series dataset of the temperature control device and the oxygenation device in the shrimp farming water body based on a set time period, including historical temperature control time series data, historical oxygenation control time series data, historical water temperature time series data, and historical dissolved oxygen concentration time series data. The collaborative analysis unit is used to input historical collaborative control time series datasets into a pre-trained collaborative analysis model to extract the control coefficient set; The collaborative control analysis unit is used to obtain the current values of water temperature and dissolved oxygen concentration in shrimp farming water and the preset target values, and to analyze the temperature collaborative control quantity and oxygenation collaborative control quantity of shrimp farming water in combination with the control coefficient set. The control execution unit is used to drive the preset temperature control device and oxygenation device to adjust and treat the shrimp farming water body according to the temperature coordination control amount and the oxygenation coordination control amount.