Intelligent aquaculture management system and method integrating water quality monitoring

Abstract

The invention discloses an intelligent aquaculture management system and method integrating water quality monitoring, and relates to the field of aquaculture, and the method comprises the steps: building a three-dimensional monitoring network through the deployment of a multi-parameter sensor array, and completing the data cleaning and fusion through edge calculation; in combination with the breeding variety growth feature library and the dynamic threshold model, water quality state evaluation and 24-hour risk prediction are realized by adopting an LSTM-random forest fusion algorithm; an optimization regulation and control instruction is generated based on the three-level strategy library, and the execution equipment cluster is driven to cooperatively work and form closed-loop control; and finally, establishing a full-life-cycle data file, continuously optimizing the model through transfer learning and outputting a management report. The method has the advantages that through three-dimensional accurate monitoring and intelligent model analysis and pre-judgment, the precision and intelligence of breeding management are realized by combining personalized dynamic regulation and control with equipment collaborative linkage, and the effect can be continuously optimized through data iteration.

Description

Technical Field

[0001] This invention relates to the field of aquaculture, and in particular to an intelligent aquaculture management system and method integrating water quality monitoring. Background Technology

[0002] Aquaculture management methods are gradually shifting towards factory-style recirculating aquaculture systems (RAS) to address the problems of high resource consumption, severe environmental pollution, and high disease risks associated with traditional farming models. This model relies on a closed-loop recirculating aquaculture system, achieving water recycling through physical filtration, biological purification, and sterilization, significantly reducing water waste and wastewater discharge. The application of intelligent technologies is key, including real-time water quality monitoring via IoT sensors, automatic feed and water change frequency adjustment via AI algorithms, and big data prediction of disease risks linked to precise medication, thereby improving stocking density and survival rates.

[0003] Most current aquaculture management methods employ single-point monitoring, resulting in non-standard sensor calibration, data susceptibility to interference, and asynchronous data transmission, hindering comprehensive and accurate monitoring across the entire aquatic area. Data processing largely relies on cloud computing, leading to high latency, crude outlier filtering, and feature extraction, making it difficult to create standardized, integrated datasets. Analysis models are mostly based on single algorithms, failing to deeply explore the temporal correlations and synergistic effects of parameters, lacking dynamic threshold systems, and only providing static assessments of water quality, thus failing to accurately predict future risks. Control strategies are rigid, lacking tiered response mechanisms, making them difficult to adapt to different aquaculture scenarios and growth stages. Furthermore, equipment is mostly decentralized, lacking collaborative linkage and closed-loop feedback, resulting in poor control effectiveness and high energy consumption. In addition, the lack of a full lifecycle data archive makes it difficult to iteratively optimize models through data, leading to a crude and poorly adaptable management approach. Summary of the Invention

[0004] To improve existing systems and methods, this paper provides an intelligent aquaculture management system and method that integrates water quality monitoring. This method constructs a closed-loop intelligent system covering the entire process of "monitoring-analysis-control-optimization". Through three-dimensional precise monitoring, intelligent model analysis and prediction, combined with personalized dynamic control and equipment collaborative linkage, it achieves precise and intelligent aquaculture management and can continuously optimize the effect through data iteration.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A smart aquaculture management method integrating water quality monitoring includes:

[0007] Multi-parameter water quality and environmental sensor arrays are deployed in various functional areas of the aquaculture waters. Through standard solution calibration and on-site blank control, a time-series synchronization mechanism is established to build a three-dimensional monitoring network covering the entire water area and its surrounding environment.

[0008] The monitoring data is transmitted to the edge computing node, where data standardization and parsing, sliding window noise removal, time-series alignment and compression are performed sequentially to obtain a standardized water quality-environment fusion dataset.

[0009] Construct a database of growth characteristics throughout the entire life cycle of aquaculture species, integrate growth data with historical cases, establish a correlation model between growth stages and water quality parameters, dynamically adjust thresholds based on actual stocking density and feeding amount, and generate a personalized dynamic water quality threshold system.

[0010] The standardized dataset is input into the LSTM and random forest fusion model to explore the temporal correlation and synergistic effect of parameters. The current water quality status is assessed by combining dynamic thresholds, and the water quality change and risk level are predicted in the next 24 hours through the trend prediction sub-model to obtain the risk time window.

[0011] Based on the assessment and prediction results, the system calls upon a three-level control strategy library containing daily, early warning, and emergency measures to generate corresponding control measures for different water quality states. The system optimizes parameters through fuzzy control to balance control effectiveness and energy consumption.

[0012] The control strategy is converted into standard control commands and transmitted to the cluster of execution equipment. A collaborative linkage mechanism is established to coordinate equipment operation, and equipment status feedback is collected in real time to form a closed-loop control.

[0013] All types of aquaculture data are stored in a cloud-based full lifecycle archive. Historical data is regularly mined and analyzed, and the association and analysis models are iteratively optimized through transfer learning to generate management reports.

[0014] Preferably, the deployment of multi-parameter water quality and environmental sensor arrays in various functional zones of the aquaculture area, unified calibration through standard solution calibration and on-site blank control, and the establishment of a time-series synchronization mechanism to construct a three-dimensional monitoring network covering the entire aquaculture area and its surrounding environment specifically includes:

[0015] An integrated multi-parameter water quality sensor array is deployed in different functional areas of aquaculture waters. The sensor array integrates at least a pH sensor, a dissolved oxygen sensor, an ammonia nitrogen sensor, a nitrite sensor, a water temperature sensor, a turbidity sensor, and a salinity sensor.

[0016] By combining standard solution calibration with on-site blank control, individual sensor errors and environmental interference are eliminated, a sensor data acquisition timing synchronization mechanism is established, and a three-dimensional water quality and environmental monitoring network covering the entire aquaculture area and its surrounding environment is formed.

[0017] Preferably, the step of transmitting the monitoring data to the edge computing node and sequentially performing data standardization parsing, sliding window noise removal, time-series alignment and compression to obtain a standardized water quality-environment fusion dataset specifically includes:

[0018] The water quality and environmental data collected by the sensor array are transmitted to the edge computing node in real time. The edge computing node performs preprocessing operations on the received data, including:

[0019] The data is parsed and standardized to convert heterogeneous data from different sensors into a preset data format.

[0020] The sliding window algorithm is used to remove outliers and noise from the data. By judging the gradient threshold of adjacent data points and correcting the mean, abnormal fluctuation data caused by sensor failure and water flow impact are filtered out.

[0021] The preprocessed data is time-series aligned and compressed to retain key feature parameters, thereby obtaining a standardized water quality-environment fusion dataset.

[0022] Preferably, the construction of a database of growth characteristics throughout the entire life cycle of aquaculture species, the integration of growth data and historical cases, the establishment of a growth stage-water quality parameter correlation model, and the dynamic adjustment of thresholds based on actual stocking density and feeding amount to generate a personalized dynamic water quality threshold system specifically include:

[0023] A database of growth characteristics throughout the entire life cycle of aquaculture species is constructed. The database includes physiological and metabolic parameters, feeding patterns, and suitable living environment parameters of the target aquaculture species at different growth stages.

[0024] By integrating growth characteristic data of aquaculture species with historical high-quality aquaculture case data through deep learning algorithms, a correlation model between growth stage and water quality parameters is established. The model inputs are the growth days and weight monitoring data of aquaculture species, and the outputs are the suitable threshold range, warning threshold range and dangerous threshold range of each water quality parameter under the corresponding growth stage.

[0025] Based on the actual stocking density, feeding amount, and water temperature change trends at the aquaculture site, the threshold range output by the model is dynamically corrected to generate a personalized dynamic water quality threshold system adapted to the current aquaculture scenario.

[0026] Preferably, the step of inputting the standardized dataset into the LSTM and random forest fusion model, mining the temporal correlation and synergistic effect of parameters, assessing the current water quality status in combination with dynamic thresholds, and predicting water quality changes and risk levels in the next 24 hours through a trend prediction sub-model to obtain the risk time window specifically includes:

[0027] The obtained standardized fusion dataset is input into a preset intelligent analysis model, which is a fusion model based on LSTM and random forest.

[0028] The LSTM network is used to mine the time series correlation features between water quality parameters and environmental parameters, and the random forest model is used to analyze the impact of the synergistic effect of multiple parameters on the aquaculture environment.

[0029] By combining a personalized dynamic water quality threshold system, the current water quality status is assessed in real time to determine whether it is in a suitable, warning, or dangerous state.

[0030] The trend prediction sub-model, trained based on historical data, predicts the changing trends of various water quality parameters in the next 24 hours, identifies potential risks of water quality deterioration, and outputs the risk level and the predicted time window for the risk to occur.

[0031] Preferably, the step of calling a three-level control strategy library (daily, early warning, and emergency) based on the assessment and prediction results, generating corresponding control measures to adapt to different water quality states, and optimizing parameters through fuzzy control to balance control effectiveness and energy consumption specifically includes:

[0032] Based on the water quality assessment results and risk prediction conclusions, a multi-level intelligent control strategy library is constructed, which includes three levels: daily maintenance control, early warning response control, and emergency response control.

[0033] When the water quality is suitable, only periodic routine operation control is performed on the aeration equipment and feeding equipment;

[0034] When the water quality is detected to be close to the warning threshold or a medium-to-low level risk is predicted, preventive control measures such as adjusting the power of the aeration equipment, changing a small amount of water, and optimizing the amount of feed are initiated.

[0035] When the water quality reaches the danger threshold or a high level of risk is predicted, the following measures are triggered: full-power operation of the oxygenation equipment, activation of the emergency water exchange system, and suspension of feeding.

[0036] Preferably, the control strategy is converted into standard control commands, transmitted to the cluster of execution equipment, a collaborative linkage mechanism is established to coordinate equipment operation, and equipment status feedback is collected in real time to form a closed-loop control, specifically including:

[0037] The generated intelligent control strategy is converted into standardized control commands and transmitted to the cluster of execution equipment at the breeding site;

[0038] The cluster of execution equipment includes intelligent aerators, automatic water exchange devices, precision feeders, and water quality improver dispensing equipment;

[0039] Establish a collaborative linkage mechanism for equipment, coordinate the running sequence and parameters of each device according to the priority and execution logic of the control strategy, collect the running status data of each executing device in real time during the control execution process, and feed it back to the edge computing node to form a closed-loop control of control command - device execution - status feedback.

[0040] Preferably, the step of storing various types of aquaculture data in a cloud-based full lifecycle archive, periodically mining and analyzing historical data, and iteratively optimizing the association and analysis models through transfer learning to generate management reports specifically includes:

[0041] The pre-treated water quality data, environmental data, control strategy data, equipment operation data, and monitoring data of the growth status of aquaculture species are stored in a cloud database to build a data archive of the entire life cycle of aquaculture.

[0042] Regularly mine and analyze historical data in the cloud database to extract water quality change patterns, control strategy effects, and equipment operating efficiency characteristics under different aquaculture scenarios;

[0043] Based on the data mining results, the transfer learning algorithm was used to iteratively optimize the growth stage-water quality parameter correlation model and the intelligent analysis model, update the model parameters and threshold system, and generate an aquaculture management report based on historical data.

[0044] Furthermore, an intelligent aquaculture management system integrating water quality monitoring is proposed, including:

[0045] Multi-parameter sensing and monitoring module: Deploy a multi-parameter water quality sensor array, and build a three-dimensional monitoring network through calibration and time synchronization to collect water quality and environmental data in real time;

[0046] Edge computing module: Performs data standardization, outlier filtering, temporal alignment and compression at edge nodes to generate a standardized fused dataset;

[0047] Intelligent Analysis and Risk Assessment Module: Based on the LSTM-random forest fusion model, it analyzes the time-series correlation of parameters, combines the dynamic threshold system to assess the water quality status, and predicts the risk level and time window for the next 24 hours;

[0048] Strategy generation module: integrates the growth stage-water quality correlation model, dynamically adjusts thresholds based on stocking density and feeding amount, and calls the three-level strategy library to generate optimized control instructions;

[0049] Equipment coordination control module: converts commands into control signals, links aerators and feeders, and provides real-time feedback on operating status;

[0050] Cloud-based data management module: Stores full lifecycle data, periodically mines historical patterns, optimizes model parameters through transfer learning, and generates management reports;

[0051] Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0052] Compared with the prior art, the advantages of the present invention are:

[0053] A closed-loop intelligent system encompassing monitoring, analysis, regulation, and optimization has been constructed. Through a multi-parameter sensor array and a three-dimensional monitoring network, combined with dual calibration and time-series synchronization mechanisms, the accuracy and comprehensiveness of water quality and environmental data collection are ensured. Edge computing enables efficient data preprocessing, and a fusion model of LSTM and random forest accurately identifies parameter correlations and predicts 24-hour water quality risks, providing a scientific basis for regulation. A dynamic threshold system and three-level regulation strategies based on the full life-cycle characteristics of aquaculture species can adapt to different water quality conditions to generate personalized optimization measures, balancing regulation effectiveness and energy consumption. Equipment collaboration and closed-loop control ensure the implementation of measures, while cloud-based full life-cycle archives and transfer learning iteration mechanisms continuously optimize model performance, achieving precise, intelligent, and long-term aquaculture management. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the method proposed in this invention;

[0055] Figure 2 This is a schematic diagram of the integrated water quality monitoring network proposed in this invention;

[0056] Figure 3 This is a schematic diagram of the monitoring data transmission and preprocessing proposed in this invention;

[0057] Figure 4 This is a schematic diagram illustrating the dynamic construction of water quality thresholds proposed in this invention;

[0058] Figure 5 This is a schematic diagram of the water quality-environment collaborative intelligent analysis proposed in this invention;

[0059] Figure 6 This is a schematic diagram illustrating the dynamic generation of the multi-level control strategy proposed in this invention;

[0060] Figure 7 This is a schematic diagram of the control execution and multi-device collaborative linkage proposed in this invention;

[0061] Figure 8 This is a schematic diagram of the full lifecycle management and model optimization proposed in this invention. Detailed Implementation

[0062] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0063] An intelligent aquaculture management system integrating water quality monitoring includes:

[0064] Multi-parameter sensing and monitoring module: Deploy a multi-parameter water quality sensor array, and build a three-dimensional monitoring network through calibration and time synchronization to collect water quality and environmental data in real time;

[0065] Edge computing module: Performs data standardization, outlier filtering, temporal alignment and compression at edge nodes to generate a standardized fused dataset;

[0066] Intelligent Analysis and Risk Assessment Module: Based on the LSTM-random forest fusion model, it analyzes the time-series correlation of parameters, combines the dynamic threshold system to assess the water quality status, and predicts the risk level and time window for the next 24 hours;

[0067] Strategy generation module: integrates the growth stage-water quality correlation model, dynamically adjusts thresholds based on stocking density and feeding amount, and calls the three-level strategy library to generate optimized control instructions;

[0068] Equipment coordination control module: converts commands into control signals, links aerators and feeders, and provides real-time feedback on operating status;

[0069] Cloud-based data management module: Stores full lifecycle data, periodically mines historical patterns, optimizes model parameters through transfer learning, and generates management reports;

[0070] Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0071] See Figure 1 As shown, an intelligent aquaculture management method integrating water quality monitoring includes:

[0072] Step 1: Deploy multi-parameter water quality and environmental sensor arrays in each functional area of ​​the aquaculture water area, and establish a time-series synchronization mechanism through standard solution calibration and on-site blank control to build a three-dimensional monitoring network covering the entire water area and surrounding environment.

[0073] Step 2: Transmit the monitoring data to the edge computing node, and sequentially complete the data standardization and parsing, sliding window removal of abnormal noise, time-series alignment and compression to obtain a standardized water quality-environment fusion dataset;

[0074] Step 3: Construct a database of growth characteristics throughout the entire life cycle of aquaculture species, integrate growth data with historical cases, establish a correlation model between growth stages and water quality parameters, dynamically adjust thresholds based on actual stocking density and feeding amount, and generate a personalized dynamic water quality threshold system.

[0075] Step 4: Input the standardized dataset into the LSTM and random forest fusion model to explore the temporal correlation and synergistic effect of parameters, combine dynamic thresholds to assess the current water quality status, and use the trend prediction sub-model to predict the water quality changes and risk level in the next 24 hours to obtain the risk time window;

[0076] Step 5: Based on the assessment and prediction results, call the three-level control strategy library containing daily, early warning and emergency measures, adapt to different water quality states to generate corresponding control measures, optimize parameters through fuzzy control, and balance control effect and energy consumption;

[0077] Step Six: Convert the control strategy into standard control commands, transmit them to the cluster of execution equipment, establish a collaborative linkage mechanism to coordinate equipment operation, collect equipment status feedback in real time, and form a closed-loop control.

[0078] Step 7: Store all types of aquaculture data in a cloud-based full lifecycle archive, regularly mine and analyze historical data, iteratively optimize the association and analysis models through transfer learning, and generate management reports.

[0079] See Figure 2 As shown, multi-parameter water quality and environmental sensor arrays are deployed in various functional zones of the aquaculture area. Through standard solution calibration and on-site blank control, a time-series synchronization mechanism is established to construct a three-dimensional monitoring network covering the entire aquaculture area and its surrounding environment. Specifically, this includes:

[0080] An integrated multi-parameter water quality sensor array is deployed in different functional areas of aquaculture waters. The sensor array integrates at least a pH sensor, a dissolved oxygen sensor, an ammonia nitrogen sensor, a nitrite sensor, a water temperature sensor, a turbidity sensor, and a salinity sensor.

[0081] By combining standard solution calibration with on-site blank control, individual sensor errors and environmental interference are eliminated, a sensor data acquisition timing synchronization mechanism is established, and a three-dimensional water quality and environmental monitoring network covering the entire aquaculture area and its surrounding environment is formed.

[0082] See Figure 3 As shown, the monitoring data is transmitted to the edge computing node, where data standardization and parsing, sliding window noise removal, time-series alignment and compression are performed sequentially to obtain a standardized water quality-environment fusion dataset. Specifically, this includes:

[0083] The water quality and environmental data collected by the sensor array are transmitted to the edge computing node in real time. The edge computing node performs preprocessing operations on the received data, including:

[0084] The data is parsed and standardized to convert heterogeneous data from different sensors into a preset data format.

[0085] The sliding window algorithm is used to remove outliers and noise from the data. By judging the gradient threshold of adjacent data points and correcting the mean, abnormal fluctuation data caused by sensor failure and water flow impact are filtered out.

[0086] The preprocessed data is time-series aligned and compressed to retain key feature parameters, thereby obtaining a standardized water quality-environment fusion dataset.

[0087] Specifically, edge computing nodes parse the received heterogeneous data frame by frame, extracting core fields such as the original measurement values, acquisition time, and device status codes of each sensor; based on preset industry data standards, a standardized data template is constructed to uniformly map the heterogeneous data from different sensors into a six-tuple format of "sensor ID-deployment location-acquisition time-water quality parameter type-measurement value-device status"; for environmental sensor data, additional unit conversion operations for light intensity and wind speed are performed to ensure the unit consistency between water quality data and environmental data; at the same time, data with missing fields are marked to distinguish between "signal interruption missing" and "device failure missing";

[0088] A two-stage filtering mechanism of "coarse screening-fine screening" is adopted to process data noise and outliers. In the coarse screening stage, a sliding window algorithm is used, with the window size set to 10 collection cycles. The mean and standard deviation of the data within the window are calculated, and data exceeding the range of "mean ± 3 times standard deviation" are marked as suspected anomalies. In the fine screening stage, secondary verification is carried out in combination with the characteristics of the aquaculture scenario. If the suspected anomaly data is a parameter with strong stability such as dissolved oxygen and water temperature, it is determined whether it is caused by local water flow impact or sensor failure by comparing the synchronous data of three adjacent deployment points. If only a single point is abnormal, it is judged as invalid data and removed. If multiple points are synchronously abnormal, it is retained and marked as "environmental change data". The mean correction method is used to fill the gaps in the data after filtering.

[0089] See Figure 4 As shown, a database of the full life-cycle growth characteristics of aquaculture species is constructed. This database integrates growth data with historical cases to establish a correlation model between growth stages and water quality parameters. Thresholds are dynamically adjusted based on actual stocking density and feeding amounts to generate a personalized dynamic water quality threshold system. Specifically, this includes:

[0090] A database of growth characteristics throughout the entire life cycle of aquaculture species is constructed. The database includes physiological and metabolic parameters, feeding patterns, and suitable living environment parameters of the target aquaculture species at different growth stages.

[0091] By integrating growth characteristic data of aquaculture species with historical high-quality aquaculture case data through deep learning algorithms, a correlation model between growth stage and water quality parameters is established. The model inputs are the growth days and weight monitoring data of aquaculture species, and the outputs are the suitable threshold range, warning threshold range and dangerous threshold range of each water quality parameter under the corresponding growth stage.

[0092] Based on the actual stocking density, feeding amount, and water temperature change trends at the aquaculture site, the threshold range output by the model is dynamically corrected to generate a personalized dynamic water quality threshold system adapted to the current aquaculture scenario.

[0093] Specifically, based on the constructed growth feature database, a correlation model is built by integrating deep learning and traditional statistical analysis algorithms; statistical analysis is used to screen out water quality parameters that have a significant impact on each growth stage; then, deep learning algorithms are used to mine the nonlinear temporal correlation between growth parameters and water quality parameters. The input layer consists of growth days, weight monitoring data, and historical water quality data, while the hidden layer learns feature interaction relationships through a multilayer perceptron. The output layer is the basic threshold range of water quality parameters for the corresponding growth stage. During model training, the training set and validation set are divided in a 7:3 ratio, and the model structure is iteratively adjusted using the validation set.

[0094] The basic thresholds output by the model are dynamically corrected based on real-time conditions at the aquaculture site. For example, when the stocking density is 15% higher than the normal value, the warning thresholds for ammonia nitrogen and nitrite are appropriately lowered; when the water temperature rises by more than 2°C for three consecutive days, the lower limit of the suitable threshold for dissolved oxygen is raised. The correction effect is verified through small-scale field experiments, and the feeding and growth status of aquaculture species under different threshold conditions are recorded. Finally, a personalized threshold system adapted to the current aquaculture scenario is determined, which is divided into three levels: suitable, warning, and dangerous, and the priority of the control response corresponding to each level is clarified.

[0095] The formula for the ammonia nitrogen warning threshold adjusted by stocking density is:

[0096]

[0097] in, The revised ammonia nitrogen warning threshold, This is the basic warning threshold for ammonia nitrogen output by the model. This is a correction factor for stocking density. This represents the percentage of actual stocking density that exceeds the standard density.

[0098] See Figure 5 As shown, a standardized dataset is input into an LSTM and random forest fusion model to mine the temporal correlations and synergistic effects of parameters. Combined with dynamic thresholds, the current water quality status is assessed. A trend prediction sub-model is used to predict water quality changes and risk levels over the next 24 hours. The specific risk time window includes:

[0099] The obtained standardized fusion dataset is input into a preset intelligent analysis model, which is a fusion model based on LSTM and random forest.

[0100] The LSTM network is used to mine the time series correlation features between water quality parameters and environmental parameters, and the random forest model is used to analyze the impact of the synergistic effect of multiple parameters on the aquaculture environment.

[0101] By combining a personalized dynamic water quality threshold system, the current water quality status is assessed in real time to determine whether it is in a suitable, warning, or dangerous state.

[0102] The trend prediction sub-model, trained based on historical data, predicts the changing trends of various water quality parameters in the next 24 hours, identifies potential risks of water quality deterioration, and outputs the risk level and the predicted time window for the risk to occur.

[0103] Specifically, the LSTM network layer targets the time-series characteristics of water quality-environment data to mine short-term fluctuation patterns and long-term trends of core parameters such as dissolved oxygen and ammonia nitrogen, capturing precursor features of parameter mutations; the random forest layer focuses on spatial correlation characteristics and multi-parameter coupling effects, analyzing parameter differences in different aquaculture areas, as well as the synergistic influence of water temperature, light, and dissolved oxygen, identifying the risk coupling pattern of "sudden rise in water temperature + decrease in dissolved oxygen"; combined with the constructed personalized dynamic water quality threshold system, the real-time analysis results are compared with the suitable / warning / danger thresholds for the corresponding growth stage, outputting refined water quality status assessment conclusions, clarifying the level, degree of exceedance, and scope of influence of each parameter, while recording the assessment process data;

[0104] The model's built-in trend prediction sub-model is invoked to predict the change trajectory of various water quality parameters in the next 24 hours based on historical 24-hour standardized data and real-time analysis features, generating parameter change trend curves. Combining the feature library of historical water quality deterioration cases in the full life cycle data archive, the similarity between the current predicted trend and historical risk cases is compared to determine the type of potential risk, such as the risk of a sudden drop in dissolved oxygen, the risk of excessive accumulation of ammonia nitrogen, and the risk of abnormal increase in turbidity. The risk level is divided according to the probability of occurrence and the degree of impact, the prediction time window for the risk is defined, and the basis for prediction is marked.

[0105] See Figure 6 As shown, based on the assessment and prediction results, a three-tiered control strategy library containing daily, early warning, and emergency measures is invoked to generate corresponding control measures adapted to different water quality states. Parameters are optimized through fuzzy control to balance control effectiveness and energy consumption. Specifically, this includes:

[0106] Based on the water quality assessment results and risk prediction conclusions, a multi-level intelligent control strategy library is constructed, which includes three levels: daily maintenance control, early warning response control, and emergency response control.

[0107] When the water quality is suitable, only periodic routine operation control is performed on the aeration equipment and feeding equipment;

[0108] When the water quality is detected to be close to the warning threshold or a medium-to-low level risk is predicted, preventive control measures such as adjusting the power of the aeration equipment, changing a small amount of water, and optimizing the amount of feed are initiated.

[0109] When the water quality reaches the danger threshold or a high level of risk is predicted, the following measures are triggered: full-power operation of the oxygenation equipment, activation of the emergency water exchange system, and suspension of feeding.

[0110] Specifically, the daily maintenance and control strategy targets suitable water quality conditions, pre-setting periodic operating parameters for aerators, precise feeding amounts for feeders, and a daily surface water exchange plan according to the growth stages of aquaculture species to maintain stable water quality. The early warning response control strategy targets low to medium-risk levels, initiating preventative parameter adjustments. For example, when dissolved oxygen approaches the warning threshold, aerator power is increased by 20%-30%, and when ammonia nitrogen accumulation shows a clear trend, feeding is reduced by 15% and intermittent water exchanges are initiated, with a small amount of water quality improver added simultaneously. The emergency response control strategy targets high-risk levels, triggering mandatory measures such as full-power aeration, emergency water exchanges, and suspension of feeding, while generating targeted treatment suggestions. All strategy parameters are optimized through fuzzy control algorithms to balance control effectiveness with energy consumption and cost, and parameter adjustment records are archived in real time to the full life cycle database.

[0111] See Figure 7 As shown, the control strategy is converted into standard control commands, transmitted to the cluster of execution equipment, and a collaborative linkage mechanism is established to coordinate equipment operation. Real-time equipment status feedback is collected to form a closed-loop control system, specifically including:

[0112] The generated intelligent control strategy is converted into standardized control commands and transmitted to the cluster of execution equipment at the breeding site;

[0113] The cluster of execution equipment includes intelligent aerators, automatic water exchange devices, precision feeders, and water quality improver dispensing equipment;

[0114] Establish a collaborative linkage mechanism for equipment, coordinate the running sequence and parameters of each device according to the priority and execution logic of the control strategy, collect the running status data of each executing device in real time during the control execution process, and feed it back to the edge computing node to form a closed-loop control of control command - device execution - status feedback.

[0115] Specifically, throughout the entire control process, the sampling frequency will be increased to once every minute, focusing on monitoring the changing trends of core water quality parameters and comparing them in real time with the predicted improvement trends. If the parameter improvement rate is lower than the predicted value, such as dissolved oxygen increasing by less than 5% within 30 minutes, or if new parameter anomalies occur, such as a sudden change in salinity caused by water exchange, a dynamic fine-tuning process will be initiated. This involves combining fine-tuning cases of similar scenarios in the full life cycle database to adjust the control parameters of relevant equipment, such as increasing the power of the aerator by 10% or reducing the water exchange flow rate and extending the water exchange time. The parameters after fine-tuning, the reasons for the adjustment, and the effect data are all archived in real time, forming a real-time data chain of "dynamic control - effect feedback".

[0116] See Figure 8As shown, various types of aquaculture data are stored in a cloud-based full lifecycle archive. Historical data is regularly mined and analyzed, and the association and analysis models are iteratively optimized through transfer learning to generate management reports, including:

[0117] The pre-treated water quality data, environmental data, control strategy data, equipment operation data, and monitoring data of the growth status of aquaculture species are stored in a cloud database to build a data archive of the entire life cycle of aquaculture.

[0118] Regularly mine and analyze historical data in the cloud database to extract water quality change patterns, control strategy effects, and equipment operating efficiency characteristics under different aquaculture scenarios;

[0119] Based on the data mining results, the transfer learning algorithm was used to iteratively optimize the growth stage-water quality parameter correlation model and the intelligent analysis model, update the model parameters and threshold system, and generate an aquaculture management report based on historical data.

[0120] Specifically, statistical analysis and feature extraction algorithms are used to uncover the changing patterns of water quality parameters at different growth stages and the influence weight of environmental parameters on water quality; the implementation effects of different control strategies under similar water quality problems are compared to extract cost-effective control parameter combinations and efficient time-series logic for coordinated equipment operation; the final aquaculture results such as aquaculture yield, survival rate, and disease incidence are correlated to construct a correlation model of "water quality parameters-control strategies-aquaculture effects" and identify key influencing factors; the uncovered patterns and features are organized into a structured knowledge graph and updated to the growth feature database and control strategy library;

[0121] Using full life-cycle data mining and newly added aquaculture samples as the core, the parameters of the growth stage-water quality parameter correlation model and the "LSTM+random forest" intelligent analysis model were updated. A transfer learning algorithm was employed, using model parameters from historical high-quality aquaculture scenarios as initial values, integrating them with newly added scenario data to optimize the model's feature extraction weights and threshold judgment logic, thereby improving the model's adaptability to diverse aquaculture scenarios. For scenarios with high model prediction bias, corresponding sample data was supplemented to strengthen the model's learning ability for special scenarios. After iteration, small-scale verification experiments were conducted in typical aquaculture zones to compare the water quality assessment accuracy, risk prediction timeliness, and control effect improvement rate of the models before and after optimization, ensuring a significant performance improvement in the optimized model. After successful verification, a model iteration version log was generated and archived in the full life-cycle database.

[0122] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0123] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0124] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart aquaculture management method integrating water quality monitoring, characterized in that, include: Multi-parameter water quality and environmental sensor arrays are deployed in various functional areas of the aquaculture waters. Through standard solution calibration and on-site blank control, a time-series synchronization mechanism is established to build a three-dimensional monitoring network covering the entire water area and its surrounding environment. The monitoring data is transmitted to the edge computing node, where data standardization and parsing, sliding window noise removal, time-series alignment and compression are performed sequentially to obtain a standardized water quality-environment fusion dataset. Construct a database of growth characteristics throughout the entire life cycle of aquaculture species, integrate growth data with historical cases, establish a correlation model between growth stages and water quality parameters, dynamically adjust thresholds based on actual stocking density and feeding amount, and generate a personalized dynamic water quality threshold system. The standardized dataset is input into the LSTM and random forest fusion model to explore the temporal correlation and synergistic effect of parameters. The current water quality status is assessed by combining dynamic thresholds, and the water quality change and risk level are predicted in the next 24 hours through the trend prediction sub-model to obtain the risk time window. Based on the assessment and prediction results, the system calls upon a three-level control strategy library containing daily, early warning, and emergency measures to generate corresponding control measures for different water quality states. The system optimizes parameters through fuzzy control to balance control effectiveness and energy consumption. The control strategy is converted into standard control commands and transmitted to the cluster of execution equipment. A collaborative linkage mechanism is established to coordinate equipment operation, and equipment status feedback is collected in real time to form a closed-loop control. All types of aquaculture data are stored in a cloud-based full lifecycle archive. Historical data is regularly mined and analyzed, and the association and analysis models are iteratively optimized through transfer learning to generate management reports.

2. The intelligent aquaculture management method integrating water quality monitoring according to claim 1, characterized in that, The deployment of multi-parameter water quality and environmental sensor arrays in various functional zones of the aquaculture area, unified calibration through standard solution calibration and on-site blank control, establishment of a time-series synchronization mechanism, and construction of a three-dimensional monitoring network covering the entire water area and its surrounding environment specifically includes: An integrated multi-parameter water quality sensor array is deployed in different functional areas of aquaculture waters. The sensor array integrates at least a pH sensor, a dissolved oxygen sensor, an ammonia nitrogen sensor, a nitrite sensor, a water temperature sensor, a turbidity sensor, and a salinity sensor. By combining standard solution calibration with on-site blank control, individual sensor errors and environmental interference are eliminated, a sensor data acquisition timing synchronization mechanism is established, and a three-dimensional water quality and environmental monitoring network covering the entire aquaculture area and its surrounding environment is formed.

3. The intelligent aquaculture management method integrating water quality monitoring according to claim 1, characterized in that, The process of transmitting monitoring data to an edge computing node, sequentially performing data standardization and parsing, sliding window noise removal, time-series alignment and compression, to obtain a standardized water quality-environment fusion dataset specifically includes: The water quality and environmental data collected by the sensor array are transmitted to the edge computing node in real time. The edge computing node performs preprocessing operations on the received data, including: The data is parsed and standardized to convert heterogeneous data from different sensors into a preset data format. The sliding window algorithm is used to remove outliers and noise from the data. By judging the gradient threshold of adjacent data points and correcting the mean, abnormal fluctuation data caused by sensor failure and water flow impact are filtered out. The preprocessed data is time-series aligned and compressed to retain key feature parameters, thereby obtaining a standardized water quality-environment fusion dataset.

4. The intelligent aquaculture management method integrating water quality monitoring according to claim 1, characterized in that, The construction of a database of growth characteristics throughout the entire life cycle of aquaculture species, integrating growth data with historical cases, establishing a correlation model between growth stages and water quality parameters, and dynamically adjusting thresholds based on actual stocking density and feeding amounts to generate a personalized dynamic water quality threshold system specifically includes: A database of growth characteristics throughout the entire life cycle of aquaculture species is constructed. The database includes physiological and metabolic parameters, feeding patterns, and suitable living environment parameters of the target aquaculture species at different growth stages. By integrating growth characteristic data of aquaculture species with historical high-quality aquaculture case data through deep learning algorithms, a correlation model between growth stage and water quality parameters is established. The model inputs are the growth days and weight monitoring data of aquaculture species, and the outputs are the suitable threshold range, warning threshold range and dangerous threshold range of each water quality parameter under the corresponding growth stage. Based on the actual stocking density, feeding amount, and water temperature change trends at the aquaculture site, the threshold range output by the model is dynamically corrected to generate a personalized dynamic water quality threshold system adapted to the current aquaculture scenario.

5. The intelligent aquaculture management method integrating water quality monitoring according to claim 1, characterized in that, The process of inputting standardized datasets into an LSTM and random forest fusion model to mine the temporal correlation and synergistic effects of parameters, combining dynamic thresholds to assess the current water quality status, and using a trend prediction sub-model to predict water quality changes and risk levels over the next 24 hours, specifically obtaining the risk time window, includes: The obtained standardized fusion dataset is input into a preset intelligent analysis model, which is a fusion model based on LSTM and random forest. The LSTM network is used to mine the time series correlation features between water quality parameters and environmental parameters, and the random forest model is used to analyze the impact of the synergistic effect of multiple parameters on the aquaculture environment. By combining a personalized dynamic water quality threshold system, the current water quality status is assessed in real time to determine whether it is in a suitable, warning, or dangerous state. The trend prediction sub-model, trained based on historical data, predicts the changing trends of various water quality parameters in the next 24 hours, identifies potential risks of water quality deterioration, and outputs the risk level and the predicted time window for the risk to occur.

6. The intelligent aquaculture management method integrating water quality monitoring according to claim 1, characterized in that, Based on the assessment and prediction results, the process involves calling upon a three-tiered control strategy library (daily, early warning, and emergency) to generate corresponding control measures tailored to different water quality conditions. Specifically, this includes optimizing parameters through fuzzy control to balance control effectiveness and energy consumption: Based on the water quality assessment results and risk prediction conclusions, a multi-level intelligent control strategy library is constructed, which includes three levels: daily maintenance control, early warning response control, and emergency response control. When the water quality is suitable, only periodic routine operation control is performed on the aeration equipment and feeding equipment; When the water quality is detected to be close to the warning threshold or a medium-to-low level risk is predicted, preventive control measures such as adjusting the power of the aeration equipment, changing a small amount of water, and optimizing the amount of feed are initiated. When the water quality reaches the danger threshold or a high level of risk is predicted, the following measures are triggered: full-power operation of the oxygenation equipment, activation of the emergency water exchange system, and suspension of feeding.

7. The intelligent aquaculture management method integrating water quality monitoring according to claim 1, characterized in that, The process of converting the control strategy into standard control commands, transmitting them to the cluster of execution devices, establishing a collaborative linkage mechanism to coordinate device operation, and collecting real-time device status feedback to form a closed-loop control specifically includes: The generated intelligent control strategy is converted into standardized control commands and transmitted to the cluster of execution equipment at the breeding site; The cluster of execution equipment includes intelligent aerators, automatic water exchange devices, precision feeders, and water quality improver dispensing equipment; Establish a collaborative linkage mechanism for equipment, coordinate the running sequence and parameters of each device according to the priority and execution logic of the control strategy, collect the running status data of each executing device in real time during the control execution process, and feed it back to the edge computing node to form a closed-loop control of control command - device execution - status feedback.

8. The intelligent aquaculture management method integrating water quality monitoring according to claim 1, characterized in that, The process of storing various types of aquaculture data in the cloud to construct a full lifecycle archive, regularly mining and analyzing historical data, and iteratively optimizing the association and analysis models through transfer learning to generate management reports specifically includes: The pre-treated water quality data, environmental data, control strategy data, equipment operation data, and monitoring data of the growth status of aquaculture species are stored in a cloud database to build a data archive of the entire life cycle of aquaculture. Regularly mine and analyze historical data in the cloud database to extract water quality change patterns, control strategy effects, and equipment operating efficiency characteristics under different aquaculture scenarios; Based on the data mining results, the transfer learning algorithm was used to iteratively optimize the growth stage-water quality parameter correlation model and the intelligent analysis model, update the model parameters and threshold system, and generate an aquaculture management report based on historical data.

9. An intelligent aquaculture management system integrating water quality monitoring, used to implement the intelligent aquaculture management method integrating water quality monitoring as described in any one of claims 1-8, characterized in that, include: Multi-parameter sensing and monitoring module: Deploy a multi-parameter water quality sensor array, and build a three-dimensional monitoring network through calibration and time synchronization to collect water quality and environmental data in real time; Edge computing module: Performs data standardization, outlier filtering, temporal alignment and compression at edge nodes to generate a standardized fused dataset; Intelligent Analysis and Risk Assessment Module: Based on the LSTM-random forest fusion model, it analyzes the time-series correlation of parameters, combines the dynamic threshold system to assess the water quality status, and predicts the risk level and time window for the next 24 hours; Strategy generation module: integrates the growth stage-water quality correlation model, dynamically adjusts thresholds based on stocking density and feeding amount, and calls the three-level strategy library to generate optimized control instructions; Equipment coordination control module: converts commands into control signals, links aerators and feeders, and provides real-time feedback on operating status; Cloud-based data management module: Stores full lifecycle data, periodically mines historical patterns, optimizes model parameters through transfer learning, and generates management reports; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images