A method for predicting dissolved oxygen in water quality based on deep learning in complex aquaculture environment

Abstract

The present application provides a method for predicting dissolved oxygen in water quality based on deep learning in a complex aquaculture environment, including: obtaining ocean current data in a sea bass aquaculture area, including flow velocity, flow direction, temperature, salinity, and dissolved oxygen data at different water depths; evaluating the volatility and irregular change degree of the ocean current data based on the historical ocean current data of the sea bass aquaculture area, and predicting the unstable changes of the ocean current in the aquaculture area during the sea bass aquaculture period; predicting the ocean current data in the aquaculture area during the sea bass aquaculture period based on the historical seasonal ocean current data of the location of the sea bass aquaculture area; predicting the dissolved oxygen content in the aquaculture area during the sea bass aquaculture period based on a prediction model for dissolved oxygen content at different water depths in the sea bass aquaculture area, and screening aquaculture areas that meet the requirements of sea bass aquaculture; and adjusting the dissolved oxygen content in the sea bass aquaculture cages using openable vents and vertically lifting aquaculture cages according to the required dissolved oxygen content at different aquaculture stages of the sea bass.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for predicting dissolved oxygen in water quality based on deep learning in a complex aquaculture environment. Background Art

[0002] Sea bass farming is an industry with great potential. However, dissolved oxygen content is one of the key factors affecting the growth and development of sea bass. Unreasonable dissolved oxygen levels in aquaculture waters can cause fish to suffocate or even die. Therefore, it is crucial to accurately predict the dissolved oxygen content in sea bass farming areas. However, this prediction faces some problems and difficulties. First, ocean currents are one of the important factors affecting dissolved oxygen content. The velocity and direction of ocean currents directly affect the transport and distribution of dissolved oxygen. Therefore, accurate acquisition of ocean current data is a prerequisite for predicting dissolved oxygen content. However, ocean currents have large fluctuations and irregular changes, which makes it difficult to obtain and evaluate ocean current data. Secondly, during sea bass farming, the unstable changes in ocean currents will also affect the dissolved oxygen content. The instability of ocean currents may lead to uneven distribution of dissolved oxygen, making the dissolved oxygen content in the farming area difficult to predict and regulate. In addition, seawater temperature and salinity are also important factors affecting dissolved oxygen content. The increase in seawater temperature and changes in salinity will affect the solubility of dissolved oxygen. Therefore, accurate prediction of seawater temperature and salinity is crucial for predicting dissolved oxygen content. In summary, there are many problems and difficulties in predicting the dissolved oxygen content in sea bass farming areas. The acquisition and evaluation of ocean current data, the unstable changes of ocean currents, and the prediction of seawater temperature and salinity all need to be solved. Only by overcoming these problems can the dissolved oxygen content be accurately predicted and adjusted accordingly according to demand to ensure that the dissolved oxygen level in the farming waters meets the requirements of sea bass farming. These are all unresolved issues at present; Summary of the invention

[0003] The present invention provides a method for predicting dissolved oxygen in water quality based on deep learning in a complex aquaculture environment, which mainly includes:

[0004] Obtain the ocean current data in the sea bass breeding area, including flow velocity, flow direction, temperature, salinity, and dissolved oxygen data at different water depths; establish a prediction model for dissolved oxygen content at different water depths in the sea bass breeding area based on the ocean current data in the sea bass breeding area and the dissolved oxygen content data at different water depths, and predict the dissolved oxygen content at different water depths in the sea bass breeding area; evaluate the volatility and irregular change degree of the ocean current data based on the historical ocean current data in the sea bass breeding area, and predict the unstable changes of the ocean current in the breeding area during the sea bass breeding period; predict the ocean current data in the breeding area during the sea bass breeding period based on the historical seasonal ocean current data at the location of the sea bass breeding area; based on the prediction model for dissolved oxygen content at different water depths in the sea bass breeding area, predict Measure the dissolved oxygen content in the breeding area during the sea bass breeding period, and select the breeding area that meets the requirements of sea bass breeding; according to the dissolved oxygen content required in different breeding stages of sea bass, use openable vents and vertical lifting breeding cages to adjust the dissolved oxygen content in the sea bass breeding cages; according to the fish density distribution in the sea bass breeding cages and the real-time horizontal distribution monitoring results of the dissolved oxygen content, adjust the distribution of bait to achieve the purpose of changing the dissolved oxygen distribution in the sea bass breeding cages; according to the historical monitoring data of water quality parameters at different water depths when the Kuroshio comes, establish a water quality model for the sea bass breeding area, predict the dissolved oxygen content at different water depths in the breeding area when the Kuroshio comes, and adjust the depth of the sea bass breeding cages, the size and number of vents.

[0005] Further optionally, the obtaining of ocean current data in the sea bass breeding area, including flow velocity, flow direction, temperature, salinity, and dissolved oxygen data at different water depths, comprises:

[0006] The velocity and direction of ocean currents in the aquaculture area are obtained through ocean monitoring stations; the seawater temperature data in the aquaculture area is obtained through water temperature sensors; the salinity data in the aquaculture area is obtained through salinity sensors; and the dissolved oxygen content data in the water body is obtained through dissolved oxygen sensors at different water depths in the aquaculture area.

[0007] Further optionally, the prediction model of dissolved oxygen content at different water depths in the sea bass breeding area is established according to the ocean current data of the sea bass breeding area and the dissolved oxygen content data at different water depths, and the prediction of dissolved oxygen content at different water depths in the sea bass breeding area includes:

[0008] Obtain ocean current data in the sea bass breeding area, including flow velocity, direction, temperature, salinity, and dissolved oxygen content at different water depths; perform quality control on the acquired data, including removing outliers and processing missing values; determine flow velocity, direction, temperature, salinity, and water depth as input variables, and take dissolved oxygen content as the target variable; perform feature selection on the data, and select features with high correlation with dissolved oxygen content; use random partitioning to divide the data set into a training set and a test set; use the support vector machine algorithm for training, using flow velocity, direction, temperature, salinity, and water depth as input, to establish a sea bass breeding area. A prediction model for dissolved oxygen content in different water depths in a seabass farming area was developed; during the training process, the model parameters were optimized according to the loss function to improve the prediction accuracy of the model; the trained model was verified using a test set, and the accuracy and performance of the model were evaluated by calculating the error indicators between the predicted values ​​and the true values, including the root mean square error and the mean absolute error; based on the evaluation results, the model was adjusted and optimized to further improve the prediction accuracy and generalization ability; the trained model was used to predict the dissolved oxygen content in different water depths in the seabass farming area based on the input flow velocity, flow direction, temperature, salinity and water depth data.

[0009] Further optionally, the step of evaluating the volatility and irregular change of the ocean current data based on the historical ocean current data of the sea bass breeding area, and predicting the unstable change of the ocean current in the breeding area during the sea bass breeding period includes:

[0010] Obtain current velocity, temperature and salinity data from historical current observation records; calculate the mean and standard deviation of current velocity, temperature and salinity, and further calculate their coefficient of variation; determine the volatility of current data based on the size of the coefficient of variation; if the coefficient of variation is greater than the preset threshold, it means that the current in the sea bass breeding area during the sea bass breeding period has unstable changes, and the area does not meet the requirements for sea bass breeding; if the coefficient of variation is less than the preset threshold, it means that the current in the sea bass breeding area during the sea bass breeding period is relatively stable, and the area meets the requirements for sea bass breeding; obtain time series data based on current observation data or historical current data; use the seasonal decomposition method to decompose the time series data to obtain three parts: long-term trend, seasonal change and random fluctuation; determine seasonal changes cycle, and by analyzing the decomposed data, determine the seasonal variation pattern of the ocean current; observe the decomposed data, determine the change of the long-term trend, and determine whether the long-term trend of the ocean current is gradually increasing or decreasing; analyze the random fluctuation part in the decomposed data, observe the impact of random fluctuations, and determine whether the random fluctuation of the ocean current is affected by weather changes and the marine environment; based on the analysis results of seasonal changes and long-term trends, predict the trend and periodicity of the ocean current changes in the breeding area during the sea bass breeding period; based on the analysis results of the volatility, seasonal changes and long-term trends of the ocean current data, preliminarily screen out the breeding areas that meet the requirements of sea bass breeding; also includes: based on the seasonal decomposition method, predict the periodicity and change trend of the ocean current in the breeding area during the sea bass breeding period.

[0011] The seasonal decomposition method is used to predict the periodicity and change trend of the ocean currents in the aquaculture area during the sea bass aquaculture period, specifically including:

[0012] Obtain a time series containing historical ocean current data. Perform data preprocessing, including removing outliers and filling missing values. Use the moving average method for seasonal decomposition. First, obtain the trend component by calculating the mean in the sliding window; then, subtract the trend component from the original data to obtain the detrended data. Calculate the average value within each season to obtain the seasonal component; finally, subtract the trend component and seasonal component from the original data to calculate the residual; perform seasonal decomposition using the X-11 seasonal adjustment method. Based on the seasonal components extracted from the historical data, infer the data on the changes in ocean currents in the same season during the seabass farming period. Obtain the forecast results by multiplying the historical seasonal components with the time series during the seabass farming period. Predict the long-term change trend based on the trend component. Predict the overall change trend of ocean currents in the farming area during the seabass farming period by fitting a linear model of the trend component.

[0013] Further optionally, the predicting of the ocean current data of the breeding area during the sea bass breeding period according to the historical seasonal ocean current data of the location of the sea bass breeding area includes:

[0014] The historical seasonal ocean current data of the location of the sea bass breeding area are obtained from relevant marine scientific research institutions or marine observation stations, including information on flow velocity, flow direction, temperature and salinity; based on the historical ocean current data, time attributes are extracted, including year, month, day and time information; the geographical location information of the breeding area, including longitude and latitude, is obtained as the geographical attributes of the ocean current; the average flow velocity at each time point is calculated to determine the flow velocity attribute of the ocean current; the average flow direction at each time point is calculated to determine the flow direction attribute of the ocean current; the average temperature at each time point is calculated to obtain the temperature attribute of the ocean current; the average salinity at each time point is calculated to obtain the salinity attribute of the ocean current; the ARIMA algorithm is used to take time and geographical attributes as input features and the flow velocity, flow direction, temperature and salinity of the ocean current as output variables to establish an ocean current prediction model for the breeding area during the sea bass breeding period; the historical ocean current data and attributes are used as training sets to train and optimize the model; during the training process, the model parameters are optimized according to the loss function to improve the prediction accuracy of the model; the established ocean current prediction model for the breeding area during the sea bass breeding period is used to predict the ocean current data during the breeding period.

[0015] Further optionally, the prediction model of dissolved oxygen content at different water depths in the sea bass breeding area is used to predict the dissolved oxygen content in the breeding area during the sea bass breeding period, and the breeding area that meets the sea bass breeding requirements is screened, including:

[0016] The flow velocity, flow direction, temperature and salinity data of the sea bass breeding area during the sea bass breeding period are obtained by means of ocean observation stations and buoy observations; the obtained data are quality controlled and cleaned to remove outliers and missing values; the flow velocity, flow direction, temperature and salinity data of the sea bass breeding area are used as input, and the dissolved oxygen content prediction model of the sea bass breeding area at different water depths is used to predict the dissolved oxygen content in the breeding area during the sea bass breeding period; according to the dissolved oxygen content predicted by the model, it is judged whether it meets the required dissolved oxygen concentration for sea bass breeding; according to the predicted dissolved oxygen content in the breeding area during the sea bass breeding period, it is judged whether it meets the required dissolved oxygen concentration for sea bass breeding; for breeding areas that meet the requirements, they are screened out as feasible sea bass breeding areas.

[0017] Further optionally, the step of adjusting the dissolved oxygen content in the sea bass breeding cages by using openable vents and vertical lifting breeding cages according to the dissolved oxygen content required at different breeding stages of sea bass comprises:

[0018] Obtain the required dissolved oxygen content at different breeding stages of seabass through relevant research reports on seabass breeding, including the hatching stage, seedling stage, young fish stage, grow-out fish stage and adult fish stage; determine the required dissolved oxygen content at each stage as the target value; use the dissolved oxygen content prediction model at different water depths in the seabass breeding area to predict the dissolved oxygen content at different water depths according to the required dissolved oxygen content at different breeding stages of seabass; compare the predicted dissolved oxygen content with the known dissolved oxygen content data to find the water depth where the predicted result is consistent with the actual data; adjust the depth of seabass breeding cages in stages according to the water depth that is consistent with the required dissolved oxygen content at different breeding stages of seabass, and ensure that the water depth in the seabass breeding area meets the requirements of different breeding stages of seabass; use openable vents when adjusting the depth of seabass breeding cages causes the living space of seabass to be less than the preset threshold; adjust the size and number of cage vents according to the required dissolved oxygen content; also include: modeling according to the dissolved oxygen content outside the seabass breeding cages, the size and number of cage vents and the dissolved oxygen content inside the seabass breeding cages, and adjusting the size and number of seabass breeding cage vents.

[0019] The modeling is performed according to the dissolved oxygen content outside the sea bass breeding cage, the size and number of the cage ventilation holes, and the dissolved oxygen content inside the sea bass breeding cage, and the size and number of the sea bass breeding cage ventilation holes are adjusted, specifically including:

[0020] Obtain a data set of dissolved oxygen content outside the sea bass breeding cage and a data set of dissolved oxygen content inside the sea bass breeding cage; use the linear regression analysis method to take the dissolved oxygen content outside the sea bass breeding cage and the dissolved oxygen content inside the sea bass breeding cage as input variables, and the size and number of cage vents as output variables to establish a regression model. Determine the parameters and equations of the regression model to obtain a prediction model for the size and number of vents related to the dissolved oxygen content; fit the dependent variable and the explanatory variable, and use the F test to judge the significance of the regression model. Calculate the regression sum of squares and the residual sum of squares, then calculate the F value based on the degrees of freedom, and judge the significance of the F value through the F distribution table or statistical software; calculate the coefficient of determination by calculating the ratio of the regression sum of squares to the total variance, and calculate the coefficient of determination and the adjusted coefficient of determination to evaluate the goodness of fit of the regression model. Combine the results of the significance test and the goodness of fit evaluation to determine whether the established regression model is reasonable. By inputting the dissolved oxygen content outside the sea bass breeding cage and the dissolved oxygen content inside the sea bass breeding cage, the size and number of vents in the sea bass breeding cage are predicted using the established regression model.

[0021] Further optionally, adjusting the distribution of bait delivery to change the distribution of dissolved oxygen in the sea bass breeding cages according to the fish density distribution and the real-time horizontal distribution monitoring results of the dissolved oxygen content in the sea bass breeding cages comprises:

[0022] Through the cameras and sensors in the sea bass breeding cages, the density distribution of the fish in the cages is monitored in real time and the data is output; through the dissolved oxygen sensors in the sea bass breeding cages, the horizontal distribution of the dissolved oxygen content in the cages is monitored in real time and the data is output; based on the obtained fish density distribution and the horizontal distribution data of the dissolved oxygen content, the high-density area and the low-oxygen area are judged; based on the fish density distribution data and the measured value of the dissolved oxygen content, a correlation analysis is performed, the correlation coefficient between the two is calculated, and the degree of linear correlation is judged; based on the fish density distribution data and the distribution of bait delivery, the adjustment plan for the bait delivery position and quantity is determined; based on the adjusted bait delivery distribution and the fish density distribution data, it is determined whether the fish density has changed; based on the real-time monitoring results of the horizontal distribution of the dissolved oxygen content, it is determined whether the dissolved oxygen content has changed; based on the correlation between the fish density and the dissolved oxygen content, it is determined whether the horizontal distribution of dissolved oxygen corresponds to the fish density distribution; based on the fish density distribution data and the real-time monitoring results of the horizontal distribution of the dissolved oxygen content, it is determined whether the purpose of adjusting the dissolved oxygen distribution has been achieved.

[0023] Further optionally, the water quality model of the sea bass breeding area is established based on the monitoring data of water quality parameters at different water depths when the Kuroshio comes in history, the dissolved oxygen content at different water depths in the breeding area is predicted when the Kuroshio comes, and the depth of the cages, the size and number of ventilation holes for sea bass breeding are adjusted in real time, including:

[0024] By accessing the database of relevant marine scientific research institutions or marine observatories, the historical ocean current data of the sea bass breeding area when the Kuroshio comes, including flow rate, flow direction, temperature and salinity data, are obtained; the dissolved oxygen content data of different water depths in the sea bass breeding area are extracted from the historical monitoring data; the support vector machine algorithm is used to train the water quality model using the historical monitoring data, with flow rate, flow direction, temperature, salinity and water depth as input variables and dissolved oxygen content as the target variable; the trained model is evaluated and the accuracy and performance of the model are calculated by the cross-validation method; the established water quality model is used to input the flow rate, flow direction, temperature, salinity and water depth data of the breeding area when the Kuroshio comes, and the dissolved oxygen content at different water depths is predicted; based on the prediction results, the dissolved oxygen content at different water depths in the sea bass breeding area is obtained; cameras and sensor equipment are installed in the sea bass breeding cages to monitor the density distribution of the fish schools in the cages and the horizontal distribution of the dissolved oxygen content in real time; the output of the camera and sensor equipment is connected to the data acquisition system to record and store the real-time monitored water quality parameter data; the real-time monitoring is analyzed. The fish density distribution data and the horizontal distribution data of dissolved oxygen content obtained by the measurement are used; the correlation coefficient between the fish density and the dissolved oxygen content is used to determine whether the depth of the sea bass breeding cage needs to be adjusted; based on the real-time monitored water quality parameter data, including the dissolved oxygen content and temperature, it is determined whether the size and number of the vents of the sea bass breeding cage need to be adjusted; by adjusting the size and number of the vents, the entry and circulation of oxygen are controlled to meet the needs of sea bass breeding; the changes in water quality parameters in the sea bass breeding cages are continuously monitored, including the fish density distribution and the horizontal distribution of the dissolved oxygen content; the changes in water quality parameters are observed and recorded, and whether the cage depth and the size and number of the vents need to be further adjusted according to the trend of changes; based on the real-time monitored water quality parameter changes and the previous adjustment results, it is determined whether the depth, size and number of the vents of the sea bass breeding cages need to be further adjusted; according to the changes in the water quality parameters, the corresponding cage depth and vents are adjusted to maintain a suitable breeding environment; according to the real-time monitoring data and the adjustment results, it is iterated and optimized to further improve the water quality and breeding environment.

[0025] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0026] The invention discloses a dissolved oxygen prediction technology based on ocean current data and dissolved oxygen content data at different water depths in a sea bass breeding area. By acquiring ocean current data in a sea bass breeding area, including flow velocity, flow direction, temperature, and salinity, and acquiring dissolved oxygen data at different water depths, a prediction model for dissolved oxygen content at different water depths in a sea bass breeding area is established. The model is used to predict dissolved oxygen content at different water depths in a sea bass breeding area, and breeding areas that meet breeding requirements are screened. By evaluating historical ocean current data in a sea bass breeding area, the volatility and irregular change degree of ocean current data are analyzed, and the unstable change of ocean current in the breeding area during the sea bass breeding period is predicted. According to the historical seasonal ocean current data at the location of the sea bass breeding area, the change of ocean current data during the breeding period is predicted. Based on the prediction model for dissolved oxygen content at different water depths in a sea bass breeding area, the dissolved oxygen content in the breeding area during the breeding period is predicted. According to the prediction results, a breeding area that meets the sea bass breeding requirements is screened. According to the required dissolved oxygen content at different breeding stages of the sea bass, an openable vent and a vertical lifting breeding cage are used to adjust the dissolved oxygen content in the sea bass breeding cage. According to the fish density distribution in the sea bass farming cages and the real-time horizontal distribution monitoring results of dissolved oxygen content, the distribution of bait is adjusted to change the distribution of dissolved oxygen in the sea bass farming cages. Based on the historical monitoring data of water quality parameters at different water depths when the Kuroshio came, a water quality model for the sea bass farming area was established to predict the dissolved oxygen content at different water depths in the farming area when the Kuroshio came, and to adjust the depth of the sea bass farming cages, the size and number of vents in real time. In summary, this technology integrates multiple technologies such as dissolved oxygen prediction model, bait placement adjustment and water quality model to effectively predict the dissolved oxygen conditions in the farming area during the sea bass farming period, and realize scientific management and optimization of farming. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of a method for predicting dissolved oxygen in water quality based on deep learning in a complex aquaculture environment according to the present invention.

[0028] Figure 2 This is a schematic diagram of a method for predicting dissolved oxygen in water quality based on deep learning in a complex aquaculture environment according to the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0030] In this embodiment, a method for predicting dissolved oxygen in water quality based on deep learning under a complex aquaculture environment may specifically include:

[0031] Step 101, obtaining ocean current data in the sea bass breeding area, including flow velocity, flow direction, temperature, salinity, and dissolved oxygen data at different water depths.

[0032] The current velocity and direction of the aquaculture area are obtained through the ocean monitoring station; the seawater temperature data of the aquaculture area is obtained through the water temperature sensor; the salinity data of the aquaculture area is obtained through the salinity sensor; the dissolved oxygen content data in the water body is obtained through the dissolved oxygen sensors at different water depths in the aquaculture area. For example, a current measuring instrument is installed at the ocean monitoring station. The current velocity obtained by measuring the movement of the water flow to the buoy or speed meter is 2 knots, about 3.7 kilometers per hour, and the flow direction is north. Through the water temperature sensor. The seawater temperature monitored is 18 degrees Celsius. The salinity monitored by the salinity sensor is 35‰, that is, 35 grams of salt per kilogram of water. Finally, the dissolved oxygen content in the water body is measured by the dissolved oxygen sensor. Assume that the dissolved oxygen content monitored at a surface water depth of 1 meter is 6 mg / L, and the dissolved oxygen content monitored at a bottom water depth of 5 meters is 4 mg / L. In summary, through the ocean monitoring station, it was obtained that the current velocity in the aquaculture area was 2 knots, the flow direction was north, the sea water temperature was 18 degrees Celsius, the salinity was 35‰, the dissolved oxygen content in the surface water was 6 mg / L, and the dissolved oxygen content in the bottom water was 4 mg / L.

[0033] Step 102, based on the ocean current data of the sea bass breeding area and the dissolved oxygen content data at different water depths, a prediction model for the dissolved oxygen content at different water depths in the sea bass breeding area is established to predict the dissolved oxygen content at different water depths in the sea bass breeding area.

[0034] The ocean current data of the sea bass breeding area were obtained, including flow velocity, flow direction, temperature, salinity, and dissolved oxygen content at different water depths; the quality control of the obtained data was performed, including removing outliers and processing missing values. The flow velocity, flow direction, temperature, salinity and water depth were determined as input variables, and the dissolved oxygen content was used as the target variable. Feature selection was performed on the data, and features with high correlation with dissolved oxygen content were selected. The data set was divided into a training set and a test set using the random partitioning method. The support vector machine algorithm was used for training, and the prediction model of dissolved oxygen content at different water depths in the sea bass breeding area was established with flow velocity, flow direction, temperature, salinity and water depth as inputs. During the training process, the model parameters were optimized according to the loss function to improve the prediction accuracy of the model. The trained model was verified using the test set, and the accuracy and performance of the model were evaluated by calculating the error indicators between the predicted value and the true value, including the root mean square error and the mean absolute error. According to the evaluation results, the model was adjusted and optimized to further improve the prediction accuracy and generalization ability. Using the trained model, the dissolved oxygen content at different water depths in the sea bass breeding area is predicted based on the input flow velocity, flow direction, temperature, salinity and water depth data. For example, the current data of the sea bass breeding area is obtained: Sample 1: flow velocity = 5m / s, flow direction = 120°, temperature = 20°C, salinity = 35ppt, water depth = 10m, dissolved oxygen content = 7mg / L Sample 2: flow velocity = 8m / s, flow direction = 180°, temperature = 18°C,

[0035] Salinity = 36ppt, water depth = 15m, dissolved oxygen content = 5mg / LSample 3: flow rate = 4m / s, flow direction = 90°, temperature = 21℃, salinity = 34ppt, water depth = 12m, dissolved oxygen content = 8mg / L; Divide the above data set into a training set and a test set. Assume that the training set contains samples 1 and 2, and the test set contains sample 3. Use the support vector machine algorithm for training, take flow rate, flow direction, temperature, salinity and water depth as input variables, and take dissolved oxygen content as the target variable to establish a prediction model. Training set data: Input variables: [5,120,20,35,10], [8,180,18,36,15], target variables: 8, 5, use the above training set data to train the support vector machine algorithm. After training, use the test set data for verification to evaluate the accuracy and performance of the model. Input the test set data into the trained model, and the predicted dissolved oxygen content is 7mg / L.

[0036] Step 103, based on the historical ocean current data of the sea bass breeding area, assess the volatility and irregular change degree of the ocean current data, and predict the unstable change of the ocean current in the breeding area during the sea bass breeding period.

[0037] Obtain velocity, temperature and salinity data from historical current observation records. Calculate the mean and standard deviation of velocity, temperature and salinity, and further calculate their coefficient of variation. Determine the volatility of current data based on the size of the coefficient of variation. If the coefficient of variation is greater than the preset threshold, it means that the current in the sea bass breeding area during the sea bass breeding period is unstable and the area does not meet the requirements for sea bass breeding; if the coefficient of variation is less than the preset threshold, it means that the current in the sea bass breeding area during the sea bass breeding period is relatively stable and the area meets the requirements for sea bass breeding. Obtain time series data based on current observation data or historical current data. Decompose the time series data using the seasonal decomposition method to obtain three parts: long-term trend, seasonal change and random fluctuation. Determine the cycle of seasonal changes, and determine the seasonal change law of the current by analyzing the decomposed data. Observe the decomposed data, determine the changes in the long-term trend, and determine whether the long-term trend of the current is gradually increasing or decreasing. Analyze the random fluctuation part in the decomposed data, observe the impact of random fluctuations, and determine whether the random fluctuations of the current are affected by weather changes and the marine environment. According to the analysis results of seasonal changes and long-term trends, the trend and periodicity of the current changes in the breeding area during the sea bass breeding period are predicted. According to the analysis results of the volatility, seasonal changes and long-term trends of the current data, the breeding areas that meet the requirements of sea bass breeding are preliminarily screened out. For example, the average value of the flow velocity is calculated to be 2m / s, and the standard deviation is 0.5m / s, then the coefficient of variation of the flow velocity is 0.5 / 2=0.25. Assuming that the preset threshold is 0.3, since the coefficient of variation is less than the threshold, it is preliminarily judged that the flow velocity is relatively stable during the breeding period. Similarly, the average value, standard deviation and coefficient of variation of temperature and salinity are calculated and compared with the preset threshold. If the coefficient of variation of the flow velocity, temperature and salinity are all less than the preset threshold, it is preliminarily judged that there is no unstable change in the current in the breeding area, which meets the requirements of sea bass breeding. However, for the case where the coefficient of variation is greater than the preset threshold, the trend and periodicity of the current are further analyzed. The seasonal decomposition method is used to decompose the time series data into three parts: long-term trend, seasonality and random fluctuation, and the trend and periodicity of the current are predicted. Suppose that the ocean current data is observed for one year, and it is found that the fluctuation of the current velocity is large and the coefficient of variation exceeds the preset threshold. The seasonal decomposition of this time series data is performed to obtain three parts: long-term trend, seasonality and random fluctuation. By analyzing seasonality and random fluctuations, it is determined that the ocean current has large fluctuations in a certain season or cycle, which may be caused by seasonal changes or other factors. If it is found that the volatility of the ocean current is mainly concentrated in a certain season or cycle, and this volatility has an adverse effect on sea bass farming, then the farming area may not meet the requirements for sea bass farming.

[0038] Based on the seasonal decomposition method, the periodicity and changing trend of ocean currents in the breeding area during the seabass breeding period are predicted.

[0039] Obtain a time series containing historical ocean current data. Perform data preprocessing, including removing outliers and filling missing values. Perform seasonal decomposition using the moving average method. First, obtain the trend component by calculating the mean within a sliding window; then, subtract the trend component from the original data to obtain detrended data. Calculate the mean within each season to obtain the seasonal component; finally, calculate the residual by subtracting the trend component and seasonal component from the original data; perform seasonal decomposition using the X-11 seasonal adjustment method. Based on the seasonal component extracted from the historical data, infer the data on the changes in ocean currents in the same season during the seabass farming period. Obtain the forecast results by multiplying the historical seasonal component with the time series during the seabass farming period. Predict long-term changes based on the trend component. Predict the overall trend of ocean currents in the farming area during the seabass farming period by fitting a linear model of the trend component. For example, there are monthly ocean current velocity data for the past 5 years. First, perform data preprocessing to remove any outliers and fill missing values. Then, calculate the trend component using a sliding window of 3 months. Average the ocean current velocity for each month and subtract it from the original data to obtain detrended data. Next, calculate the average value for each month to obtain the seasonal component. For example, calculate the average value for January and the average value for February. Then, subtract the trend component and seasonal component from the original data to calculate the residual. Use the X-11 seasonal adjustment method to perform seasonal decomposition to extract the seasonal component more accurately. Based on the seasonal components extracted from the historical data, infer the data on the changes in ocean currents in the same season during the seabass farming period. For example, if seabass farming is being carried out in the summer, multiply the seasonal component of the historical summer by the time series during the seabass farming period to obtain the forecast result. Finally, predict the long-term change trend based on the trend component. Use a linear model to fit the trend component and predict the overall change trend of ocean currents in the farming area during the seabass farming period.

[0040] Step 104, predicting the ocean current data of the breeding area during the sea bass breeding period based on the historical seasonal ocean current data of the location of the sea bass breeding area.

[0041] Obtain historical seasonal ocean current data of the location of the sea bass breeding area from relevant marine scientific research institutions or marine observatories, including information on velocity, direction, temperature and salinity. Extract time attributes, including year, month, day and time information, based on historical ocean current data. Obtain geographical location information of the breeding area, including longitude and latitude, as the geographical attributes of the ocean current. Calculate the average velocity at each time point to determine the velocity attribute of the ocean current. Calculate the average direction at each time point to determine the direction attribute of the ocean current. Calculate the average temperature at each time point to obtain the temperature attribute of the ocean current. Calculate the average salinity at each time point to obtain the salinity attribute of the ocean current. Using the ARIMA algorithm, taking time and geographic attributes as input features and the velocity, direction, temperature and salinity of the ocean current as output variables, establish an ocean current prediction model for the breeding area during the sea bass breeding period. Use historical ocean current data and attributes as training sets to train and optimize the model. During the training process, optimize the model parameters according to the loss function to improve the prediction accuracy of the model. The established current prediction model for the aquaculture area during the sea bass aquaculture period is used to predict the current data during the aquaculture period. For example, three years of current data in the aquaculture area were obtained from a marine scientific research institution, including flow velocity, flow direction, temperature and salinity. Assume that the geographical location of the aquaculture area is longitude 40°N and latitude 130°E. For example, time attributes: year: 2018, 2019, 2020, month: January to December, date: 1st to 31st, time: recorded once every hour, geographical attributes: longitude: 40°N, latitude: 130°E, according to the provided current data, the average flow velocity, average flow direction, average temperature and average salinity at each time point are calculated. For example, at 00:00 on January 1, 2018, the average flow velocity calculated according to the current data is 0.5m / s, the average flow direction is 150°, the average temperature is 15℃, and the average salinity is 35ppt. The ARIMA algorithm is used to take time and geographic information as input features, and the velocity, direction, temperature and salinity of the ocean current as output to establish an ocean current prediction model for the aquaculture area during the sea bass aquaculture period. The model predicts the velocity, direction, temperature and salinity of future ocean currents based on the time and geographic attributes of historical data. For example, the model is used to predict the ocean current data at 00:00 on January 1, 2021, and the predicted velocity is 0.6m / s, the direction is 160°, the temperature is 16°C, and the salinity is 36ppt.

[0042] Step 105, based on the prediction model of dissolved oxygen content at different water depths in the sea bass breeding area, predict the dissolved oxygen content in the breeding area during the sea bass breeding period, and select the breeding area that meets the sea bass breeding requirements.

[0043] The flow velocity, flow direction, temperature and salinity data of the sea bass breeding area during the sea bass breeding period are obtained through ocean observation stations and buoy observations. The acquired data are quality controlled and cleaned to remove outliers and missing values. The flow velocity, flow direction, temperature and salinity data of the sea bass breeding area are used as input, and the dissolved oxygen content prediction model of the sea bass breeding area at different water depths is used to predict the dissolved oxygen content in the breeding area during the sea bass breeding period. According to the dissolved oxygen content predicted by the model, it is judged whether it meets the required dissolved oxygen concentration for sea bass breeding. According to the predicted dissolved oxygen content in the breeding area during the sea bass breeding period, it is judged whether it meets the required dissolved oxygen concentration for sea bass breeding. For breeding areas that meet the requirements, they are screened out as feasible sea bass breeding areas. For example, the flow velocity in the sea bass breeding area is 5m / s, the flow direction is northeast, the temperature is 20℃, and the salinity is 30ppt. According to the prediction model of dissolved oxygen content at different water depths in the area, the following data are obtained: in the area with a water depth of 1 meter, the predicted dissolved oxygen content is 6 mg / L; in the area with a water depth of 5 meters, the predicted dissolved oxygen content is 8 mg / L; in the area with a water depth of 10 meters, the predicted dissolved oxygen content is 5 mg / L. According to the predicted dissolved oxygen content, evaluate whether it meets the dissolved oxygen concentration required for sea bass farming. Assuming that the dissolved oxygen concentration required for sea bass farming is 7 mg / L, according to the predicted data: in the area with a water depth of 1 meter, the dissolved oxygen content is 6 mg / L, which does not meet the requirements; in the area with a water depth of 5 meters, the dissolved oxygen content is 8 mg / L, which meets the requirements; in the area with a water depth of 10 meters, the dissolved oxygen content is 5 mg / L, which meets the requirements. Therefore, the breeding areas that meet the requirements of sea bass breeding are screened out as areas with water depths of 5 meters and 10 meters.

[0044] Step 106, according to the dissolved oxygen content required in different breeding stages of sea bass, openable vents and vertical lifting breeding cages are used to adjust the dissolved oxygen content in the sea bass breeding cages.

[0045] Obtain the required dissolved oxygen content of sea bass at different breeding stages through relevant research reports on sea bass breeding, including the hatching stage, seedling stage, young fish stage, grow-out fish stage and adult fish stage. Determine the required dissolved oxygen content at each stage as the target value. Use the prediction model of dissolved oxygen content at different water depths in the sea bass breeding area to predict the dissolved oxygen content at different water depths according to the required dissolved oxygen content at different breeding stages of sea bass. Compare the predicted dissolved oxygen content with the known dissolved oxygen content data to find the water depth where the predicted result is consistent with the actual data. According to the water depth that matches the required dissolved oxygen content at different breeding stages of sea bass, adjust the depth of sea bass breeding cages in stages to ensure that the water depth in the sea bass breeding area meets the requirements of sea bass at different breeding stages. When the adjustment of the depth of sea bass breeding cages causes the living space of sea bass to be less than the preset threshold, use openable vents. Adjust the size and number of cage vents according to the required dissolved oxygen content. For example, according to the required dissolved oxygen content of sea bass in different breeding stages obtained from the relevant research report on sea bass breeding, sea bass in the hatching stage needs at least 6 mg of dissolved oxygen per liter of water, sea bass in the seedling stage needs at least 7 mg of dissolved oxygen per liter of water, sea bass in the juvenile stage needs at least 8 mg of dissolved oxygen per liter of water, sea bass in the breeding stage needs at least 9 mg of dissolved oxygen per liter of water, and sea bass in the adult stage needs at least 10 mg of dissolved oxygen per liter of water. According to the above required dissolved oxygen content, the dissolved oxygen content prediction model of different water depths in the sea bass breeding area is used to predict the dissolved oxygen content at different water depths. For example, in a breeding area with a water depth of 1 meter, the prediction model predicts that the dissolved oxygen content at this depth is 5 mg / L; in a breeding area with a water depth of 2 meters, the prediction model predicts that the dissolved oxygen content at this depth is 2 mg / L; in a breeding area with a water depth of 3 meters, the prediction model predicts that the dissolved oxygen content at this depth is 8 mg / L. Then, the predicted dissolved oxygen content is compared with the known dissolved oxygen content data to find the water depth where the predicted result is consistent with the actual data. Assume that according to the actual measurement data, the dissolved oxygen content at a depth of 1 meter is 7 mg / L, the dissolved oxygen content at a depth of 2 meters is 0 mg / L, and the dissolved oxygen content at a depth of 3 meters is 5 mg / L. By comparison, it is found that the water depths where the dissolved oxygen content predicted by the prediction model is consistent with the actual data are 1 meter and 3 meters. Therefore, in the hatching stage and the fish breeding stage, the depth of the sea bass breeding cage is adjusted to 1 meter; in the seedling stage, the young fish stage and the adult fish stage, the depth of the sea bass breeding cage is adjusted to 3 meters. When the adjustment of the depth of the sea bass breeding cage causes the living space of the sea bass to be less than the preset threshold, for example, in the breeding area at a depth of 2 meters, the living space of the sea bass is less than the preset threshold. Adopt openable vents, adjust the size and number of cage vents as needed, improve oxygen supply, and achieve the required dissolved oxygen content in different breeding stages of sea bass.

[0046] A model was built based on the dissolved oxygen content outside the sea bass farming cages, the size and number of the cage ventilation holes, and the dissolved oxygen content inside the sea bass farming cages, and the size and number of the sea bass farming cage ventilation holes were adjusted.

[0047] Obtain a data set of dissolved oxygen content outside the sea bass breeding cage and a data set of dissolved oxygen content inside the sea bass breeding cage; use the linear regression analysis method to take the dissolved oxygen content outside the sea bass breeding cage and the dissolved oxygen content inside the sea bass breeding cage as input variables, and the size and number of cage vents as output variables to establish a regression model. Determine the parameters and equations of the regression model to obtain a prediction model for the size and number of vents related to the dissolved oxygen content; fit the dependent variable and the explanatory variable, and use the F test to judge the significance of the regression model. Calculate the regression sum of squares and the residual sum of squares, then calculate the F value based on the degrees of freedom, and judge the significance of the F value through the F distribution table or statistical software; calculate the coefficient of determination by calculating the ratio of the regression sum of squares to the total variance, and calculate the coefficient of determination and the adjusted coefficient of determination to evaluate the goodness of fit of the regression model. Combine the results of the significance test and the goodness of fit evaluation to determine whether the established regression model is reasonable. By inputting the dissolved oxygen content outside the sea bass breeding cage and the dissolved oxygen content inside the sea bass breeding cage, the size and number of vents in the sea bass breeding cage are predicted using the established regression model. For example, the dissolved oxygen content outside the sea bass farming cages: dissolved oxygen content (mg / L): [5,2,3,9,8], cage vent size and number: vent size (cm): [20,25,18,22,19], number of vents: [2,3,2,2,1], the dissolved oxygen content inside the sea bass farming cages: dissolved oxygen content (mg / L): [1,12,8,5,3], cage vent size and number: vent size (cm): [30,35,28,32,29], number of vents: [3,4,3,3,2], the linear regression analysis method was used to establish the prediction model. First, we took the dissolved oxygen content as the input variable and the vent size and number as the output variable. For the data set outside the sea bass breeding cage, the following linear regression model was obtained: vent size = 5 * dissolved oxygen content - 6 vent number = 8 * dissolved oxygen content + 1, for the data set inside the sea bass breeding cage, the following linear regression model was obtained: vent size = 2 * dissolved oxygen content - 12 vent number = 2 * dissolved oxygen content - 1; the F test was used to judge the significance of the entire regression model; the significance level was assumed to be 05, so when the F value was greater than 10, the regression model was considered significant. For the data set outside the sea bass breeding cage, the determination coefficient was calculated to be 84 and the adjusted determination coefficient was 75. For the data set inside the sea bass breeding cage, the determination coefficient was calculated to be 75 and the adjusted determination coefficient was 60. According to the results of the significance test and the goodness of fit evaluation, it was judged whether the established regression model was reasonable. If the regression model was significant and the goodness of fit was good, the model was trusted and used to predict the size and number of vents in the sea bass breeding cage. If the dissolved oxygen content outside the sea bass farming cage is 2 mg / L and the dissolved oxygen content inside the sea bass farming cage is 8 mg / L, the regression model predicts that the size of the ventilation holes of the sea bass farming cage is 16 cm and the number of ventilation holes is 96.

[0048] Step 107, according to the fish density distribution in the sea bass breeding cage and the horizontal distribution monitoring result of the real-time dissolved oxygen content, the bait delivery distribution is adjusted to achieve the purpose of changing the dissolved oxygen distribution in the sea bass breeding cage.

[0049] Through the cameras and sensors in the sea bass breeding cages, the density distribution of the fish in the cages is monitored in real time, and the data is output. Through the dissolved oxygen sensor in the sea bass breeding cages, the horizontal distribution of the dissolved oxygen content in the cages is monitored in real time, and the data is output. According to the obtained fish density distribution and dissolved oxygen content horizontal distribution data, the high-density area and the low-oxygen area are judged. According to the fish density distribution data and the dissolved oxygen content measurement value, correlation analysis is performed, the correlation coefficient between the two is calculated, and the degree of linear correlation is judged. According to the fish density distribution data and the distribution of bait delivery, the adjustment plan for the bait delivery position and quantity is determined. According to the adjusted bait delivery distribution and fish density distribution data, it is judged whether the fish density has changed. According to the real-time dissolved oxygen content horizontal distribution monitoring results, it is determined whether the dissolved oxygen content has changed. According to the correlation between the fish density and the dissolved oxygen content, it is judged whether the dissolved oxygen horizontal distribution corresponds to the fish density distribution. According to the fish density distribution data and the real-time dissolved oxygen content horizontal distribution monitoring results, it is determined whether the purpose of adjusting the dissolved oxygen distribution has been achieved. For example, there are 10 fish in a sea bass breeding cage. Through the camera and sensor, the number of fish in different positions is monitored in real time as follows: Position 1: 20 fish, Position 2: 10 fish, Position 3: 30 fish, Position 4: 40 fish. After outputting these data, it can be seen that the fish density at position 4 is the highest. At the same time, the dissolved oxygen content levels at different positions are monitored in real time through the dissolved oxygen sensor as follows: Position 1: 8mg / L, Position 2: 7mg / L, Position 3: 6mg / L, Position 4: 5mg / L. After outputting these data, it can be seen that the dissolved oxygen content at position 4 is the lowest. According to the fish density distribution data and the dissolved oxygen content measurement value, a correlation analysis is performed to calculate the correlation coefficient between the two. Assume that the calculation result is -0.8, indicating that the fish density and dissolved oxygen content are strongly negatively correlated. According to the distribution of bait placement, it is assumed that the bait is mainly concentrated in positions 1 and 3, and the amount of bait placed is 100g and 200g respectively. According to the fish density distribution data and the distribution of bait placement, it is determined whether the bait placement position and amount need to be adjusted. Assume that the distribution of bait after adjustment is as follows: Position 1: 150g, Position 2: 50g, Position 3: 100g, Position 4: 200g. According to the fish density distribution data, determine whether the fish density has changed. According to the horizontal distribution monitoring results of the real-time dissolved oxygen content, determine whether the dissolved oxygen content has changed. According to the correlation between fish density and dissolved oxygen content, determine whether the horizontal distribution of dissolved oxygen corresponds to the fish density distribution. According to the fish density distribution data and the horizontal distribution monitoring results of the real-time dissolved oxygen content, determine whether the purpose of adjusting the dissolved oxygen distribution has been achieved.

[0050] Step 108, based on the historical monitoring data of water quality parameters at different water depths when the Kuroshio comes, a water quality model of the sea bass breeding area is established to predict the dissolved oxygen content at different water depths in the breeding area when the Kuroshio comes, and the depth of the sea bass breeding cages, the size and number of ventilation holes are adjusted in real time.

[0051] By accessing the database of relevant marine scientific research institutions or marine observatories, historical ocean current data of the sea bass breeding area when the Kuroshio arrives, including flow velocity, flow direction, temperature and salinity data, are obtained. Dissolved oxygen content data at different water depths in the sea bass breeding area are extracted from historical monitoring data. The support vector machine algorithm is used to train the water quality model using historical monitoring data, with flow velocity, flow direction, temperature, salinity and water depth as input variables and dissolved oxygen content as the target variable. The trained model is evaluated, and the accuracy and performance of the model are calculated by cross-validation method. Using the established water quality model, the flow velocity, flow direction, temperature, salinity and water depth data of the breeding area when the Kuroshio arrives are input to predict the dissolved oxygen content at different water depths. According to the prediction results, the dissolved oxygen content at different water depths in the sea bass breeding area is obtained. Cameras and sensor equipment are installed in sea bass breeding cages to monitor the density distribution of fish schools and the horizontal distribution of dissolved oxygen content in the cages in real time. The output of the camera and sensor equipment is connected to the data acquisition system to record and store the real-time monitored water quality parameter data. The fish school density distribution data and the horizontal distribution data of dissolved oxygen content obtained by real-time monitoring are analyzed. The correlation coefficient between fish density and dissolved oxygen content is used to determine whether the depth of the sea bass breeding cage needs to be adjusted. Based on the real-time monitored water quality parameter data, including dissolved oxygen content and temperature, determine whether the size and number of vents in the sea bass breeding cage need to be adjusted. By adjusting the size and number of vents, the entry and circulation of oxygen can be controlled to meet the needs of sea bass breeding. Continuously monitor the changes in water quality parameters in sea bass breeding cages, including the distribution of fish density and the horizontal distribution of dissolved oxygen content. Observe and record the changes in water quality parameters, and analyze whether the cage depth and the size and number of vents need to be further adjusted based on the trend of changes. Combined with the real-time monitored water quality parameter changes and the previous adjustment results, determine whether the depth, size and number of vents of the sea bass breeding cage need to be further adjusted. According to the changes in water quality parameters, make corresponding cage depth and vent adjustments to maintain a suitable breeding environment. According to the real-time monitoring data and adjustment results, iterate and optimize the adjustments to further improve the water quality and breeding environment. For example, by accessing the database of relevant marine scientific research institutions or marine observatories, historical current data of the sea bass breeding area when the Kuroshio arrives can be obtained, such as the flow rate of 2 knots, the flow direction is north, the temperature is 22 degrees Celsius, and the salinity is 32ppt. The dissolved oxygen content data of different water depths in the sea bass breeding area are extracted from the historical monitoring data, such as the dissolved oxygen content at a depth of 10 meters is 6 mg / L, and the dissolved oxygen content at a depth of 20 meters is 5 mg / L. The support vector machine algorithm is used to train the water quality model using historical monitoring data, such as using flow rate, flow direction, temperature, salinity and water depth as input variables and dissolved oxygen content as the target variable. The trained model is evaluated, and the accuracy and performance of the model are calculated by the cross-validation method, such as the model accuracy is 90%.Using the established water quality model, the flow velocity, flow direction, temperature, salinity and water depth data of the aquaculture area when the Kuroshio comes are input to predict the dissolved oxygen content at different water depths. For example, the dissolved oxygen content at a water depth of 10 meters is predicted to be 8 mg / L, and the dissolved oxygen content at a water depth of 20 meters is predicted to be 3 mg / L. Cameras and sensor equipment are installed in sea bass aquaculture cages to monitor the density distribution of fish in the cages and the horizontal distribution of dissolved oxygen content in real time. For example, the fish density distribution data recorded by the camera and sensor equipment is that there are 100 fish per cubic meter of cages, and the horizontal distribution data of dissolved oxygen content is that the dissolved oxygen content in the surface layer of the cage is 2 mg / L, and the dissolved oxygen content in the bottom layer is 7 mg / L. Analyze the fish density distribution data and the horizontal distribution data of dissolved oxygen content obtained by real-time monitoring. For example, the correlation coefficient between fish density and dissolved oxygen content is calculated to be -8, indicating that the dissolved oxygen content will decrease when the fish density increases. According to the correlation coefficient between fish density and dissolved oxygen content, it is judged whether the depth of sea bass aquaculture cages needs to be adjusted. For example, if the correlation coefficient is -8, which exceeds the set threshold of -6, it is judged that the depth of the sea bass breeding cage needs to be adjusted. According to the real-time monitored water quality parameter data, including dissolved oxygen content and temperature, it is judged whether the size and number of the vents of the sea bass breeding cage need to be adjusted. When the dissolved oxygen content is lower than the set threshold, it is judged that the size and number of the vents need to be increased. By adjusting the size and number of the vents, the entry and circulation of oxygen can be controlled to meet the needs of sea bass breeding. For example, the size and number of vents can be increased to improve the oxygen supply. Continuously monitor the changes in water quality parameters in the sea bass breeding cage, including the fish density distribution and the horizontal distribution of dissolved oxygen content. For example, record the fish density distribution and the horizontal distribution data of dissolved oxygen content every day. Observe and record the changes in water quality parameters, and analyze whether the cage depth and the size and number of vents need to be further adjusted according to the trend of changes. For example, when the dissolved oxygen content continues to decrease, the cage depth and the size and number of vents are further adjusted. Combined with the changes in the real-time monitored water quality parameters and the previous adjustment results, it is judged whether the depth, size and number of the sea bass breeding cage need to be further adjusted. When the dissolved oxygen content continues to decline and the previous adjustments have not significantly improved, it is judged that the depth, size and number of vents of the sea bass farming cages need to be further adjusted. According to the changes in water quality parameters, make corresponding adjustments to the cage depth and vents to maintain a suitable farming environment. According to the decline in dissolved oxygen content monitored in real time, increase the cage depth and the size and number of vents. According to the real-time monitoring data and adjustment results, iterate and optimize the adjustments to further improve the water quality and farming environment. According to the real-time monitoring data and adjustment results, further optimize the cage depth and the size and number of vents to improve the dissolved oxygen content and farming effect.

[0052] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for predicting dissolved oxygen in water quality based on deep learning in a complex aquaculture environment, characterized in that: The method comprises: Obtain the ocean current data in the sea bass breeding area, including flow velocity, flow direction, temperature, salinity, and dissolved oxygen data at different water depths; establish a prediction model for dissolved oxygen content at different water depths in the sea bass breeding area based on the ocean current data in the sea bass breeding area and the dissolved oxygen content data at different water depths, and predict the dissolved oxygen content at different water depths in the sea bass breeding area; evaluate the volatility and irregular change degree of the ocean current data based on the historical ocean current data in the sea bass breeding area, and predict the unstable changes of the ocean current in the breeding area during the sea bass breeding period; predict the ocean current data in the breeding area during the sea bass breeding period based on the historical seasonal ocean current data at the location of the sea bass breeding area; based on the prediction model for dissolved oxygen content at different water depths in the sea bass breeding area, predict Measure the dissolved oxygen content in the breeding area during the sea bass breeding period, and select the breeding area that meets the requirements of sea bass breeding; according to the required dissolved oxygen content in different breeding stages of sea bass, use openable vents and vertical lifting breeding cages to adjust the dissolved oxygen content in the sea bass breeding cages; according to the fish density distribution in the sea bass breeding cages and the real-time horizontal distribution monitoring results of the dissolved oxygen content, adjust the distribution of bait to achieve the purpose of changing the dissolved oxygen distribution in the sea bass breeding cages; according to the historical monitoring data of water quality parameters at different water depths when the Kuroshio comes, establish a water quality model for the sea bass breeding area, predict the dissolved oxygen content at different water depths in the breeding area when the Kuroshio comes, and adjust the depth of the sea bass breeding cages, the size and number of vents; The method comprises establishing a prediction model for dissolved oxygen content at different water depths in the sea bass breeding area according to the ocean current data in the sea bass breeding area and the dissolved oxygen content data at different water depths, and predicting the dissolved oxygen content at different water depths in the sea bass breeding area, including: Obtain ocean current data in the sea bass breeding area, including flow velocity, direction, temperature, salinity, and dissolved oxygen content at different water depths; perform quality control on the acquired data, including removing outliers and processing missing values; determine flow velocity, direction, temperature, salinity, and water depth as input variables, and take dissolved oxygen content as the target variable; perform feature selection on the data, and select features with high correlation with dissolved oxygen content; use random partitioning to divide the data set into a training set and a test set; use the support vector machine algorithm for training, using flow velocity, direction, temperature, salinity, and water depth as input, to establish a sea bass breeding area A prediction model for dissolved oxygen content in different water depths in a seabass farming area was developed; during the training process, the model parameters were optimized according to the loss function to improve the prediction accuracy of the model; the trained model was verified using a test set, and the accuracy and performance of the model were evaluated by calculating the error indicators between the predicted values ​​and the true values, including the root mean square error and the mean absolute error; based on the evaluation results, the model was adjusted and optimized to further improve the prediction accuracy and generalization ability; the trained model was used to predict the dissolved oxygen content in different water depths in the seabass farming area based on the input flow velocity, flow direction, temperature, salinity and water depth data.

2. The method according to claim 1, wherein: The acquisition of ocean current data in the sea bass breeding area, including flow velocity, flow direction, temperature, salinity, and dissolved oxygen data at different water depths, includes: The velocity and direction of ocean currents in the aquaculture area are obtained through ocean monitoring stations; the seawater temperature data in the aquaculture area is obtained through water temperature sensors; the salinity data in the aquaculture area is obtained through salinity sensors; and the dissolved oxygen content data in the water body is obtained through dissolved oxygen sensors at different water depths in the aquaculture area.

3. The method according to claim 1, wherein: The method of evaluating the volatility and irregularity of the ocean current data based on the historical ocean current data of the sea bass breeding area and predicting the unstable changes of the ocean current in the breeding area during the sea bass breeding period includes: Obtain the velocity, temperature and salinity data from the historical ocean current observation records; calculate the mean and standard deviation of the velocity, temperature and salinity, and further calculate their coefficient of variation; determine the volatility of the ocean current data according to the size of the coefficient of variation; if the coefficient of variation is greater than the preset threshold, it means that the ocean current in the sea bass breeding area during the sea bass breeding period is unstable and the area does not meet the requirements for sea bass breeding; if the coefficient of variation is less than the preset threshold, it means that the ocean current in the sea bass breeding area during the sea bass breeding period is relatively stable and the area meets the requirements for sea bass breeding; obtain time series data according to ocean current observation data or historical ocean current data; Use seasonal decomposition method to decompose time series data and obtain three parts: long-term trend, seasonal change and random fluctuation; determine the cycle of seasonal change, and judge the seasonal change law of ocean current by analyzing the decomposed data; Observe the decomposed data, judge the changes in the long-term trend, and determine whether the long-term trend of the current is gradually increasing or decreasing; analyze the random fluctuations in the decomposed data, observe the impact of random fluctuations, and determine whether the random fluctuations of the current are affected by weather changes and the marine environment; predict the trend and periodicity of the current changes in the breeding area during the sea bass breeding period based on the analysis results of seasonal changes and long-term trends; Based on the analysis results of the volatility, seasonal changes and long-term trends of ocean current data, the breeding areas that meet the requirements of sea bass breeding are preliminarily screened; It also includes: predicting the periodicity and changing trend of ocean currents in the breeding area during the seabass breeding period based on the seasonal decomposition method.

4. The method according to claim 1, wherein: The method of predicting the ocean current data of the aquaculture area during the sea bass aquaculture period based on the historical seasonal ocean current data of the location of the sea bass aquaculture area includes: Obtain historical seasonal ocean current data of the location of the sea bass farming area from relevant marine scientific research institutions or marine observation stations, including information on flow velocity, flow direction, temperature and salinity; extract time attributes based on historical ocean current data, including year, month, day and time information; obtain geographical location information of the farming area, including longitude and latitude, as geographical attributes of the ocean current; Calculate the average velocity at each time point and determine the velocity properties of the current; Calculate the average flow direction at each time point and determine the flow direction properties of the current; Calculate the average temperature at each time point to obtain the temperature properties of the ocean current; Calculate the average salinity at each time point to obtain the salinity properties of the ocean current; The ARIMA algorithm is used to take time and geographical attributes as input features, and the velocity, direction, temperature and salinity of the current as output variables to establish a current prediction model for the breeding area during the seabass breeding period. The historical current data and attributes are used as training sets to train and optimize the model. During the training process, the model parameters are optimized according to the loss function to improve the prediction accuracy of the model. The established current prediction model for the breeding area during the seabass breeding period is used to predict the current data during the breeding period.

5. The method according to claim 1, wherein: The prediction model of dissolved oxygen content at different water depths in the sea bass breeding area is used to predict the dissolved oxygen content in the breeding area during the sea bass breeding period, and to screen the breeding area that meets the sea bass breeding requirements, including: The flow velocity, flow direction, temperature and salinity data of the sea bass breeding area during the sea bass breeding period are obtained by means of ocean observation stations and buoy observations; the obtained data are quality controlled and cleaned to remove outliers and missing values; the flow velocity, flow direction, temperature and salinity data of the sea bass breeding area are used as input, and the dissolved oxygen content prediction model of the sea bass breeding area at different water depths is used to predict the dissolved oxygen content in the breeding area during the sea bass breeding period; according to the dissolved oxygen content predicted by the model, it is judged whether it meets the required dissolved oxygen concentration for sea bass breeding; according to the predicted dissolved oxygen content in the breeding area during the sea bass breeding period, it is judged whether it meets the required dissolved oxygen concentration for sea bass breeding; for breeding areas that meet the requirements, they are screened out as feasible sea bass breeding areas.

6. The method according to claim 1, wherein: The method comprises: adjusting the dissolved oxygen content in the sea bass breeding cage by using an openable vent and a vertical lifting breeding cage according to the dissolved oxygen content required in different breeding stages of the sea bass, including: Obtain the required dissolved oxygen content of sea bass at different breeding stages through relevant research reports on sea bass breeding, including the hatching stage, seedling stage, young fish stage, grow-out fish stage and adult fish stage; determine the required dissolved oxygen content of each stage as the target value; use the dissolved oxygen content prediction model of different water depths in the sea bass breeding area to predict the dissolved oxygen content at different water depths according to the required dissolved oxygen content at different breeding stages of sea bass; compare the predicted dissolved oxygen content with the known dissolved oxygen content data to find the water depth where the predicted result is consistent with the actual data; adjust the depth of sea bass breeding cages in stages according to the water depth that is consistent with the required dissolved oxygen content at different breeding stages of sea bass, and ensure that the water depth in the sea bass breeding area meets the requirements of sea bass at different breeding stages; use openable vents when the adjustment of the depth of sea bass breeding cages causes the living space of sea bass to be less than the preset threshold; Adjust the size and number of cage vents according to the required dissolved oxygen content; It also includes: modeling according to the dissolved oxygen content outside the sea bass breeding cages, the size and number of the cage ventilation holes and the dissolved oxygen content inside the sea bass breeding cages, and adjusting the size and number of the sea bass breeding cage ventilation holes.

7. The method according to claim 1, wherein: The method of adjusting the distribution of bait according to the fish density distribution in the sea bass breeding cage and the horizontal distribution monitoring results of the real-time dissolved oxygen content to achieve the purpose of changing the dissolved oxygen distribution in the sea bass breeding cage comprises: Through the cameras and sensors in the sea bass breeding cages, the density distribution of the fish in the cages is monitored in real time and the data is output; through the dissolved oxygen sensors in the sea bass breeding cages, the horizontal distribution of the dissolved oxygen content in the cages is monitored in real time and the data is output; based on the obtained fish density distribution and the horizontal distribution data of the dissolved oxygen content, the high-density area and the low-oxygen area are judged; based on the fish density distribution data and the measured value of the dissolved oxygen content, a correlation analysis is performed, the correlation coefficient between the two is calculated, and the degree of linear correlation is judged; based on the fish density distribution data and the distribution of bait delivery, the adjustment plan for the bait delivery position and quantity is determined; based on the adjusted bait delivery distribution and the fish density distribution data, it is determined whether the fish density has changed; based on the real-time monitoring results of the horizontal distribution of the dissolved oxygen content, it is determined whether the dissolved oxygen content has changed; based on the correlation between the fish density and the dissolved oxygen content, it is determined whether the horizontal distribution of dissolved oxygen corresponds to the fish density distribution; based on the fish density distribution data and the real-time monitoring results of the horizontal distribution of the dissolved oxygen content, it is determined whether the purpose of adjusting the dissolved oxygen distribution has been achieved.

8. The method according to claim 1, wherein: The method includes establishing a water quality model for the sea bass breeding area based on the monitoring data of water quality parameters at different water depths when the Kuroshio comes in history, predicting the dissolved oxygen content at different water depths in the breeding area when the Kuroshio comes, and adjusting the cage depth, ventilation hole size and number of the sea bass breeding in real time, including: By accessing the database of relevant marine scientific research institutions or marine observatories, the historical ocean current data of the sea bass breeding area when the Kuroshio comes, including flow rate, flow direction, temperature and salinity data, are obtained; the dissolved oxygen content data of different water depths in the sea bass breeding area are extracted from the historical monitoring data; the support vector machine algorithm is used to train the water quality model using the historical monitoring data, with flow rate, flow direction, temperature, salinity and water depth as input variables and dissolved oxygen content as the target variable; the trained model is evaluated and the accuracy and performance of the model are calculated by the cross-validation method; the established water quality model is used to input the flow rate, flow direction, temperature, salinity and water depth data of the breeding area when the Kuroshio comes, and the dissolved oxygen content at different water depths is predicted; based on the prediction results, the dissolved oxygen content at different water depths in the sea bass breeding area is obtained; cameras and sensor equipment are installed in the sea bass breeding cages to monitor the density distribution of the fish schools in the cages and the horizontal distribution of the dissolved oxygen content in real time; the output of the camera and sensor equipment is connected to the data acquisition system to record and store the real-time monitored water quality parameter data; the real-time monitoring is analyzed. The fish density distribution data and the horizontal distribution data of dissolved oxygen content obtained by the measurement are used; the correlation coefficient between the fish density and the dissolved oxygen content is used to determine whether the depth of the sea bass breeding cage needs to be adjusted; based on the real-time monitored water quality parameter data, including the dissolved oxygen content and temperature, it is determined whether the size and number of the vents of the sea bass breeding cage need to be adjusted; by adjusting the size and number of the vents, the entry and circulation of oxygen are controlled to meet the needs of sea bass breeding; the changes in water quality parameters in the sea bass breeding cages are continuously monitored, including the fish density distribution and the horizontal distribution of the dissolved oxygen content; the changes in water quality parameters are observed and recorded, and whether the cage depth and the size and number of the vents need to be further adjusted according to the trend of changes; based on the real-time monitored water quality parameter changes and the previous adjustment results, it is determined whether the depth, size and number of the vents of the sea bass breeding cages need to be further adjusted; according to the changes in the water quality parameters, the corresponding cage depth and vents are adjusted to maintain a suitable breeding environment; according to the real-time monitoring data and the adjustment results, it is iterated and optimized to further improve the water quality and breeding environment.

Images