Water quality adaptability simulation control method based on fishery breeding scene

Abstract

The invention relates to the technical field of simulation control, in particular to a water quality adaptability simulation control method based on a fishery breeding scene. The method comprises the following steps: acquiring real-time water quality monitoring data, cultured organism growth state data and environmental meteorological data corresponding to a fishery culture water area, performing standardized preprocessing, and constructing a water quality-organism-environment multi-dimensional incidence matrix; constructing a water quality adaptability simulation model by using the water quality-biology-environment multi-dimensional incidence matrix, simulating corresponding water quality parameter evolution processes under different environment variable combinations, and generating a water quality dynamic simulation data set; determining an optimal water quality parameter interval based on the water quality dynamic simulation data set; according to the optimal water quality parameter interval, a dynamic regulation and control strategy is formulated for comparative analysis, and a water quality regulation and control deviation value is generated; based on the water quality regulation and control deviation value, driving water quality regulation equipment corresponding to the fishery breeding water area to execute adaptive control operation and display. According to the invention, the water quality control efficiency of fishery breeding can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of analog control, in particular to a water quality adaptability simulation control method based on a fishery breeding scene. BACKGROUND

[0002] In recent years, with the rapid development of fishery breeding towards intensification and high density, and the continuous improvement of consumers' requirements for the quality and safety of aquatic products, water quality, as a core factor affecting the growth, survival rate and product quality of breeding organisms, its precision of dynamic regulation, adaptability to biological growth and response speed to environmental changes have become key demands.

[0003] At present, some fishery breeding water quality control methods have been proposed. These methods mostly collect water quality parameters (such as dissolved oxygen, ammonia nitrogen) through a single sensor, trigger oxygenation, water exchange and other regulation operations combined with preset fixed thresholds, and then determine the water quality regulation effect according to experience. However, the existing methods are insufficient in multi-dimensional data fusion analysis and dynamic simulation, and are affected by factors such as single detection parameter, not considering the correlation between breeding organism growth state and environmental factors, which easily leads to regulation timing lag or parameter matching deviation. At the same time, the method also lacks the simulation capability of water quality evolution process, can only passively regulate based on real-time data, and is difficult to predict the influence of environmental changes (such as heavy rain, high temperature) on water quality in advance, thereby reducing the precision of water quality control. SUMMARY

[0004] Therefore, it is necessary to provide a water quality adaptability simulation control method based on a fishery breeding scene to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a water quality adaptability simulation control method based on a fishery breeding scene comprises the following steps: Step S1: obtaining real-time water quality monitoring data, breeding organism growth state data and environmental meteorological data corresponding to the fishery breeding water area, and standardizing and preprocessing the real-time water quality monitoring data, breeding organism growth state data and environmental meteorological data to generate standardized water quality feature data set, biological growth feature data set and environmental influence factor data set; Step S2: extracting basic water quality features including dissolved oxygen dynamic distribution, pH value fluctuation trend and ammonia nitrogen concentration gradient change based on the standardized water quality feature data set, combining the biological growth feature data set to mine the correlation between breeding organism activity intensity and water quality parameters, and introducing the environmental influence factor data set to construct a water quality-organism-environment multi-dimensional correlation matrix; Step S3: Construct a water quality adaptability simulation model using the water quality-biological-environment multi-dimensional correlation matrix, simulate the corresponding water quality parameter evolution process under different combinations of environmental variables based on the water quality adaptability simulation model, and generate a water quality dynamic simulation data set; based on the coupling analysis of the water quality dynamic simulation data set and the biological growth characteristic data set, determine the optimal water quality parameter interval for the survival of the cultured organisms; Step S4: According to the optimal water quality parameter interval, a dynamic control strategy is formulated, and based on the dynamic control strategy, real-time water quality monitoring data is compared and analyzed with the optimal water quality parameter interval to generate a water quality control deviation value; based on the water quality control deviation value, the corresponding water quality adjusting equipment of the fishery culture water area is driven to perform adaptive control operation, and the water quality control effect and biological growth state prediction result are displayed through a visual interface.

[0006] The beneficial effects of the present application are: The fishery breeding scene-based water quality adaptability simulation control method has the beneficial effects that, through multi-dimensional data acquisition and standardized preprocessing, the core limitations of the existing water quality control method, i.e., single detection parameter and uneven data quality, are completely solved, and a comprehensive and reliable data foundation is laid for subsequent accurate regulation. The existing method only relies on a single water quality sensor (such as a dissolved oxygen sensor) to collect data, ignores the correlation between the growth state of the cultured organisms and the environmental meteorology, and does not perform standardized processing on the data, which is prone to analysis deviation due to non-uniform parameter units and data noise interference, thereby affecting the accuracy of regulation. In this step, real-time water quality data (dissolved oxygen, pH value, etc.), biological growth data (activity intensity, growth rate, etc.), and environmental meteorological data (temperature, rainfall, etc.) are simultaneously obtained, comprehensively covering the three key dimensions of water quality changes, and making up for the information blind area of a single parameter; at the same time, the three types of data are standardized and preprocessed to eliminate data format differences and noise interference, ensure data consistency and reliability, provide a unified data benchmark for subsequent multi-dimensional correlation analysis, and avoid regulation misjudgment caused by data quality problems, thereby improving the accuracy of water quality control from the source. Secondly, by extracting basic features such as dissolved oxygen dynamic distribution and pH value fluctuation trend from the standardized water quality data, the dynamic changes of water quality are intuitively reflected; by mining the correlation between biological activity intensity and water quality parameters (such as determining the optimal dissolved oxygen range corresponding to the peak of biological activity), the response mechanism of organisms to water quality is clarified; and by introducing environmental impact factors to construct a multi-dimensional correlation matrix, the regulation effect of environmental factors (such as rainstorm-induced pH value drop) on the water quality-organism relationship is quantified. This multi-dimensional correlation analysis can not only reveal the deep driving logic of water quality changes, but also provide core correlation basis for subsequent construction of an accurate water quality simulation model. Then, through water quality adaptability simulation model construction and optimal parameter interval determination, the core bottleneck of the existing water quality control method, i.e., lack of water quality evolution simulation capability and only passive regulation, is broken through, realizing the leap from "real-time response" to "advance prediction", and greatly improving the forward-looking of regulation. The existing method only triggers regulation based on real-time water quality data, cannot simulate future water quality evolution trends, is prone to water quality deterioration due to regulation lag, especially in environmental mutations (such as rainstorm and high temperature), and does not clearly define the optimal water quality range for biological survival, so the regulation parameters are only set according to experience, which is prone to parameter matching deviation.In this step, the water quality adaptability simulation model based on the multi-dimensional correlation matrix can simulate the water quality evolution process under different combinations of environmental variables, generate dynamic simulation data sets, and enable aquaculture personnel to grasp the direction of water quality change in advance. Through coupling analysis of the simulation data set and biological growth data, the optimal water quality parameter interval for the survival of the cultured organisms (such as the optimal dissolved oxygen range for a certain fish being 5-8 mg / L and the pH value being 7.5-8.5) is determined, rather than relying on empirical thresholds. This "simulation prediction + optimal interval" design not only solves the problem that existing methods cannot cope with environmental mutations, but also provides precise target basis for subsequent regulation and avoids over-regulation or under-regulation due to ambiguous parameters. Finally, through dynamic regulation strategy development, device linkage control, and visual display, the limitations of existing water quality control methods, such as "regulation timing lag and non-intuitive effect feedback", are solved, forming a complete closed loop of "prediction-regulation-feedback" and significantly improving the timeliness and operability of water quality control. Existing methods only passively trigger regulation when water quality parameters exceed fixed thresholds, which can easily miss the best intervention time, and there is a lack of intuitive effect display and biological growth prediction after regulation, making it difficult for aquaculture personnel to determine whether the regulation is effective. In this step, a dynamic regulation strategy is developed based on the optimal water quality parameter interval, a deviation value is generated by comparing real-time water quality data with the optimal interval, and adaptive operations are performed by driving water quality regulation devices (such as starting an oxygenator to restore dissolved oxygen to 5 mg / L), achieving precise and timely proactive regulation. At the same time, the regulation effect (such as the real-time change curve of dissolved oxygen) and the biological growth state prediction results (such as the biological growth rate after 3 days based on the current water quality) are displayed on the visual interface, enabling aquaculture personnel to intuitively understand the regulation effectiveness and future trends. This not only avoids the passive lag problem of existing methods, but also helps aquaculture personnel optimize subsequent regulation strategies through intuitive feedback, forming a virtuous cycle and ultimately improving the accuracy of water quality control and aquaculture efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0007] Other features, objects, and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments, made with reference to the accompanying drawings: Figure 1 The following is a step flow diagram of the water quality adaptability simulation control method of the present application based on the fishery aquaculture scene: Figure 2 The following is a detailed step flow diagram of step S1 in the present application. Figure 1 DETAILED DESCRIPTION

[0008] The technical method of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.​

[0009] Furthermore, the accompanying drawings are included to provide a further understanding of the present application, and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application. In the drawings:

[0010] It should be understood that, although the terms“first,”“second,” etc. can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element without departing from the scope of the example embodiments. The term“and / or” as used herein includes any and all combinations of one or more of the associated listed items.

[0011] To achieve the above object, there is provided Figures 1 to 2 The present application provides a water quality adaptability simulation control method based on a fishery breeding scene. In an embodiment of the present application, please refer to Figure 1 The present application provides a water quality adaptability simulation control method based on a fishery breeding scene. In an embodiment of the present application, please refer to Step S1: acquiring real-time water quality monitoring data, breeding biological growth state data and environmental meteorological data corresponding to a fishery breeding water area, and performing standardization preprocessing on the real-time water quality monitoring data, the breeding biological growth state data and the environmental meteorological data to generate a standardized water quality feature data set, a biological growth feature data set and an environmental influence factor data set; In the embodiment of the present application, real-time water quality monitoring data of a 10 mu grass carp culture pond is collected by a distributed water quality sensor array (3x3 grid, 50 m apart), including dissolved oxygen (0-20 mg / L, accuracy ±0.1 mg / L, 5 minutes / time, such as 9:007.8 mg / L), pH value (5.5-9.5, accuracy ±0.01 pH, 5 minutes / time, such as 9:007.5 pH), ammonia nitrogen concentration (0-10 mg / L, accuracy ±0.05 mg / L, 10 minutes / time, such as 9:000.42 mg / L), water temperature (0-40℃, accuracy ±0.1℃, 5 minutes / time, such as 9:0025.3℃); the growth state data of the cultured organisms is obtained by underwater imaging equipment (1920x1080 resolution, 30 minutes / shot) and RFID biological sensors, including grass carp swimming trajectory (X-Y coordinate accuracy ±1 cm, such as a certain tail fish 9:00 from (10m, 15m) to (25m, 20m)), feeding frequency (counted per hour, such as 9:00-10:0028 times), growth size (image calibration 1 cm corresponds to 10 pixels, such as a certain tail fish 35 cm), stress response (swimming speed >0.5m / s is determined, such as 9:151 times); environmental meteorological data is collected by a shore meteorological station (installation height 2m), including light intensity (0-200000 lux, accuracy ±100 lux, 10 minutes / time, such as 9:0085000 lux), precipitation (0-50 mm / h, accuracy ±0.1 mm, 10 minutes / time, such as 9:000 mm), wind speed (0-30 m / s, accuracy ±0.1 m / s, 5 minutes / time, such as 9:002.3 m / s), air pressure (800-1100 hPa, accuracy ±0.1 hPa, 10 minutes / time, such as 9:001013.2 hPa). The water quality data is subjected to 3σ criterion to remove outliers (such as dissolved oxygen 15 mg / L exceeding the mean value ±3σ to determine abnormality, replaced by interpolation), and 5-point moving average smoothing; the biological data is normalized to the interval of 0-1 by linear transformation (growth size 30-40 cm is mapped to 0-1, such as 35 cm corresponding to 0.5); the environmental data is aligned by timestamp (10 minutes / time uniform frequency), generating standardized water quality feature dataset (containing 4 types of parameters, 24 hours 288 groups), biological growth feature dataset (containing 4 types of indicators, 24 hours 48 groups), environmental influence factor dataset (containing 4 types of parameters, 24 hours 288 groups).

[0012] Step S2: Based on the standardized water quality feature dataset, basic water quality features including dissolved oxygen dynamic distribution, pH value fluctuation trend and ammonia nitrogen concentration gradient change are extracted, and the correlation between biological activity intensity and water quality parameters is mined in combination with the biological growth feature dataset, and a water quality-biology-environment multidimensional correlation matrix is constructed by introducing the environmental influence factor dataset; In the embodiment of the present application, by extracting the basic water quality characteristics from the standardized water quality characteristic dataset: dissolved oxygen dynamic distribution (calculate the 24-hour day-night fluctuation amplitude, the maximum value of 9.2 mg / L in the daytime, the minimum value of 7.0 mg / L at night, and the difference of 2.2 mg / L), pH value fluctuation trend (linear fitting drift rate-0.0015 pH / h, -0.06 pH mutation point appears at 18:00), ammonia nitrogen concentration gradient change (calculate the concentration difference of adjacent monitoring points, the difference of 0.03 mg / L between No. 1 and No. 2 monitoring points, and the attenuation rate of 6×10^-4 mg / (L·m) per unit distance). Combined with the biological growth characteristic dataset, the correlation rule is mined: the correlation coefficient of the activity intensity index (0-1, such as 9:000.65) and the ammonia nitrogen change rate (0.02h⁻¹) is calculated by cross-correlation analysis, the optimal lag time of 10 minutes is determined, the mapping model y=0.8x-0.334 (R²=0.82) is constructed, and the biological-water quality correlation factor 0.945 is generated. The environmental influence factor dataset is introduced, and the influence weight of each environmental parameter on water quality is calculated by partial least squares regression: light 0.2 (promote dissolved oxygen to rise), precipitation 0.1 (dilute ammonia nitrogen), wind speed-0.15 (accelerate water quality mixing), and air pressure 0.05 (affect the solubility of dissolved oxygen). A 3×4 water quality-biological-environmental multi-dimensional correlation matrix is constructed: the row represents the space (monitoring point mean value), time (24-hour mean value), and biological (larval fish period) dimensions, and the column represents dissolved oxygen, pH value, ammonia nitrogen, and comprehensive water quality. The matrix element is the correlation strength in the corresponding dimension (such as the space-ammonia nitrogen element=ammonia nitrogen gradient weight 0.3×wind speed weight-0.15×biological correlation factor 0.945≈-0.042), accurate to three decimal places.

[0013] Step S3: using the water quality-biological-environmental multi-dimensional correlation matrix to construct a water quality adaptability simulation model, simulating the evolution process of the corresponding water quality parameters under different environmental variable combinations based on the water quality adaptability simulation model, and generating a water quality dynamic simulation dataset; based on the coupling analysis of the water quality dynamic simulation dataset and the biological growth characteristic dataset, determining the optimal water quality parameter interval for the survival of the cultured organisms; In the embodiment of the present application, by constructing a water quality adaptability simulation model based on a water quality-biology-environment multi-dimensional correlation matrix, a physical diffusion module (improved Fick model, input ammonia nitrogen comprehensive diffusion coefficient 1.125 m² / s, simulate spatial diffusion), a biological metabolism module (input metabolic rate 1.15 mg / (kg·h), simulate biological ammonia nitrogen emission), and an environmental impact module (input environmental weight, simulate the effect of light and wind speed on water quality) are included. Latin hypercube sampling is used to generate 625 sets of environmental variable combinations (light 50000-150000 lux, precipitation 0-4 mm / h, etc., 5 levels per variable), which are input into the model to simulate for 30 days (time step 1 hour), and the water quality parameter evolution curve of each combination (such as the diurnal fluctuation of dissolved oxygen and the convergence of ammonia nitrogen to 0.3 mg / L) is output. Statistical analysis of the curve is performed to calculate the steady-state value (dissolved oxygen 8.0 mg / L), fluctuation period (24 hours), and convergence rate (0.008 mg / (L·h)), and a water quality dynamic simulation dataset (625 records) is generated. The dataset is coupled with biological growth characteristic data (growth rate 5 g / d, survival rate 98%) to construct a water quality-biology response surface using a radial basis function neural network, and the optimal water quality parameter interval is determined through grid search optimization: dissolved oxygen 8.5-9.5 mg / L, pH 7.8-8.0, ammonia nitrogen ≤0.35 mg / L, and ammonia nitrogen convergence rate 0.006-0.008 mg / (L·h), within which the biological growth index is ≥0.8 (growth rate 6-7 g / d, survival rate 97%-99%).

[0014] Step S4: Formulate a dynamic control strategy according to the optimal water quality parameter interval, compare and analyze real-time water quality monitoring data with the optimal water quality parameter interval based on the dynamic control strategy, generate water quality control deviation values, and drive the corresponding water quality adjusting equipment of the fish farming water area to perform adaptive control operations based on the water quality control deviation values, while displaying the water quality control effect and biological growth state prediction results through a visual interface.

[0015] In the embodiment of the present application, three-level dynamic regulation strategies are formulated according to the optimal water quality parameter interval: first-level deviation (single parameter exceeding 10% of the interval, such as dissolved oxygen 8.0-8.5 mg / L) corresponds to low-intensity measures (oxygenation 0.8 kW + pH adjustment 0.15 L / h, priority 3); second-level deviation (single parameter exceeding 10%-20% or two parameter deviations, such as dissolved oxygen 7.5-8.0 mg / L) corresponds to medium-intensity measures (oxygenation 1.5 kW + pH adjustment 0.3 L / h + circulation 15 m³ / h, priority 2); third-level deviation (single parameter exceeding 20% or three parameter deviations, such as dissolved oxygen <7.5 mg / L) corresponds to high-intensity measures (oxygenation 2.0 kW + pH adjustment 0.5 L / h + circulation 30 m³ / h, priority 1). Real-time collection of current water quality parameters (dissolved oxygen 7.8 mg / L, pH 7.5, ammonia nitrogen 0.42 mg / L), calculation of deviation values (dissolved oxygen -13.3%, pH -5.1%, ammonia nitrogen 20%) and comprehensive deviation index -0.85%, fuzzy clustering determination of second-level deviation. Match the second-level measure set, and get the optimal combination through multi-objective optimization (max compliance rate, min energy consumption): oxygenation 1.4 kW, pH addition 0.28 L / h, circulation 16 m³ / h, which is converted into control instructions (oxygenation 10:00-14:00, pH adjustment 10:00-14:00, circulation 10:00-16:00) and issued to the equipment for execution. Simultaneously, the visualization interface is displayed: the power change of the equipment is presented by the line chart, the water quality before and after regulation is compared by the column chart (dissolved oxygen 7.8→8.6 mg / L, pH 7.5→7.8, ammonia nitrogen 0.42→0.37 mg / L), the survival rate of 98% is displayed by the pie chart, the growth rate of 5→5.8 g / d is displayed by the line chart, the evaluation report is generated, and the regulation effectiveness and subsequent maintenance suggestions are clear.

[0016] Further, as an embodiment of the present application, referring to Figure 2 shown, the step S1 includes the following steps: Figure 1 The detailed step flowchart of the step S1 in the embodiment, the step S1 includes the following steps: Step S11: collecting real-time water quality monitoring data corresponding to dissolved oxygen, pH value, ammonia nitrogen concentration, water temperature and salinity in the fishery breeding water area through a distributed water quality sensor array; In the embodiment of the present application, the distributed water quality sensor array is arranged in a "3x3" grid in a 10 mu circular culture pond (radius about 126 m), the sensor spacing is 50 m, covering all culture areas from the center of the pond to the shore, each sensor node integrates dissolved oxygen, pH value, ammonia nitrogen concentration, water temperature, salinity detection module. The dissolved oxygen module measures 0-20 mg / L, the accuracy is ±0.1 mg / L, and the data is collected every 5 minutes (such as 9:00 collection value 8.2 mg / L); the pH value module measures 5.5-9.5, the accuracy is ±0.01 pH, and the collection frequency is 5 minutes / time (such as 9:00 collection value 7.85 pH); the ammonia nitrogen concentration module measures 0-10 mg / L, the accuracy is ±0.05 mg / L, and the collection frequency is 10 minutes / time (such as 9:00 collection value 0.3 mg / L); the water temperature module measures 0-40℃, the accuracy is ±0.1℃, and the collection frequency is 5 minutes / time (such as 9:00 collection value 25.3℃); the salinity module measures 0-35‰, the accuracy is ±0.1‰, and the collection frequency is 10 minutes / time (such as 9:00 collection value 15.2‰). All sensors upload data to the data collection gateway in real time through underwater LoRa communication (transmission distance ≤1km, anti-interference level IP68), and generate real-time water quality monitoring data set containing time stamp, sensor number, 5 types of parameters.

[0017] Step S12: acquiring the growth state data of the cultured organisms corresponding to the swimming trajectory, feeding frequency, growth size and stress response of the cultured organisms in the fishery culture water area by using the underwater imaging device and the biological sensor; In the embodiment of the present application, the underwater imaging device adopts a high-definition waterproof camera (resolution 1920x1080, frame rate 25 fps, field of view 120°), which is installed at a density of 1 per 2 mu at 1.5 m underwater in the pond, with the lens facing the intensive activity area of the cultured organisms, and takes a 5-minute video data every 30 minutes; the biological sensor adopts an implantable RFID tag (working frequency 13.56 MHz, reading distance 0.5 m), which is worn by 500 grass carp (body weight 500-800 g) in the pond one by one, and the tag information is read in real time through an underwater card reader (laid at an interval of 30 m). The swimming trajectory of the cultured organisms is extracted from the video data: the grass carp contour is identified by frame difference method, and the coordinates of each fish per second (accuracy ±1 cm) are recorded to generate the X-Y plane swimming path (such as a certain fish from (10m, 15m) to (25m, 20m) from 9:00 to 9:05); the feeding frequency is counted: the opening and closing action of the fish mouth is identified, and the number of opening and closing per hour is recorded (such as a certain fish feeding 28 times from 9:00 to 10:00); the growth size is measured: the fish body length is calculated by image calibration (1 cm corresponds to 10 pixel numbers) (such as a certain fish length 35 cm); the stress reaction is judged: when the swimming speed suddenly exceeds 0.5 m / s or the stationary time exceeds 5 minutes, it is judged as stress state (such as a certain fish appears 1 stress reaction at 9:15). All data are integrated to generate the growth state data set of the cultured organisms, and each record contains the organism ID, time stamp and 4 types of state parameters.

[0018] Step S13: Collecting environmental and meteorological data corresponding to light intensity, precipitation, wind speed and air pressure in the fishery culture water area through the weather station and environmental monitoring equipment; In the embodiment of the present application, a small weather station (installation height 2 m, protection level IP65) is arranged 10 m away from the bank of the breeding pond, which integrates light intensity, precipitation, wind speed, and air pressure detection modules; at the same time, one environmental monitoring device (installation height 1.5 m) is arranged in each of the five corners around the pond to assist in collecting local environmental data. The light intensity module measures in the range of 0-200000 lux, with an accuracy of ±100 lux, and collects data every 10 minutes (for example, the value collected at 9:00 is 85000 lux); the precipitation module measures in the range of 0-50 mm / h, with an accuracy of ±0.1 mm, and collects data every 10 minutes (for example, the value collected at 9:00 is 0 mm); the wind speed module measures in the range of 0-30 m / s, with an accuracy of ±0.1 m / s, and collects data every 5 minutes (for example, the value collected at 9:00 is 2.3 m / s); the air pressure module measures in the range of 800-1100 hPa, with an accuracy of ±0.1 hPa, and collects data every 10 minutes (for example, the value collected at 9:00 is 1013.2 hPa). The data collected by the environmental monitoring device and the weather station are averaged (for example, the average wind speed of the five devices is 2.2 m / s, and the average wind speed of the weather station is 2.3 m / s, which is averaged to 2.25 m / s), to generate environmental and meteorological data containing time stamp, four types of parameters, to ensure that the data covers the overall environmental conditions of the pond.

[0019] Step S14: Abnormal value detection and smoothing processing are performed on the real-time water quality monitoring data, the growth state data of the breeding organisms are normalized, and the environmental meteorological data are spatio-temporally aligned to generate standardized water quality feature data set, organism growth feature data set, and environmental influence factor data set.

[0020] In the embodiments of the present application, by performing outlier detection on real-time water quality monitoring data: adopting 3σ criterion, calculating the mean and standard deviation of parameters such as dissolved oxygen and pH value (for example, the mean of dissolved oxygen is 8.0 mg / L, and the standard deviation is 0.5 mg / L), and determining the data (for example, 10.0 mg / L) exceeding the range of mean ± 3σ as outliers and replacing it with linear interpolation of adjacent time data; then smoothing by 5-point moving average method (for example, dissolved oxygen data 8.2, 8.3, 8.1, 8.4, 8.2, average value 8.24 mg / L), to generate a standardized water quality feature data set (parameter range: dissolved oxygen 0-20 mg / L, pH 5.5-9.5, etc., retaining two decimal places). The growth state data of cultured organisms are normalized: the swimming track length (0-100 m), feeding frequency (0-100 times / hour), growth size (0-50 cm), and stress reaction frequency (0-10 times / hour) are mapped to the 0-1 interval by linear transformation (for example, 35 cm growth size normalized value 0.7), to generate a biological growth feature data set. The environmental meteorological data are spatio-temporally aligned: taking the water quality data collection timestamp (5 minutes / time) as the basis, the 10 minutes / time of light, precipitation, and air pressure data are interpolated by time (for example, 9:05 light value is calculated by linearly calculating the data of 9:00 and 9:10), and matched with the 5 minutes / time of wind speed data to ensure that all parameter timestamps are consistent, to generate an environmental impact factor data set (parameters retaining two decimal places, time interval 5 minutes).

[0021] Further, step S2 includes the following steps: Step S21: extracting the water quality parameter time series of dissolved oxygen, pH value, and ammonia nitrogen concentration corresponding to each monitoring point from the standardized water quality feature data set, and dividing them into multiple analysis periods according to time granularity; based on multiple analysis periods, dividing the dissolved oxygen time series into day and night cycles to calculate the difference between the maximum and minimum values of dissolved oxygen in each period, obtaining the dissolved oxygen day and night fluctuation amplitude; at the same time, analyzing the periodic change rule corresponding to the fluctuation amplitude, to generate dissolved oxygen fluctuation feature parameters; In the embodiment of the present application, by extracting the dissolved oxygen time sequence of the 9 monitoring points (such as 24-hour data of No. 1 monitoring point: 9:008.2 mg / L, 9:058.3 mg / L, …, 8:557.9 mg / L of the next day) from the standardized water quality feature data set (5 minutes / time dissolved oxygen, pH value, 10 minutes / time ammonia nitrogen concentration, 24 hours, 288 groups of dissolved oxygen data, 288 groups of pH data, 144 groups of ammonia nitrogen data), the time granularity of 24 hours is divided into 1 analysis period (if the data amount is sufficient, 7 periods can be divided to form weekly analysis). Based on the period, the day and night are divided: 6:00-18:00 is day (12 hours, 144 groups of data), 18:00-6:00 of the next day is night (12 hours, 144 groups of data); the difference between the maximum value (14:009.5 mg / L) and the minimum value (4:006.8 mg / L) of the dissolved oxygen in the period is calculated =9.5-6.8=2.7 mg / L, which is the diurnal fluctuation range of the dissolved oxygen. Analyze the periodic variation law: the fluctuation ranges of the continuous 7-day period are 2.7, 2.6, 2.8, 2.5, 2.9, 2.7, 2.6 mg / L, the mean value is 2.7 mg / L, the standard deviation is 0.12 mg / L, the dissolved oxygen fluctuation feature parameter (diurnal fluctuation range 2.7 mg / L, weekly mean value 2.7 mg / L, weekly standard deviation 0.12 mg / L) is generated, and the parameter is kept to one decimal place.

[0022] Step S22: linear fitting and trend analysis are performed on the pH value time sequence to calculate the change slope of the pH value corresponding to the unit time, obtain the pH value drift rate, and capture the mutation point corresponding to the pH value drift rate through the sliding window to generate the pH value dynamic drift feature parameter; In the embodiment of the present application, by linear fitting on the time sequence of pH value of the first monitoring point (24 hours 288 groups of data: 9:00 7.85, 9:05 7.86…8:55 the next day 7.83), taking time as abscissa (unit: hour, 9:00 as 0, 8:55 the next day as 23.92) and pH value as ordinate, the linear equation y=-0.0012x+7.852 is obtained by least square fitting. The slope of the equation-0.0012 pH / h is the pH value drift rate per unit time, and the negative value indicates that the pH value slowly decreases with time. A 30-minute sliding window (including 6 groups of data) is set to capture the mutation point: calculate the pH value change in each window, when the pH value in a window (such as 18:00-18:30) decreases from 7.82 to 7.75, the change is-0.07 pH, which is out of the normal fluctuation range (±0.02 pH), and is determined as a mutation point, and the mutation time 18:15 and the mutation amplitude-0.07 pH are recorded. Integrate the drift rate and the mutation point information to generate the pH value dynamic drift characteristic parameters (drift rate-0.0012 pH / h, 24-hour mutation times 1, maximum mutation amplitude-0.07 pH), and the parameters are kept to four decimal places or two decimal places.

[0023] Step S23: based on the spatial distribution corresponding to the time sequence of ammonia nitrogen concentration, the diffusion model is used to calculate the diffusion coefficient of ammonia nitrogen concentration in the fish farming water area, and the concentration gradient difference corresponding to different monitoring points is combined to generate the ammonia nitrogen concentration spatial diffusion characteristic parameter; at the same time, the dissolved oxygen fluctuation characteristic parameter, the pH value dynamic drift characteristic parameter and the ammonia nitrogen concentration spatial diffusion characteristic parameter are subjected to dimension normalization processing, and the characteristic fusion is carried out according to the preset weight to generate the basic water quality characteristics; In the embodiment of the application, the diffusion coefficient is calculated by using the improved Fick diffusion model based on the spatial distribution of the time sequence of ammonia nitrogen concentration (9 monitoring points 144 group data, such as 9:00 concentration [0.3, 0.31, 0.33, 0.29, 0.3, 0.32, 0.28, 0.29, 0.31] mg / L). Taking monitoring points No. 1 and No. 2 as an example, the concentration difference is 0.01 mg / L, the interval is 50 m, and the water flow velocity is 0.05 m / s. The molecular diffusion coefficient is 1.106*10^-9 m² / s, the turbulent diffusion coefficient is 1.25 m² / s, and the comprehensive diffusion coefficient is 1.125 m² / s. The concentration gradient difference of each monitoring point is calculated: the gradient of No. 1 and No. 5 is 1.35*10^-4 mg / (L·m), the gradient of No. 2 and No. 6 is 1.4*10^-4 mg / (L·m), and the average is 1.375*10^-4 mg / (L·m). The ammonia nitrogen concentration spatial diffusion characteristic parameters (comprehensive diffusion coefficient 1.125 m² / s, average concentration gradient 1.375*10^-4 mg / (L·m), and diffusion direction X axis positive direction) are generated. The three types of characteristic parameters are dimensionally normalized (mapped to 0-1): the dissolved oxygen fluctuation amplitude is 2.7 mg / L (normalized 0.68), the pH drift rate is -0.0012 pH / h (normalized 0.32), and the ammonia nitrogen diffusion coefficient is 1.125 m² / s (normalized 0.75). According to the preset weight (dissolved oxygen 0.4, pH 0.3, and ammonia nitrogen 0.3), the weighted fusion is performed, the basic water quality characteristic = 0.68*0.4+0.32*0.3+0.75*0.3=0.272+0.096+0.225=0.593, and the three decimal places are retained.

[0024] Step S24: based on the biological growth characteristic data set, the growth stage division of the cultured organism is analyzed, and the activity intensity index and metabolic rate parameter corresponding to each growth stage are calculated; based on the activity intensity index and metabolic rate parameter corresponding to each growth stage, the water quality time sequence correlation analysis of the basic water quality characteristic is performed to mine the lag response relationship between the activity intensity of the cultured organism and the change of the water quality parameter, and the biological-water quality correlation factor is generated. In the embodiment of the present application, by dividing the growth stage based on the biological growth characteristic data set (grass carp weight 500-800g, swimming trajectory, feeding frequency, etc.): juvenile stage (500-600g), middle fish stage (600-700g), adult fish stage (700-800g), the current is the juvenile stage. Calculate the activity intensity index of the juvenile stage: 24-hour average 0.65 (range 0-1), metabolic rate parameter 1.15 mg / (kg h) (basal metabolism 0.5 mg / (kg h), activity metabolic multiple 2.3). Use cross-correlation analysis to perform water quality time series correlation: take the activity intensity index (lag 10 minutes) and the ammonia nitrogen change rate as variables, calculate the correlation coefficient 0.85, construct the mapping model y=0.8x-0.334 (R²=0.82), generate preliminary correlation factors under different activity intensities (high intensity 0.970, medium intensity 0.945, low intensity 0.920), take the average 0.945 as the biological-water quality correlation factor, keep three decimal places, the factor reflects the influence intensity of biological activity on water quality change.

[0025] Step S25: Perform partial least squares regression analysis on the corresponding illumination, precipitation, wind speed and air pressure parameters in the environmental impact factor data set and the basic water quality characteristics to calculate the influence weight coefficients of each environmental factor on the water quality parameters; fuse the biological-water quality correlation factor and the influence weight coefficients of each environmental factor on the water quality parameters to construct a water quality-biology-environment multi-dimensional correlation matrix containing spatial dimension, time dimension and biological dimension.

[0026] In the embodiment of the present application, by performing partial least squares regression analysis on the environmental impact factor data set (illumination 85000 lux, precipitation 0 mm, wind speed 2.3 m / s, air pressure 1013.2 hPa, 288 groups of data in 24 hours) and the basic water quality characteristics (0.593), a regression model is constructed: basic water quality characteristics=0.2×illumination+0.1×precipitation-0.15×wind speed+0.05×air pressure+0.1 (constant term). Calculate the influence weight coefficients of each environmental factor: illumination 0.2, precipitation 0.1, wind speed-0.15, air pressure 0.05 (weight sum 0.2), the greater the absolute value of the weight, the stronger the influence, and the negative value represents the inhibition effect. Fuse the biological-water quality correlation factor (0.945) and the environmental weight coefficients to construct a 3x4 water quality-biology-environment multi-dimensional correlation matrix: the rows represent the spatial dimension (average of 9 monitoring points), the time dimension (24-hour average), and the biological dimension (juvenile stage), the columns represent the dissolved oxygen, pH value, ammonia nitrogen, and comprehensive water quality, and the matrix elements are the correlation intensities under the corresponding dimensions (such as the spatial-ammonia nitrogen element=ammonia nitrogen diffusion coefficient weight×environmental wind speed weight×biological correlation factor=0.3×(-0.15)×0.945≈-0.042), the matrix elements keep three decimal places, and the correlation relationship among the multi-dimensions is fully presented.

[0027] Further, the ammonia nitrogen concentration time sequence in step S23 corresponds to the spatial distribution, and the diffusion model is used to calculate the diffusion coefficient of ammonia nitrogen concentration in the aquaculture water area, and the concentration gradient difference corresponding to different monitoring points is combined to generate ammonia nitrogen concentration spatial diffusion characteristic parameters, including the following steps: The ammonia nitrogen concentration time sequence and spatial coordinate distribution information corresponding to each monitoring point are extracted from the standardized water quality feature data set, and the ammonia nitrogen concentration time-space distribution matrix is constructed based on the ammonia nitrogen concentration time sequence and spatial coordinate distribution information. In the embodiment of the present application, from the standardized water quality feature data set (dissolved oxygen, pH, etc. 5 minutes / time, ammonia nitrogen concentration 10 minutes / time), the ammonia nitrogen concentration time sequence (total 144 groups of data, such as 9:000.3mg / L, 9:100.32mg / L) of 9 monitoring points (3×3 grid, numbered 1-9, coordinates such as No. 1 (0m, 0m), No. 2 (50m, 0m), No. 3 (100m, 0m) and the like) for 24 hours and the spatial coordinate distribution information of each monitoring point (accurate to 1m) are extracted. Take time as the vertical axis (24 hours, divided into 144 time nodes at 10 minute intervals), monitoring point number as the horizontal axis (9 monitoring points), and ammonia nitrogen concentration value as the matrix element to construct a 144×9 ammonia nitrogen concentration space-time distribution matrix. For example, the first row (9:00) element of the matrix is [0.3, 0.31, 0.33, 0.29, 0.3, 0.32, 0.28, 0.29, 0.31], which corresponds to the ammonia nitrogen concentration of the 9 monitoring points, and the matrix element retains two decimal places, which fully presents the change rule of ammonia nitrogen concentration with time and space.

[0028] Further, based on the ammonia nitrogen concentration space-time distribution matrix, the ammonia nitrogen concentration difference between adjacent monitoring points is calculated, and the concentration decay rate per unit distance is calculated combined with the spatial distance of each monitoring point to generate preliminary concentration gradient parameters. In the embodiment of the present application, the ammonia nitrogen concentration difference between adjacent monitoring points is calculated based on the ammonia nitrogen concentration spatial-temporal distribution matrix: the concentration difference of 9:00 between horizontally adjacent monitoring points (such as No. 1 and No. 2, with a distance of 50 m) = 0.31-0.3=0.01 mg / L, the concentration difference of 9:00 between vertically adjacent monitoring points (such as No. 1 and No. 4, with a distance of 50 m) = 0.29-0.3=-0.01 mg / L, and the concentration difference of 9:00 between diagonally adjacent monitoring points (such as No. 1 and No. 5, with a distance of 50√2≈70.71 m) = 0.3-0.3=0 mg / L. According to the formula “concentration decay rate per unit distance = |concentration difference| / space distance”, the horizontal decay rate = 0.01 mg / L ÷ 50 m = 2×10-4mg / (L·m), the vertical decay rate = 0.01 mg / L ÷ 50 m = 2×10-4mg / (L·m), and the diagonal decay rate = 0 mg / L ÷ 70.71 m = 0 mg / (L·m). The decay rate of each time node (such as 9:00) is calculated for all adjacent monitoring points, and the average value is taken as the preliminary concentration gradient parameter of the time node. The preliminary gradient parameter of 9:00 is (2×10-4+2×10-4+0) / 3≈1.33×10-4mg / (L·m), and a preliminary concentration gradient parameter sequence of 144 time nodes within 24 hours is generated.

[0029] Further, the preliminary concentration gradient parameter is corrected by introducing the water flow velocity vector data corresponding to the aquaculture water area, so as to calculate the flow field-concentration coupling analysis calculation water flow disturbance coefficient, and the preliminary concentration gradient parameter is corrected and evaluated based on the water flow disturbance coefficient, so as to obtain the corrected actual concentration gradient parameter. In the embodiment of the present application, the underwater flow velocity sensor (disposed near 9 monitoring points, measuring range 0-1 m / s, accuracy ±0.01 m / s, 10 minutes / each collection) is used to obtain the flow velocity vector data of the aquaculture water area, such as the flow velocity of 9:001 monitoring point is 0.05 m / s, and the direction is along the positive direction of X axis (towards 2 monitoring point). Based on the flow field-concentration coupling analysis, the formula "flow disturbance coefficient=flow velocity×concentration difference direction coefficient" is used for calculation, the concentration difference direction is consistent with the flow direction, the coefficient is 1, the opposite is-1, and the vertical is 0, the disturbance coefficient between 1 and 2 monitoring points is 0.05 m / s×1=0.05 m / s. According to the formula "actual concentration gradient parameter=preliminary concentration gradient parameter×(1+flow disturbance coefficient)", the actual gradient parameter between 1 and 2 is 2×10-4×(1+0.05)=2.1×10-4mg / (L·m); the flow direction between 1 and 4 is along the negative direction of Y axis (opposite to the concentration difference direction), the disturbance coefficient is 0.03 m / s×(-1)=-0.03 m / s, and the actual gradient parameter is 2×10-4×(1-0.03)=1.94×10-4mg / (L·m). After correction of all adjacent monitoring points, the average value is taken, and the actual concentration gradient parameter of 9:00 is (2.1×10-4+1.94×10-4+0) / 3≈1.35×10-4mg / (L·m), which ensures that the gradient parameter reflects the influence of flow on concentration diffusion.

[0030] Further, the actual concentration gradient parameter is input into the improved Fick diffusion model to calculate the ammonia nitrogen molecular diffusion coefficient and the turbulent diffusion coefficient, and the comprehensive diffusion coefficient is generated by fusion. In the embodiment of the application, by inputting the actual concentration gradient parameter (such as 9:001.35x10^-4mg / (L.m)) into the improved Fick diffusion model, the model introduces a water flow disturbance correction term, and the formula is "diffusion flux = comprehensive diffusion coefficient x actual concentration gradient". The basic value of the ammonia nitrogen molecular diffusion coefficient calibrated in the laboratory is 1x10^-9m² / s, combined with the water temperature of 25.3℃ (9:00 data) correction, the molecular diffusion coefficient increases by 2% per 1℃ increase in temperature, and the corrected molecular diffusion coefficient = 1x10^-9x(1+(25.3-20) x 2%) = 1x10^-9x1.106≈1.106x10^-9m² / s. Based on the actual concentration gradient and the water flow velocity, the turbulent diffusion coefficient is calculated, and the formula is "turbulent diffusion coefficient = water flow velocity x monitoring point spacing x 0.5", and the turbulent diffusion coefficient at 9:00 is 0.05m / s x 50m x 0.5 = 1.25m² / s. According to the weight fusion (molecular diffusion coefficient weight 0.1, turbulent diffusion coefficient weight 0.9), the comprehensive diffusion coefficient = 1.106x10^-9x0.1 + 1.25x0.9≈0.0000000001106 + 1.125≈1.125m² / s, and the comprehensive diffusion coefficient sequence of each time node within 24 hours is generated by retaining three decimal places.

[0031] Further, based on the comprehensive diffusion coefficient, an ammonia nitrogen diffusion simulation model is constructed to simulate the concentration field distribution at different times and calculate the spatial heterogeneity index, and the ammonia nitrogen concentration spatial diffusion characteristic parameters including diffusion intensity, direction and range are generated combined with the actual concentration gradient parameter.

[0032] In the embodiment of the application, by constructing an ammonia nitrogen diffusion simulation model based on the comprehensive diffusion coefficient (such as 9:00 1.125m² / s), the model takes the aquaculture pond as a two-dimensional plane (126m x 126m), the grid division precision is 5m x 5m, and the initial ammonia nitrogen concentration is assigned according to the 9:00 time and space distribution matrix. The concentration field distribution at different times (9:00-10:00, output once every 10 minutes) is simulated, and the spatial heterogeneity index is calculated: the coefficient of variation is used for representation, and the formula is "coefficient of variation = standard deviation / mean", the ammonia nitrogen concentration mean at 9:00 is 0.3mg / L, the standard deviation is 0.015mg / L, and the heterogeneity index = 0.015 / 0.3 = 0.05. Combined with the actual concentration gradient parameter (1.35x10^-4mg / (L.m)), the diffusion intensity (the higher the comprehensive diffusion coefficient 1.125m² / s, the greater the intensity), the direction (diffusion is faster along the water flow direction X axis positive direction), and the range (1 hour within the 10m range from the 1st monitoring point to the 2nd monitoring point) are analyzed, and the ammonia nitrogen concentration spatial diffusion characteristic parameters are generated: diffusion intensity 1.125m² / s, diffusion direction X axis positive direction, diffusion range 10m / hour, spatial heterogeneity index 0.05, which provides diffusion law basis for subsequent water quality adaptability control.

[0033] Further, the water quality time sequence correlation analysis of the base water quality characteristics based on the activity intensity index and the metabolic rate parameter corresponding to each growth stage in step S24 to mine the lag response relationship between the activity intensity of the cultured organism and the water quality parameter change comprises the following steps: The base water quality characteristics are segmented into multiple analysis windows according to the time sequence, and the corresponding water quality parameter change rate in each window is calculated, and the activity intensity index corresponding to each growth stage is also segmented into the same window, so as to calculate the activity intensity mean value in the corresponding window; In the embodiment of the present application, the ammonia nitrogen concentration is extracted from the standardized water quality characteristic data set (10 minutes / time ammonia nitrogen concentration, 24 hours for a total of 144 groups of data) as the base water quality characteristic, segmented according to 1 hour as the analysis window, and 24 hours are divided into 24 windows (each window contains 6 groups of data). The ammonia nitrogen concentration change rate in each window is calculated: taking the first window (9:00-10:00) as an example, the data is 0.3, 0.32, 0.33, 0.35, 0.36, 0.37 mg / L, the change rate = (final value-initial value) / initial value / time = (0.37-0.3) / 0.3 / 1≈0.233 h⁻¹. At the same time, the activity intensity index of grass carp fry stage (body weight 500-600 g) is extracted from the biological growth characteristic data set (30 minutes / time activity intensity index, range 0-1, 24 hours for a total of 48 groups of data), and is also segmented according to 1 hour window (each window contains 2 groups of data), and the activity intensity mean value in the window is calculated: the first window data 0.6, 0.7, the mean value = (0.6+0.7) / 2=0.65. All 24 windows are processed in this way to generate a corresponding list of “window number-ammonia nitrogen change rate-activity intensity mean value”, ensuring that the time dimensions of the two are aligned.

[0034] Further, the corresponding correlation coefficient of the water quality parameter change rate and the activity intensity mean value in each window under different time lags is calculated by cross-correlation analysis, and the optimal lag time corresponding to the two is determined based on the correlation coefficient; the expected change amount of the water quality parameter corresponding to different activity intensities is calculated based on the optimal lag time, and a preliminary correlation factor is generated; In the embodiment of the present application, by performing cross-correlation analysis on each window, taking the 1st window (ammonia nitrogen change rate 0.233h⁻¹, average activity intensity 0.65) as an example, the time lag range is set to 0-30 minutes (interval 10 minutes), and the correlation coefficients under different lags are calculated: lag 0 minutes (activity intensity and water quality change at the same period) correlation coefficient 0.62; lag 10 minutes (activity intensity 10 minutes ahead of water quality change) correlation coefficient 0.85; lag 20 minutes correlation coefficient 0.71; lag 30 minutes correlation coefficient 0.58. Take the lag of 10 minutes corresponding to the maximum correlation coefficient 0.85 as the optimal lag time. Based on the optimal lag time, the fitting equation is established: ammonia nitrogen expected change amount = 0.35×average activity intensity (lag 10 minutes)-0.01, substitute the average activity intensity 0.65 of the 1st window, the expected change amount = 0.35×0.65-0.01≈0.218h⁻¹, the ratio of this value to the actual change rate 0.233h⁻¹ is 0.218 / 0.233≈0.936, which is used as the preliminary correlation factor. The 24 windows are calculated in this way to generate the preliminary correlation factor sequence (range 0.85-0.95).

[0035] Further, by introducing the mapping relationship between the metabolic rate parameter corresponding to each growth stage and the corresponding activity intensity index, and calculating the standardized metabolic rate index corresponding to each analysis window according to the mapping relationship; In the embodiment of the present application, the metabolic rate parameters of grass carp at different growth stages are obtained through biological experiments: the basic metabolic rate of juvenile stage (500-600g) is 0.5mg / (kg·h), and the activity metabolic rate is 1.5-3 times of the basic metabolic rate (increases with activity intensity). The mapping relationship between metabolic rate and activity intensity index is established: metabolic rate = 0.5×(1+2×activity intensity index), for example, when the activity intensity is 0.65, the metabolic rate = 0.5×(1+2×0.65)=0.5×2.3=1.15mg / (kg·h). For each analysis window, divide the calculated metabolic rate by the maximum metabolic rate of the growth stage (3×0.5=1.5mg / (kg·h)) to obtain the standardized metabolic rate index: the 1st window = 1.15 / 1.5≈0.767 (range 0-1). The 24 windows are processed in this way to generate the standardized metabolic rate index sequence, which ensures that the index can represent the metabolic difference under different activity intensities.

[0036] Further, the standardized metabolic rate index is taken as a weight factor, and the preliminary correlation factor is weighted and corrected based on the weight factor to obtain an intermediate correlation factor considering metabolic difference; historical breeding data are acquired, and the intermediate correlation factor considering metabolic difference is error-compensated based on the historical breeding data to generate a compensated intermediate correlation factor, which is normalized to generate a biological-water quality correlation factor.

[0037] In the embodiment of the application, the preliminary correlation factor is weighted and corrected by taking the standardized metabolic rate index as a weight factor: intermediate correlation factor = preliminary correlation factor x (1 + standardized metabolic rate index), and the first window = 0.936 x (1 + 0.767) = 0.936 x 1.767 = 1.654. The historical breeding data of the past three months (same pond, same growth stage) are called, the prediction error of the intermediate correlation factor in the same period and the actual correlation is calculated: the historical average error is 0.08, the intermediate correlation factor is corrected according to the formula "compensated intermediate correlation factor = intermediate correlation factor - 0.08", and the first window = 1.654 - 0.08 = 1.574. All the compensated intermediate correlation factors are mapped to the interval of 0-1 through linear transformation: the maximum value is 1.7, the minimum value is 1.5, and the normalized value of the first window = (1.574 - 1.5) / (1.7 - 1.5) = 0.074 / 0.2 = 0.372. The 24 windows are all normalized to generate a biological-water quality correlation factor sequence (range 0.3-0.4), which can accurately reflect the correlation strength of the activity, metabolism and ammonia nitrogen change of grass carp, and provide a basis for water quality control.

[0038] Further, the generating the preliminary correlation factor comprises the following steps: The corresponding water quality parameter change rate and the average activity intensity in each window are time-axis aligned, and multiple groups of lag time candidate values are set, and the Pearson correlation coefficient and the partial correlation coefficient corresponding to each group of lag time candidate values are calculated to construct a corresponding correlation coefficient matrix; In the embodiment of the application, by selecting the first window (9:00-10:00) from the 24 analysis windows, the ammonia nitrogen concentration change rate sequence (6 groups of data, time stamps 9:00, 9:10, 9:20, 9:30, 9:40, 9:50, corresponding change rates 0.067, 0.031, 0.061, 0.028, 0.027, 0.028 h-1) and the activity intensity mean sequence (2 groups of data, time stamps 9:00, 9:30, corresponding mean values 0.6, 0.7) in the window are time axis aligned, and the activity intensity mean is extended to 6 groups of data (9:000.6, 9:100.62, 9:200.64, 9:300.7, 9:400.68, 9:500.66) by time interpolation. The candidate values of the lag time are set to 0, 10, 20, and 30 minutes (a total of 4 groups), and the Pearson correlation coefficient and the partial correlation coefficient of the two sequences are calculated for each candidate value: when the lag is 0 minute, the Pearson correlation coefficient is 0.62 and the partial correlation coefficient is 0.58; when the lag is 10 minutes (the activity intensity sequence is shifted by 10 minutes, and is aligned with the water quality sequence 9:10-10:00), the Pearson correlation coefficient is 0.85 and the partial correlation coefficient is 0.81; when the lag is 20 minutes, the Pearson correlation coefficient is 0.71 and the partial correlation coefficient is 0.67; and when the lag is 30 minutes, the Pearson correlation coefficient is 0.58 and the partial correlation coefficient is 0.54. Taking the lag time as the row and the correlation coefficient type as the column, a 4x2 correlation coefficient matrix is constructed, and the matrix elements are accurate to two decimal places, and the correlation strength under different lags is completely presented.

[0039] Further, based on the correlation coefficient matrix, a significance test is performed to screen out the lag time candidate values corresponding to the correlation coefficients that pass the confidence test, and to calculate the information gain values corresponding to the lag time candidate values that pass the test, so as to determine the optimal lag time with the maximum information gain value as the target; In the embodiment of the application, by performing a significance test based on the correlation coefficient matrix, the confidence level is set to 95%, the corresponding significance level is α=0.05, the sample size n=6 (the number of aligned data groups), and the critical value t0. 025(4)=2.776. According to the formula "t statistic = correlation coefficient x √(n-2) / √(1-correlation coefficient 2)", the Pearson correlation coefficient t statistic of lag 0 minutes = 0.62 x √4 / √(1-0.62 2) = 0.62 x 2 / √0.6156≈1.24 / 0.785≈1.58<2.776, does not pass the test; the t statistic of lag 10 minutes = 0.85 x 2 / √(1-0.85 2) = 1.7 / √0.2775≈1.7 / 0.527≈3.23>2.776, passes the test; the t statistic of lag 20 minutes = 0.71 x 2 / √(1-0.71 2) = 1.42 / √0.4959≈1.42 / 0.704≈2.02<2.776, does not pass the test; the t statistic of lag 30 minutes = 0.58 x 2 / √(1-0.58 2) = 1.16 / √0.6636≈1.16 / 0.815≈1.42<2.776, does not pass the test. Only the lag 10 minutes candidate value is retained, and the information gain value thereof is calculated: taking the average activity intensity as a feature and the water quality change rate as a target variable, the information gain = entropy(water quality change rate) - conditional entropy(water quality change rate | average activity intensity lag 10 minutes), the entropy value is 0.693, the conditional entropy is 0.215, and the information gain = 0.693-0.215 = 0.478. This value is the information gain of the only candidate value that passes the test, so the optimal lag time is determined to be 10 minutes.

[0040] Further, based on the optimal lag time, an activity intensity-water quality response function is constructed to take the average activity intensity as an input variable and the water quality parameter change rate as an output variable, and the parameters corresponding to the activity intensity-water quality response function are trained through historical data to generate a mapping relationship model of activity intensity and water quality parameter change. In the embodiment of the present application, by constructing the activity intensity-water quality response function based on the optimal lag time of 10 minutes, a linear function form is selected: y=ax+b, wherein x is the activity intensity mean (lagged for 10 minutes), y is the ammonia nitrogen concentration change rate, a and b are the parameters to be trained. The historical breeding data of nearly 3 months (the same grass carp juvenile stage, a total of 720 analysis window data, each window contains x, y corresponding values) are called, and the least square method is used to train the parameters: the mean of x is 0.68, the mean of y is 0.21h⁻¹, the covariance of x and y is 0.012, the variance of x is 0.015, the parameter a=co variance / variance=0.012 / 0.015=0.8, the parameter b=y mean-a*x mean=0.21-0.8*0.68=0.21-0.544=-0.334. The response function y=0.8x-0.334 is obtained, the determination coefficient R²=1-Σ(y actual-y predicted)² / Σ(y actual-y mean)² is calculated, R²=0.82 is obtained, which indicates that the model can explain 82% of the water quality change rate variation, the mapping relationship model of activity intensity and water quality parameter change is generated, and the model parameters are accurate to three decimal places.

[0041] Further, the different activity intensity interval values are input into the mapping relationship model, the corresponding water quality parameter change rate predicted values are calculated, and the water quality parameter expected change amount under different activity intensities is calculated by combining the water quality parameter reference value, to generate the preliminary correlation factor containing the time lag characteristic.

[0042] In the embodiment of the present application, by dividing the activity intensity interval into low (0.2-0.4), medium (0.4-0.6), and high (0.6-0.8) three intervals, the midpoint value in each interval is taken as the input: low intensity 0.3, medium intensity 0.5, and high intensity 0.7. The midpoint value is input into the mapping relationship model y=0.8x-0.334, and the water quality parameter change rate prediction value is calculated: low intensity y=0.8x0.3-0.334=0.24-0.334=-0.094h-1, medium intensity y=0.8x0.5-0.334=0.4-0.334=0.066h-1, and high intensity y=0.8x0.7-0.334=0.56-0.334=0.226h-1. The ammonia nitrogen concentration reference value is set to 0.3 mg / L (fishery safety threshold), and the expected change amount is calculated according to the formula "expected change amount=reference valuexchange rate prediction valuextime (1 hour)": low intensity expected change amount=0.3x(-0.094)x1≈-0.028 mg / L, medium intensity=0.3x0.066x1≈0.020 mg / L, and high intensity=0.3x0.226x1≈0.068 mg / L. The ratio of the expected change amount to the actual change rate is taken as the preliminary correlation factor (for example, the actual change rate under high intensity is 0.233h-1, and the ratio is 0.226 / 0.233≈0.970), and a preliminary correlation factor sequence containing time lag characteristics (10 minutes) is generated, with a factor range of 0.92-0.98, accurately reflecting the expected correlation of water quality changes under different activity intensities.

[0043] Further, step S3 comprises the following steps: Step S31: constructing a water quality adaptability simulation model corresponding to the physical diffusion module, biological metabolism module, and environmental impact module based on the water quality-biology-environment multidimensional correlation matrix; In the embodiment of the present application, by constructing a water quality adaptability simulation model based on a water quality-biology-environment multi-dimensional correlation matrix (3x4 matrix, containing space, time, biological dimensions and correlation strength of dissolved oxygen, pH, ammonia nitrogen, and comprehensive water quality), three core modules are included. Physical diffusion module: an improved Fick diffusion model is used, inputting the spatial diffusion characteristic parameters of ammonia nitrogen concentration (comprehensive diffusion coefficient 1.125 m² / s, average concentration gradient 1.375x10^-4 mg / (L.m)), simulating the spatial diffusion process of ammonia nitrogen in the water area, and outputting the ammonia nitrogen concentration values at different time points; biological metabolism module: based on the biological-water quality correlation factor (0.945), inputting the grass carp juvenile metabolic rate parameter (1.15 mg / (kg.h)) and activity intensity index (0.65), simulating the influence of biological metabolism on water quality, and outputting the ammonia nitrogen concentration increment caused by metabolism; environmental impact module: inputting the environmental impact factor weight coefficient (light 0.2, wind speed-0.15, etc.), simulating the light on the promotion of dissolved oxygen (for every 10000 lux increase in light, the dissolved oxygen rises by 0.5 mg / L), and the wind speed on the promotion of water quality mixing (for every 1 m / s increase in wind speed, the ammonia nitrogen diffusion coefficient increases by 0.1 m² / s), and outputting the water quality parameter changes caused by environmental factors. The three modules are operated cooperatively to realize the dynamic simulation of water quality parameters under the action of physics, biology, and environment.

[0044] Step S32: generate multiple sets of environmental variable combinations by the Latin hypercube sampling method, and input the multiple sets of environmental variable combinations into the corresponding water quality adaptability simulation model for simulation operation to obtain the corresponding water quality parameter evolution curves under different environmental variable combinations; In the embodiment of the present application, by the Latin hypercube sampling method, the environmental variables (light, precipitation, wind speed, and air pressure) are sampled, and 5 levels are set for each variable: light 50000-150000 lux (interval 25000 lux), precipitation 0-4 mm / h (interval 1 mm / h), wind speed 1-5 m / s (interval 1 m / s), and air pressure 1000-1020 hPa (interval 5 hPa), a total of 5^4=625 sets of environmental variable combinations (ensuring coverage of the entire variable domain and uniform sampling). Each set of combination is input into the water quality adaptability simulation model for simulation operation, the simulation period is set to 30 days, the time step is 1 hour, and the dissolved oxygen, pH value, and ammonia nitrogen concentration evolution curves at 24 time points per day are output. For example, the simulation result of a set of combination (light 80000 lux, precipitation 1 mm / h, wind speed 2 m / s, and air pressure 1010 hPa) is: the dissolved oxygen curve fluctuates day and night (rising to 9.0 mg / L during the day and falling to 7.0 mg / L at night), the ammonia nitrogen concentration curve gradually converges from the initial 0.5 mg / L to 0.3 mg / L, and the pH value curve slowly decreases (from 7.9 to 7.7). 625 sets of combinations correspond to 625 sets of water quality parameter evolution curves.

[0045] Step S33: Perform water quality simulation statistical analysis on the corresponding water quality parameter evolution curves under different environmental variable combinations to calculate the steady-state values, fluctuation periods, and convergence rates of the water quality parameter evolution curves, and generate a water quality dynamic simulation dataset; In the embodiment of the present application, through simulation statistical analysis on the water quality parameter evolution curves of 625 environmental variable combinations, the dissolved oxygen steady-state value is calculated: taking the average value of dissolved oxygen in the last 7 days (such as the average value of 8.0 mg / L in the last 7 days of a certain curve); the pH value fluctuation period is calculated: through Fourier transform analysis of the periodicity of the curve (such as the fluctuation period of 24 hours of a certain curve, which is consistent with the circadian rhythm); the ammonia nitrogen concentration convergence rate is calculated: taking the reciprocal of the time required for ammonia nitrogen to decrease from the initial value to 90% of the steady-state value (such as 100 hours required for a certain curve to decrease from 0.5 mg / L to 0.33 mg / L, and the convergence rate is 0.0017 mg / (L·h)). The above indexes are calculated for each combination of dissolved oxygen, pH value, and ammonia nitrogen concentration, respectively, to generate a water quality dynamic simulation dataset containing 625 records, each record containing environmental variable combinations, steady-state values, fluctuation periods, and convergence rates of three types of water quality parameters, ensuring that the dataset can reflect the water quality dynamic characteristics under different environments.

[0046] Step S34: Coupling analysis of the water quality dynamic simulation dataset and the corresponding growth rate and survival rate indexes in the biological growth characteristic dataset is performed to construct a water quality parameter-biological growth response surface; In the embodiment of the present application, the grass carp growth rate (range 3-7 g / d) and survival rate (range 90%-99%) of 625 corresponding scenes are extracted from the biological growth characteristic dataset, and are converted into standardized biological growth indexes (growth index = (growth rate-3) / 4, survival rate index = (survival rate-90) / 9) in the 0-1 interval through linear normalization. Coupling analysis is performed on the key parameters (dissolved oxygen steady-state value, ammonia nitrogen convergence rate) of the water quality dynamic simulation dataset and the standardized biological growth indexes: taking dissolved oxygen and ammonia nitrogen rate as independent variables and growth index as dependent variable, the response surface is constructed by the method of steps S344-S345. First, the comprehensive water quality influence factor of dissolved oxygen (marginal contribution coefficient 0.12) and ammonia nitrogen rate (marginal contribution coefficient -0.0085) is calculated, and discrete data points (a total of 625×5+625=3125 points) are supplemented through three-dimensional interpolation, and the dimension is reduced through principal component analysis (extracting the first two principal components, explaining 88% of the variance), and the continuous water quality parameter-biological growth response surface is generated by inputting the radial basis function neural network training, and the surface shows that when the dissolved oxygen increases and the ammonia nitrogen rate decreases, the growth index significantly increases (such as dissolved oxygen 8.0 mg / L, ammonia nitrogen rate 0.008 mg / L corresponding to growth index 0.5).

[0047] Step S35: Determine the water quality parameter combination that makes the biological growth index optimal by water quality parameter-biological growth response surface optimization, and generate the optimal water quality parameter interval for the survival of the cultured organism.

[0048] In the embodiment of the present application, by using grid search method to optimize on the water quality parameter-biological growth response surface, dissolved oxygen (6.0-10.0 mg / L, step size 0.1 mg / L), ammonia nitrogen convergence rate (0.001-0.012 mg / (L·h), step size 0.0001 mg / (L·h)) are divided into grids, the standardized biological growth index (the average of growth index and survival rate index) of each grid point is calculated, and the parameter combination corresponding to the maximum index is found. The optimization result shows that when the dissolved oxygen is 8.5-9.5 mg / L and the ammonia nitrogen convergence rate is 0.006-0.008 mg / L, the biological growth index is above 0.8 (growth rate 6-7 g / d, survival rate 97%-99%), which is the optimal growth state. Further combined with the optimal range of pH value (7.8-8.0), the optimal water quality parameter interval for the survival of the cultured organism is integrated: dissolved oxygen 8.5-9.5 mg / L, pH value 7.8-8.0, ammonia nitrogen concentration ≤0.35 mg / L, ammonia nitrogen convergence rate 0.006-0.008 mg / (L·h), which provides a clear target for water quality regulation and ensures the growth of the cultured organism in the optimal water quality environment.

[0049] Further, step S34 includes the following steps: Step S341: Extract the key water quality parameter time series data under each simulation scenario from the water quality dynamic simulation data set, including the steady-state value of dissolved oxygen, the pH value fluctuation period, the ammonia nitrogen concentration convergence rate and the water temperature change amplitude; calculate the time weighted mean of each water quality parameter based on the time series data, and generate the water quality parameter feature set; at the same time, extract the growth rate and survival rate of the cultured organism in the corresponding simulation period from the biological growth feature data set, and convert the above biological indexes into standardized biological growth index in the interval of 0-1 through normalization algorithm to form the biological growth feature set; In the embodiment of the present application, when extracting the time series data of key water quality parameters of each simulation scenario (such as low oxygen, high ammonia nitrogen, and normal water quality) from the water quality dynamic simulation data set (simulation period 30 days, 24 groups of data per day, containing dissolved oxygen, pH value, ammonia nitrogen concentration, water temperature): the steady-state value of dissolved oxygen (the middle value of the fluctuation range per day, such as 8.0 mg / L on the first day, 7.9 mg / L on the second day, …, 8.1 mg / L on the thirtieth day of the normal scenario), the pH value fluctuation period (calculated by Fourier transform, 24 hours for the normal scenario and 18 hours for the low oxygen scenario), the ammonia nitrogen concentration convergence rate (the rate from the initial 0.5 mg / L to the steady-state 0.3 mg / L, 0.008 mg / (L·h) for the normal scenario), and the water temperature change amplitude (the difference between the highest and lowest temperature per day, 2.5℃ for the normal scenario). The time-weighted mean formula (the weight is the number of data collection times per day, 24 times / day) is used to calculate: the time-weighted mean of dissolved oxygen = (8.0×24+7.9×24+…+8.1×24) / (30×24) = 8.0 mg / L, the mean pH value fluctuation period is 24 hours, the mean ammonia nitrogen convergence rate is 0.008 mg / (L·h), and the mean water temperature amplitude is 2.5℃, generating a water quality parameter feature set. The grass carp 30-day growth rate (from 500 g to 650 g, with an average daily increase of 5 g / d) and survival rate (98%) are extracted from the biological growth characteristic data set, and linear normalization (growth rate 0-10 g / d mapped to 0-1, survival rate 90%-100% mapped to 0-1) is performed: standardized growth index = 5 / 10 = 0.5, standardized survival rate = (98-90) / (100-90) = 0.8, forming a biological growth characteristic set (growth index 0.5, survival rate 0.8).

[0050] Step S342: Spatiotemporal matching of the water quality parameter feature set and the biological growth characteristic set is performed to associate water quality parameters and biological growth indexes within the same simulation period based on timestamps, and a water quality parameter-biological growth correlation data set is constructed; Pearson correlation analysis is used to calculate the linear correlation coefficient of each water quality parameter and each standardized biological growth index, and water quality parameters with an absolute correlation coefficient greater than a preset threshold are selected as key influence parameters to generate a key water quality parameter subset; In this embodiment of the invention, a water quality parameter-biological growth association dataset of 30 sets of data is constructed by spatiotemporally matching the water quality parameter feature set (dissolved oxygen 8.0 mg / L, pH cycle 24h, ammonia nitrogen rate 0.008 mg / (L・h), water temperature range 2.5℃) with the biological growth feature set (growth index 0.5, survival rate 0.8) and using "simulated days" as the timestamp, the water quality parameters and biological indicators of each day are associated (e.g., dissolved oxygen 8.0 mg / L on day 1 corresponds to growth index 0.48 and survival rate 0.79; 7.9 mg / L on day 2 corresponds to 0.47, 0.78... and 8.1 mg / L on day 30 corresponds to 0.52, 0.81). Pearson correlation analysis was used to calculate linear correlation coefficients: dissolved oxygen was correlated with growth index at 0.82 and with survival rate at 0.75; pH cycle was correlated with growth index at 0.31 and with survival rate at 0.28; ammonia nitrogen rate was correlated with growth index at -0.78 and with survival rate at -0.72; and water temperature range was correlated with growth index at 0.45 and with survival rate at 0.41. A threshold of 0.6 for the absolute value of the correlation coefficients was preset, and dissolved oxygen (0.82) and ammonia nitrogen rate (-0.78) were selected as key influencing parameters, generating a subset of key water quality parameters (time-series data of dissolved oxygen and ammonia nitrogen rate) to ensure that subsequent analyses focus on high-impact parameters.

[0051] Step S343: Based on a subset of key water quality parameters, use orthogonal experimental design to divide the water quality parameter combination intervals, calculate the mean and variance of the standardized biological growth index within each interval, and generate a water quality parameter interval-biological growth mean mapping table; combine the data in the water quality parameter interval-biological growth mean mapping table, and construct a preliminary correlation model between water quality parameters and biological growth index through a local weighted regression algorithm, and output the marginal contribution coefficient of each key water quality parameter to the biological growth index; In the embodiment of the present application, by using the orthogonal test design method based on the key water quality parameter subset, 3 levels (6.0, 8.0, 10.0 mg / L) of dissolved oxygen and 3 levels (0.004, 0.008, 0.012 mg / (L.h)) of ammonia nitrogen rate are set, and 3*3=9 water quality parameter combination intervals (such as interval 1: dissolved oxygen 6.0 mg / L+ammonia nitrogen 0.004 mg / (L.h); interval 2: 6.0 mg / L+0.008 mg / (L.h) …… interval 9: 10.0 mg / L+0.012 mg / (L.h)) are constructed. The mean and variance of the standardized biological growth index in each interval are calculated: the mean of the growth index in interval 5 (8.0 mg / L+0.008 mg / (L.h)) is 0.5, and the variance is 0.01; the mean in interval 1 (6.0 mg / L+0.004 mg / (L.h)) is 0.3, and the variance is 0.02; the mean in interval 9 (10.0 mg / L+0.012 mg / (L.h)) is 0.4, and the variance is 0.015, to generate a water quality parameter interval-biological growth mean mapping table. Combined with the mapping table data, a preliminary correlation model is constructed by using the local weighted regression algorithm (the weight function is a Gaussian kernel function, and the bandwidth is 0.5): growth index=0.12* dissolved oxygen-8.5* ammonia nitrogen rate+0.05. The marginal contribution coefficient is calculated: when the ammonia nitrogen rate is fixed at 0.008 mg / (L.h), the growth index increases by 0.12 (marginal contribution coefficient 0.12) for every 1 mg / L increase in dissolved oxygen; when the dissolved oxygen is fixed at 8.0 mg / L, the growth index decreases by 0.0085 (marginal contribution coefficient-0.0085) for every 0.001 mg / (L.h) increase in ammonia nitrogen rate, and the marginal contribution coefficient of the key water quality parameter to the biological growth index is output, providing a quantitative basis for water quality regulation.

[0052] Step S344: The marginal contribution coefficient is used as a weight to weight and fuse the key water quality parameters to generate a comprehensive water quality influence factor; based on the comprehensive water quality influence factor and the standardized biological growth index, a three-dimensional interpolation algorithm is used to generate discrete comprehensive water quality influence factor-biological growth index data points; the discrete data points are processed by principal component analysis for dimension reduction, and the first two principal components are extracted as the horizontal and vertical coordinate axes of the response surface, and the data points with a principal component explanation variance ratio greater than a preset proportion are retained to form a response surface basic data set; In the embodiment of the present application, by the marginal contribution coefficient of the previously acquired key water quality parameters: dissolved oxygen marginal contribution coefficient 0.12 (positive value indicating promoting growth), ammonia nitrogen concentration convergence rate marginal contribution coefficient-0.0085 (negative value indicating inhibiting growth). The absolute value of the marginal contribution coefficient is taken as the weight, and the normalization processing is carried out first: dissolved oxygen weight=0.12 / (0.12+0.0085)≈0.12 / 0.1285≈0.934, ammonia nitrogen rate weight=0.0085 / 0.1285≈0.066. The key water quality parameters are weighted and fused, and the comprehensive water quality influence factor=(dissolved oxygen value×dissolved oxygen weight)+(ammonia nitrogen rate value×ammonia nitrogen rate weight×(-1))(negative sign conversion inhibitory effect into positive vectorization), such as interval 5 (dissolved oxygen 8.0 mg / L, ammonia nitrogen rate 0.008 mg / (L·h)): comprehensive water quality influence factor=8.0×0.934+(0.008×0.066×(-1))×(-1)≈7.472+0.0005≈7.4725. Based on the comprehensive water quality influence factor (range 6.0-10.0) and the standardized biological growth index (range 0.3-0.5), a linear three-dimensional interpolation algorithm is adopted, and 5 interpolation points are generated between every two adjacent orthogonal test intervals, such as interval 1 (comprehensive factor 6.0, growth index 0.3) and interval 2 (comprehensive factor 6.05, growth index 0.32), 6.01 / 0.304, 6.02 / 0.308, 6.03 / 0.312, 6.04 / 0.316, 6.05 / 0.32 data points are generated, a total of 9×5+9=54 discrete data points. The discrete data points are subjected to principal component analysis, the covariance matrix is calculated, and the first two principal components are extracted: the first principal component (PC1) accounts for 85% of the variance, and the second principal component (PC2) accounts for 12% of the variance. The preset variance explained threshold is 80%, and the PC1 and PC2 corresponding to the 54 data points are retained to form a response surface basic data set (each data point contains PC1 value, PC2 value, and standardized growth index).

[0053] Step S345: input the response surface basic data set into the improved radial basis function neural network, take the principal components as the input variables and the standardized biological growth index as the output variable, train the neural network model; optimize the model parameters through cross-validation to ensure that the model prediction error is less than the preset error threshold; based on the trained model, a continuous water quality parameter-biological growth response surface is generated.

[0054] In the embodiment of the present application, the response surface basic data set is divided into a training set (38 data points) and a verification set (16 data points) in a ratio of 7:3, and an improved radial basis function neural network is input. The neural network is set to have a three-layer structure: 2 neurons in the input layer (corresponding to PC1 and PC2), 15 neurons in the hidden layer (using a Gaussian radial basis function, and the kernel function width is set to 0.8), and 1 neuron in the output layer (corresponding to the normalized biological growth index). The model is trained using the training set data: the network weights are initialized, the radial basis function output of each hidden layer neuron is calculated, the predicted value is obtained through the linear output layer mapping, the mean square error (MSE) between the predicted value and the actual value is minimized using the gradient descent method, the learning rate is set to 0.01, and the iteration is performed 500 times until the MSE is stable below 0.001. The model parameters are optimized through cross-validation of the verification set: the kernel function width is adjusted to 0.75, the learning rate is adjusted to 0.008, the verification set MSE is reduced to 0.0008 after retraining, which is less than the preset error threshold 0.001, and the model converges. Based on the trained model, the normalized biological growth index prediction value is calculated point by point on the grid interval (step size 0.01) of PC1 (range -2.0-2.0) and PC2 (range -1.0-1.0), 10000 continuous data points are generated, the data points are connected by a three-dimensional surface drawing tool, and a continuous water quality parameter-biological growth response surface is generated. The surface clearly presents the dynamic relationship between PC1, PC2 combination and biological growth index, and can intuitively query the expected biological growth corresponding to any water quality parameter combination, providing a visual decision basis for water quality adaptability control.

[0055] Further, step S4 includes the following steps: Step S41: According to the optimal water quality parameter interval and the water quality dynamic simulation data set, a corresponding hierarchical control strategy is formulated to clearly define the control measures and execution priority corresponding to different water quality deviation levels; In the embodiment of the present application, by the optimal water quality parameter interval (dissolved oxygen 8.5-9.5 mg / L, pH 7.8-8.0, ammonia nitrogen ≤0.35 mg / L) and water quality dynamic simulation data set, a three-level regulation strategy is formulated. First deviation (slight deviation): single parameter exceeds the optimal interval within 10% (such as dissolved oxygen 8.0-8.5 mg / L, pH 7.6-7.8, ammonia nitrogen 0.35-0.4 mg / L), the regulation measures are "low intensity oxygenation (oxygenation machine power 0.8 kW) + trace pH adjustment (dosing amount 0.15 L / h)", and the execution priority is 3 (low priority, non-emergency response); second deviation (moderate deviation): single parameter exceeds 10%-20% or two parameters slight deviation (such as dissolved oxygen 7.5-8.0 mg / L, pH 7.4-7.6, ammonia nitrogen 0.4-0.45 mg / L), the regulation measures are "medium intensity oxygenation (1.5 kW) + conventional pH adjustment (0.3 L / h) + low flow circulation (15 m³ / h)", and the execution priority is 2 (medium priority, response within 4 hours); third deviation (serious deviation): single parameter exceeds 20% or three parameters deviation (such as dissolved oxygen <7.5 mg / L, pH <7.4, ammonia nitrogen >0.45 mg / L), the regulation measures are "high intensity oxygenation (2.0 kW) + high dose pH adjustment (0.5 L / h) + high flow circulation (30 m³ / h)", and the execution priority is 1 (high priority, emergency response within 1 hour). The strategy clearly corresponds the measures to the deviation, ensuring the regulation pertinence.

[0056] Step S42: The current water quality parameters corresponding to the fishery breeding water area are collected in real time, and the current water quality parameters are compared and analyzed based on the optimal water quality parameter interval to calculate the deviation values of each water quality parameter and the comprehensive deviation index, and a water quality regulation deviation value is generated; In the embodiment of the present application, the current water quality parameters are collected in real time by the distributed water quality sensor array: dissolved oxygen 7.8 mg / L, pH 7.5, ammonia nitrogen 0.42 mg / L. The deviation values of each parameter are calculated: dissolved oxygen deviation value = (7.8-9.0) / 9.0x100% ≈-13.3% (negative sign indicates lower than the optimal interval), pH deviation value = (7.5-7.9) / 7.9x100% ≈-5.1%, ammonia nitrogen deviation value = (0.42-0.35) / 0.35x100% ≈20%. The comprehensive deviation index is calculated by weighted summation: dissolved oxygen weight 0.4, pH 0.3, ammonia nitrogen 0.3, comprehensive deviation index = (-13.3% x 0.4) + (-5.1% x 0.3) + (20% x 0.3) ≈-5.32%-1.53%+6% ≈-0.85% (the smaller the absolute value, the lighter the deviation). The water quality regulation deviation value report is generated: single parameter deviation value (dissolved oxygen-13.3%, pH-5.1%, ammonia nitrogen 20%), comprehensive deviation index-0.85%, providing a quantitative basis for deviation level determination.

[0057] Step S43: Based on the water quality regulation deviation value, fuzzy clustering analysis is performed to determine the deviation level to which the current water quality state belongs; according to the deviation level, a corresponding set of regulation measures is matched from the hierarchical regulation strategy, and the expected regulation effect value of each regulation measure is calculated; a multi-objective optimization model is constructed based on the expected regulation effect value and the equipment operation energy consumption parameter, and the optimal regulation measure combination is obtained by solving; the optimal regulation measure combination is converted into specific operation parameters of the water quality regulation equipment, and a control instruction sequence including execution time, operation intensity and duration is generated; In the embodiment of the present application, by performing fuzzy clustering analysis based on the water quality regulation deviation value, the clustering centers are set to be first level (comprehensive deviation index ± 1%), second level (± 3%) and third level (± 5%), the Euclidean distance between the current comprehensive deviation index -0.85% and each center is calculated: 0.15 from the first level center, 1.15 from the second level center, and 3.15 from the third level center. The closest clustering center is the first level, but the ammonia nitrogen deviation is 20%, reaching the second level standard. Combined with the parameter deviation degree correction, it is determined as the second level deviation. The second level regulation measure set (medium intensity oxygenation 1.5kW, conventional pH regulation 0.3L / h, low flow circulation 15m³ / h) is matched from the hierarchical strategy, and the expected regulation effect value is calculated: through the water quality simulation model prediction, the dissolved oxygen is increased to 8.5mg / L, the pH is 7.8, and the ammonia nitrogen is 0.38mg / L after the execution of the measures, and the standard reaching rate is increased from the current 60% to 92%. A multi-objective optimization model (max standard reaching rate, min energy consumption) is constructed, and the measure parameters are substituted to solve, and the optimal regulation measure combination is obtained: oxygenation machine 1.4kW, pH dosage 0.28L / h, circulation pump 16m³ / h (energy consumption 22kWh / d, standard reaching rate 93%). The combination is converted into a control instruction sequence: "execution time 2024060110:00, oxygenation machine 1.4kW (for 4 hours), pH dosage 0.28L / h (for 4 hours), circulation pump 16m³ / h (for 6 hours)", and the instruction clearly executes the details.

[0058] Step S44: The control instruction sequence is issued to the water quality regulation equipment corresponding to the fishery breeding water area to perform adaptive control operation, and the equipment operation state, real-time water quality change data and biological growth prediction results are recorded synchronously; In the embodiment of the application, the control instruction sequence is issued to the corresponding water quality adjusting equipment through the data transmission module: the oxygenator receives the "1.4 kW, 10:00-14:00" instruction, starts and maintains the power on time; the pH regulator dosing pump receives the "0.28 L / h, 10:00-14:00" instruction, accurately controls the flow of the reagent; the circulating pump receives the "16 m³ / h, 10:00-16:00" instruction, and stably runs. The running state is recorded synchronously through the equipment state sensor: the oxygenator voltage is 220 V, the current is 6.4 A (power 1.4 kW), the dosing pump flow is 0.28 L / h, and the circulating pump speed is 1500 r / min (flow 16 m³ / h); the water quality sensor collects data every 10 minutes, records the changes after the regulation and control: 10:00 7.8 mg / L→12:00 8.3 mg / L→14:00 8.6 mg / L, pH 7.5→7.7→7.8, ammonia nitrogen 0.42→0.4→0.37; the biological growth model predicts the results: the grass carp growth rate is increased from 5 g / d to 5.8 g / d, and all data are stored in the local database in real time to ensure that the regulation and control process is traceable.

[0059] Step S45: The device running state, real-time water quality change data and biological growth prediction results before and after the regulation and control are multi-dimensionally displayed through data visualization technology to show the water quality regulation and control effect and the biological growth state prediction results, and a water quality regulation and control effect evaluation report is generated.

[0060] In the embodiment of the application, a multi-dimensional display interface is constructed by using data visualization technology: the device running state interface displays the time changes of the oxygenator power (10:00-14:00 stably 1.4 kW), the dosing pump flow (0.28 L / h) and the circulating pump flow (16 m³ / h) in a line chart; the water quality change interface compares the parameters (dissolved oxygen 7.8→8.6 mg / L, pH 7.5→7.8, ammonia nitrogen 0.42→0.37 mg / L) before and after the regulation and control in a column chart; and the biological growth prediction interface displays the survival rate (98%) in a pie chart and the growth rate trend (5→5.8 g / d) in a line chart. The water quality regulation and control effect evaluation report is generated: the deviation level before the regulation and control is second level, and the deviation level is reduced to first level 12 hours after the regulation and control, the compliance rate is increased from 60% to 93%, the energy consumption is 22 kWh / d (lower than the upper limit 30 kWh / d), and the biological growth index is increased by 16%; the report includes data charts, deviation level changes, energy consumption analysis, growth prediction, and clearly concludes that "the regulation and control is effective, and it is suggested to maintain the current measures for 24 hours and then switch to regular monitoring", which provides a reference for subsequent regulation and control.

[0061] Further, the multi-objective optimization model constructed based on the expected regulation and control effect value and the device running energy consumption parameter in step S43 includes the following steps: The expected regulation effect value is taken as the benefit index, and the equipment operation energy consumption parameter is taken as the cost index to construct a multi-objective optimization function; In the embodiment of the present application, the expected regulation effect value is taken as the benefit index, and is defined as the "water quality parameter compliance rate", that is, the time proportion (unit: %) of the dissolved oxygen, pH value and ammonia nitrogen concentration in the optimal interval (dissolved oxygen 8.5-9.5 mg / L, pH 7.8-8.0, ammonia nitrogen ≤0.35 mg / L) after regulation, the higher the compliance rate, the higher the benefit, and the value range is 0-100%. The equipment operation energy consumption parameter is taken as the cost index, covering the energy consumption of the oxygenation machine, pH regulator dosing pump and water quality circulating pump, and is defined as the "daily average energy consumption" (unit: kWh / d), the lower the energy consumption, the lower the cost, and the value range is 10-50 kWh / d. A multi-objective optimization function is constructed: objective 1 (maximize benefit) maxf1(x)=water quality parameter compliance rate; objective 2 (minimize cost) minf2(x)=daily average energy consumption, wherein x is the regulation decision variable (oxygenation machine power, pH regulator dosage, circulating pump flow), the function needs to consider both water quality compliance effect and operation economy, and avoid single pursuit of benefit leading to high energy consumption or single control of cost leading to non-compliance of water quality.

[0062] Further, the regulation effect threshold and the upper limit of energy consumption are set as constraint conditions to determine the decision variables and value ranges corresponding to the multi-objective optimization function; In the embodiment of the present application, the regulation effect threshold (constraint condition 1) is set: the water quality parameter compliance rate is greater than or equal to 90%, to ensure that the water quality after regulation is stable in the optimal interval and meets the growth needs of the breeding organisms; the upper limit of energy consumption (constraint condition 2) is set: the daily average energy consumption is less than or equal to 30 kWh / d, to control the operation cost within the range that the farm can bear. The decision variables and value ranges are determined: decision variable 1 "oxygenation machine power" (x1) corresponds to the oxygenation capacity, and the value range is 0.5-2.0 kW (the higher the power, the faster the dissolved oxygen is increased, and the higher the energy consumption); decision variable 2 "pH regulator dosage" (x2) uses sodium hydroxide solution for adjustment, and the value range is 0.1-0.5 L / h (the higher the dosage, the faster the pH is increased, and the higher the energy consumption associated with the consumption of the reagent); decision variable 3 "circulating pump flow" (x3) affects the water quality mixing efficiency, and the value range is 10-30 m³ / h (the higher the flow, the faster the ammonia nitrogen is diffused, and the higher the energy consumption). The values of all decision variables need to satisfy the constraint conditions at the same time to ensure that the optimization direction meets the actual application needs.

[0063] Further, a non-dominated sorting genetic algorithm is used to solve the multi-objective optimization function to obtain a Pareto optimal solution set; In the embodiment of the present application, the multi-objective optimization function is solved by using the non-dominated sorting genetic algorithm (NSGA-II). The algorithm parameters are set: population size 50 (50 groups of decision variable combinations are generated each time), iteration number 100 (to ensure convergence), crossover probability 0.8 (gene recombination probability), and mutation probability 0.05 (gene mutation probability). When initializing the population, 50 groups of decision variable combinations are randomly generated (such as x1=1.2kW, x2=0.3L / h, x3=20m³ / h), each group is substituted into the water quality regulation simulation model, and the corresponding f1 (compliance rate) and f2 (energy consumption) are calculated. Through non-dominated sorting (dividing the population into different non-dominated levels, the solution is better as the level is earlier) and crowding degree calculation (maintaining the diversity of solutions), the Pareto optimal solution set is obtained after 100 iterations, which contains 12 groups of non-dominated solutions, such as solution 1 (compliance rate 98%, energy consumption 28kWh / d), solution 2 (compliance rate 95%, energy consumption 22kWh / d), solution 3 (compliance rate 90%, energy consumption 18kWh / d), etc. Each solution cannot improve one target without reducing the other target, forming an optimal trade-off scheme set.

[0064] Further, the fuzzy analytic hierarchy process is used to comprehensively evaluate each scheme in the Pareto optimal solution set and calculate the comprehensive score of the scheme. In the embodiment of the present application, the 12 schemes in the Pareto optimal solution set are comprehensively evaluated based on the fuzzy analytic hierarchy process (FAHP). The evaluation index system is constructed: the first-level index "benefit dimension" (weight 0.6) and "cost dimension" (weight 0.4), the benefit dimension is set with "compliance rate stability" (weight 0.7) and "regulation response speed" (weight 0.3), and the cost dimension is set with "daily energy consumption" (weight 0.8) and "equipment maintenance cost" (weight 0.2) (maintenance cost is positively correlated with energy consumption, which is converted according to the proportion of energy consumption). Five fishery experts are invited to score each scheme (1-10 points, the higher the score, the better), such as solution 1 compliance rate stability 9 points, response speed 8 points, daily energy consumption 6 points, and maintenance cost 6 points. The index weight is calculated by fuzzy consistent matrix, and the comprehensive score is calculated by combining the score: solution 1 comprehensive score = (9x0.7+8x0.3)x0.6+(6x0.8+6x0.2)x0.4 = (6.3+2.4)x0.6+(4.8+1.2)x0.4 = 8.7x0.6+6x0.4 = 5.22+2.4 = 7.62 points; the scores of other schemes are calculated in the same way, solution 2 is 7.25 points, solution 3 is 6.88 points, etc. The higher the score, the better the comprehensive performance of the scheme.

[0065] Further, the scheme with the highest comprehensive score is selected as the optimal regulation measure combination, and the corresponding regulation parameter setting is output.

[0066] In the embodiment of the present application, by comparing the comprehensive scores of 12 groups of schemes in the Pareto optimal solution set, solution 1 (98% compliance rate, 28 kWh / d energy consumption) ranks first with 7.62 points, with the highest compliance rate (98%), although the energy consumption (28 kWh / d) is close to the upper limit but does not exceed, and the control response speed is fast (it only takes 30 minutes for the dissolved oxygen to rise from 7.0 mg / L to 8.5 mg / L), the maintenance cost is controllable (the average daily maintenance cost is about 5 yuan), and the comprehensive demand of "high compliance rate + reasonable energy consumption" is met. Select this scheme as the optimal control measure combination, and output the corresponding control parameter settings: oxygenation machine power 1.8 kW (to ensure that the dissolved oxygen is stable at about 9.0 mg / L), pH adjusting agent dosage 0.35 L / h (to ensure that the pH is stable at about 7.9), circulating pump flow 25 m³ / h (to ensure that the ammonia nitrogen concentration is stable at about 0.32 mg / L). The parameter settings need to be specific in numerical value, so as to facilitate the direct debugging of the equipment by the farm and realize the accurate control of water quality.

[0067] Therefore, from any viewpoint, the embodiments should be considered as exemplary and non-limiting, the scope of the present application being defined by the appended claims and not by the above description, and it is intended to encompass all variations falling within the meaning and the scope of the equivalent elements of the application file.

[0068] The above description is only a specific implementation of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A water quality adaptability simulation control method based on aquaculture scenarios, characterized in that, Includes the following steps: Step S1: Obtain real-time water quality monitoring data, aquaculture organism growth status data, and environmental meteorological data corresponding to the aquaculture water area, and perform standardized preprocessing on the real-time water quality monitoring data, aquaculture organism growth status data, and environmental meteorological data to generate standardized water quality characteristic dataset, organism growth characteristic dataset, and environmental impact factor dataset. Step S2: Based on the standardized water quality feature dataset, extract basic water quality features including the dynamic distribution of dissolved oxygen, the trend of pH fluctuation, and the gradient change of ammonia nitrogen concentration. Combine the biological growth feature dataset to explore the correlation between the activity intensity of aquaculture organisms and water quality parameters. At the same time, introduce the environmental impact factor dataset to construct a multidimensional correlation matrix of water quality-organisms-environment. Step S3: Construct a water quality adaptability simulation model using a multidimensional correlation matrix of water quality, organisms, and environment. Simulate the evolution of water quality parameters under different combinations of environmental variables based on the water quality adaptability simulation model to generate a dynamic water quality simulation dataset. Based on the coupling analysis of the dynamic water quality simulation dataset and the biological growth characteristic dataset, determine the optimal water quality parameter range for the survival of aquaculture organisms. Step S4: Formulate a dynamic control strategy based on the optimal water quality parameter range, and compare and analyze the real-time water quality monitoring data with the optimal water quality parameter range based on the dynamic control strategy to generate a water quality control deviation value. The system drives the water quality regulation equipment in the aquaculture area to perform adaptive control operations based on the water quality regulation deviation value, and displays the water quality regulation effect and the prediction results of biological growth status through a visual interface.

2. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Collect real-time water quality monitoring data, including dissolved oxygen, pH value, ammonia nitrogen concentration, water temperature and salinity, in aquaculture waters using a distributed water quality sensor array; Step S12: Use underwater imaging equipment and biosensors to acquire data on the swimming trajectory, feeding frequency, growth size, and stress response of cultured organisms in aquaculture waters. Step S13: Collect environmental meteorological data for aquaculture waters, including light intensity, precipitation, wind speed and air pressure, through meteorological stations and environmental monitoring equipment; Step S14: Perform outlier detection and smoothing on real-time water quality monitoring data, normalize the features of aquaculture organism growth status data, and perform spatiotemporal alignment on environmental meteorological data to generate standardized water quality feature datasets, organism growth feature datasets, and environmental impact factor datasets.

3. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Extract the time series of water quality parameters (dissolved oxygen, pH, and ammonia nitrogen concentration) corresponding to each monitoring point from the standardized water quality characteristic dataset, and divide them into multiple analysis periods according to time granularity; divide the dissolved oxygen time series into diurnal cycles based on multiple analysis periods to calculate the difference between the maximum and minimum dissolved oxygen values ​​in each cycle, thereby obtaining the diurnal fluctuation amplitude of dissolved oxygen; at the same time, analyze the periodic change pattern corresponding to the fluctuation amplitude to generate dissolved oxygen fluctuation characteristic parameters; Step S22: Perform linear fitting and trend analysis on the pH time series to calculate the slope of pH change per unit time, obtain the pH drift rate, and capture the abrupt change point corresponding to the pH drift rate through a sliding window to generate pH dynamic drift characteristic parameters. Step S23: Based on the spatial distribution corresponding to the time series of ammonia nitrogen concentration, the diffusion coefficient of ammonia nitrogen concentration in aquaculture waters is calculated using a diffusion model. Combined with the concentration gradient differences at different monitoring points, spatial diffusion characteristic parameters of ammonia nitrogen concentration are generated. At the same time, the dimensionality of dissolved oxygen fluctuation characteristic parameters, pH dynamic drift characteristic parameters, and ammonia nitrogen concentration spatial diffusion characteristic parameters is normalized, and the features are fused according to preset weights to generate basic water quality characteristics. Step S24: Analyze the growth stage division of aquaculture organisms based on the biological growth characteristic dataset, and calculate the activity intensity index and metabolic rate parameters corresponding to each growth stage; perform water quality time series correlation analysis on the basic water quality characteristics based on the activity intensity index and metabolic rate parameters corresponding to each growth stage, so as to explore the lag response relationship between the activity intensity of aquaculture organisms and the changes in water quality parameters, and generate biological-water quality correlation factors. Step S25: Perform partial least squares regression analysis on the corresponding light, precipitation, wind speed and air pressure parameters in the environmental impact factor dataset and the basic water quality characteristics to calculate the influence weight coefficients of each environmental factor on the water quality parameters; integrate the biological-water quality correlation factor and the influence weight coefficients of each environmental factor on the water quality parameters to construct a multidimensional correlation matrix of water quality-biology-environment that includes spatial, temporal and biological dimensions.

4. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 3, characterized in that, The step S23, which involves calculating the diffusion coefficient of ammonia nitrogen concentration in aquaculture waters based on the spatial distribution of ammonia nitrogen concentration time series and combining the concentration gradient differences at different monitoring points to generate spatial diffusion characteristic parameters of ammonia nitrogen concentration, includes the following steps: By extracting the time series and spatial coordinate distribution information of ammonia nitrogen concentration at each monitoring point from the standardized water quality characteristic dataset, and constructing the corresponding spatiotemporal distribution matrix of ammonia nitrogen concentration based on the time series and spatial coordinate distribution information of ammonia nitrogen concentration; The ammonia nitrogen concentration difference between adjacent monitoring points is calculated based on the spatiotemporal distribution matrix of ammonia nitrogen concentration, and the concentration attenuation rate per unit distance is calculated in combination with the spatial distance between each monitoring point to generate preliminary concentration gradient parameters. By introducing the water flow velocity vector data corresponding to the aquaculture water area, the initial concentration gradient parameters are corrected, the water flow disturbance coefficient is calculated through flow field-concentration coupling analysis, and the initial concentration gradient parameters are corrected and evaluated based on the water flow disturbance coefficient to obtain the corrected actual concentration gradient parameters. The actual concentration gradient parameters are input into the improved Fick diffusion model to calculate the molecular diffusion coefficient of ammonia nitrogen and the turbulent diffusion coefficient, and then the combined diffusion coefficient is generated. A simulation model of ammonia nitrogen diffusion is constructed based on the comprehensive diffusion coefficient. The concentration field distribution at different time periods is simulated and the spatial heterogeneity index is calculated. Combined with the actual concentration gradient parameters, spatial diffusion characteristic parameters of ammonia nitrogen concentration including diffusion intensity, direction and range are generated.

5. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 3, characterized in that, Step S24, which involves performing a time-series correlation analysis of basic water quality characteristics based on the activity intensity index and metabolic rate parameters corresponding to each growth stage, to uncover the hysteresis response relationship between the activity intensity of aquaculture organisms and changes in water quality parameters, includes the following steps: The basic water quality characteristics are divided into multiple analysis windows according to the time series and the change rate of water quality parameters in each window is calculated. At the same time, the activity intensity index corresponding to each growth stage is divided into the same window to calculate the mean activity intensity in the corresponding window. Cross-correlation analysis was used to calculate the correlation coefficient between the rate of change of water quality parameters and the mean activity intensity within each window at different time lags, and the optimal lag time was determined based on the correlation coefficient. Based on the optimal lag time, the expected change of water quality parameters under different activity intensities was calculated, and preliminary correlation factors were generated. By introducing the metabolic rate parameters corresponding to each growth stage and the corresponding activity intensity index, a mapping relationship between metabolic rate and activity intensity is established, and the standardized metabolic rate index corresponding to each analysis window is calculated based on the mapping relationship. The standardized metabolic rate index is used as a weighting factor, and the preliminary correlation factors are weighted and corrected based on the weighting factor to obtain the intermediate correlation factors that take into account metabolic differences. Historical aquaculture data is obtained, and the prediction error is calculated based on the historical aquaculture data to compensate for the error of the intermediate correlation factors that take into account metabolic differences, so as to generate the compensated intermediate correlation factors. At the same time, normalization is performed to generate the biological-water quality correlation factors.

6. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 5, characterized in that, The generation of preliminary correlation factors includes the following steps: Align the rate of change of water quality parameters with the mean value of activity intensity within each window on the time axis, set multiple sets of lag time candidate values, and calculate the Pearson correlation coefficient and partial correlation coefficient of the two sequences for each set of lag time candidate values ​​to construct the corresponding correlation coefficient matrix. Significance tests are performed based on the correlation coefficient matrix to screen out candidate lag times corresponding to the correlation coefficients that pass the confidence test, and the information gain value corresponding to the candidate lag times that pass the test is calculated. The optimal lag time is determined with the goal of maximizing the information gain value. Based on the optimal lag time, an activity intensity-water quality response function is constructed, with the mean activity intensity as the input variable and the rate of change of water quality parameters as the output variable. The parameters corresponding to the activity intensity-water quality response function are trained through historical data to generate a mapping relationship model between activity intensity and changes in water quality parameters. By inputting different activity intensity ranges into the mapping relationship model, the predicted values ​​of the corresponding water quality parameter change rates are calculated. Combined with the water quality parameter benchmark values, the expected changes in water quality parameters under different activity intensities are calculated, generating preliminary correlation factors that include time lag characteristics.

7. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Construct a water quality adaptability simulation model based on the water quality-biology-environment multidimensional correlation matrix, which includes a physical diffusion module, a biological metabolism module, and an environmental impact module. Step S32: Generate multiple sets of environmental variable combinations using the Latin hypercube sampling method, and input the multiple sets of environmental variable combinations into the corresponding water quality adaptability simulation model for simulation calculation to obtain the water quality parameter evolution curves under different environmental variable combinations. Step S33: Perform water quality simulation statistical analysis on the water quality parameter evolution curves corresponding to different combinations of environmental variables to calculate the steady-state value, fluctuation period and convergence rate corresponding to the water quality parameter evolution curves, and generate a water quality dynamic simulation dataset. Step S34: Couple the water quality dynamic simulation dataset with the corresponding growth rate and survival rate indicators in the biological growth characteristic dataset to construct the water quality parameter-biological growth response surface; Step S35: Determine the optimal combination of water quality parameters that enables the biological growth index to reach the optimal value through water quality parameter-biological growth response surface optimization, and generate the optimal water quality parameter range for the survival of cultured organisms.

8. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 7, characterized in that, Step S34 includes the following steps: Step S341: Extract time-series data of key water quality parameters under each simulation scenario from the water quality dynamic simulation dataset, including dissolved oxygen steady-state value, pH fluctuation period, ammonia nitrogen concentration convergence rate, and water temperature change amplitude; calculate the time-weighted average of each water quality parameter based on the time-series data to generate a water quality parameter feature set; simultaneously extract the growth rate and survival rate of cultured organisms in the corresponding simulation period from the biological growth feature dataset, and convert the above biological indicators into standardized biological growth indices in the 0-1 range using a normalization algorithm to form a biological growth feature set; Step S342: Perform spatiotemporal matching on the water quality parameter feature set and the biological growth feature set to associate water quality parameters and biological growth indices within the same simulation period based on timestamps, and construct a water quality parameter-biological growth association dataset; use Pearson correlation analysis to calculate the linear correlation coefficient between each water quality parameter and each standardized biological growth index, and select water quality parameters with an absolute value of correlation coefficient greater than a preset threshold as key influencing parameters to generate a subset of key water quality parameters; Step S343: Based on a subset of key water quality parameters, use orthogonal experimental design to divide the water quality parameter combination intervals, calculate the mean and variance of the standardized biological growth index within each interval, and generate a water quality parameter interval-biological growth mean mapping table; combine the data in the water quality parameter interval-biological growth mean mapping table, and construct a preliminary correlation model between water quality parameters and biological growth index through a local weighted regression algorithm, and output the marginal contribution coefficient of each key water quality parameter to the biological growth index; Step S344: Using the marginal contribution coefficient as a weight, the key water quality parameters are weighted and fused to generate a comprehensive water quality impact factor; based on the comprehensive water quality impact factor and the standardized biological growth index, a three-dimensional interpolation algorithm is used to generate discrete comprehensive water quality impact factor-biological growth index data points; the discrete data points are dimensionality reduced by principal component analysis, the first two principal components are extracted as the horizontal and vertical axes of the response surface, and data points whose variance explained by the principal components is greater than the preset proportion are retained to form the basic dataset of the response surface; Step S345: Input the basic dataset of the response surface into the improved radial basis function neural network, with principal components as input variables and standardized biological growth index as output variables, and train the neural network model; optimize the model parameters through cross-validation to ensure that the model prediction error is less than the preset error threshold; generate a continuous water quality parameter-biological growth response surface based on the trained model.

9. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Formulate corresponding graded control strategies based on the optimal water quality parameter range and water quality dynamic simulation dataset to clarify the control measures and implementation priorities corresponding to different water quality deviation levels. Step S42: By collecting the current water quality parameters corresponding to the aquaculture water area in real time, and comparing and analyzing the current water quality parameters with the optimal water quality parameter range, the deviation value and comprehensive deviation index corresponding to each water quality parameter are calculated, and the water quality control deviation value is generated. Step S43: Perform fuzzy clustering analysis based on the water quality control deviation value to determine the deviation level of the current water quality status; match the corresponding control measure set from the graded control strategy according to the deviation level, and calculate the expected control effect value of each control measure; construct a multi-objective optimization model based on the expected control effect value and equipment operating energy consumption parameters, and solve for the optimal control measure combination; transform the optimal control measure combination into specific operating parameters of the water quality control equipment, and generate a control command sequence including execution time, operating intensity, and duration. Step S44: Send the control command sequence to the water quality regulation equipment corresponding to the aquaculture area to perform adaptive control operations, and simultaneously record the equipment operation status, real-time water quality change data and biological growth prediction results; Step S45: Using data visualization technology, the equipment operating status before and after regulation, real-time water quality change data, and biological growth prediction results are displayed in multiple dimensions to demonstrate the water quality regulation effect and biological growth status prediction results, and a water quality regulation effect evaluation report is generated.

10. The water quality adaptability simulation control method based on aquaculture scenarios according to claim 9, characterized in that, Step S43, which involves constructing a multi-objective optimization model based on the expected control effect value and equipment operating energy consumption parameters, and solving for the optimal combination of control measures, includes the following steps: Using the expected control effect as the benefit indicator and the equipment operating energy consumption parameter as the cost indicator, a multi-objective optimization function is constructed. Set the control effect threshold and energy consumption upper limit as constraints to determine the decision variables and their value ranges corresponding to the multi-objective optimization function; A non-dominated sorting genetic algorithm is used to solve the multi-objective optimization function to obtain the Pareto optimal solution set; The comprehensive evaluation of each scheme in the Pareto optimal solution set is carried out based on the fuzzy hierarchical analysis method, and the comprehensive score of the scheme is calculated. Select the scheme with the highest comprehensive score as the optimal combination of control measures, and output the corresponding control parameter settings.

Images