AI water quality regulation and control-based oyster culture water quality regulation and control method and system

Abstract

The invention discloses an oyster culture water quality regulation and control method and system based on AI water quality regulation and control, and relates to the technical field of water quality monitoring and intelligent regulation and control. Carrying out score calculation and weighted integration on the smooth parameter sequence to generate a comprehensive evaluation quantity; identifying the change direction and the change amplitude in the fixed window and generating a trend record; and combining the comprehensive evaluation quantity, the trend record and the smoothing parameter sequence according to a time index, and determining a regulation and control level and regulation and control parameters. According to the method, a staged evaluation system based on smooth water quality parameters, a trend identification mechanism and a regulation and control quantity generation method are constructed, so that regulation and control decisions of continuous expression, staged sensitivity analysis and time sequence consistency of the oyster culture water body are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water quality monitoring and intelligent regulation, in particular to an oyster aquaculture water quality regulation method and system based on AI water quality regulation. BACKGROUND

[0002] With the acceleration of the process of large-scale and factory-like aquaculture, real-time monitoring and water quality regulation technology of aquaculture water gradually changes from manual experience management to data-driven mode. Traditional water quality monitoring methods mainly rely on fixed frequency sensor sampling, and whether the key parameters such as dissolved oxygen, pH, salinity, water temperature, ammonia nitrogen, etc. are in the safe range is judged by single-point threshold value. However, with the improvement of the life cycle management awareness of the aquaculture objects, the industry has begun to pay attention to the differentiated rules of water quality demand at different growth stages, and attempts to combine big data analysis, sequence modeling and intelligent regulation mechanism to realize the refinement of water quality management. In addition, some researches begin to explore trend identification based on time series information and comprehensive evaluation method based on multi-index combination in order to improve the responsiveness and adaptability of regulation, but the existing methods are still rough in cross-stage modeling, dynamic parameter updating and regulation signal generation.

[0003] The existing water quality regulation technology generally adopts single-parameter threshold judgment or multi-index parallel judgment method, lacks a dynamic evaluation system matched with the growth stages of the aquaculture organisms, and cannot adjust the target value, allowable deviation and index importance according to the stage difference, resulting in lack of stage sensitivity of regulation basis. In addition, the existing researches mostly take time series as static input without structuring the change pattern of evaluation quantity and physical parameters, making it difficult to identify the direction and amplitude of the continuous change of parameters through window analysis, so as to provide prospective judgment basis for regulation. Further, the existing technology usually only focuses on the water quality state before the regulation action, but lacks a state re-estimation mechanism after the regulation execution; the water quality data and trend information after regulation are often not backfilled to the evaluation link according to time index, so that the system cannot form a continuous and consistent state expression. Therefore, under the existing regulation framework, it is difficult to realize the determination of regulation level based on the linkage of comprehensive evaluation quantity, trend record and smoothed water quality parameter sequence, and it is also impossible to establish an integrated state updating process before and after the execution of regulation action. SUMMARY

[0004] In view of the above problems, the present application is proposed.

[0005] Therefore, the technical problems solved by the present application are that the existing aquaculture water quality regulation method cannot construct a dynamic evaluation system according to the growth stage, cannot identify the continuous trend change based on the time window structure, cannot update and backfill the data before and after the regulation execution according to the time index, and how to generate executable regulation level and regulation parameters through smoothed water quality parameter sequence, stage evaluation model and trend record.

[0006] To solve the above technical problems, the application provides the following technical solutions: an oyster culture water quality regulation method based on AI water quality regulation, comprising collecting original water quality data in a fixed sampling period, and generating a smooth water quality parameter sequence through preprocessing; constructing growth stage target values, stage allowable deviations, and index weights based on the smooth water quality parameter sequence, and performing score calculation and weighted integration on the smooth water quality parameter sequence to generate a comprehensive evaluation quantity; arranging the comprehensive evaluation quantity and the smooth parameter in chronological order, identifying the change direction and change amplitude in a fixed window, and generating a trend record; combining the comprehensive evaluation quantity, the trend record, and the smooth water quality parameter sequence according to the time index to determine the regulation level and the regulation parameter; after the regulation action is executed, continue to collect and process the water quality data, and update the smooth water quality parameter sequence, the comprehensive evaluation quantity, and the trend record according to the corresponding relationship before and after the execution and write them into a backfill record.

[0007] As a preferred scheme of the oyster culture water quality regulation method based on AI water quality regulation, wherein: the original water quality data is collected in a fixed sampling period, and the smooth water quality parameter sequence is generated through preprocessing, which includes recording dissolved oxygen, pH, salinity, water temperature and ammonia nitrogen according to a unified timestamp when collecting the original water quality data, and performing de-duplication processing on repeated records appearing at the same timestamp; for data with missing fields, values out of range, and violating the set upper and lower limits, mark them as abnormal data and replace them by linear interpolation, forward filling or backward filling; after forming continuous valid data, a sliding window with a fixed length of 5 is used to generate a smooth water quality parameter sequence for each type of data, so that the smooth value at each time point is output by the same type of parameter in the continuous five sampling periods in an arithmetic average manner.

[0008] As a preferred scheme of the oyster culture water quality regulation method based on AI water quality regulation, wherein: the smooth water quality parameter sequence is used to construct growth stage target values, stage allowable deviations, and index weights, and the smooth water quality parameter sequence is used to perform score calculation and weighted integration to generate a comprehensive evaluation quantity, which includes setting a unique stage identifier for each growth stage when constructing the growth stage target values, the stage allowable deviations, and the index weights, and selecting a fixed number of preprocessed smooth water quality parameter sequences from the historical database as reference data; for each index, the arithmetic average method is used to calculate the target value in the same stage, and the standard deviation is generated by the dispersion degree of the sample deviating from the target value; the standard deviation is combined with a preset multiple to form the stage allowable deviation; the weights of all indexes are normalized to form a weight vector with a fixed dimension.

[0009] As a preferred scheme of the oyster culture water quality regulation method based on AI water quality regulation, wherein: the arrangement of the comprehensive evaluation quantity and the smoothing parameter in time sequence, the identification of the change direction and the change amplitude in the fixed window and the generation of the trend record include: when the smoothing parameter is evaluated, the smoothing parameter is matched with the target value and the allowable deviation of the stage, the deviation between the real-time value and the target value is judged whether it is within the allowable deviation range for segmented processing, the deviation within the range is calculated by linear proportion, and the deviation exceeding the range is set to zero; the score values of each smoothing index at the same time point are weighted and combined according to the index weight to form a fixed-scale comprehensive evaluation quantity sequence.

[0010] As a preferred scheme of the oyster culture water quality regulation method based on AI water quality regulation, wherein: the arrangement of the comprehensive evaluation quantity and the smoothing parameter in time sequence, the identification of the change direction and the change amplitude in the fixed window and the generation of the trend record include: when the smoothing parameter is evaluated, the smoothing parameter is matched with the target value and the allowable deviation of the stage, the deviation between the real-time value and the target value is judged whether it is within the allowable deviation range for segmented processing, the deviation within the range is calculated by linear proportion, and the deviation exceeding the range is set to zero; the score values of each smoothing index at the same time point are weighted and combined according to the index weight to form a fixed-scale comprehensive evaluation quantity sequence.

[0011] As a preferred scheme of the oyster culture water quality regulation method based on AI water quality regulation, wherein: the arrangement of the comprehensive evaluation quantity and the smoothing parameter in time sequence, the identification of the change direction and the change amplitude in the fixed window and the generation of the trend record include: when the smoothing parameter is evaluated, the smoothing parameter is matched with the target value and the allowable deviation of the stage, the deviation between the real-time value and the target value is judged whether it is within the allowable deviation range for segmented processing, the deviation within the range is calculated by linear proportion, and the deviation exceeding the range is set to zero; the score values of each smoothing index at the same time point are weighted and combined according to the index weight to form a fixed-scale comprehensive evaluation quantity sequence.

[0012] As a preferred scheme of the oyster culture water quality regulation method based on AI water quality regulation, wherein: the water quality data is continuously collected and processed after the regulation action is performed, and the smooth water quality parameter sequence, the comprehensive evaluation quantity and the trend record are updated and written into the backfill record according to the correspondence before and after the execution; after the regulation action is performed, the smooth parameters in the sampling period before and after the execution are compared in time sequence, and the updated entries are written in the record format; the comprehensive evaluation quantity at the same time point before and after the execution is regenerated according to the predetermined calculation method, and the trend record of the recalculation is formed through the trend identification process; the smooth parameter update record, the comprehensive evaluation quantity update record and the trend update record are uniformly written into the backfill record table.

[0013] Another object of the present application is to provide an oyster culture water quality regulation system based on AI water quality regulation, which can solve the problems that the current culture water quality regulation method contains the method based on the time window structure to identify the continuous trend change and cannot update and backfill the data before and after the regulation execution according to the time index by constructing the growth stage target value, the stage allowable deviation and the index weight based on the smooth water quality parameter sequence, and generating the comprehensive evaluation quantity through the score calculation and the weighted integration of the smooth water quality parameter sequence.

[0014] As a preferred scheme of the oyster culture water quality regulation system based on AI water quality regulation, wherein: a water quality data preprocessing module, a stage water quality evaluation modeling module and a regulation determination and state updating module are included; the water quality data preprocessing module is used for collecting original water quality data, and generating smooth water quality parameter sequence through exception elimination, time alignment and sliding average; the stage water quality evaluation modeling module is used for constructing target value, allowable deviation and index weight of different growth stages of oysters through historical samples, and generating time-updated comprehensive water quality evaluation quantity through score calculation and weighted integration of real-time smooth parameters; the regulation determination and state updating module is used for generating regulation level and regulation parameters based on comprehensive evaluation quantity and trend record, and continuously collecting and processing water quality data after regulation execution, updating and backfilling smooth parameters, comprehensive evaluation quantity and trend record.

[0015] A computer device includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the steps of the oyster culture water quality regulation method based on AI water quality regulation.

[0016] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the steps of the oyster culture water quality regulation method based on AI water quality regulation.

[0017] The oyster aquaculture water quality regulation method based on AI water quality regulation provided by the application has the beneficial effects that: the oyster aquaculture water quality regulation method based on AI water quality regulation provided by the application performs abnormal elimination, time alignment and sliding average processing on original water quality data, so that the data input into the subsequent modeling link has uniformity and stability; on this basis, the target value, allowable deviation and index weight are constructed by taking the growth stage as an index, the real-time multi-dimensional water quality parameters can be mapped into comprehensive evaluation quantities suitable for different stages, and a clear scale is provided for parameter interpretation, deviation identification and subsequent decision-making. Further, the continuous change direction and change amplitude of the evaluation quantity and each index are identified through a fixed window, the time evolution process of the water quality state is structured and expressed, so that the regulation and determination are based on the current state and the stage change mode. In the regulation amount generation link, the comprehensive evaluation quantity, trend record and smoothing parameter are combined according to the time index, so that the determination of the regulation level and the regulation parameter is based on the complete data context, and isolated decision-making is avoided. After the regulation is executed, the smoothing parameter, comprehensive evaluation quantity and trend record are backfilled and updated, so that the data chain is continuously closed, and the automatic evaluation of the next cycle can be directly supported. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1 The overall flowchart of the oyster aquaculture water quality regulation method based on AI water quality regulation provided by the first embodiment of the application. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned purposes, features and advantages of the application more apparent and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are only a part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the application.

[0021] Embodiment 1, refer to Figure 1 For an embodiment of the application, an oyster aquaculture water quality regulation method based on AI water quality regulation is provided, comprising: S1: collecting original water quality data in a fixed sampling period, and generating a smoothing water quality parameter sequence through preprocessing.

[0022] Furthermore, several monitoring points were set up within the oyster farming area. At each point, a dissolved oxygen probe, pH probe, conductivity-salinity meter, water temperature probe, and online ammonia nitrogen analyzer were installed. All sensor probes were fixed at a depth of 0.5–1.0 meters above the water surface and stably secured using supports or floats. The dissolved oxygen probe was an electrochemical or optical device with a range of 0–20 mg / L; the pH probe range was set to pH 4–10; the conductivity-salinity meter range covered a conversion range of 0–40 ppt; the water temperature probe range was set to 0–40℃; and the online ammonia nitrogen analyzer range was set to 0–10 mg / L. Each sensor underwent zero-point calibration and range setting according to the manufacturer's instructions, and the calibration parameters were recorded in a configuration file and did not change automatically during operation.

[0023] Sensors at each measuring point are connected to data acquisition devices on the shore or buoys via shielded signal cables or industrial buses. These devices have built-in unified clocks, synchronized via GPS or network time services. The data acquisition cycle is fixed at 1 minute. Within each cycle, instantaneous measurements of dissolved oxygen, pH, salinity, water temperature, and ammonia nitrogen are read sequentially. Each set of measurements, along with its corresponding timestamp and measuring point number, forms a raw record. These raw records are appended to a local cache in chronological order and written to a data file or database table in a fixed format (e.g., the order of "measuring point number, time, DO, pH, Sal, Temp, NH3").

[0024] The original records in the buffer are subjected to integrity checks. Records missing any field are marked as invalid and skipped during the preprocessing stage. For multiple records at the same time stamp for the same measurement point, only the first record written is retained; the remaining records are treated as duplicates. Each field in a valid record is validated. Fields with dissolved oxygen less than 0 or greater than 20 mg / L, pH less than 4 or greater than 10, salinity less than 0 or greater than 40 ppt, water temperature less than 0 or greater than 40℃, or ammonia nitrogen less than 0 or greater than 10 mg / L are marked as outliers. For records with only one outlier field and all other fields normal, only the outlier field is interpolated. When three or more fields in the same record are simultaneously outlier, the entire record is marked as invalid and skipped in subsequent calculations.

[0025] It should be noted that for a single field marked as abnormal, the two most recent valid values ​​from adjacent timestamps of the measurement point are selected, and a replacement value for the current time point is calculated using linear interpolation based on chronological order. For example, when interpolating the dissolved oxygen field at a certain moment, the most recent valid dissolved oxygen value before and after that moment is read, and the replacement value for the current moment is calculated based on the time interval ratio. This replacement value then overwrites the original abnormal value in the data table. For abnormal fields where two valid values ​​cannot be found within a reasonable time window, the abnormality marker is retained, and the field is ignored during subsequent averaging.

[0026] After anomaly removal and interpolation, time alignment was performed on the water quality parameter sequences for each measuring point on the time axis to ensure that all five fields—dissolved oxygen, pH, salinity, water temperature, and ammonia nitrogen—were filled with valid values ​​or explicitly marked as unavailable at the same timestamp. Timestamps marked as unavailable were not included in any subsequent statistical calculations. For each type of water quality parameter at the same measuring point, a sliding window of length 5 was constructed in chronological order. The arithmetic mean of the same parameter values ​​over five consecutive collection periods was calculated, and the average result was defined as the smoothed parameter value for that measuring point at the current time. The smoothing operation was performed sequentially on the time axis in 1-minute increments, generating a smoothed time series for each of dissolved oxygen, pH, salinity, water temperature, and ammonia nitrogen.

[0027] The smoothed water quality parameter sequences from all monitoring points are summarized. At each timestamp, the smoothed dissolved oxygen, pH, salinity, water temperature, and ammonia nitrogen values ​​from different monitoring points are combined according to the monitoring point number to form a data record with a timestamp, monitoring point number, and five smoothed parameters, which is then written into a data table. This data table only retains records that have undergone integrity verification, anomaly handling, time alignment, and moving average processing. After the above processing, the water quality status of the aquaculture water body at any given time point can be uniquely represented by the five smoothed parameter values ​​corresponding to each monitoring point.

[0028] S2: Based on the smoothed water quality parameter sequence, construct the target value, stage allowable deviation and index weight for the growth stage, and calculate and weight the smoothed water quality parameter sequence to generate a comprehensive evaluation quantity.

[0029] Furthermore, after completing the collection, anomaly removal, time alignment, and smoothing process using a sliding window of length 5 for the original water quality data, the resulting smoothed water quality parameter sequence is used as input data. Considering the varying water quality requirements of oysters at different growth stages, statistical modeling is performed on each type of smoothed water quality parameter, generating target values ​​for each growth stage, stage deviation ranges, indicator weights, real-time scores, and comprehensive evaluation quantities. To ensure the determinism of the model parameters, all calculations are based on a fixed-size, well-structured dataset, and all variables are uniquely determined through mathematical relationships, without any empirical settings or uncertainties.

[0030] Considering that oysters have different suitable ranges for water quality indicators such as dissolved oxygen, pH, salinity, water temperature, and ammonia nitrogen during the seedling, growth, and fattening stages, stage markers are set for each growth stage. ,in Corresponding to the seedling stage, Corresponding growth period, Corresponding to the fattening period. For each growth stage, 1000 smoothed water quality records, processed using a completely consistent data preprocessing procedure (including anomaly removal, time alignment, and moving average), are pre-selected from the historical database. These 1000 records are used as the high-quality historical sample set for that stage. For the [missing information - likely a specific stage or period]... Five water quality indicators are used, where 1 is dissolved oxygen, 2 is pH, 3 is salinity, 4 is water temperature, and 5 is ammonia nitrogen. These 1000 samples are represented as... , … Sample number The value range is 1 to 1000, consistent with the sample size.

[0031] Based on this historical sample set, a target value is calculated for each stage and each indicator. This target value is determined by the arithmetic mean of the sample set, expressed as:

[0032] in, Indicates that the oyster is in At that time, the first Target values ​​for each water quality indicator; This represents historical smoothed sample data for the same period and indicator. The target value is obtained by calculating the arithmetic mean of 1000 samples.

[0033] It should be noted that, subsequently, the standard deviation was calculated for 1000 samples of the same period and the same indicator to describe the degree of fluctuation of each sample around the target value. The standard deviation is expressed as:

[0034] in, Indicates the first Water quality indicators at The standard deviation below.

[0035] Based on standard deviation Define this indicator in The allowable deviation range is defined as follows. To ensure the stability and certainty of the allowable deviation, it is defined as twice the standard deviation, i.e.:

[0036] in, Indicates in Below, allow smoothing values relative target value The maximum deviation.

[0037] After determining the target value and allowable deviation, weighting coefficients are further assigned to different indicators. The weights are calculated using the reciprocal of the standard deviation, ensuring that indicators with smaller historical sample fluctuations have larger weights. The weights are expressed as:

[0038] in, Indicates the first Water quality indicators at The weighting coefficients in the text; This represents a water quality index. This represents a smooth parameter sequence.

[0039] After obtaining the target value, tolerance, and weights, use the following expression to smooth the real-time value. The score is calculated. The score function uses the absolute value of the deviation as the measurement variable and is defined using a piecewise structure:

[0040] in, Indicates real-time smoothing value Oysters are in The score is calculated as follows: When the deviation is less than the allowable deviation, a score between 0 and 1 is obtained based on the linear proportion; when the deviation exceeds or equals the allowable deviation, the score is directly set to 0. This scoring method relies only on three types of variables: the smoothing parameter value, the stage target value, and the allowable deviation, all of which can be calculated using existing data and fixed formulas.

[0041] Finally, the scores of all indicators are weighted according to their corresponding weights to obtain the oyster's... Next, Time The comprehensive evaluation metric is expressed as:

[0042] in, Indicates the overall evaluation quantity; The weights of each indicator; These are the scores for each indicator.

[0043] S3: Organize the comprehensive evaluation quantities and smoothing parameters in chronological order, identify the direction and magnitude of change within a fixed window, and generate trend records.

[0044] Furthermore, after obtaining the output comprehensive evaluation quantity, the variation patterns of this evaluation quantity and each smoothed water quality parameter on the time axis are organized, and a trend analysis process for subsequent control preparation is constructed. The comprehensive evaluation quantity records of all measuring points at a unified timestamp are arranged in chronological order to form a continuous time series. Simultaneously, the obtained five types of smoothed water quality parameter sequences are correlated according to the same time structure, ensuring consistency between the two types of data in the time dimension. In this way, any point in time can be represented by smoothed dissolved oxygen, smoothed pH, smoothed salinity, smoothed water temperature, smoothed ammonia nitrogen, and the corresponding comprehensive evaluation quantity, giving all input data a clear time index relationship. Based on this, a fixed-length time window is selected to identify recent trends. In the implementation, this window length is chosen to be 60 minutes, corresponding to 60 continuous data records sampled minute by minute. The window is advanced along the time axis using a sliding method, and the smoothed parameters and comprehensive evaluation quantities within each window are sorted by time, ensuring that each sequence maintains a strictly equal-interval structure.

[0045] After establishing continuous data within a fixed window, the changing directions of multiple water quality parameters and comprehensive evaluation quantities within the window are identified chronologically. For example, within the same window, the values ​​of smoothed dissolved oxygen, smoothed pH, smoothed salinity, smoothed water temperature, and smoothed ammonia nitrogen are extracted at the window's starting point, middle, and ending points. Their magnitude relationships are used to determine the changing direction of each indicator within the window. Simultaneously, the numerical sequence of the comprehensive evaluation quantity within the window is sorted chronologically. By analyzing its relative magnitude at the window's beginning, middle, and end positions, the changing pattern of the comprehensive evaluation quantity during that time period—whether it is increasing, decreasing, or nearly stable—is identified. After identifying the changing direction, the magnitude of change for each water quality indicator within the window is quantified. The difference between the end value and the beginning value of the window is calculated, and this difference is used as a reference indicator of the change intensity. For cases where the change intensity is insufficient to represent the significance of the trend, supplementary calculations are performed using the differences in point values ​​at multiple locations within the window, ensuring that trend judgment is based on multi-point sampling.

[0046] After obtaining the direction and magnitude of change, the variation characteristics of the five water quality parameters within the same window are correlated with the variation characteristics of the comprehensive evaluation quantity. The comprehensive evaluation quantity already reflects the overall suitability of the five indicators at the current growth stage; therefore, the direction and magnitude of change of the comprehensive evaluation quantity within the window can serve as the main reference quantity for trend analysis. During trend analysis, the comprehensive evaluation quantity sequence within the window is used as the main sequence, and the smoothed water quality parameter sequences of the five categories are used as auxiliary sequences. The main sequence determines the overall trend, and the auxiliary sequences identify the specific parameter changes that led to the trend. To avoid interference from single fluctuations, multiple consecutive windows are analyzed synchronously in trend analysis; that is, changes are assessed using a multi-window overlap method within the same time period. By comparing the consistency of trend results between adjacent windows, the final trend identification result is made more stable. In this process, when multiple windows show a continuous and consistent direction of change for the same indicator, this direction is taken as the main trend of that indicator in the current aquaculture area; conversely, if multiple windows show a continuous decrease or increase in the comprehensive evaluation quantity, this direction of change is taken as the trend result of the comprehensive evaluation quantity.

[0047] It should be noted that during trend analysis, data from different monitoring points need to be aggregated spatially. Since aquaculture areas typically have multiple sampling points, and water quality changes at each point may differ across time periods, the trend identification results are merged. Within the same time window, the direction and magnitude of change are extracted from multiple monitoring points and arranged in order of their numbers. Subsequently, the overall trend is determined using the majority consensus principle for all monitoring points; when a majority of monitoring points show the same direction of change, that direction is taken as the final trend for that window; otherwise, the trend is marked as having no clear direction. Regarding the magnitude of change, the median magnitude of change across all monitoring points is calculated and used as a reference value for the overall magnitude of change.

[0048] After trend identification is completed, the trend direction and magnitude are arranged chronologically to generate a time trend record. This record has a sliding structure with a fixed-length time window, consistent with the original sampling clock, and can be directly aligned with the comprehensive evaluation quantity in time. After the trend record is generated, it is used together with the comprehensive evaluation quantity sequence as data input for subsequent control processes, enabling the control process to make judgments not only based on the current comprehensive evaluation quantity but also to utilize trend information as a basis for advance judgment.

[0049] S4: Combine the comprehensive evaluation quantity, trend record and smoothed water quality parameter sequence according to time index to determine the control level and control parameters.

[0050] Furthermore, after identifying water quality change trends, the trend records are linked to the obtained comprehensive evaluation quantities according to a time index. This ensures that real-time comprehensive evaluation quantities, trend direction, trend magnitude, water quality smoothing parameters, and historical reference data on the same time axis together constitute a data input set that can be used for regulatory decision-making. To ensure consistent input structure, decision input records are generated at each time point according to a fixed field order of comprehensive evaluation quantity, trend direction, trend magnitude, and five types of smoothed water quality parameters, so as to maintain the parsability of fields during the regulation generation process. Based on this data structure, the water body status is evaluated point-by-point, forming regulatory decision quantities for subsequent execution layers.

[0051] When generating decision parameters for each time point, the output comprehensive evaluation value is first read and used as the primary reference value for judging the current water body state. Simultaneously, the trend direction and magnitude corresponding to that time point are read to determine whether the water body state is trending towards improvement, deterioration, or stability. Subsequently, to fully describe the real-time condition of the water body, the smoothed dissolved oxygen, smoothed pH, smoothed salinity, smoothed water temperature, and smoothed ammonia nitrogen values ​​for the current time point are read and integrated into the same decision input record in a preset order. In this way, each time point can be fully described by the aforementioned multidimensional data, and all data sources are clear and traceable.

[0052] After constructing the decision input records, the input fields are structured and interpreted to establish reference relationships during the generation of control quantities. For example, the comprehensive evaluation quantity is interpreted as absolute magnitude, the trend direction is interpreted as a three-value structure (rising, falling, or no significant change), and the trend amplitude is interpreted quantitatively, enabling the control generation process to distinguish the significance of trends. Simultaneously, five types of smoothed water quality parameters are used as raw physical quantities in the construction of decision quantities to pinpoint deviations in specific indicators. To ensure a consistent logical structure in the decision quantity generation process, the input fields are treated as a set of fixed-dimensional time-series inputs. The comprehensive evaluation quantity and trend field are processed as univariate variables, while the five types of smoothed water quality parameters are processed as indicator groups. All variables are updated according to the sampling period.

[0053] It should be noted that, based on the aforementioned data structure, several fixed rules are set for the generation of control quantities. These rules are defined according to the magnitude relationship between data values. For example, multiple intervals are set for the comprehensive evaluation quantity, each interval corresponding to a different control intensity level; a fixed Boolean interpretation is set for the trend direction to trigger the control level conversion; and several intervals are set for the trend amplitude to identify the significance of the change, thereby determining the lead time for control actions. In this rule framework, when the trend direction at a certain moment is downward, the trend field is written into the rule system as a priority trigger condition; when the trend direction is upward, this information is used as a condition to suppress the escalation of control actions; when the trend has no obvious direction, the judgment is made solely based on the comprehensive evaluation quantity and real-time water quality parameters.

[0054] Within the rule system, specific data thresholds are set for five types of smoothed water quality parameters to pinpoint the sources of deviation for specific indicators. For example, when the smoothed dissolved oxygen value is significantly lower than the target value obtained based on historical statistics, this indicator is marked in the decision input record; when the smoothed ammonia nitrogen value shows a continuous upward trend and the comprehensive evaluation value decreases, these two types of information are combined to generate a specific control quantity. After all rule judgments are completed, the comprehensive evaluation value range, trend direction judgment, trend amplitude range determination, and the positional relationship of the five types of real-time smoothed parameters are integrated, and a control level value is output according to preset rules. This level value is represented by integers and has a fixed range division, such as 0 indicating no control required, 1 indicating slight control, 2 indicating moderate control, and 3 indicating high control.

[0055] Once the control level is determined, specific control action parameters are assigned, including the target control intensity and control duration. The control intensity value is assigned according to a fixed level-intensity mapping table and does not change with the data; the control duration is assigned according to a fixed number of sampling periods, ensuring that the control action has a standardized duration. During the decision quantity generation process, all input variables retain their original units, and all levels are expressed in integer form, without generating fuzzy values, enabling the generated control quantities to be directly used in the execution equipment.

[0056] After the control level and control parameters are finally determined, they are written into the decision record table by timestamp and stored together with the comprehensive evaluation quantity, trend direction, trend amplitude and five types of smooth water quality parameters at the same time point, so that they can be read by the subsequent execution layer.

[0057] S5: After the control action is executed, continue to collect and process water quality data, and update the smooth water quality parameter sequence, comprehensive evaluation quantity and trend record according to the correspondence before and after the execution and write them into the backfill record.

[0058] Furthermore, after generating the control level and control parameters, these levels and parameters are recorded and written into the control execution queue according to the update cycle of the acquisition device, enabling the execution equipment to read and execute the corresponding control commands in subsequent cycles. The reading process of control commands is completed by the execution equipment in chronological order, without any human intervention. After receiving the control level and control parameters, the execution equipment completes the corresponding oxygenation, water exchange, stirring, or related water treatment operations according to the pre-specified action intensity and duration in the control parameters. During execution, the sampling cycle remains unchanged, and five original water quality parameters—dissolved oxygen, pH, salinity, water temperature, and ammonia nitrogen—continue to be collected from the water body at a fixed frequency, ensuring that real-time water changes are fully recorded during the control execution period.

[0059] During and after the control action is executed, the sampling device continues to generate a new round of raw data records. These new data are processed using anomaly removal, time alignment, and moving average methods to ensure that the smoothed water quality parameters after the control action maintain the same data structure as those before the control action, allowing for direct comparison of the two over time. After the smoothed data is generated, the smoothed values ​​of the same indicators before and after the control action are correlated over time, establishing a clear relationship between each control action record and the smoothed values ​​of multiple sampling periods before and after the action. After correlation is established, changes in dissolved oxygen, pH, salinity, water temperature, and ammonia nitrogen are extracted based on the differences before and after the control action, including the differences in value, direction, and duration of the differences.

[0060] Furthermore, to ensure continuous traceability throughout the entire regulatory process, after the regulatory directive is executed, all fields required for calculating the comprehensive evaluation quantity are updated using the same calculation method, including recalculating the individual scores and comprehensive evaluation quantities of the five categories of indicators at the same time point. This method allows for the acquisition of changes in the comprehensive evaluation quantity after regulatory implementation, ensuring that the impact of regulation can be represented through a unified calculation system. Regarding trend direction recording, based on trend identification methods, trend records are re-formed for the comprehensive evaluation quantity and the five smoothing indicators after regulatory implementation, directly linking these trend records to the regulatory level, regulatory parameters, and regulatory implementation time. Updating these trend records allows for the use of natural and continuous trend information in subsequent decision-making processes without introducing new data structures.

[0061] Subsequently, the differences in comprehensive evaluation values ​​before and after the implementation of the control measures were compiled, and these differences, along with the trend update values, were written into the control effect record table. The control effect record table uses a timestamp as the primary index and includes fields such as the comprehensive evaluation value before implementation, the comprehensive evaluation value after implementation, the difference in value, the type of trend change, the control level, and the control parameters, ensuring that each control implementation generates an independent effect record. After the record is written, the reference data used to generate future control values ​​is updated based on this record, so that the water state after the control implementation can automatically serve as input for subsequent judgments in the next sampling period.

[0062] During data backfilling, to avoid judgment biases caused by local fluctuations in the short term, multiple time points before and after the implementation of the control measures are used as time windows. The smoothing parameters, comprehensive evaluation quantities, and trend records within the windows are combined, so that the control results depend not only on single-point data after implementation but also on continuous data from several sampling periods after implementation. The window data is arranged in natural chronological order to form continuous state segments, which are used to provide continuous data input for trend interpretation, weight judgment, and comprehensive evaluation judgment in the next generation of control quantities.

[0063] It should be noted that after all update records are written, the data chain of the control results is automatically established, forming a complete time-series linkage from window smoothing parameters to comprehensive evaluation quantities, then to trend records, and finally to control level records and control effect records. In the next sampling period, all these updated data will participate in the control generation together with the new sampled data, ensuring that the entire process is continuously updated and recorded without data breakpoints.

[0064] Example 2, an embodiment of the present invention, provides a method for controlling water quality in oyster farming based on AI water quality control. To verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0065] First, this embodiment selected six oyster culture ponds with similar water exchange conditions and bottom sediment types within the same coastal aquaculture area, numbered as traditional control pond A, traditional control pond B, traditional control pond C, and the control ponds of this invention A, this invention B, and this invention C. Each culture pond had an area of ​​approximately 0.5 hm², with a water depth maintained between 1.8 and 2.2 m. Oyster larvae of the same batch and size were stocked at a density of 100,000 oysters per pond. All culture ponds used the same feed source and feeding regime to ensure consistency in nutritional conditions, allowing for a comparison of the impact of differences in water quality control strategies during the experiment.

[0066] Dissolved oxygen, conductivity and salinity meters, pH probes, water temperature probes, and online ammonia nitrogen analyzers were installed in all aquaculture ponds. The sensor sampling depth was uniformly set to 0.8m above the water surface, and the sampling period was fixed at 10 minutes. All sensor signals were uploaded to the data storage unit via on-site acquisition devices, and the raw data was recorded with a uniform timestamp. The raw data from all six aquaculture ponds underwent preprocessing according to the same procedure, including outlier removal, time alignment, and calculation of a smoothed parameter sequence using a moving average with a length of 5 sampling periods, so that each time point could be uniquely represented by five smoothed water quality parameters.

[0067] In traditional control ponds A to C, the industry-standard single-index threshold combined with manual experience-based water quality management is adopted. That is, fixed upper and lower limits are set based on dissolved oxygen and ammonia nitrogen. When the monitored values ​​exceed the limits, intervention is carried out through artificial oxygenation or water exchange. Once the values ​​return to within the threshold, the operation is stopped. No comprehensive evaluation is constructed, the oyster growth stage is not distinguished, and the correspondence between water quality changes and growth performance before and after control is not recorded. In the control ponds A to C of this invention, the oyster farming water quality control method based on AI proposed in this invention is adopted: First, the historical smoothed parameter sequence is organized, and target values, stage tolerance deviations, and index weights are constructed according to the oyster's growth stage information. Then, the online-collected smoothed parameter sequence is scored and weighted to generate a comprehensive water quality evaluation quantity that updates over time. Within a fixed time window, the direction and magnitude of change of the comprehensive evaluation quantity and each smoothed parameter are identified to form a trend record. Next, the comprehensive evaluation quantity, trend record, and smoothed parameters are combined with a time index, and the control level, corresponding control duration, and intensity are determined through rule logic to control the aeration equipment, water replenishment, and drainage equipment. After the control is implemented, water quality data continues to be recorded, and the smoothed parameter sequence, comprehensive evaluation quantity, and trend record are updated accordingly before and after implementation and written into the backfill record for subsequent evaluation and parameter correction. The experimental period is set to 30 days. During the experiment, no other differential measures are introduced, and the water quality process data and farming results of each farming pond are recorded and compared synchronously.

[0068]

[0069] As shown in Table 1, under identical stocking densities, feed conditions, and environmental backgrounds, the three aquaculture ponds using the method of this invention exhibit significant differences from traditional control ponds in terms of the stability of key water quality indicators and overall condition characterization. Firstly, regarding dissolved oxygen levels, the average dissolved oxygen in traditional control ponds A–C ranged from 5.1 to 5.4 mg / L, while the average dissolved oxygen in control ponds A–C using the method of this invention remained stable at around 6.0 mg / L, with the overall value closer to the suitable range. More importantly, the intraday standard deviation of dissolved oxygen was 1.0–1.3 mg / L in the traditional control ponds, while it was 0.5–0.6 mg / L in the control ponds using the method of this invention, indicating a significant reduction in dissolved oxygen fluctuations. This demonstrates that the control levels and parameters generated based on comprehensive evaluation quantities and trend records can better constrain dissolved oxygen changes within a narrower range, which is substantially different from traditional control methods that rely solely on threshold triggers.

[0070] Regarding ammonia nitrogen levels, the average ammonia nitrogen level in traditional control tanks is maintained at 0.40–0.45 mg / L, while the control tank of this invention maintains it within the range of 0.28–0.30 mg / L, and the duration of ammonia nitrogen exceedance is reduced from 3.0–3.8 h / day to approximately 1.0 h / day. Traditional methods mainly operate only after the value exceeds the threshold, lacking the utilization of trends and feedforward adjustment mechanisms. The method of this invention identifies the direction and magnitude of change through time window analysis based on smoothing parameters and comprehensive evaluation quantities, allowing the control level to be adjusted when ammonia nitrogen approaches the unfavorable range, thereby significantly shortening the duration of exceedance.

[0071] At the multi-indicator comprehensive level, the comprehensive water quality evaluation mean S can be seen as the result of uniformly quantifying the five smoothed water quality parameters under the current growth stage. Data shows that the comprehensive water quality evaluation mean of traditional control ponds A-C is 0.70-0.73, while that of the control ponds of this invention is 0.84-0.86, indicating a significant difference in overall performance. More telling is the row showing the number of days with an evaluation value <0.6: 5-8 days in traditional control ponds, and only 0-1 days in the control ponds of this invention, indicating that the method of this invention significantly reduces the duration of unfavorable comprehensive conditions throughout the entire experimental period.

[0072] Further analysis of the aquaculture results shows that the average daily oyster mortality rate was 9.8–11.4 oysters / 10,000 in the traditional control pond, and 5.7–6.3 oysters / 10,000 in the control pond of this invention; the average individual weight gain was 0.37–0.39 g / day in the traditional control pond, and 0.46–0.48 g / day in the control pond of this invention; the survival rate at the end of the experimental period was 91.8%–93.1% in the traditional control pond, while it reached 96.8%–97.5% in the control pond of this invention. Although this embodiment does not directly constrain growth indicators with formulas in the technical solution, the data results indirectly reflect that the method chain based on stage-based water quality assessment modeling, trend identification, and post-control state backfilling has a continuous impact on the macro-aquaculture effect. Traditional methods cannot incorporate growth stage information into water quality assessment, nor can they backfill new time-series data into the same assessment framework after control, thus having significant limitations in terms of control strategy updates and parameter optimization.

[0073] In summary, the comparative data of this embodiment shows that the present invention does not merely adjust the parameters of traditional water quality monitoring or control processes, but rather forms a new methodological path in several key aspects such as evaluation system construction, trend utilization, regulation quantity generation, and state backfilling. As a result, it demonstrates technical characteristics and implementation effects that are different from traditional methods in terms of water quality stability and aquaculture results, providing quantifiable support for the demonstration of inventiveness and novelty.

[0074] Example 3, an embodiment of the present invention, provides an oyster farming water quality control system based on AI water quality control, including a water quality data preprocessing module, a stage water quality evaluation modeling module, and a control determination and status update module.

[0075] The water quality data preprocessing module is used to collect raw water quality data and generate a smooth water quality parameter sequence through anomaly removal, time alignment, and moving average. The stage water quality evaluation modeling module is used to construct target values, allowable deviations, and index weights for different growth stages of oysters based on historical samples, and to calculate and weight the real-time smoothing parameters to generate a comprehensive water quality evaluation quantity that updates over time. The regulation judgment and status update module is used to generate regulation levels and regulation parameters based on the comprehensive evaluation quantity and trend records. After the regulation is implemented, water quality data continues to be collected and processed, and the smoothing parameters, comprehensive evaluation quantity, and trend records are updated and backfilled.

[0076] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0077] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0078] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0079] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for controlling water quality in oyster farming based on AI-based water quality control, characterized in that, include: Raw water quality data are collected within a fixed sampling period and a smooth water quality parameter sequence is generated through preprocessing. Based on the smoothed water quality parameter sequence, the target value, allowable deviation and index weight of the growth stage are constructed, and the score of the smoothed water quality parameter sequence is calculated and weighted to generate a comprehensive evaluation quantity. Organize the comprehensive evaluation quantities and smoothing parameters in chronological order, identify the direction and magnitude of change within a fixed window, and generate trend records; The comprehensive evaluation quantity, trend record and smoothed water quality parameter sequence are combined according to time index to determine the control level and control parameters; After the control measures are implemented, water quality data will continue to be collected and processed. The smoothed water quality parameter sequence, comprehensive evaluation quantity and trend record will be updated according to the correspondence before and after the implementation and written into the backfill record.

2. The oyster farming water quality control method based on AI water quality control as described in claim 1, characterized in that: The process of collecting raw water quality data within a fixed sampling period and generating a smooth water quality parameter sequence through preprocessing includes recording dissolved oxygen, pH, salinity, water temperature and ammonia nitrogen according to a uniform timestamp when collecting raw water quality data, and deduplicating duplicate records that appear at the same timestamp. For data with missing fields, values ​​exceeding the range, or data that violates the set upper and lower limits, mark them as abnormal data and replace them using linear interpolation, forward padding, or backward padding. After generating continuous and valid data, a smooth water quality parameter sequence is generated for each type of data using a sliding window of fixed length 5, so that the smooth value at each time point is output by the arithmetic average of the same type of parameters from five consecutive sampling periods.

3. The oyster farming water quality control method based on AI water quality control as described in claim 2, characterized in that: The process of constructing growth stage target values, stage allowable deviations, and index weights based on smooth water quality parameter sequences, and calculating and weighting the smooth water quality parameter sequences to generate a comprehensive evaluation quantity includes setting a unique stage identifier for each growth stage when constructing growth stage target values, stage allowable deviations, and index weights, and selecting a fixed number of preprocessed smooth water quality parameter sequences from the historical database as reference data. For each type of indicator, the target value is calculated using the arithmetic mean at the same stage, and the standard deviation is generated based on the dispersion of the sample from the target value. The standard deviation is combined with a preset multiple to form the stage allowable deviation; The weights of all indicators are normalized so that all weights form a weight vector with fixed dimensions.

4. The oyster farming water quality control method based on AI water quality control as described in claim 3, characterized in that: The process of organizing the comprehensive evaluation quantity and smoothing parameter in chronological order, identifying the direction and magnitude of change within a fixed window and generating trend records includes matching the smoothing parameter with the target value and allowable deviation of the stage when evaluating the smoothing parameter. By judging whether the deviation between the real-time value and the target value is within the allowable deviation range, the process is segmented so that when the deviation is within the range, the score is calculated according to a linear ratio, and when the deviation exceeds the range, the score is set to zero. The scores of each smoothing indicator at the same time point are weighted and combined according to the indicator weights to form a comprehensive evaluation sequence with a fixed scale.

5. The oyster farming water quality control method based on AI water quality control as described in claim 4, characterized in that: The step of sorting the comprehensive evaluation quantity and smoothing parameter in chronological order, identifying the direction and magnitude of change within a fixed window, and generating trend records includes, when sorting the comprehensive evaluation quantity and smoothing parameter to generate trend records, dividing the continuous time series into segments according to a fixed-length time window, and arranging various smoothing parameters and comprehensive evaluation quantities in chronological order within the window; The data at the start, middle, and end positions within the window are compared to identify the direction of change of the values ​​within the window. The magnitude of change is constructed using the difference between different time points within the window. The direction and magnitude of change within the same window are combined to form a trend field.

6. The oyster farming water quality control method based on AI water quality control as described in claim 5, characterized in that: The process of combining comprehensive evaluation quantities, trend records, and smoothed water quality parameter sequences according to time indexes to determine the control level and control parameters includes combining comprehensive evaluation quantities, trend directions, trend amplitudes, and real-time smoothing parameters into a decision input structure according to a unified time label when determining the control level and control parameters. By dividing the comprehensive evaluation quantity into fixed numerical intervals, interpreting the trend field according to a preset three-value structure, judging the trend amplitude according to a quantified interval, matching the input fields according to preset rules, and determining the control level accordingly. After the control level is determined, the parameters of control intensity and duration are obtained through a fixed mapping table, and the control level, control parameters and corresponding timestamps are written into the control record table.

7. The oyster farming water quality control method based on AI water quality control as described in claim 6, characterized in that: The process of continuing to collect and process water quality data after the control action is executed, and updating the smooth water quality parameter sequence, comprehensive evaluation quantity and trend record according to the correspondence before and after the execution and writing them into the backfill record includes comparing the smooth parameters in the sampling period before and after the execution in chronological order after the control action is executed, and writing them into the update entry according to the record format. The comprehensive evaluation values ​​at the same time point before and after execution are regenerated according to the established calculation method, and a recalculated trend record is formed through the trend identification process; Write the smoothing parameter update records, comprehensive evaluation quantity update records, and trend update records into the backfill record table.

8. A system employing the AI-based water quality control method for oyster farming as described in any one of claims 1 to 7, characterized in that: It includes a water quality data preprocessing module, a stage water quality assessment modeling module, and a regulation determination and status update module; The water quality data preprocessing module is used to collect raw water quality data and generate a smooth water quality parameter sequence through anomaly removal, time alignment and moving average; The phased water quality evaluation modeling module is used to construct target values, allowable deviations and index weights for different growth stages of oysters from historical samples, and to calculate and weight and integrate the scores of real-time smoothing parameters to generate a comprehensive water quality evaluation quantity that is updated over time. The regulation determination and status update module is used to generate regulation levels and regulation parameters based on comprehensive evaluation quantities and trend records. After regulation is executed, it continues to collect and process water quality data, and updates and backfills the smoothing parameters, comprehensive evaluation quantities and trend records.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the oyster farming water quality control method based on AI water quality control as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the oyster farming water quality control method based on AI water quality control as described in any one of claims 1 to 7.

Images