Water quality monitoring and early warning method for aquaculture

Abstract

The invention discloses a water quality monitoring and early warning method for aquaculture, relates to the technical field of aquaculture, and particularly relates to a water quality monitoring and early warning method for aquaculture by acquiring real-time water quality data and real-time behavior data and combining historical data to analyze whether corresponding parameters in the water quality data and corresponding parameters in the behavior data have a statistical stable coupling structure or not in a non-disjunction state. And forming a sample pair, inputting the sample pair with specific time sequence into a long short-term memory network to output a predicted value of a behavior supervision learning target sample, triggering a verification early warning instruction, starting the verification early warning instruction, and judging whether a current abnormal region has a trend of overall abnormal aggregation or not. According to the method, whether systematic water quality coupling failure exists in the current aquaculture area or not is recognized, the decoupling parameters are locked, the original threshold values of the decoupling parameters are dynamically corrected, and timeliness of water quality monitoring and early warning is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of aquaculture, and in particular to a method for monitoring and early warning of water quality in aquaculture. Background Art

[0002] The present invention relates to the technical field of aquaculture environmental monitoring and risk warning, and in particular to an intelligent water quality warning method for aquaculture ecological environment management. More specifically, it relates to a water quality monitoring and warning method based on multi-source data linkage and dynamic feedback mechanism. The method is suitable for complex water environments such as high-density ponds, factory farming, and recirculating aquaculture for aquatic economic animals such as fish and shrimp. The method focuses on real-time state recognition, abnormal behavior correlation analysis, and parameter adjustment feedback of aquaculture water bodies during aquaculture, with particular attention paid to the complex state recognition problem where water quality indicators are normal but aquaculture animals have exhibited abnormal behavior.

[0003] Currently, conventional aquaculture water quality monitoring systems generally rely on real-time data collection and threshold comparison of basic physical and chemical parameters such as dissolved oxygen, ammonia nitrogen, and pH to provide risk warnings. However, these methods often lack a mechanism for modeling the temporal coupling between water quality changes and the behavior of farmed animals. This makes it difficult to determine whether water quality parameters truly reflect potential risks when farmed animals exhibit obvious behavioral anomalies, such as group agitation or sudden changes in feeding habits.

[0004] The data-phenomenon disconnect problem mainly stems from the fact that the current system ignores the coupling structure between water quality and behavior during modeling, and lacks a linkage reasoning mechanism between behavioral deviations and water quality trends. Once behavioral anomalies cannot be explained by existing water quality parameters, that is, an effective early warning mechanism cannot be triggered, which may lead to missing the risk window. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method for monitoring and early warning of water quality for aquaculture, which solves the problems in the above-mentioned background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for monitoring and early warning of water quality for aquaculture, comprising the following steps:

[0007] S1: Acquire real-time water quality data and real-time behavior data, and combine them with historical data to analyze whether the corresponding parameters in the water quality data and the corresponding parameters in the behavior data have a statistically stable coupling structure in a non-disconnected state, so as to obtain normal coupling feature pairs;

[0008] S2: The water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the behavior parameters belonging to the behavior state are marked as supervised learning target samples to form sample pairs. The sample pairs with time series are input into the long short-term memory network to output the predicted value of the behavior supervised learning target sample. The predicted value of the behavior supervised learning target sample is used to determine whether there is a risk of disconnection between data and phenomena in the current aquaculture area, and to trigger verification warning instructions;

[0009] S3: Start the verification warning instruction to determine whether there is an overall abnormal aggregation trend in the current abnormal area to identify whether there is a systematic water quality coupling failure in the current aquaculture area, and lock the decoupling parameters to dynamically correct the original threshold of the decoupling parameters to achieve timeliness of water quality monitoring and early warning.

[0010] Preferably, several groups of monitoring equipment are deployed in advance in the aquaculture area to monitor the water quality and the behavior of the aquaculture objects in the aquaculture area to obtain real-time water quality data and real-time behavior data;

[0011] The real-time water quality data and the real-time behavior data are preprocessed to construct a standardized vector set under a synchronized time axis. The standardized vector set includes the real-time water quality data and the real-time behavior data. The standardized vector set is used to reflect the real-time water quality data under different behavior conditions.

[0012] Preferably, the historical data under normal conditions is subjected to a grey correlation analysis method to analyze the trend coupling relationship between each parameter in the normal water quality data and each parameter in the normal behavior data, so as to obtain a grey correlation coefficient;

[0013] Combined with historical data, the average grey correlation coefficient between each parameter in normal water quality data and each parameter in normal behavior data is accumulated. The average grey correlation coefficient is used to measure the sensitivity of water quality data to behavior data.

[0014] The nonlinear dependence between the parameters in the normal water quality data and the parameters in the normal behavior data was analyzed using the probability density evaluation method to obtain the mutual information.

[0015] According to the grey correlation coefficient and mutual information accumulated in the historical period, combined with the mean-standard deviation method, the correlation threshold and dependence threshold are obtained respectively. Based on the correlation threshold and dependence threshold, the normal coupling feature pair is identified. Specifically, if the grey correlation coefficient exceeds the correlation threshold and the mutual information exceeds the dependence threshold, it indicates that the corresponding parameters in the water quality data and the corresponding parameters in the behavior data have a statistically stable coupling structure in a non-disconnected state, and the two parameters are recorded as a set of normal coupling feature pairs.

[0016] Preferably, the water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the behavior parameters belonging to the behavior state are marked as supervised learning target samples to form sample pairs, and the sample pairs with time sequence are input into the long short-term memory network to output the predicted value of the behavior supervised learning target sample, including:

[0017] The water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the supervision sample time series is obtained; the behavior parameters belonging to the behavior state in the normal coupling feature pair are marked as supervised learning target samples;

[0018] For the supervised samples at each time point in the supervised sample time series, the supervised learning target samples at the next moment are extracted from the historical data, so as to combine the supervised samples at each time point with the supervised learning target samples at the next moment to form a sample pair;

[0019] According to the supervised sample time series, obtain the corresponding sample pair time series;

[0020] The sample pair time series is used as a training set and input into the long short-term memory network. The long short-term memory network is trained through the input layer, hidden layer and output layer respectively. According to the standardized vector set, the parameters of the supervised samples in the real-time water quality data are input into the trained long short-term memory network to output the predicted value of the target sample of behavioral supervised learning.

[0021] Preferably, the predicted value of the target sample of the behavior supervision learning is used to determine whether there is a risk of data being out of sync with the phenomenon in the current aquaculture area, and to trigger a verification warning instruction, including:

[0022] Wait for the next moment to obtain the true value of the supervised learning target sample;

[0023] The deviation degree is obtained by subtracting the true value of the supervised learning target sample from the predicted value and taking the absolute value. If the deviation degree exceeds the preset deviation threshold, it indicates that there is a risk of disconnection between data and phenomena in the current aquaculture area, and a verification warning instruction will be triggered.

[0024] If the deviation does not exceed the preset deviation threshold, it indicates that there is no risk of data being out of sync with phenomena in the current aquaculture area, and the status in the aquaculture area will continue to be monitored.

[0025] Preferably, the verification warning instruction is started to determine whether there is a trend of overall abnormal aggregation in the current abnormal area to identify whether there is a systematic water quality coupling failure in the current aquaculture area, including:

[0026] Start the verification warning command, divide the aquaculture area into sub-regions, and identify whether there is a risk of disconnection between data and phenomena in each sub-region to obtain the number of abnormal areas;

[0027] Based on the global Moran index method of the spatial weight matrix, the autocorrelation of each abnormal area in the aquaculture area is measured to obtain the clustering index. According to the clustering index value, it is judged whether there is an overall abnormal clustering trend in the current abnormal area. If there is an overall abnormal clustering trend, it indicates that there is a systematic water quality coupling failure in the current aquaculture area, and the corresponding behavior parameter is marked as an abnormal behavior parameter.

[0028] Preferably, the decoupling parameter is locked to dynamically modify the original threshold of the decoupling parameter to achieve the timeliness of water quality monitoring and early warning, including:

[0029] When there is a systematic water quality coupling failure in the aquaculture area, the normal coupling feature pair to which the abnormal behavior parameter belongs is locked, and the water quality parameters belonging to the water quality state are extracted from the normal coupling feature pair to be marked as decoupling parameters;

[0030] Determine the abnormal behavior coverage based on the number of abnormal areas and the total number of sub-areas.

[0031] Preferably, according to the decoupling parameter, an original threshold value corresponding to the decoupling parameter is determined;

[0032] According to the abnormal behavior coverage, the original threshold of the decoupling parameter is dynamically modified. The specific modification method is: θ dyn =θ ori -η*(1-e -λ*Ls ), where θ dyn is the corrected threshold, θ ori is the original threshold of the decoupling parameter, η is the maximum adjustable amplitude, e is the natural base, λ is the response speed coefficient, and Ls is the abnormal behavior coverage.

[0033] The present invention provides a method for monitoring and early warning of water quality for aquaculture, which has the following beneficial effects:

[0034] (1) Real-time water quality data and real-time behavior data are obtained through step S1, and combined with historical samples, the normal coupling structure between behavior parameters and water quality parameters is mined, and a multivariate association system is constructed to effectively identify the hidden risks of abnormal behaviors or phenomena although water quality parameters have not exceeded the limit, further improving the preemptiveness and accuracy of abnormal perception. In step S2, water quality parameters are used as supervision samples and behavior characteristics are used as supervised learning target samples to form sample pairs, and a time series deep learning network is introduced to perform predictive modeling on them. The biological behavior trend can be predicted in advance and compared with the real-time behavior data, and potential coupling failure areas can be identified in time, providing a quantitative indicator basis for the determination of the disconnection state. The threshold is adaptively adjusted dynamically. In step S3, the spatial clustering analysis of the abnormal behavior area is used to identify the systematic coupling failure and further lock the decoupling parameter; a dynamic correction function is constructed based on the abnormal behavior coverage rate, and the original water quality threshold is nonlinearly adjusted, so that the early warning strategy has adaptive adjustment capabilities, effectively compensates for the hysteresis of the static threshold warning, and improves the real-time and accuracy of the early warning response. In summary, this method builds a complete feedback loop of coupled perception, prediction bias, verification aggregation, and threshold correction based on the construction of a behavior-water quality linkage analysis system to further improve complex water quality management and risk prevention and control in aquaculture scenarios.

[0035] (2) The present invention proposes a dual judgment mechanism that combines grey correlation analysis with mutual information evaluation. It evaluates the normal coupling strength between water quality parameters and behavioral parameters from the perspectives of trend sensitivity and nonlinear dependence, measures the degree to which water quality changes follow behavioral changes through the grey correlation coefficient, and then uses mutual information to evaluate the nonlinear coupling degree between parameters. This ensures that the selected normal coupling feature pairs not only have statistical correlation but also have causal coupling explanatory power, providing a stable benchmark model for subsequent "decoupling identification".

[0036] (3) After determining the risk of local disconnection, this method constructs a global Moran index based on the spatial weight matrix and conducts a quantitative clustering analysis of the distribution pattern of abnormal areas within the aquaculture area. By statistically analyzing the spatial autocorrelation of abnormal areas, it is possible to distinguish between "isolated anomalies" and "clustered anomalies" and accurately identify whether there is a systemic risk of overall failure of regional water quality-behavior coupling, thereby improving the macro-judgment ability and response depth of the early warning mechanism. When the systemic coupling failure is identified, the system can automatically lock the corresponding normal coupling feature pair that produces abnormal behavior and track it to the water quality parameter that affects the behavior, thereby marking the parameter as a decoupling parameter. This reverse tracking mechanism not only clarifies the specific intervention target of water quality control, but also lays a causal foundation for subsequent dynamic correction, breaking through the defect of traditional methods that are difficult to locate the source of the fault. In summary, this method can effectively solve the technical shortcomings of traditional water quality monitoring systems, such as the rigidity of parameter threshold setting, weak regional anomaly judgment, and no response to disconnection, and improve the accuracy, timeliness and control intelligence level of water quality early warning in complex aquaculture environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a method for monitoring and early warning water quality for aquaculture according to the present invention;

[0038] Figure 2 This is a logic diagram of a water quality monitoring and early warning method for aquaculture according to the present invention. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] Example 1

[0041] See also Figure 1 and Figure 2 The present invention provides a method for monitoring and warning water quality for aquaculture, comprising the following steps:

[0042] S1: Acquire real-time water quality data and real-time behavior data, and combine them with historical data to analyze whether the corresponding parameters in the water quality data and the corresponding parameters in the behavior data have a statistically stable coupling structure in a non-disconnected state, so as to obtain normal coupling feature pairs;

[0043] S2: The water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the behavior parameters belonging to the behavior state are marked as supervised learning target samples to form sample pairs. The sample pairs with time series are input into the long short-term memory network to output the predicted value of the behavior supervised learning target sample. The predicted value of the behavior supervised learning target sample is used to determine whether there is a risk of disconnection between data and phenomena in the current aquaculture area, and to trigger verification warning instructions;

[0044] S3: Start the verification warning instruction to determine whether there is an overall abnormal aggregation trend in the current abnormal area to identify whether there is a systematic water quality coupling failure in the current aquaculture area, and lock the decoupling parameters to dynamically correct the original threshold of the decoupling parameters to achieve timeliness of water quality monitoring and early warning.

[0045] In this embodiment, through step S1, the system, under normal aquaculture conditions, uses gray correlation and mutual information analysis based on historical water quality data and behavioral data to identify parameter combinations with stable linkage relationships under non-disconnected conditions (i.e., normal ecological conditions) and construct normal coupling feature pairs. This feature pair serves as the basis for subsequent model building, effectively reducing the modeling variable space and improving prediction accuracy and training efficiency. For example, in a certain fish farm, the system identified that dissolved oxygen concentration and fish swimming speed have a high gray correlation and high mutual information, indicating that the two have a stable coupled response relationship under normal conditions, which can be used to construct a supervised prediction model.

[0046] Through step S2, the system uses the water quality parameters in the normal coupling feature pairs as supervision samples and the behavioral parameters as learning targets, constructs time series sample pairs, and inputs them into the long short-term memory network LSTM for training.

[0047] The trained model is used to predict the next behavior based on the current water quality and compare it with the actual observed behavior. If the deviation between the predicted and actual values ​​exceeds a threshold, a disconnect between abnormal behavior and normal water quality can be identified, enabling earlier risk awareness.

[0048] After identifying the abnormal area in step S3, the system initiates a verification warning command. By dividing the aquaculture area into subregions and combining spatial statistical methods (such as the global Moran index based on a spatial weight matrix), it determines whether there is a clustering trend of abnormal behavior. If a systematic coupling failure is detected (i.e., multiple regions exhibit abnormal behavior simultaneously and the coupling model has failed), the system further identifies the feature pair to which the abnormal behavior parameter belongs, and marks the corresponding water quality parameter as a decoupled parameter.

[0049] The system then dynamically adjusts the warning threshold for this decoupling parameter based on the coverage of abnormal areas, enabling it to more sensitively respond to potential water quality risks and avoid significant aquaculture losses due to delayed judgment. For example, if the system determines that unexplained behavioral anomalies are occurring simultaneously in 60% of the sub-areas within the current region, it sets the pH value associated with feeding frequency as the decoupling parameter and dynamically lowers the original threshold of 7.5 to 7.2. This allows for proactive intervention of control strategies, such as activating the aeration system or replacing some water bodies, to effectively control the spread of the risk.

[0050] Normal coupling feature pairs refer to pairs of water quality parameters and behavioral parameters with a stable linkage relationship under normal aquaculture conditions, which are used for subsequent behavior prediction and deviation identification.

[0051] In the supervised samples and the learning target samples, the former are input variables (water quality parameters) and the latter are variables to be predicted (behavioral parameters), which constitute the input of time series modeling.

[0052] The long short-term memory network (LSTM) model is used to predict behavioral trends based on historical water quality sequences and is suitable for deep learning structures that capture long-term dependencies and nonlinear relationships.

[0053] Deviation identification mechanism: used to compare the predicted behavior value with the actual behavior value. If the deviation exceeds the threshold, it indicates that the system is in a potential "decoupling" state.

[0054] Abnormal behavior clustering trend refers to the simultaneous existence of abnormal behavior in multiple sub-areas that cannot be explained by water quality data, indicating that systemic problems are occurring.

[0055] In summary, this method forms a complete closed-loop water quality monitoring and early warning mechanism with predictive, responsive and adaptive characteristics by constructing a coupling structure, behavioral prediction deviation detection, systematic failure identification, and dynamic adjustment of decoupling parameters. It further enhances the early warning system's ability to identify and respond to hidden risks in complex aquaculture environments.

[0056] Example 2

[0057] Please refer to Figure 1 Specifically: deploy several groups of monitoring equipment in the aquaculture area in advance to monitor the water quality and behavior of the aquaculture objects in the aquaculture area to obtain real-time water quality data and real-time behavior data;

[0058] The monitoring equipment includes, but is not limited to, dissolved oxygen sensors, conductivity sensors, pH sensors, potential sensors, turbidity sensors, gas sensors, and image acquisition devices; the monitoring equipment is used to monitor in real time whether the corresponding water quality parameters exceed the threshold value and if an abnormality occurs, triggering the corresponding early warning instructions when an abnormality occurs;

[0059] The real-time water quality data and real-time behavior data are preprocessed. The preprocessing includes noise removal, missing value filling and data smoothing operations. Among them, the missing value filling methods include mean filling, median filling, interpolation filling and regression filling. The preprocessed data are time-aligned to construct a standardized vector set under a synchronized time axis. The standardized vector set includes real-time water quality data and real-time behavior data. The standardized vector set is used to reflect real-time water quality data under different behavioral conditions.

[0060] The grey correlation analysis method is used to analyze the trend coupling relationship between each parameter in the normal water quality data and each parameter in the normal behavior data to obtain the grey correlation coefficient.

[0061] Among them, the grey correlation coefficient is obtained as follows:

[0062]

[0063] Among them, r ij (t) is the behavior parameter B at time t j and water quality parameters P i The grey correlation coefficient, B j (t) is the value of the jth behavior parameter at time t, P i (t) is the value of the i-th water quality parameter at time t, ρ is the resolution coefficient, the value range is [0, 1], and 0.5 is usually used; t is the time index, i is the water quality parameter index, j is the behavior parameter index, min i，t The absolute difference between all water quality parameters and all time points is the minimum, max i，t The maximum absolute difference among all water quality parameters and all time points is taken;

[0064] Molecular part min i，t |B j (t)-P i (t)|+ρ*max i，t |B j (t)-P i (t)| represents the minimum difference + resolution coefficient × maximum difference at the global scale. As a global comparison benchmark, it can be understood as a normalized reference frame that reflects the boundary between the maximum and minimum coupling. i，t and min i，t The purpose is to make the formula global and uniform, that is, to scale the difference between the current behavior and water quality combination, so that each pair of P i With B j The comparisons are all within a unified evaluation range (the minimum is 0 and the maximum is 1) to prevent the incomparability between indicators with different dimensions or ranges.

[0065] Denominator|B j (t)-P i (t)|+ρ*max i，t |B j (t)-P i (t)| represents the actual difference structure between a certain behavior parameter and the water quality parameter at the current moment. The larger the denominator, the greater the current difference and the smaller the grey correlation coefficient (the weaker the coupling). ij The closer (t) is to 1, the stronger the synchronization or trend coupling between the behavioral parameter and the water quality parameter at that moment. On the contrary, the closer it is to 0, the weaker the coupling relationship is, and the behavioral fluctuation is not explained by the water quality indicator.

[0066] The resolution coefficient is a constant used to adjust the sensitivity of the grey correlation coefficient. It controls the distribution range of the grey correlation coefficient and affects the relative correlation strength between different variables. Its value can be obtained through empirical setting method. In most practical applications, the resolution coefficient = 0.5 is the standard value recommended in grey system theory. Its advantages are good stability, strong adaptability, and not easy to cause extreme deviation.

[0067] The grey correlation coefficient can only reflect the consistency of the changing trends (increase and decrease direction and amplitude) of the two time series to score, that is, it focuses on the synchronous trend of numerical changes. If the changing trends of the two variables are highly consistent, the grey correlation coefficient is high; if the trends are opposite or inconsistent, the grey correlation coefficient is low. However, it does not consider whether the response mechanism between the variables is nonlinear, delayed, or threshold-triggered. For example, if the response of fish to dissolved oxygen concentration is not linear, when the dissolved oxygen concentration varies in the range of 6-4 mg / L, the feeding behavior of the fish remains almost unchanged, but when the dissolved oxygen concentration drops below 3 mg / L, the feeding behavior suddenly drops sharply. Therefore, although there is a clear coupled response relationship between dissolved oxygen concentration and feeding frequency, the value given by the grey correlation coefficient will be low. Therefore, a low grey correlation coefficient does not mean that there is no coupling, but that the grey correlation coefficient cannot capture the "threshold triggering" feature. If only the grey correlation coefficient is considered, it may be mistakenly judged that "dissolved oxygen concentration is unrelated to feeding frequency" and this parameter is discarded. By considering the mutual information, the miscoupling situation can be effectively reduced.

[0068] Combined with historical data, the average grey correlation coefficient between each parameter in normal water quality data and each parameter in normal behavior data is accumulated. The average grey correlation coefficient is used to measure the sensitivity of water quality data to behavior data.

[0069] Historical data refers to the water quality data and behavior data collected during the historical period. Normal water quality data and normal behavior data refer to the state data in which no abnormalities are found in the water quality data or the behavior data within the aquaculture area.

[0070] The nonlinear dependence between the parameters in the normal water quality data and the parameters in the normal behavior data was analyzed using the probability density evaluation method to obtain the mutual information.

[0071] The mutual information is obtained as follows:

[0072]

[0073] Among them, MI ij is the behavior parameter B j and water quality parameters P i The mutual information between i , b j ) is the combination of water quality and behavior parameter values ​​(p i , b j ), which represents the probability of the behavior and water quality occurring together in a certain combination state, P(p i ) Water quality parameters P i The value is p i The marginal probability density, P(b j ) is the behavior parameter B j The value is b j The marginal probability density, p i 、b j are the discretized values ​​of water quality parameters and behavior parameters, respectively. log(*) is a logarithmic function with a natural base of approximately 2.71828. P(p i )*P(b j ) is the joint probability when the two are assumed to be independent of each other, It represents the information gain of the relative independence hypothesis of the joint probability. If the ratio is large, it means that the two variables are not independent at this point and have information coupling;

[0074] P(p i , b j )、P(p i )、P(b j ) can be obtained by the histogram estimation method, and B j (t) and P i (t) Perform discrete binning, then count the frequency of each value or combination, and then normalize it into probability to obtain it. Of course, kernel density estimation can also be used to obtain it;

[0075] Discretize p i with b j It can be obtained by equal-interval binning (for example, into 10 bins), equal-frequency binning or kernel density estimation;

[0076] Specifically, the calculation of the mutual information is based on the joint probability distribution modeling of water quality parameters and behavioral parameters in historical samples. The marginal distribution and joint distribution of variables are obtained by probability density estimation and discretization frequency normalization, and finally the coupling strength is calculated through the mutual information function. The larger the mutual information value, the stronger the nonlinear correlation between the variables, which serves as the basis for selecting important parameters for the subsequent establishment of the coupling model.

[0077] According to the grey correlation coefficient and mutual information accumulated in the historical period, combined with the mean-standard deviation method, the correlation threshold and dependence threshold are obtained respectively, and based on the correlation threshold and dependence threshold, the normal coupling feature pair is identified. Specifically, if the grey correlation coefficient exceeds the correlation threshold and the mutual information exceeds the dependence threshold, it indicates that the corresponding parameters in the water quality data and the corresponding parameters in the behavior data have a statistically stable coupling structure in a non-disconnected state, and the two parameters are recorded as a set of normal coupling feature pairs; or, if the grey correlation coefficient exceeds the correlation threshold but the mutual information exceeds the dependence threshold, it indicates that the corresponding parameters in the water quality data and the corresponding parameters in the behavior data have a statistically stable coupling structure in a non-disconnected state, and the two parameters are recorded as a set of normal coupling feature pairs;

[0078] Among them, the non-disconnected state is the normal state;

[0079] In this embodiment, the method realizes the pre-identification of potential water quality risks in the aquaculture environment through multi-dimensional data modeling and parameter sensitivity quantification. By pre-deploying multiple sets of water quality monitoring equipment (such as pH detectors, ammonia nitrogen sensors) and behavior monitoring devices (such as high-definition cameras, sonar sensors) in the aquaculture area, the water quality status and the behavior status of the aquaculture objects can be collected synchronously.

[0080] Subsequently, the system time-aligns, denoises and normalizes all monitoring data to construct a standardized vector set, which can be used to dynamically characterize the real-time water quality status under a specific behavioral background, realizing homologous analysis and feature retention of the data.

[0081] Based on historical data collected about normal conditions, the system uses gray correlation analysis to measure the trend consistency between water quality parameters (such as dissolved oxygen) and behavioral characteristics (such as swimming speed). This method calculates the gray correlation coefficient, which reflects the degree to which a particular water quality indicator explains a specific behavior over time. For example, in historical data from multiple aquaculture cycles, it was found that the gray correlation coefficient between fish feeding activity and ammonia nitrogen concentration in the water reached 0.85, indicating a high degree of synchronization between the two.

[0082] Specifically, the grey correlation coefficient measures the similarity in the trends of two variables; larger values ​​indicate stronger coupling. However, trend consistency alone is often insufficient to determine underlying dependencies. Therefore, this method further introduces mutual information as a nonlinear dependency indicator, estimating the degree of deviation from the joint distribution of water quality and behavioral variables using probability density. If a water quality parameter and a behavioral characteristic not only have similar trends but also exhibit mutual information greater than the dependency threshold, they are considered to have a stable, nonlinear statistical coupling structure.

[0083] Mutual information is used to measure the degree of information sharing between two variables. It can identify nonlinear correlations and is an important tool for distinguishing between synchronization and intrinsic causal coupling. After obtaining all gray correlation coefficients and mutual information, the system uses the mean-standard deviation method to set the discrimination threshold and identify parameter pairs that meet the double-high condition. These are defined as normal coupling feature pairs. These parameter pairs represent stable coupling relationships in a non-disconnected state and are used for subsequent predictive modeling. For example, if (dissolved oxygen, swimming speed) is identified as constituting a normal coupling feature pair, it means that under normal conditions, fluctuations in dissolved oxygen will be reflected in swimming behavior. If significant behavioral abnormalities occur in the future while dissolved oxygen remains unchanged, a data-phenomenon disconnection is suspected.

[0084] Through the linked implementation of the above steps, this method identifies the coupling sensitivity of key water quality parameters and establishes a water quality-behavior response system with statistical stability, providing a model basis and quantitative basis for abnormal but non-abnormal data states. Compared with the traditional single-parameter judgment model based on a fixed threshold, the method of the present invention can effectively identify the risk of coupling failure and greatly enhance the perception of non-explicit threats such as hidden water quality changes and environmental toxin fluctuations, thereby providing early warning and avoiding significant losses caused by biological anomalies.

[0085] Example 3

[0086] Please refer to Figure 1 Specifically: the water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the behavior parameters belonging to the behavior state are marked as supervised learning target samples to form sample pairs. By inputting the sample pairs with time sequence into the long short-term memory network, the predicted value of the behavior supervised learning target sample is output, including,

[0087] The water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the supervision sample time series is obtained; the behavior parameters belonging to the behavior state in the normal coupling feature pair are marked as supervised learning target samples;

[0088] For the supervised samples at each time point in the supervised sample time series, the supervised learning target samples at the next moment are extracted from the historical data, so as to combine the supervised samples at each time point with the supervised learning target samples at the next moment to form a sample pair;

[0089] According to the supervised sample time series, obtain the corresponding sample pair time series;

[0090] The sample pair time series is used as a training set and input into the long short-term memory network (LSTM). The LSTM is trained through the input layer, hidden layer and output layer respectively. According to the standardized vector set, the parameters of the supervised samples in the real-time water quality data are input into the trained LSTM network to output the predicted value of the target sample of behavioral supervised learning.

[0091] The predicted value of the target sample of behavioral supervision learning is used to determine whether there is a risk of disconnection between data and phenomena in the current aquaculture area, and trigger verification warning instructions, including:

[0092] Wait for the next moment to obtain the true value of the supervised learning target sample;

[0093] The deviation degree is obtained by subtracting the true value of the supervised learning target sample from the predicted value and taking the absolute value. If the deviation degree exceeds the preset deviation threshold, it indicates that there is a risk of disconnection between data and phenomena in the current aquaculture area, and a verification warning instruction will be triggered.

[0094] If the deviation does not exceed the preset deviation threshold, it indicates that there is no risk of data being out of sync with phenomena in the current aquaculture area, and the status in the aquaculture area will continue to be monitored.

[0095] After identifying the normal coupling structure, the system uses the water quality parameters within this structure as the primary modeling variables to construct a behavioral prediction model. Specifically, an LSTM (long short-term memory) time series network is used as the model's main framework to enhance its ability to fit the delayed behavioral responses of water quality changes. The system uses the water quality parameter sequence within a sliding time window as the model input and the behavioral feature vector as the prediction target, training a neural network model for predicting future behavioral trends. The model training process optimizes weights by minimizing the mean squared error loss function between the predicted behavioral vector and the true observation vector. To determine whether the coupling relationship has failed, the system constructs a coupling deviation index to quantify the degree of deviation between the current behavioral observation and the predicted value. When this deviation exceeds a set threshold, and all water quality parameters involved in the coupled modeling are within the threshold during this period (i.e., the system does not generate a regular alarm), it is inferred that the behavioral anomaly cannot be explained by water quality changes. In other words, the behavior has shifted but the water quality has not fluctuated, indicating coupling failure and a data-phenomenon disconnect.

[0096] In this embodiment, the method proposes using the water quality parameters in the normal coupling feature pairs as supervisory samples and their corresponding behavioral parameters as supervised learning target samples. The two are combined to form a set of sample pairs. By extracting the historical time series of supervisory samples and pairing the corresponding behavioral targets at the next moment at each time point, a sequence of time-series sample pairs is formed. This sequence is then input into a long-short-term memory (LSTM) network as a training set. The model learns the temporal evolution of the relationship between water quality and behavior through a memory gate mechanism (including input gate, forget gate, and output gate). For example, using dissolved oxygen as the supervisory sample and fish activity as the target sample, the model can learn the statistical structure of the range in which fish activity should be at the next moment under certain water quality conditions. After training is complete, real-time water quality data is fed into the trained LSTM network as new input to obtain a predicted value for the behavioral state at the next moment. Subsequently, by waiting for the actual observed values ​​of the behavioral parameters, an absolute difference calculation is performed between the two to obtain a deviation degree. If this deviation exceeds a preset threshold, a verification warning instruction is triggered, indicating that the system currently faces a potential data-phenomenon disconnect risk. This process is equivalent to dynamically monitoring the stability of the coupled relationship between behavior and water quality, allowing for early detection of ecological fluctuations.

[0097] Deviation is the core criterion. It not only determines whether the behavior is beyond the expectations of water quality, but also has the function of explaining the source of the behavior: if the deviation exists for a long time and the water quality parameters fluctuate very little, it means that the system has entered a non-physical abnormal state (such as disease, microecological imbalance, etc.), providing a decision-making basis for subsequent response strategies. On the contrary, if the water quality fluctuates violently, it is more likely that the behavior change is caused by water quality drift. Therefore, this mechanism is not only used for disconnection identification, but also lays the foundation for abnormal source tracing and early warning model optimization. The supervised sample pair refers to a binary learning unit consisting of a water quality parameter (input variable) and the next behavior parameter (output target) at the corresponding time.

[0098] Deviation is used to measure the difference between behavior prediction results and actual values, and is the core indicator for determining system coupling failure.

[0099] By introducing a supervised sample-driven LSTM behavioral prediction model and using prediction deviation as a risk trigger mechanism, an expected comparison system between "data and phenomena" was effectively established. This not only enhanced the system's ability to predict behavioral responses caused by water quality changes, but also achieved accurate identification of non-water quality-dominated anomalies, breaking through the limitations of traditional threshold-based judgment models that cannot reflect complex ecological states, and ultimately improving the intelligence and response sensitivity of the aquaculture early warning system.

[0100] Example 4

[0101] Please refer to Figure 1Specifically: Start the verification warning instruction, and judge whether there is an overall abnormal clustering trend in the current abnormal area to identify whether there is a systematic water quality coupling failure in the current aquaculture area, including:

[0102] Start the verification warning command, divide the aquaculture area into sub-regions, and identify whether there is a risk of disconnection between data and phenomena in each sub-region to obtain the number of abnormal areas;

[0103] Based on the global Moran index method of the spatial weight matrix, the autocorrelation of each abnormal area in the aquaculture area is measured to obtain the clustering index. According to the clustering index value, it is judged whether there is an overall abnormal clustering trend in the current abnormal area. If there is an overall abnormal clustering trend, it indicates that there is a systematic water quality coupling failure in the current aquaculture area, and the corresponding behavior parameter is marked as an abnormal behavior parameter.

[0104] The clustering index value is obtained by the following formula:

[0105]

[0106] Where I is the clustering index value, N is the number of sub-regions, and x g and x f are the deviations of sub-region g and sub-region f respectively, is the average deviation of all sub-regions, g and f are sub-region numbers; w gf is the spatial weight value, that is, whether the gth area is adjacent to the fth area (if adjacent, it is 1, if not adjacent, it is 0), where w gf This constitutes a spatial weight matrix, a matrix used to represent the adjacency relationship between each region and other regions.

[0107] If the clustering index value is greater than 1, it means that similar values ​​are clustered together, which is positively correlated and presents a clustered distribution. At this time, it is judged that the current abnormal area has an overall abnormal clustering trend. If it is less than 0, it means that similar values ​​are dispersed with each other, which is negatively correlated and presents an anti-clustering or discrete distribution. Between 0 and 1, it means that there is no obvious spatial clustering and presents a random distribution.

[0108] The existence of systematic water quality coupling failures indicates that there is a risk of disconnection between data and phenomena in current aquaculture areas;

[0109] Lock the decoupling parameters to dynamically modify the original thresholds of the decoupling parameters to achieve timely water quality monitoring and early warning, including:

[0110] When there is a systematic water quality coupling failure in the aquaculture area, the normal coupling feature pair to which the abnormal behavior parameter belongs is locked, and the water quality parameters belonging to the water quality state are extracted from the normal coupling feature pair to be marked as decoupling parameters;

[0111] The abnormal behavior coverage rate is determined based on the number of abnormal areas and the total number of sub-areas. The abnormal behavior coverage rate can be used to determine whether the current aquaculture area is globally disturbed or locally disturbed, that is, it can reflect the degree of disturbance.

[0112] The abnormal behavior coverage ratio is the ratio of the number of abnormal areas to the total number of sub-areas, where the total number of sub-areas refers to the total number of sub-areas within the aquaculture area;

[0113] According to the decoupling parameter, an original threshold value corresponding to the decoupling parameter is determined;

[0114] According to the abnormal behavior coverage, the original threshold of the decoupling parameter is dynamically modified. The specific modification method is: θ dyn =θ ori -η*(1-e -λ*Ls ), where θ dyn is the corrected threshold, θ ori is the original threshold of the decoupling parameter, η is the maximum adjustable range, which can be obtained by the safety margin method by setting a fixed ratio, such as 10% to 20% of the original threshold, e is the natural base, the value is about 2.71828, λ is the response speed coefficient, which is used to control the sensitivity of the threshold correction to the coverage rate. The larger the value, the steeper the curve, indicating that the system is more sensitive to the decoupling coverage rate, Ls is the abnormal behavior coverage rate; η*(1-e -λ*Ls ) is the dynamic downward adjustment caused by coverage.

[0115] The response speed coefficient can be tested through simulation or historical data playback to test the performance of the correction curve for different λ under different abnormal behavior coverage. The value that relatively fits the system response speed is selected. The common range is: 2 to 6. It can also be set by the user according to the situation.

[0116] In this embodiment, the present invention introduces the spatial clustering analysis of abnormal behavior areas as the second level of verification for disconnection judgment, that is, after the deviation index triggers the preliminary disconnection judgment, the aquaculture area is further divided into multiple sub-areas, the number of areas with abnormal behavior is counted, and the global Moran index is used to determine whether these abnormal areas show clustering in space. For example: if the abnormal areas are only randomly distributed in corners or a small number of ponds, rather than concentrated in patches, it may be an occasional disturbance; but if multiple adjacent areas continuously trigger disconnection behavior, the Moran index tends to be positive, indicating that the abnormality has a significant clustering trend in space and is highly credible. This spatial autocorrelation analysis, as the core of verifying the early warning instruction, makes up for the false alarm problem that may be caused by relying solely on single-point deviation to trigger the early warning, and effectively improves the robustness and credibility of disconnection identification.

[0117] After clustering is established, this method further traces back the normal coupling feature pairs corresponding to the abnormal behavior parameters and marks the water quality parameters as decoupled parameters, that is, parameters where the current behavior is abnormal but the corresponding water quality does not produce the expected response. This identification mechanism can help the system accurately lock in the culprit of "systemic failure", help guide farmers to pay attention to the sensor accuracy or pollution source changes of the decoupling parameters, provide direct objects for subsequent dynamic adjustment, and form a closed-loop traceability link "from spatial identification to parameter backtracking".

[0118] This method uses an abnormal behavior coverage metric—the ratio of the current number of abnormal areas to the total number of sub-areas—to characterize the prevalence of abnormalities. This metric dynamically adjusts the alarm threshold for the corresponding decoupling parameter based on this information. The correction formula utilizes a nonlinear exponential correction model. This mechanism offers significant advantages over fixed threshold systems. When there are relatively few abnormal areas, the alarm threshold is only fine-tuned to avoid overreaction. When abnormalities are significantly concentrated, the alarm threshold is significantly lowered, enhancing sensitivity. For example, if the original dissolved oxygen threshold is 4.0 mg / L, but current testing reveals a significant decrease in fish feeding frequency and a high concentration index in 70% of the area, the coverage-driven mechanism automatically lowers the threshold to 3.6 mg / L, thereby promptly capturing water quality risks caused by the spread of pollution sources and the potential presence of microcystins.

[0119] In summary, this invention further improves the accuracy and responsiveness of systemic coupling failures by constructing a continuous chain mechanism of disconnection detection, cluster verification, parameter tracking, and adaptive threshold adjustment. Especially in the early stages of water quality deterioration, when signals are not significant, it can reversely guide water quality indicators through behavioral clustering anomalies, breaking the bottleneck of traditional "passive alarms and delayed responses" and providing a more intelligent, forward-looking, and reliable early warning method for aquaculture water quality management.

[0120] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and early warning of water quality for aquaculture, characterized by: The following steps are included: S1: Acquire real-time water quality data and real-time behavior data, and combine them with historical data to analyze whether the corresponding parameters in the water quality data and the corresponding parameters in the behavior data have a statistically stable coupling structure in a non-disconnected state, so as to obtain normal coupling feature pairs; S2: The water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the behavior parameters belonging to the behavior state are marked as supervised learning target samples to form sample pairs. The sample pairs with time series are input into the long short-term memory network to output the predicted value of the behavior supervised learning target sample. The predicted value of the behavior supervised learning target sample is used to determine whether there is a risk of disconnection between data and phenomena in the current aquaculture area, and to trigger verification warning instructions; S3: Start the verification warning instruction to determine whether there is an overall abnormal aggregation trend in the current abnormal area to identify whether there is a systematic water quality coupling failure in the current aquaculture area, and lock the decoupling parameters to dynamically correct the original threshold of the decoupling parameters to achieve timeliness of water quality monitoring and early warning.

2. The method for monitoring and early warning of water quality for aquaculture according to claim 1, characterized in that: Deploy several groups of monitoring equipment in the aquaculture area in advance to monitor the water quality and behavior of the aquaculture objects in the aquaculture area to obtain real-time water quality data and real-time behavior data; The real-time water quality data and the real-time behavior data are preprocessed to construct a standardized vector set under a synchronized time axis. The standardized vector set includes the real-time water quality data and the real-time behavior data. The standardized vector set is used to reflect the real-time water quality data under different behavior conditions.

3. The method for monitoring and early warning of water quality for aquaculture according to claim 2, characterized in that: The grey correlation analysis method is used to analyze the trend coupling relationship between each parameter in the normal water quality data and each parameter in the normal behavior data to obtain the grey correlation coefficient. Combined with historical data, the average grey correlation coefficient between each parameter in normal water quality data and each parameter in normal behavior data is accumulated. The average grey correlation coefficient is used to measure the sensitivity of water quality data to behavior data. The nonlinear dependence between each parameter in the normal water quality data and each parameter in the normal behavior data is analyzed using the probability density evaluation method to obtain the mutual information. According to the grey correlation coefficient and mutual information accumulated in the historical period, combined with the mean-standard deviation method, the correlation threshold and dependence threshold are obtained respectively. Based on the correlation threshold and dependence threshold, the normal coupling feature pair is identified. Specifically, if the grey correlation coefficient exceeds the correlation threshold and the mutual information exceeds the dependence threshold, it indicates that the corresponding parameters in the water quality data and the corresponding parameters in the behavior data have a statistically stable coupling structure in a non-disconnected state, and the two parameters are recorded as a set of normal coupling feature pairs.

4. The method for monitoring and early warning of water quality for aquaculture according to claim 3, characterized in that: The water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the behavior parameters belonging to the behavior state are marked as supervised learning target samples to form sample pairs. By inputting the sample pairs with time sequence into the long short-term memory network, the predicted values ​​of the behavior supervised learning target samples are output, including: The water quality parameters belonging to the water quality state in the normal coupling feature pair are marked as supervision samples, and the supervision sample time series is obtained; Mark the behavioral parameters belonging to the behavioral state in the normal coupling feature pair as the supervised learning target sample; For the supervised samples at each time point in the supervised sample time series, the supervised learning target samples at the next moment are extracted from the historical data, so as to combine the supervised samples at each time point with the supervised learning target samples at the next moment to form a sample pair; According to the supervised sample time series, obtain the corresponding sample pair time series; The sample pair time series is used as a training set and input into the long short-term memory network. The long short-term memory network is trained through the input layer, hidden layer and output layer respectively. According to the standardized vector set, the parameters of the supervised samples in the real-time water quality data are input into the trained long short-term memory network to output the predicted value of the target sample of behavioral supervised learning.

5. The method for monitoring and early warning of water quality for aquaculture according to claim 4, characterized in that: The predicted value of the target sample of behavioral supervision learning is used to determine whether there is a risk of disconnection between data and phenomena in the current aquaculture area, and to trigger verification warning instructions. include, Wait for the next moment to obtain the true value of the supervised learning target sample; The deviation degree is obtained by subtracting the true value of the supervised learning target sample from the predicted value and taking the absolute value. If the deviation degree exceeds the preset deviation threshold, it indicates that there is a risk of disconnection between data and phenomena in the current aquaculture area, and a verification warning instruction will be triggered. If the deviation does not exceed the preset deviation threshold, it indicates that there is no risk of data being out of sync with phenomena in the current aquaculture area, and the status in the aquaculture area will continue to be monitored.

6. A method for monitoring and early warning of water quality for aquaculture according to claim 5, characterized in that: Start the verification warning instruction to determine whether there is a trend of overall abnormal aggregation in the current abnormal area to identify whether there is a systematic water quality coupling failure in the current aquaculture area, including: Start the verification warning command, divide the aquaculture area into sub-regions, and identify whether there is a risk of disconnection between data and phenomena in each sub-region to obtain the number of abnormal areas; Based on the global Moran index method of the spatial weight matrix, the autocorrelation of each abnormal area in the aquaculture area is measured to obtain the clustering index. According to the clustering index value, it is judged whether there is an overall abnormal clustering trend in the current abnormal area. If there is an overall abnormal clustering trend, it indicates that there is a systematic water quality coupling failure in the current aquaculture area, and the corresponding behavior parameter is marked as an abnormal behavior parameter.

7. A method for monitoring and early warning of water quality for aquaculture according to claim 6, characterized in that: Lock the decoupling parameters to dynamically modify the original thresholds of the decoupling parameters to achieve timely water quality monitoring and early warning, including: When there is a systematic water quality coupling failure in the aquaculture area, the normal coupling feature pair to which the abnormal behavior parameter belongs is locked, and the water quality parameters belonging to the water quality state are extracted from the normal coupling feature pair to be marked as decoupling parameters; Determine the abnormal behavior coverage based on the number of abnormal areas and the total number of sub-areas.

8. The method for monitoring and early warning of water quality for aquaculture according to claim 7, characterized in that: According to the decoupling parameter, an original threshold value corresponding to the decoupling parameter is determined; According to the abnormal behavior coverage, the original threshold of the decoupling parameter is dynamically modified. The specific modification method is: θ dyn =θ ori -η*(1-e -λ*Ls ), where θ dyn is the corrected threshold, θ ori is the original threshold of the decoupling parameter, η is the maximum adjustable amplitude, e is the natural base, λ is the response speed coefficient, and Ls is the abnormal behavior coverage.

Images