Aquaculture water quality pollution early warning system and method

Abstract

The invention provides an aquaculture water quality pollution early warning system and method, relates to the field of water quality monitoring, and adopts a data processing and analysis technology based on deep learning. The real-time water basic property data and the water nutrient solution concentration data in the culture water environment are comprehensively analyzed to judge whether the water quality of the water in the culture water environment reaches the standard or not. Therefore, whether the water quality reaches the standard or not can be more comprehensively judged, and the water quality pollution problem can be timely and accurately warned.

Description

Technical Field

[0001] The present application relates to the field of water quality monitoring, and more specifically, to an aquaculture water pollution early warning system and method. Background Art

[0002] As an important part of agriculture, aquaculture is of great significance to ensuring food safety and promoting farmers' income. However, with the promotion of intensive farming models and the increase in production, water pollution has become increasingly prominent and has become one of the main factors restricting the healthy development of aquaculture. Water quality not only directly affects the growth rate, survival rate and economic benefits of aquatic animals, but also affects the quality and safety of aquatic products. Some farmers do not pay enough attention to water quality monitoring and the monitoring frequency is insufficient, which leads to the failure to timely detect water quality problems and take measures to rectify them. Moreover, the existing water quality monitoring system often only focuses on a few key indicators (such as pH value, dissolved oxygen, etc.), while ignoring other indicators that have an important impact on water quality (such as ammonia nitrogen, nitrite, etc.). This incomplete monitoring method is difficult to fully reflect the true condition of water quality.

[0003] Therefore, an optimized early warning scheme for aquaculture water pollution is needed. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an aquaculture water pollution early warning system and method, which adopts data processing and analysis technology based on deep learning, and comprehensively analyzes the real-time basic water property data and water nutrient solution concentration data in the aquaculture water environment to judge whether the water quality in the aquaculture water environment meets the standard. In this way, it is possible to more comprehensively judge whether the water quality meets the standard, and to timely and accurately warn of water pollution problems.

[0005] According to one aspect of the present application, there is provided an aquaculture water pollution early warning system, comprising: Aquaculture water environment water quality detection parameter data acquisition module, used to obtain the pH value, dissolved oxygen content value, water temperature value, ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value of the aquaculture water environment at multiple predetermined time points within a predetermined time period; A water body basic property parameter time series feature extraction module is used to perform time series analysis on the pH value, dissolved oxygen content value, and water temperature value at the plurality of predetermined time points to obtain a water body basic property parameter time series feature vector; A water body nutrient solution concentration time series emphasis feature extraction module is used to perform time series analysis on the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points to obtain a water body nutrient solution concentration time series emphasis feature vector; A water body water quality pollution monitoring parameter time series feature fusion module, used to fuse the water body basic property parameter time series feature vector and the water body nutrient solution concentration time series emphasis feature vector to obtain a water body water quality pollution monitoring parameter time series feature vector; The water quality monitoring result generation module is used to determine whether the water quality of the aquaculture water environment meets the standard based on the time series feature vector of the water quality pollution monitoring parameter.

[0006] According to another aspect of the present application, there is provided an early warning method for aquaculture water pollution, comprising: Obtaining pH value, dissolved oxygen content value, water temperature value, ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value of the aquaculture water environment at multiple predetermined time points within a predetermined time period; Performing time series analysis on the pH values, dissolved oxygen content values, and water temperature values ​​at the plurality of predetermined time points to obtain a time series characteristic vector of basic property parameters of the water body; Performing time series analysis on the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points to obtain a time series emphasized feature vector of the nutrient solution concentration in the water body; The water body basic property parameter time series feature vector and the water body nutrient solution concentration time series emphasis feature vector are integrated to obtain the water body water quality pollution monitoring parameter time series feature vector; Based on the time series characteristic vector of the water quality pollution monitoring parameters, it is judged whether the water quality of the aquaculture water environment meets the standards.

[0007] Compared with the prior art, the aquaculture water pollution early warning system and method provided by this application adopts data processing and analysis technology based on deep learning, and comprehensively analyzes the real-time basic water property data and water nutrient solution concentration data in the aquaculture water environment to judge whether the water quality in the aquaculture water environment meets the standard. In this way, it is possible to more comprehensively judge whether the water quality meets the standard, and to timely and accurately warn of water pollution problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 This is a system block diagram of an aquaculture water pollution early warning system according to an embodiment of the present application.

[0009] Figure 2It is a block diagram of a module for extracting time series characteristics of basic water property parameters in an aquaculture water pollution early warning system according to an embodiment of the present application.

[0010] Figure 3 It is a block diagram of a water nutrient solution concentration time series emphasis feature extraction module in an aquaculture water pollution early warning system according to an embodiment of the present application.

[0011] Figure 4 This is a block diagram of a water quality monitoring result generation module in an aquaculture water pollution early warning system according to an embodiment of the present application.

[0012] Figure 5 Flow chart of an aquaculture water pollution early warning method according to an embodiment of the present application. DETAILED DESCRIPTION

[0013] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0014] As mentioned in the above background technology, aquaculture, as a key link in the agricultural field, plays a vital role in ensuring food supply security and increasing farmers' income. However, with the popularization of intensive farming methods and the increase in production, water pollution has gradually become a major obstacle to the sustainable development of the industry. The quality of water is directly related to the growth rate, survival rate and economic benefits of aquatic organisms, and has a profound impact on the quality and safety of aquatic products. At present, some farmers do not pay enough attention to water quality monitoring and the monitoring frequency is not high, which makes it difficult for them to identify water quality problems in time and take corresponding improvement measures. In addition, the existing water quality monitoring system often only focuses on a few core parameters (such as pH value, dissolved oxygen, etc.), while ignoring other equally important water quality indicators (such as ammonia nitrogen, nitrite, etc.). This one-sided monitoring method cannot fully reveal the true condition of water quality. Therefore, an optimized aquaculture water pollution early warning scheme is expected.

[0015] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for early warning of water pollution in aquaculture.

[0016] Figure 1 FIG. 1 is a system block diagram of an aquaculture water pollution early warning system according to an embodiment of the present application. Figure 1As shown, in the aquaculture water pollution early warning system 100, it includes: aquaculture water environment water quality detection parameter data acquisition module 110, which is used to obtain the pH value, dissolved oxygen content value, water temperature value, ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value of the aquaculture water environment at multiple predetermined time points within a predetermined time period; a water body basic property parameter time series feature extraction module 120, which is used to perform time series analysis on the pH value, dissolved oxygen content value, and water temperature value at the multiple predetermined time points to obtain a water body basic property parameter time series feature vector; a water body nutrient solution concentration time series emphasis feature extraction module 130 , used for performing time series analysis on the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points to obtain a time series emphasis feature vector of the water body nutrient solution concentration; a water body water quality pollution monitoring parameter time series feature fusion module 140, used for fusing the water body basic property parameter time series feature vector and the water body nutrient solution concentration time series emphasis feature vector to obtain a water body water quality pollution monitoring parameter time series feature vector; a water body water quality monitoring result generation module 150, used for judging whether the water quality of the aquaculture water environment meets the standard based on the water body water quality pollution monitoring parameter time series feature vector.

[0017] In the embodiment of the present application, the aquaculture water environment water quality detection parameter data acquisition module 110 is used to obtain the pH value, dissolved oxygen content value, water temperature value, ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value of the aquaculture water environment at multiple predetermined time points within a predetermined time period. It should be understood that the pH value reflects the acidity and alkalinity of the water body, and different aquatic organisms have a specific adaptation range to the pH of the water body; dissolved oxygen is a substance necessary for aquatic animals to breathe, and whether there is enough dissolved oxygen in the water is directly related to whether aquatic organisms can breathe normally; water temperature has a significant impact on aquatic organisms, it will affect the metabolic rate, feeding intensity and reproduction cycle of aquatic organisms, etc. When the water temperature is too high or too low, it will put aquatic organisms in a state of stress, affecting their normal growth and development, and may also change the activity of microorganisms in the water body and the rate of various chemical reactions; ammonia nitrogen mainly comes from the excrement, leftover bait, etc. of aquatic organisms. Decomposition of nitrogen organic matter, when the concentration of ammonia nitrogen in water is too high, it is highly toxic to aquatic animals, which will damage their gill tissue, affect their breathing and osmotic pressure regulation functions, inhibit their growth and even cause death; nitrite is an intermediate product in the process of ammonia nitrogen conversion, and it is also highly toxic to aquatic animals. It can oxidize ferrous hemoglobin in the blood into methemoglobin, causing it to lose its oxygen-carrying capacity, causing aquatic animals to be hypoxic and poisoned, and seriously affecting the health of aquatic organisms; excessive nitrate concentrations will lead to eutrophication of water bodies, causing problems such as large-scale reproduction of algae, destroying the ecological balance of water bodies, affecting water quality stability and the living environment of aquatic organisms. In general, these index values ​​reflect the chemical properties, biological suitability and degree of pollution of water bodies from different aspects. Only when the values ​​of various indicators are in the range suitable for the survival of corresponding aquatic organisms and meet the requirements of water quality standards can the water quality of the aquaculture water environment be judged to be up to standard. That is, these data are interrelated and influence each other, and together constitute the basis for judging whether the water quality meets the standard. Considering that the aquaculture water environment is a dynamic system, it is affected by many factors, such as weather changes (sunny, cloudy, rainy, etc. will affect water temperature, dissolved oxygen, etc.), the growth stage of aquaculture organisms (different excrement and food intake at different stages affect water quality indicators), feeding conditions (feeding amount, feeding time, etc. affect the nutrient content of water bodies), etc. Only obtaining data at a single time point can only reflect the water body conditions at that time, and it is impossible to know the trend of water quality changes over time. By obtaining data at multiple time points, the fluctuations of various water quality indicators at different times can be observed, and the real dynamic changes of water bodies can be more comprehensively and accurately grasped, providing a more sufficient basis for accurately judging whether the water quality meets the standards.

[0018] In the embodiment of the present application, the water body basic property parameter time series feature extraction module 120 is used to perform time series analysis on the pH value, dissolved oxygen content value, and water temperature value at the plurality of predetermined time points to obtain a water body basic property parameter time series feature vector. Specifically, Figure 2FIG. 1 is a block diagram of a water body basic property parameter time series feature extraction module in an aquaculture water pollution early warning system according to an embodiment of the present application. Figure 2 As shown, the water body basic property parameter time series feature extraction module 120 includes: a water body basic property parameter data sorting unit 121, which is used to arrange the pH value, dissolved oxygen content value, and water temperature value of the multiple predetermined time points according to the time dimension to obtain a water body basic property parameter time series input matrix; a water body basic property parameter time series feature generation unit 122, which is used to pass the water body basic property parameter time series input matrix through a water body basic property parameter time series change observer to obtain a water body basic property parameter time series feature matrix; a water body basic property parameter time series feature dimensionality reduction unit 123, which is used to expand the water body basic property parameter time series feature matrix to obtain the water body basic property parameter time series feature vector.

[0019] In an embodiment of the present application, the water body basic property parameter data sorting unit 121 is used to arrange the pH value, dissolved oxygen content value, and water temperature value at the multiple predetermined time points according to the time dimension to obtain a water body basic property parameter time series input matrix. Accordingly, considering that these water quality index data (pH value, dissolved oxygen, water temperature) themselves change dynamically over time, by arranging them according to the time dimension, the changes of various indicators over time can be clearly and intuitively displayed, which helps to explore the water quality change rules hidden behind the data. And through arrangement, the scattered data can be integrated into a unified structure, making the data organization more orderly and standardized, and facilitating subsequent overall data management and comprehensive analysis operations. Different rows in the water body basic property parameter time series input matrix obtained here correspond to data at different time points, and different columns correspond to different water quality basic property parameters (pH value, dissolved oxygen, water temperature).

[0020] In the embodiment of the present application, the water body basic property parameter time series feature generation unit 122 is used to pass the water body basic property parameter time series input matrix through the water body basic property parameter time series change attention device to obtain the water body basic property parameter time series feature matrix. Accordingly, considering that the water body basic property parameter time series input matrix obtained by arrangement only contains the direct record of the basic parameters of water quality at each time point, the deep-level and more representative characteristic laws contained therein have not yet been extracted. For example, it is difficult to directly see the periodic change characteristics of these water quality parameters over a long period of time, the complex interactive change relationship between different parameters, etc. from the original data alone. In order to be able to transform and abstract these original data, dig out the potential characteristic information hidden behind the data that is more valuable for judging whether the water quality meets the standard, and provide strong support for subsequent accurate water pollution warning, it is necessary to encode the water body basic property parameter time series input matrix. The encoding processing method is to input the water body basic property parameter time series input matrix into the water body basic property parameter time series change attention device. In this way, we can capture the trend characteristics of the basic property parameters of each water body changing over time, such as whether it is linear growth, periodic fluctuations or random changes, and at the same time, we can also explore the correlation characteristics between the basic property parameters of different water bodies.

[0021] In particular, the water body basic property parameter time series change observer described in the present application is a convolutional neural network model in which adjacent layers use mutually transposed convolution kernels. In a convolutional neural network, the convolution kernel is one of its core operating elements. It slides on the input data (here, the water body basic property parameter time series input matrix) and calculates according to specific convolution operation rules to extract features at different positions and scales. In a conventional convolutional neural network, each layer uses an independently designed convolution kernel to gradually extract features, while in this special structure, adjacent layers use mutually transposed convolution kernels. Specifically, when the convolution kernel of the previous layer performs convolution operations on the input matrix, it extracts local features according to certain rules such as step size and padding, and the next layer uses the transposed convolution kernel to perform reverse convolution operations on the output of the previous layer (functionally similar to a restoration and reconstruction, but not a simple inverse process). In this way, the two layers form a mutually cooperative and mutually restrictive relationship through this special convolution kernel configuration, aiming to capture the time series feature information in the input data more finely, especially for water quality parameter matrices with time series characteristics, it can better mine their change characteristics along the time dimension. In this way, the continuous change characteristics and periodic change laws of water quality parameters at different time points can be accurately captured, which is more effective for mining the dynamic evolution characteristics of parameters such as water temperature and dissolved oxygen over time, thereby providing more targeted feature information for accurately judging whether the water quality meets the standards.

[0022] In an embodiment of the present application, the water body basic property parameter time series feature dimensionality reduction unit 123 is used to expand the water body basic property parameter time series feature matrix to obtain the water body basic property parameter time series feature vector. It should be understood that although the data in matrix form can well reflect the two-dimensional structure and mutual relationship of each water quality basic property parameter in the time dimension, it is relatively complex in some subsequent calculation scenarios. In order to make the representation of feature data more concise and intuitive, the original two-dimensional matrix structure can be converted into a one-dimensional vector structure through an expansion operation, which can reduce the complexity of the data dimension and facilitate subsequent calculations.

[0023] In the embodiment of the present application, the water body nutrient solution concentration time series emphasis feature extraction module 130 is used to perform time series analysis on the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points to obtain a water body nutrient solution concentration time series emphasis feature vector. Specifically, Figure 3 FIG. 1 is a block diagram of a water body nutrient solution concentration time series emphasis feature extraction module in an aquaculture water pollution early warning system according to an embodiment of the present application. Figure 3 As shown, the water body nutrient solution concentration time series emphasis feature extraction module 130 includes: a water body nutrient solution concentration data sorting unit 131, which is used to arrange the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the multiple predetermined time points according to the time dimension to obtain a water body nutrient solution concentration time series input matrix; a water body nutrient solution concentration time series feature generation unit 132, which is used to pass the water body nutrient solution concentration time series input matrix through a water body nutrient solution concentration time series change capturer to obtain a water body nutrient solution concentration time series feature matrix; a water body nutrient solution concentration time series feature emphasis unit 133, which is used to pass the water body nutrient solution concentration time series feature matrix through a water body nutrient solution concentration time series change filter to obtain a water body nutrient solution concentration time series emphasis feature matrix; a water body nutrient solution concentration time series emphasis feature dimensionality reduction unit 134, which is used to expand the water body nutrient solution concentration time series emphasis feature matrix to obtain the water body nutrient solution concentration time series emphasis feature vector.

[0024] In an embodiment of the present application, the water body nutrient solution concentration data collating unit 131 is used to arrange the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the multiple predetermined time points according to the time dimension to obtain a water body nutrient solution concentration time series input matrix. It should be understood that the concentrations of ammonia nitrogen, nitrite, and nitrate are not fixed in the aquaculture water environment. They will change dynamically over time due to many factors such as the growth metabolism of aquaculture organisms, the feeding of bait, microbial activities, and changes in the external environment. Arranging these data according to the time dimension just fits their time series characteristics, and this arrangement method can clearly show the trend of changes in the concentration parameters of each water body nutrient solution over time. And in the aquaculture water environment, there is a complex mutual conversion relationship between ammonia nitrogen, nitrite, and nitrate, and their concentration changes do not occur in isolation. By arranging the data into a matrix according to the time dimension, it is convenient to study the correlation between the parameters over time, such as observing how the concentrations of nitrite and nitrate change accordingly after the concentration of ammonia nitrogen increases. Understanding the interaction mechanism between these parameters is of great significance for a deep understanding of the ecological processes within the aquaculture water environment and for more accurately judging whether the water quality meets the standards. This can also provide a scientific basis for formulating precise water quality control strategies.

[0025] In an embodiment of the present application, the water body nutrient solution concentration time series feature generation unit 132 is used to pass the water body nutrient solution concentration time series input matrix through the water body nutrient solution concentration time series change capturer to obtain the water body nutrient solution concentration time series feature matrix. In particular, the water body nutrient solution concentration time series change capturer in the present application is a convolutional neural network model using a two-dimensional convolution kernel. It should be understood that in the aquaculture water environment, the concentrations of ammonia nitrogen, nitrite, and nitrate will continue to change over time, and this change is crucial for judging the water quality. The convolutional neural network model of the two-dimensional convolution kernel is good at capturing the local characteristics of data in two-dimensional space and the time series characteristics. For the water body nutrient solution concentration time series input matrix, it itself contains the changes in the concentrations of different nutrients in the time dimension (reflected as data at different time points), and can also be regarded as a data "plane" with a certain structure (different nutrients as different dimensions). The model can focus on the characteristics of the changes in the concentration of these nutrients over time, and accurately extract the time-series change characteristics such as the trend of concentration change (whether it is rising, falling or fluctuating), the difference in the amplitude of change in different time stages, and the synergistic or antagonistic relationship between the changes in the concentration of different nutrients, which can better grasp the dynamic evolution of the nutrient status of water bodies. Compared with directly using the original water nutrient solution concentration time series input matrix, the water nutrient solution concentration time series feature matrix obtained after processing by the convolutional neural network model with a two-dimensional convolution kernel has significantly enhanced its feature expression ability. Some subtle but key change information in the original data is amplified and highlighted through the convolution, pooling and other operations of the model, and some messy information is filtered out, so that the final feature matrix can more concisely and accurately reflect the essential characteristics and changes of the water nutrient solution concentration, providing a more convincing feature basis for the subsequent more accurate judgment of whether the water quality meets the standard, and improving the accuracy of the entire water pollution early warning system in judging the water quality status.

[0026] In an embodiment of the present application, the water body nutrient solution concentration time series feature emphasis unit 133 is used to pass the water body nutrient solution concentration time series feature matrix through a water body nutrient solution concentration time series change filter to obtain a water body nutrient solution concentration time series emphasis feature matrix. Here, the water body nutrient solution concentration time series change filter in the present application refers to a convolutional neural network model using a spatial attention mechanism. It should be understood that in the water body nutrient solution concentration time series feature matrix obtained by the water body nutrient solution concentration time series change capturer, although a lot of feature information about the changes in ammonia nitrogen, nitrite, and nitrate concentrations over time has been extracted, the importance of these features is not completely consistent. Some of these features play a more critical role in accurately judging whether the water quality of the aquaculture water environment meets the standards, while some features may be relatively minor or have a certain degree of redundancy. A convolutional neural network model with a spatial attention mechanism is used as a filter for the time-series changes in the concentration of nutrient solution in water bodies. It can automatically focus on key features that have an important impact on water quality judgment, highlight these key feature information through the screening mechanism, and weaken or even eliminate less important information. This allows subsequent water quality analysis based on these features to focus more on the core points and improve the accuracy and efficiency of judgment.

[0027] In an embodiment of the present application, the water body nutrient solution concentration time series emphasized feature dimensionality reduction unit 134 is used to expand the water body nutrient solution concentration time series emphasized feature matrix to obtain the water body nutrient solution concentration time series emphasized feature vector. It should be understood that each element in the water body nutrient solution concentration time series emphasized feature matrix reflects the situation of different time points and different nutrient (ammonia nitrogen, nitrite, nitrate) concentration related characteristics, but in the matrix form, these elements are scattered in different rows and columns, and the correlation between the features is relatively not intuitive. Through the expansion operation, these originally scattered feature elements can be integrated into a one-dimensional vector, which can more clearly reflect the overall correlation between these features, and then show the comprehensive situation of the characteristics of the water body nutrient solution concentration in the entire time series as a whole. In addition, the expansion operation converts two-dimensional data into one dimension, which can effectively improve the efficiency of data processing.

[0028] In the embodiment of the present application, the water body water quality pollution monitoring parameter time series feature fusion module 140 is used to fuse the water body basic property parameter time series feature vector and the water body nutrient solution concentration time series emphasis feature vector to obtain the water body water quality pollution monitoring parameter time series feature vector. It should be understood that the aquaculture water environment is a complex ecosystem, and there are inextricable connections and mutual influences between the basic properties of the water body and the nutrient concentration in the water body. For example, water temperature will affect the activity of microorganisms in the water, thereby changing the decomposition rate of nitrogen-containing organic matter, thereby affecting the concentration of nutrients such as ammonia nitrogen and nitrates; the level of dissolved oxygen content will also affect chemical reactions involving nutrient conversion such as nitrification and denitrification in the water body, affecting the concentration changes of substances such as nitrite. In turn, changes in the concentration of nutrients in the water body will also affect basic properties such as the pH value of the water body. In order to capture the complex ecological characteristics reflected by these interaction relationships, better restore the real ecological processes within the aquaculture water environment, help to deeply understand the internal mechanism of water quality changes and make more accurate early warning judgments on water quality conditions, it is necessary to integrate the time series characteristic vector of the basic property parameters of the water body and the time series emphasis characteristic vector of the nutrient solution concentration of the water body in this application.

[0029] In the embodiment of the present application, the water quality monitoring result generating module 150 is used to determine whether the water quality of the aquaculture water environment meets the standard based on the water quality pollution monitoring parameter time series feature vector. Specifically, Figure 4 FIG. 1 is a block diagram of a water quality monitoring result generation module in an aquaculture water pollution early warning system according to an embodiment of the present application. Figure 4 As shown, the water body water quality monitoring result generation module 150 includes: a water body water quality pollution monitoring parameter time series feature optimization unit 151, which is used to reconstruct the feature space of the water body water quality pollution monitoring parameter time series feature vector based on decision boundary analysis to obtain an optimized water body water quality pollution monitoring parameter time series feature vector; a water body water quality pollution monitoring parameter time series feature analysis unit 152, which is used to pass the optimized water body water quality pollution monitoring parameter time series feature vector through a water body water quality classifier to obtain a classification result, and the classification result is used to indicate whether the water quality of the aquaculture water environment meets the standard.

[0030] In an embodiment of the present application, the water body water quality pollution monitoring parameter time series feature optimization unit 151 is used to reconstruct the feature space of the water body water quality pollution monitoring parameter time series feature vector based on decision boundary analysis to obtain an optimized water body water quality pollution monitoring parameter time series feature vector. In particular, considering that the original data may contain a large amount of redundant or noise information, these irrelevant information may mask key features, resulting in the classifier being unable to correctly identify the change pattern of water quality, such as a small change in ammonia nitrogen concentration may be an early signal of water quality deterioration, but these small changes are submerged by other irrelevant information, resulting in the classifier making misjudgments when making judgments based on the information in the water body water quality pollution monitoring parameter time series feature vector. Based on this, in the technical solution of the present application, it is necessary to reconstruct the feature space of the water body water quality pollution monitoring parameter time series feature vector based on decision boundary analysis to obtain an optimized water body water quality pollution monitoring parameter time series feature vector.

[0031] Specifically, in an embodiment of the present application, the water body water quality pollution monitoring parameter time series feature optimization unit includes: a classification parameter matrix extraction subunit, which is used to extract the target water body water quality monitoring classification parameter matrix of the water body water quality classifier; a matrix node decomposition subunit, which is used to node-decompose the target water body water quality monitoring classification parameter matrix in units of row vectors to obtain a set of target water body water quality monitoring classification parameter node encoding vectors; a water body water quality pollution monitoring parameter time series feature constraint subunit, which is used to use each target water body water quality monitoring classification parameter node encoding vector in the set of target water body water quality monitoring classification parameter node encoding vectors as the wandering topological space, and respectively perform topological space constraints on the water body water quality pollution monitoring parameter time series feature vectors to obtain a set of constrained water body water quality pollution monitoring parameter time series feature vectors; an optimization feature generation subunit, which is used to calculate the positional mean vector of the set of constrained water body water quality pollution monitoring parameter time series feature vectors to obtain the optimized water body water quality pollution monitoring parameter time series feature vectors.

[0032] More specifically, in an embodiment of the present application, the water body water quality pollution monitoring parameter time series characteristic constraint subunit is used to: multiply the water body water quality pollution monitoring parameter time series characteristic vector and the transposed vector of the target water body water quality monitoring classification parameter node coding vector, and then calculate the natural exponential function value of the multiplication result to obtain the water body water quality monitoring classification weighted index response weight; calculate the Euclidean distance between the water body water quality pollution monitoring parameter time series characteristic vector and the target water body water quality monitoring classification parameter node coding vector to obtain the water body water quality monitoring classification node coding distance value; perform point multiplication on the water body water quality monitoring classification node coding distance value and the water body water quality pollution monitoring parameter time series characteristic vector, and calculate the natural exponential function value for each eigenvalue of the vector after the point multiplication to obtain the water body water quality monitoring classification distance guiding index characteristic vector; perform point multiplication on the water body water quality monitoring classification weighted index response weight and the water body water quality monitoring classification distance guiding index characteristic vector to obtain the constrained water body water quality pollution monitoring parameter time series characteristic vector.

[0033] In the embodiment of the present application, specifically, the water body water quality pollution monitoring parameter time series feature optimization unit is used to: process the water body water quality pollution monitoring parameter time series feature vector according to the following optimization formula to obtain the optimized water body water quality pollution monitoring parameter time series feature vector; wherein the optimization formula is: ;in, represents the target water body water quality monitoring classification parameter matrix, The first, second, and third nodes of the set of classification parameter node encoding vectors representing the target water quality monitoring , The target water body water quality monitoring classification parameter node encoding vector, represents the transpose of a vector, represents the natural exponential function, represents the time series feature vector of water quality pollution monitoring parameters, represents matrix multiplication, It means point multiplication by position. Represents the calculation vector and vector The Euclidean distance between The first one represents the set of time series feature vectors of water quality pollution monitoring parameters after constraints. The time series characteristic vector of water quality pollution monitoring parameters after constraints, Represents the total number of sets of time series feature vectors of water quality pollution monitoring parameters after constraints, Represents the optimized time series feature vector of water quality pollution monitoring parameters.

[0034] That is, in response to the above technical problems, in the technical solution of the present application, the time series feature vector of the water body water quality pollution monitoring parameter is reconstructed in feature space based on decision boundary analysis. The process first extracts the key parameters used for decision-making from the trained water body water quality classifier. These parameters form a matrix in a high-dimensional space, and each row represents the weight or influencing factor in different dimensions. The target water body water quality monitoring classification parameter matrix can provide insight into the location and shape of the model decision boundary, and then infer which input features are most critical to the prediction results.

[0035] Next, the target water body water quality monitoring classification parameter matrix is ​​node-decomposed in units of row vectors to obtain a set of node encoding vectors of target water body water quality monitoring classification parameters. Here, each row vector is a node in graph theory, which means that each set of parameters is now regarded as an entity with potential connectivity. This transformation allows the application of methods from graph theory and network science to explore the interactions between features. The node encoding vector not only carries information about the original parameters, but also implies knowledge about the topological structure of the entire system. Node decomposition further reveals the intrinsic connection pattern or structure of the data, so that the optimized feature vector can better adapt to new task requirements.

[0036] Then, each target water body water quality monitoring classification parameter node encoding vector in the set of the target water body water quality monitoring classification parameter node encoding vector is used as the walking topological space, and the water body water quality pollution monitoring parameter time series feature vector is respectively subjected to topological space constraints to obtain a set of water body water quality pollution monitoring parameter time series feature vectors after constraints. "Walking" in the topological space defined by the node encoding vector is actually simulating an exploratory process, the purpose of which is to find those feature transformations that can best maintain the structural characteristics of the original data. Each step determines the position of the next step based on the probability distribution of the current state. The topological space constraints ensure that even in different contexts, the feature representation still retains certain invariances. At the same time, it can also promote cross-domain transfer learning because it emphasizes the universal relationship between features rather than the details of a specific field. In this way, the reconstruction of the feature space is achieved, making the optimized feature vector more compact and having better generalization ability.

[0037] Finally, the positional mean vector of the set of the constrained water body water quality pollution monitoring parameter time series feature vectors is calculated to obtain the optimized water body water quality pollution monitoring parameter time series feature vector. Calculating the mean vector is a statistical aggregation method used to integrate the optimal solution from multiple perspectives. The idea behind this step is to reduce the deviation caused by a single estimate by fusing the information provided by different sample points. The averaging process is equivalent to performing a soft voting, which enhances the expressiveness of common features and makes the optimized feature vector more stable and reliable.

[0038] In the embodiment of the present application, the water body water quality pollution monitoring parameter time series feature parsing unit 152 is used to pass the optimized water body water quality pollution monitoring parameter time series feature vector through the water body water quality classifier to obtain a classification result, and the classification result is used to indicate whether the water quality of the aquaculture water environment meets the standard. It should be understood that the optimized water body water quality pollution monitoring parameter time series feature vector contains a wealth of comprehensive feature information about the basic properties of the water body and the change of the concentration of the water body nutrient solution over time, but the specific relationship between these features and whether the water quality meets the standard is not intuitively presented, and needs to be mined and revealed by an effective data analysis method. The water body water quality classifier is such a tool, which learns the potential mapping relationship between the characteristics of each dimension of the feature vector and the water quality compliance based on a large amount of labeled training data (i.e., the time series information of the water body monitoring indicators corresponding to different water body samples with known water quality compliance), and then incorporates the optimized water body water quality pollution monitoring parameter time series feature vector to be judged into the mapping relationship for analysis, thereby extracting the information related to whether the water quality meets the standard hidden behind the feature data, and making an accurate classification judgment. A good water quality environment that meets the standards is the basis for the healthy growth and reproduction of aquatic organisms. It can ensure that the farmed organisms have a high survival rate, normal growth rate and good quality, thereby improving the economic benefits of aquaculture. Through the accurate classification results obtained by the water quality classifier, the aquaculture personnel can promptly discover the situation where the water quality does not meet the standards, and quickly take targeted measures such as changing water, adding oxygen, adjusting the water pH, and controlling the amount of feed to improve the water quality, so as to avoid the occurrence of farmed organisms getting sick or dying due to the deterioration of water quality. This can provide a strong guarantee for the sustainable and healthy development of aquaculture.

[0039] In summary, the aquaculture water pollution early warning system 100 based on the embodiment of the present application is explained, which uses data processing and analysis technology based on deep learning to comprehensively analyze the real-time basic water property data and water nutrient solution concentration data in the aquaculture water environment to determine whether the water quality in the aquaculture water environment meets the standard. In this way, it is possible to more comprehensively judge whether the water quality meets the standard, and to timely and accurately warn of water pollution problems.

[0040] Figure 5FIG. 1 is a flow chart of an aquaculture water pollution early warning method according to an embodiment of the present application. Figure 5 As shown, in the aquaculture water quality pollution early warning method, it includes: S1, obtaining the pH value, dissolved oxygen content value, water temperature value, ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value of the aquaculture water environment at multiple predetermined time points within a predetermined time period; S2, performing time series analysis on the pH value, dissolved oxygen content value, and water temperature value at the multiple predetermined time points to obtain a time series characteristic vector of basic water property parameters; S3, performing time series analysis on the ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value at the multiple predetermined time points to obtain a time series emphasis characteristic vector of water nutrient solution concentration; S4, fusing the time series characteristic vector of basic water property parameters and the time series emphasis characteristic vector of water nutrient solution concentration to obtain a time series characteristic vector of water quality pollution monitoring parameters; S5, judging whether the water quality of the aquaculture water environment meets the standard based on the time series characteristic vector of water quality pollution monitoring parameters.

[0041] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned aquaculture water pollution early warning method have been described in the above reference. Figures 1 to 4 The invention has been introduced in detail in the description of the early warning system for aquaculture water pollution, and therefore, its repeated description will be omitted.

[0042] In summary, the aquaculture water pollution early warning method based on the embodiment of the present application is explained, which uses data processing and analysis technology based on deep learning to comprehensively analyze the real-time basic water property data and water nutrient solution concentration data in the aquaculture water environment to determine whether the water quality in the aquaculture water environment meets the standard. In this way, it is possible to more comprehensively judge whether the water quality meets the standard, and to timely and accurately warn of water pollution problems.

Claims

1. An aquaculture water pollution early warning system, characterized in that: include: The aquaculture water environment water quality detection parameter data acquisition module is used to obtain the pH value, dissolved oxygen content value, water temperature value, ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value of the aquaculture water environment at multiple predetermined time points within a predetermined time period; A water body basic property parameter time series feature extraction module is used to perform time series analysis on the pH value, dissolved oxygen content value, and water temperature value at the plurality of predetermined time points to obtain a water body basic property parameter time series feature vector; A water body nutrient solution concentration time series emphasis feature extraction module is used to perform time series analysis on the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points to obtain a water body nutrient solution concentration time series emphasis feature vector; A water body water quality pollution monitoring parameter time series feature fusion module, used to fuse the water body basic property parameter time series feature vector and the water body nutrient solution concentration time series emphasis feature vector to obtain a water body water quality pollution monitoring parameter time series feature vector; The water quality monitoring result generation module is used to determine whether the water quality of the aquaculture water environment meets the standard based on the time series feature vector of the water quality pollution monitoring parameter.

2. The aquaculture water pollution early warning system according to claim 1, characterized in that: The water body basic property parameter time series feature extraction module includes: A water body basic property parameter data sorting unit, used for arranging the pH values, dissolved oxygen content values, and water temperature values ​​at the plurality of predetermined time points according to the time dimension to obtain a water body basic property parameter time series input matrix; A water body basic property parameter time series feature generation unit, used for passing the water body basic property parameter time series input matrix through the water body basic property parameter time series change attention device to obtain a water body basic property parameter time series feature matrix; The water body basic property parameter time series characteristic dimension reduction unit is used to expand the water body basic property parameter time series characteristic matrix to obtain the water body basic property parameter time series characteristic vector.

3. The aquaculture water pollution early warning system according to claim 2 is characterized in that: The water body nutrient solution concentration time series emphasized feature extraction module includes: A water body nutrient solution concentration data sorting unit, used for arranging the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points according to the time dimension to obtain a water body nutrient solution concentration time series input matrix; A water body nutrient solution concentration time series characteristic generating unit, used for passing the water body nutrient solution concentration time series input matrix through a water body nutrient solution concentration time series change capture device to obtain a water body nutrient solution concentration time series characteristic matrix; a water body nutrient solution concentration time series feature emphasis unit, used for passing the water body nutrient solution concentration time series feature matrix through a water body nutrient solution concentration time series change filter to obtain a water body nutrient solution concentration time series emphasis feature matrix; The water body nutrient solution concentration time series emphasis feature dimensionality reduction unit is used to expand the water body nutrient solution concentration time series emphasis feature matrix to obtain the water body nutrient solution concentration time series emphasis feature vector.

4. The aquaculture water pollution early warning system according to claim 3, characterized in that: The water body basic property parameter temporal change attention device is a convolutional neural network model in which adjacent layers use mutually transposed convolution kernels, the water body nutrient solution concentration temporal change capture device is a convolutional neural network model using two-dimensional convolution kernels, and the water body nutrient solution concentration temporal change filter is a convolutional neural network model using a spatial attention mechanism.

5. The aquaculture water pollution early warning system according to claim 4, characterized in that: The water quality monitoring result generating module comprises: A water body water quality pollution monitoring parameter time series feature optimization unit, used for performing feature space reconstruction based on decision boundary analysis on the water body water quality pollution monitoring parameter time series feature vector to obtain an optimized water body water quality pollution monitoring parameter time series feature vector; The water body pollution monitoring parameter time series feature analysis unit is used to pass the optimized water body pollution monitoring parameter time series feature vector through a water body quality classifier to obtain a classification result, and the classification result is used to indicate whether the water quality of the aquaculture water environment meets the standard.

6. The aquaculture water pollution early warning system according to claim 5, characterized in that: The water body water quality pollution monitoring parameter time series characteristic optimization unit comprises: A classification parameter matrix extraction subunit, used to extract a target water body water quality monitoring classification parameter matrix of the water body water quality classifier; A matrix node decomposition subunit, used for performing node decomposition on the target water body water quality monitoring classification parameter matrix in units of row vectors to obtain a set of target water body water quality monitoring classification parameter node encoding vectors; A water body water quality pollution monitoring parameter time series feature constraint subunit is used to use each target water body water quality monitoring classification parameter node encoding vector in the set of target water body water quality monitoring classification parameter node encoding vectors as a wandering topological space, and to perform topological space constraints on the water body water quality pollution monitoring parameter time series feature vectors respectively to obtain a set of constrained water body water quality pollution monitoring parameter time series feature vectors; The optimization feature generation subunit is used to calculate the position mean vector of the set of the constrained water body water quality pollution monitoring parameter time series feature vectors to obtain the optimized water body water quality pollution monitoring parameter time series feature vectors.

7. The aquaculture water pollution early warning system according to claim 6, characterized in that: The water body water quality pollution monitoring parameter time series characteristic constraint subunit is used to: After multiplying the water body water quality pollution monitoring parameter time series feature vector and the transposed vector of the target water body water quality monitoring classification parameter node encoding vector, the natural exponential function value of the multiplication result is calculated to obtain the water body water quality monitoring classification weighted index response weight; Calculate the Euclidean distance between the water body water quality pollution monitoring parameter time series feature vector and the target water body water quality monitoring classification parameter node coding vector to obtain the water body water quality monitoring classification node coding distance value; Performing a dot multiplication on the water body water quality monitoring classification node encoding distance value and the water body water quality pollution monitoring parameter time series feature vector, and calculating a natural exponential function value for each feature value of the vector after the dot multiplication to obtain a water body water quality monitoring classification distance guidance index feature vector; The water body water quality monitoring classification weighted index response weight and the water body water quality monitoring classification distance guidance index feature vector are point-multiplied to obtain the constrained water body water quality pollution monitoring parameter time series feature vector.

8. A method for early warning of aquaculture water pollution, characterized in that: include: Obtaining pH value, dissolved oxygen content value, water temperature value, ammonia nitrogen concentration value, nitrite concentration value, and nitrate concentration value of the aquaculture water environment at multiple predetermined time points within a predetermined time period; Performing time series analysis on the pH values, dissolved oxygen content values, and water temperature values ​​at the plurality of predetermined time points to obtain a time series characteristic vector of basic property parameters of the water body; Performing time series analysis on the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points to obtain a time series emphasized feature vector of the nutrient solution concentration in the water body; The water body basic property parameter time series feature vector and the water body nutrient solution concentration time series emphasis feature vector are integrated to obtain the water body water quality pollution monitoring parameter time series feature vector; Based on the time series characteristic vector of the water quality pollution monitoring parameters, it is judged whether the water quality of the aquaculture water environment meets the standards.

9. The aquaculture water pollution early warning method according to claim 8, characterized in that: The pH values, dissolved oxygen content values, and water temperature values ​​at the plurality of predetermined time points are subjected to time series analysis to obtain a time series characteristic vector of basic water property parameters, including: Arrange the pH values, dissolved oxygen content values, and water temperature values ​​at the plurality of predetermined time points according to the time dimension to obtain a time series input matrix of basic water property parameters; The water body basic property parameter time series input matrix is ​​passed through the water body basic property parameter time series change observer to obtain the water body basic property parameter time series characteristic matrix; The time series characteristic matrix of the basic property parameters of the water body is expanded to obtain the time series characteristic vector of the basic property parameters of the water body.

10. The aquaculture water pollution early warning method according to claim 9, characterized in that: Performing time series analysis on the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points to obtain a time series emphasized feature vector of the nutrient solution concentration in the water body, including: Arrange the ammonia nitrogen concentration values, nitrite concentration values, and nitrate concentration values ​​at the plurality of predetermined time points according to the time dimension to obtain a water body nutrient solution concentration time series input matrix; The water body nutrient solution concentration time series input matrix is ​​passed through a water body nutrient solution concentration time series change capture device to obtain a water body nutrient solution concentration time series characteristic matrix; The water body nutrient solution concentration time series characteristic matrix is ​​passed through a water body nutrient solution concentration time series change filter to obtain a water body nutrient solution concentration time series emphasized characteristic matrix; The water body nutrient solution concentration time series emphasis feature matrix is ​​expanded to obtain the water body nutrient solution concentration time series emphasis feature vector.

Images