Pond dissolved oxygen prediction method based on Informer model

Abstract

According to the high-precision dissolved oxygen prediction method provided by the invention, the dissolved oxygen prediction accuracy in pond culture is remarkably improved through environment monitoring and an Informer model. According to the method, aiming at the problems of low dissolved oxygen and high organic matter content, environment and water quality parameters such as temperature, pH value, turbidity, salinity, ammonia nitrogen concentration, chemical oxygen demand, residual chlorine, total phosphorus, illumination intensity, wind speed, humidity and solar radiation are comprehensively monitored by arranging sensors. In addition, the relationship of dissolved oxygen among different water layers is deeply analyzed, and an accurate vertical distribution model is established. Key variables are screened through data preprocessing and a Spearman correlation coefficient method, specific dissolved oxygen vertical distribution characteristics are combined, input parameters of an Informer model are designed, and data normalization and model training are carried out. Finally, the pond dissolved oxygen is accurately predicted by using the trained model, and the efficiency of the aerator and the biological safety of the water body are improved. According to the technical scheme, the breeding efficiency can be remarkably improved, organic matter redundancy is reduced, and sustainable development of the aquaculture industry is promoted.

Description

Technical Field

[0001] This invention is part of the intersection of aquaculture technology and artificial intelligence, focusing on using advanced machine learning models, specifically the Informer model, to accurately predict dissolved oxygen concentration in ponds. This method, combined with Internet of Things technology, aims to improve the environmental sustainability and economic benefits of aquaculture processes. Background Art

[0002] Environmental management of aquaculture is a key factor in ensuring aquaculture benefits and ecological balance. Among them, the content of dissolved oxygen directly affects the healthy growth of aquatic organisms and the self-purification capacity of water bodies. The main challenges faced by traditional pond aquaculture include: the uneven distribution of dissolved oxygen concentration in deep-water aquaculture environments, and the difficulty of decomposing bottom organic matter due to lack of oxygen, which not only increases the risk of aquaculture, but also reduces water quality. At present, the real-time monitoring and prediction of dissolved oxygen by using Internet of Things technology combined with intelligent algorithms can realize the automatic adjustment of aerators, thereby maintaining the stability of dissolved oxygen concentration. This method can not only effectively reduce energy consumption and pollution, but also conform to the concept of green and low-carbon aquaculture, and is of great significance to achieving national carbon reduction goals. Despite this, there are also problems such as insufficient capture of dependencies, difficulty in parallel processing, high memory requirements and computational complexity of model training when time series prediction algorithms are processing long sequence data. Therefore, it is particularly important to develop a prediction model with high prediction accuracy, high efficiency and low computational complexity. This is not only an important basis for realizing adaptive adjustment of aerators and optimizing the dissolved oxygen concentration in ponds, but also an effective means to ensure the maximization of economic, social and ecological benefits of the aquaculture process. Summary of the invention

[0003] The present invention provides an efficient and accurate method for predicting dissolved oxygen, which uses an advanced informer model combined with real-time monitoring of pond water quality and environmental parameter data to achieve high-precision prediction of dissolved oxygen content and optimal control of aerators. This method can effectively maintain the dissolved oxygen concentration in the water at an ideal level and promote the healthy and sustainable development of aquaculture. The technical solutions adopted are as follows:

[0004] The present invention collects relevant water quality / environmental parameter data in a pond, pre-processes the data, and then inputs the data into an Informer model for learning to obtain a prediction result.

[0005] Specifically, a method for predicting dissolved oxygen in a pond based on an Informer model, the method comprising:

[0006] Step 1: Arrange sensors to collect water quality / environmental data of the pond and meteorological environmental data of the farm in real time, obtain relevant water quality and meteorological raw data, and perform data preprocessing on the relevant water quality and meteorological raw data;

[0007] Step 2: construct a vertical distribution model of dissolved oxygen based on diffusion theory to extract the characteristics of dissolved oxygen distribution; use the Spearman correlation coefficient method to perform correlation analysis on the pretreated water quality / environmental data and the dissolved oxygen concentration in the pond, determine the correlation coefficient between the relevant water quality / environmental data and the dissolved oxygen concentration in the pond, and select data parameters with a correlation coefficient greater than 0.4;

[0008] Step 3, using the vertical distribution characteristics of dissolved oxygen in step 2 and the water quality / environmental parameters with a correlation coefficient greater than 0.4 as inputs of the Informer model, and predicting the dissolved oxygen concentration value of the pond at the next moment through the Informer model;

[0009] Furthermore, the water quality / environmental data in step 1 include pond water temperature, pH value, turbidity, salinity, ammonia nitrogen concentration, residual chlorine, chemical oxygen demand, total phosphorus and dissolved oxygen concentration; the relevant meteorological environmental data include solar radiation, atmospheric pressure, temperature, humidity and light intensity.

[0010] Furthermore, the data preprocessing described in step 1 includes data cleaning and data normalization. The processed data can be used as input parameters of the Informer model. The normalization formula is as follows:

[0011]

[0012] Where X * represents the normalized data, X max is the maximum value of X among all data, X min is the minimum value of X among all data, X n is the new value obtained after normalization.

[0013] Furthermore, the dissolved oxygen vertical distribution model described in step 2 is characterized in that the established dissolved oxygen vertical distribution model based on diffusion theory is:

[0014]

[0015] Where C is the concentration of the transported substance, z is the depth of a certain point in the water body, P is the photosynthetic rate, and R mac Indicates the respiration of aquatic plants, M in Indicates the mineralization of organic matter in water. Further, after using this model to predict the vertical distribution of dissolved oxygen concentration, this output is used as the input of the subsequent Informer model. Specifically, the steps of establishing the vertical distribution model of dissolved oxygen include:

[0016] Step 301: Establish the diffusion term of the diffusion equation. Usually, the unidirectional transport of substances in water is usually described by the one-dimensional convection-diffusion equation:

[0017]

[0018] Where C is the concentration of the transported substance, A is the horizontal surface area at the depth z of the water body, and K z K is the vertical diffusion coefficient of water body, and S is the net production rate of substances transported by each water layer. z It is related to the vertical distribution of dissolved oxygen in the water body and the thermal stratification of the water body. The traditional calculation method is:

[0019] K z =α(N 2 ) -γ

[0020] Among them, α and β are constants, which are determined by the turbulent top of the water body, and N 2 It is determined by the ratio of the density of the water body at the current location to the acceleration of gravity. Assuming that in a rectangular space of the pond water body, the concentration of oxygen molecules changes over time, and the oxygen molecules follow the law of conservation of mass during the transfer process, then within a unit time τ, there will be a flux of J(x) oxygen molecules entering this space, and at the same time, there will be a flux of J(x+Δx) oxygen molecules leaving the space point. When τ→0, Δx→0, the dissolved oxygen concentration at the space point changes to:

[0021]

[0022] Combined with Fick's first law:

[0023]

[0024] Substituting x in the formula with the pond depth z, the diffusion term of the vertical diffusion equation of dissolved oxygen in the pond is:

[0025]

[0026] Step 302: Establish photosynthetic rate source term, the established model is:

[0027] P=P max ×f m (I)×B

[0028] Where P max represents the maximum daily oxygen production rate of aquatic plants, P max =a×e bT , a is the regression coefficient, b is the temperature coefficient; f m (I) is the light confinement function, I is the solar radiation intensity on the water surface, I optis the optimal solar radiation intensity for photosynthesis of aquatic plants; B is the biomass of aquatic plants.

[0029] Step 303: Establish the source term of aquatic plant respiration: The established model is:

[0030] R max =resp×B

[0031] In the formula, resp represents the limiting function of water temperature on respiration rate, resp = m×e nT , m is the exponential regression coefficient, and n is the respiration temperature correction coefficient.

[0032] Step 304: Establish the source term of organic mineralization. The established model is:

[0033]

[0034] Where min0 is the standard mineralization rate at 0°C, k T is the biochemical reaction temperature coefficient, O om is the oxygen demand per mole of organic molecules, C om is the concentration of organic matter in water.

[0035] Furthermore, in step 2, the Spearman correlation coefficient method is used to perform correlation coefficient analysis on the pre-processed relevant water quality / environmental data:

[0036]

[0037] Where cov(x,y) represents the covariance of x and y; σ(x) represents the standard deviation of x; σ(y) represents the standard deviation of y; x i Represents the variable x with serial number i; represents the average value of variable x; y i Represents the variable y with serial number i; Represents the mean value of the variable y.

[0038] Furthermore, the step 3 described in which the Informer model is used to predict the dissolved oxygen concentration at the next moment specifically includes:

[0039] Step 301: Use the Informer model to train the processed data to learn the relationship between the dissolved oxygen concentration and the selected input parameters. Specifically, the Informer model is an encoder-decoder structure.

[0040] Step 302: In the encoder, the ProbSparse self-attention mechanism is used, which can achieve O(LlogL) in terms of time complexity and memory usage, where L is the sequence length. The specific formula is

[0041]

[0042] Among them, Q represents query, K represents key, V represents value, and d represents dimension.

[0043] Step 303: Further, in the encoder, a self-attention distillation mechanism is used to apply a Conv1d layer to each attention layer result, and then add a Maxpooling layer to halve the output of each layer, highlight the dominant attention, and effectively process overly long input sequences. The process of the distilling operation from layer j to layer j+1 is as follows:

[0044]

[0045] Among them, Conv1d represents the one-dimensional convolution operation on the time series, and ELU is the activation function.

[0046] Step 304: In the decoder, a parallel generative decoder mechanism is used for prediction, which is specifically a batch generative prediction that directly outputs multi-step prediction results. The formula is as follows:

[0047]

[0048] in, Represents a placeholder (predicted value), Indicates the start character.

[0049] Step 305: After the model training is completed, the cross-test dataset is used to evaluate the performance of the model and output the single-step or multi-step prediction step size.

[0050] Furthermore, the high-precision sensor used in step 1 collects the water quality / environmental data of the pond in real time to obtain relevant water quality / environmental raw data, including:

[0051] s1: Dynamic sampling time period setting: Dynamically adjust the interval sampling time period according to actual monitoring needs, and the initial setting is 120 seconds. This dynamic adjustment mechanism allows automatic adjustment of the sampling frequency according to environmental changes and specific events, ensuring that higher density data is obtained at critical moments.

[0052] s2: In each sampling period, water quality parameters including pH value, turbidity, salinity, ammonia nitrogen concentration, chemical oxygen demand, residual chlorine, total phosphorus, dissolved oxygen concentration, and environmental parameters such as light intensity, wind speed, humidity, solar radiation, and temperature are collected. Intelligent strategies are adopted to adaptively adjust the collection density of each parameter according to the current environmental conditions and historical data analysis results, so as to improve the collection efficiency while ensuring the comprehensiveness of the data.

[0053] s3: For the data collected in each standard period, data smoothing technology is used to convert parameter values ​​with multiple data points into average values ​​within the period, ensuring that each water quality and environmental parameter is presented as a single data point in the final analysis. This step aims to reduce data noise and improve the accuracy of model predictions.

[0054] s4: The processed data is combined into a single water quality / environmental data set for model training and real-time prediction of dissolved oxygen levels. This ensures that the data from each sampling period effectively reflects the current state of the pond and provides accurate input for dissolved oxygen prediction.

[0055] Beneficial effects of the present invention:

[0056] The present invention introduces an innovative dissolved oxygen prediction mechanism by designing a pond dissolved oxygen prediction method based on the Informer model. This method cleverly combines the physics-based diffusion theory model with the data-driven Informer prediction model, aiming to deeply analyze the various factors affecting the dissolved oxygen concentration, and achieve high-precision prediction of the dissolved oxygen concentration through this fusion model. Through accurate prediction of dissolved oxygen concentration, the present invention can not only ensure the stability of the dissolved oxygen level in the pond water body and optimize water quality management, but also significantly improve the overall water quality, providing important support for the sustainable development of the aquaculture industry. In addition, the method significantly reduces energy consumption by optimizing the operation strategy of the aerator, helps to reduce carbon emissions in the aquaculture process, and demonstrates its positive impact on environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flow chart of the present invention;

[0058] Figure 2 This is a schematic diagram of the structure of the Informer prediction model; DETAILED DESCRIPTION

[0059] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0060] The present invention proposes a pond dissolved oxygen prediction method based on Informer. By deploying high-precision sensors in the pond to monitor and collect pond water quality / environmental data in real time, a pond dissolved oxygen vertical distribution model is constructed. The future dissolved oxygen concentration value is preliminarily predicted by theoretical methods. At the same time, the Informer prediction model is used, combined with the dissolved oxygen characteristics of the vertical distribution model, and the future dissolved oxygen concentration prediction value is predicted and output in real time, which provides a guarantee for the subsequent control of the aerator to adaptively increase oxygen to maintain the stability of the dissolved oxygen at the bottom of the pond, thereby maximizing the improvement of the pond water quality and saving the energy consumption of the aerator. The technical scheme adopted is as follows:

[0061] The present invention utilizes water quality / environmental data collected by sensors deployed in ponds and dissolved oxygen vertical distribution features extracted by a constructed dissolved oxygen vertical distribution model, performs data preprocessing on these data and inputs them into an Informer prediction model for learning to obtain prediction results.

[0062] Specifically, a pond dissolved oxygen prediction method based on the Informer model, the dissolved oxygen prediction method includes:

[0063] Step 1: Arrange sensors to collect water quality / environmental data of the pond and meteorological environmental data of the farm in real time, obtain relevant water quality and meteorological raw data, and perform data preprocessing on the relevant water quality and meteorological raw data;

[0064] Step 2: construct a vertical distribution model of dissolved oxygen based on diffusion theory to extract the characteristics of dissolved oxygen distribution; use the Spearman correlation coefficient method to perform correlation analysis on the pretreated water quality / environmental data and the dissolved oxygen concentration in the pond, determine the correlation coefficient between the relevant water quality / environmental data and the dissolved oxygen concentration in the pond, and select data parameters with a correlation coefficient greater than 0.4;

[0065] Step 3, using the vertical distribution characteristics of dissolved oxygen in step 2 and the water quality / environmental parameters with a correlation coefficient greater than 0.4 as inputs of the Informer model, and predicting the dissolved oxygen concentration value of the pond at the next moment through the Informer model;

[0066] Specifically, the water quality environmental data in step 1 include pond water temperature, pH value, turbidity, salinity, ammonia nitrogen, residual chlorine, chemical oxygen demand, total phosphorus and dissolved oxygen concentration; the relevant meteorological environmental data include solar radiation, atmospheric pressure, temperature, humidity and light intensity.

[0067] Specifically, the data preprocessing described in step 1 includes data cleaning and data normalization. The processed data can be used as input parameters of the Informer model. The normalization formula is as follows:

[0068]

[0069] Where X * represents the normalized data, X max is the maximum value of X among all data, X min is the minimum value of X among all data, X n is the new value obtained after normalization.

[0070] Specifically, the dissolved oxygen vertical distribution model based on diffusion theory established in step 2 is:

[0071]

[0072] Where C is the concentration of the transported substance, z is the depth of a certain point in the water body, P is the photosynthetic rate, and R mac Indicates the respiration of aquatic plants, M in Indicates the mineralization of organic matter in water. Further, after using this model to predict the vertical distribution of dissolved oxygen concentration, this output is used as the input of the subsequent Informer model. The specific steps of establishing the vertical distribution model of dissolved oxygen include:

[0073] Step 201: Establish the diffusion term of the diffusion equation. Usually, the unidirectional transport of substances in water is usually described by the one-dimensional convection-diffusion equation:

[0074]

[0075] Where C is the concentration of the transported substance, A is the horizontal surface area at the depth z of the water body, and K z K is the vertical diffusion coefficient of water body, and S is the net production rate of substances transported by each water layer. z It is related to the vertical distribution of dissolved oxygen in the water body and the thermal stratification of the water body. The traditional calculation method is:

[0076] K z =α(N 2 ) -γ

[0077] Among them, α and γ are constants, which are determined by the turbulent top of the water body, and N 2 It is determined by the ratio of the density of the water body at the current location to the acceleration of gravity. Assuming that in a rectangular space of the pond water body, the concentration of oxygen molecules changes over time, and oxygen molecules follow the law of conservation of mass during the transfer process, then in a unit time υ, there will be J(x) oxygen molecules entering this space, and at the same time, there will be J(x+Δx) oxygen molecules leaving the space point. When τ→0, Δx→0, the dissolved oxygen concentration at the space point changes to:

[0078]

[0079] Combined with Fick's first law:

[0080]

[0081] Substituting x in the formula with the pond depth z, the diffusion term of the vertical diffusion equation of dissolved oxygen in the pond is:

[0082]

[0083] Step 202: Establish photosynthetic rate source term, the established model is:

[0084] P=P max ×f m (I)×B

[0085] Where P max represents the maximum daily oxygen production rate of aquatic plants, P max =a×e bT , a is the regression coefficient, b is the temperature coefficient; f m (I) is the light confinement function, I is the solar radiation intensity on the water surface, I opt is the optimal solar radiation intensity for photosynthesis of aquatic plants; B is the biomass of aquatic plants.

[0086] Step 203: Establish the source term of aquatic plant respiration: The established model is:

[0087] R mac =resp×B

[0088] In the formula, resp represents the limiting function of water temperature on respiration rate, resp = m×e nT , m is the exponential regression coefficient, and n is the respiration temperature correction coefficient.

[0089] Step 204: Establish the source term of organic mineralization. The established model is:

[0090]

[0091] Where min0 is the standard mineralization rate at 0°C, k T is the biochemical reaction temperature coefficient, O om is the oxygen demand per mole of organic molecules, C om is the concentration of organic matter in water.

[0092] Specifically, in step 2, the Spearman correlation coefficient method is used to perform correlation coefficient analysis on the pre-processed relevant water quality / environmental data:

[0093]

[0094] Where cov(x,y) represents the covariance of x and y; σ(x) represents the standard deviation of x; σ(y) represents the standard deviation of y; x i Represents the variable x with serial number i; represents the average value of variable x; y i Represents the variable y with serial number i; Represents the mean value of the variable y.

[0095] Specifically, the use of the Informer model in step 3 to predict the dissolved oxygen concentration at the next moment specifically includes:

[0096] Step 301: Use the Informer model to train the processed data to learn the relationship between the dissolved oxygen concentration and the selected input parameters. Specifically, the Informer model is an encoder-decoder structure.

[0097] Step 302: In the encoder, the ProbSparse self-attention mechanism is used, which can achieve O(LlogL) in terms of time complexity and memory usage, where L is the sequence length. The specific formula is

[0098]

[0099] Among them, Q represents query, K represents key, V represents value, and d represents dimension.

[0100] Step 303: Further, in the encoder, a self-attention distillation mechanism is used to apply a Conv1d layer to each attention layer result, and then add a Maxpooling layer to halve the output of each layer, highlight the dominant attention, and effectively process overly long input sequences. The process of the distilling operation from layer j to layer j+1 is as follows:

[0101]

[0102] Among them, Conv1d represents the one-dimensional convolution operation on the time series, and ELU is the activation function.

[0103] Step 304: In the decoder, a parallel generative decoder mechanism is used for prediction, which is specifically a batch generative prediction that directly outputs multi-step prediction results. The formula is as follows:

[0104]

[0105] in, Represents a placeholder (predicted value), Indicates the start character.

[0106] Step 305: After the model training is completed, the cross-test dataset is used to evaluate the performance of the model and output the single-step or multi-step prediction step size.

[0107] Specifically, in step 1, high-precision sensors are used to collect pond water quality / environmental data in real time to obtain relevant water quality and meteorological raw data, including:

[0108] s1: Dynamically adjust the interval sampling time period according to actual monitoring needs, initially set to 120 seconds. This dynamic adjustment mechanism allows automatic adjustment of the sampling frequency according to environmental changes and specific events, ensuring that higher density data is obtained at critical moments.

[0109] s2: In each sampling period, water quality parameters including pH value, turbidity, salinity, ammonia nitrogen concentration, chemical oxygen demand, residual chlorine, total phosphorus, dissolved oxygen concentration, and environmental parameters such as light intensity, wind speed, humidity, solar radiation, and temperature are collected. Intelligent strategies are adopted to adaptively adjust the collection density of each parameter according to the current environmental conditions and historical data analysis results, so as to improve the collection efficiency while ensuring the comprehensiveness of the data.

[0110] s3: For the data collected in each standard period, data smoothing technology is used to convert parameter values ​​with multiple data points into average values ​​within the period, ensuring that each water quality and environmental parameter is presented as a single data point in the final analysis. This step aims to reduce data noise and improve the accuracy of model predictions.

[0111] s4: The processed data is combined into a single water quality / environmental data set for model training and real-time prediction of dissolved oxygen levels. This ensures that the data from each sampling period effectively reflects the current state of the pond and provides accurate input for dissolved oxygen prediction.

Claims

1. A method for predicting dissolved oxygen in a pond based on an Informer model, characterized in that: The dissolved oxygen time series prediction method comprises: Step 1: Arrange sensors to collect water quality / environmental data of the pond and meteorological environmental data of the farm in real time, obtain relevant water quality and meteorological raw data, and perform data preprocessing on the relevant water quality and meteorological raw data; Step 2: construct a vertical distribution model of dissolved oxygen based on diffusion theory to extract the characteristics of dissolved oxygen distribution; use the Spearman correlation coefficient method to perform correlation analysis on the pretreated water quality / environmental data and the dissolved oxygen concentration in the pond, determine the correlation coefficient between the relevant water quality / environmental data and the dissolved oxygen concentration in the pond, and select data parameters with a correlation coefficient greater than 0.4; Step 3: Use the vertical distribution characteristics of dissolved oxygen in step 2 and the water quality / environmental parameters with a correlation coefficient greater than 0.4 as inputs of the Informer model, and predict the dissolved oxygen concentration value of the pond at the next moment through the Informer model.

2. The method for predicting dissolved oxygen in a pond according to claim 1, characterized in that: The water quality environment data in step 1 include pond water temperature, pH value, turbidity, salinity, ammonia nitrogen, residual chlorine, chemical oxygen demand, total phosphorus and dissolved oxygen concentration; the relevant meteorological environment data include solar radiation, atmospheric pressure, temperature, humidity and light intensity.

3. The method for predicting dissolved oxygen in a pond according to claim 1, characterized in that: The data preprocessing described in step 1 includes data cleaning and data normalization. The processed data can be used as input parameters of the Informer model. The normalization formula is as follows: Where X * represents the normalized data, X max is the maximum value of X among all data, X min is the minimum value of C among all data, X n is the new value obtained after normalization.

4. The dissolved oxygen vertical distribution model according to claim 1, characterized in that: The vertical distribution model of dissolved oxygen based on diffusion theory established in step 2 is: Where C is the concentration of the transported substance, z is the depth of a certain point in the water body, P is the photosynthetic rate, and R mac Indicates the respiration of aquatic plants, M in Indicates the mineralization of organic matter in water. Further, after using this model to predict the vertical distribution of dissolved oxygen concentration, this output is used as the input of the subsequent Informer model. Specifically, the specific steps of establishing the vertical distribution model of dissolved oxygen include: Step 201: Establish the diffusion term of the diffusion equation. Usually, the unidirectional transport of substances in water is usually described by the one-dimensional convection-diffusion equation: Where C is the concentration of the transported substance, A is the horizontal surface area at the depth z of the water body, and K z K is the vertical diffusion coefficient of water body, and S is the net production rate of substances transported by each water layer. z It is related to the vertical distribution of dissolved oxygen in the water body and the thermal stratification of the water body. The traditional calculation method is: K z =α(N 2 ) -γ Among them, α and γ are constants, which are determined by the turbulent top of the water body, and N 2 It is determined by the ratio of the density of the water body at the current location to the acceleration of gravity. Assuming that in a rectangular space of the pond water body, the concentration of oxygen molecules changes over time, and the oxygen molecules follow the law of conservation of mass during the transfer process, then within a unit time τ, there will be a flux of J(x) oxygen molecules entering this space, and at the same time, there will be a flux of J(x+Δx) oxygen molecules leaving the space point. When τ→0, Δx→0, the dissolved oxygen concentration at the space point changes to: Combined with Fick's first law: Substituting x in the formula with the pond depth z, the diffusion term of the vertical diffusion equation of dissolved oxygen in the pond is: Step 202: Establish photosynthetic rate source term, the established model is: P=P max ×f m (I)×B Where P max represents the maximum daily oxygen production rate of aquatic plants, P max =a×e bT , a is the regression coefficient, b is the temperature coefficient; f m (I) is the light confinement function, I is the solar radiation intensity on the water surface, I opt is the optimal solar radiation intensity for photosynthesis of aquatic plants; B is the biomass of aquatic plants. Step 203: Establish the source term of aquatic plant respiration: The established model is: R mac =resp×B In the formula, resp represents the limiting function of water temperature on respiration rate, resp = m×e nT , m is the exponential regression coefficient, and n is the respiration temperature correction coefficient. Step 204: Establish the source term of organic mineralization. The established model is: Where min0 is the standard mineralization rate at 0°C, k T is the biochemical reaction temperature coefficient, O om is the oxygen demand per mole of organic molecules, C om is the concentration of organic matter in water.

5. The method for predicting dissolved oxygen in a pond according to claim 1, characterized in that: In step 2, the Spearman correlation coefficient method is used to perform correlation coefficient analysis on the pre-processed relevant water quality / environmental data: Where cov(x,y) represents the covariance of x and y; σ(x) represents the standard deviation of x; σ(y) represents the standard deviation of y; x i Represents the variable x with serial number i; represents the average value of variable x; y i Represents the variable y with serial number i; Represents the mean value of the variable y.

6. The dissolved oxygen time series prediction method according to claim 1, characterized in that: The Informer model is used in step 3 to predict the dissolved oxygen concentration at the next moment, specifically including: Step 301: Use the Informer model to train the processed data to learn the relationship between the dissolved oxygen concentration and the selected input parameters. Specifically, the Informer model is an encoder-decoder structure. Step 302: In the encoder, the ProbSparse self-attention mechanism is used, which can achieve O(LlogL) in terms of time complexity and memory usage, where L is the sequence length. The specific formula is Among them, Q represents query, K represents key, V represents value, and d represents dimension. Step 303: Further, in the encoder, a self-attention distillation mechanism is used to apply a Conv1d to the result of each attention layer, and then add a Maxpooling layer to halve the output of each layer, highlight the dominant attention, and effectively process overly long input sequences. The process of the distilling operation from layer j to layer j+1 is as follows: Among them, Conv1d represents the one-dimensional convolution operation on the time series, and ELU is the activation function. Step 304: In the decoder, a parallel generative decoder mechanism is used for prediction, which is specifically a batch generative prediction that directly outputs multi-step prediction results. The formula is as follows: in, Represents a placeholder (predicted value), Indicates the start character. Step 305: After the model training is completed, the cross-test dataset is used to evaluate the performance of the model and output the single-step or multi-step prediction step size.

7. The dissolved oxygen time series prediction method according to claim 1, characterized in that: The method described in step 1 uses a high-precision sensor to collect pond water quality / environmental data in real time to obtain relevant water quality and meteorological raw data, including the following steps: s1: Dynamic sampling time period setting: Dynamically adjust the interval sampling time period according to actual monitoring needs, and the initial setting is 120 seconds. This dynamic adjustment mechanism allows automatic adjustment of the sampling frequency according to environmental changes and specific events, ensuring that higher density data is obtained at critical moments. s2: In each sampling period, water quality parameters including pH value, turbidity, salinity, ammonia nitrogen concentration, chemical oxygen demand, residual chlorine, total phosphorus, dissolved oxygen concentration, and environmental parameters such as light intensity, wind speed, humidity, solar radiation, and temperature are collected. Intelligent strategies are adopted to adaptively adjust the collection density of each parameter according to the current environmental conditions and historical data analysis results, so as to improve the collection efficiency while ensuring the comprehensiveness of the data. s3: For the data collected in each standard period, data smoothing technology is used to convert parameter values ​​with multiple data points into average values ​​within the period, ensuring that each water quality and environmental parameter is presented as a single data point in the final analysis. This step aims to reduce data noise and improve the accuracy of model predictions. s4: The processed data is combined into a single water quality / environmental data set for model training and real-time prediction of dissolved oxygen levels. This ensures that the data from each sampling period effectively reflects the current state of the pond and provides accurate input for dissolved oxygen prediction.

Images