Water pollutant concentration intelligent detection system and detection method thereof

Abstract

The invention relates to the technical field of intelligent detection, and discloses a water pollutant concentration intelligent detection system and a detection method thereof, and the water pollutant concentration intelligent detection method comprises the following steps: obtaining optical data, chemical data, electrical data, environmental data and biological data; performing normalization and standardization processing on the obtained data through edge calculation; performing feature extraction on the acquired data; splicing the features to obtain a fused feature; the fusion features are input into a multi-layer perceptron for water pollutant concentration prediction; the diffusion conditions of pollutants at different time points in the future are predicted through a fluid mechanics formula; inputting the fusion features into a decision tree model, and judging the instant influence of the water pollutants on the cultured organisms; the breeding environment parameters are adjusted through instant influence. According to the method, through edge calculation, behaviors of aquatic organisms are combined, and the diffusion trend of water pollutants is predicted, so that more accurate water quality monitoring is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection technology, and more specifically, to an intelligent detection system for water pollutant concentration and a detection method thereof. Background Art

[0002] In the water quality monitoring of aquaculture ponds, farmed organisms need a better breeding environment to speed up the breeding process, so intelligent detection of water pollutant concentrations is crucial. However, the spread of water pollutants is relatively fast and difficult to control. Therefore, there is an urgent need for a method that can predict the diffusion trend of water pollutants and accurately detect the concentration of water pollutants in real time and combine aquatic organism behavior data to determine the immediate impact on farmed organisms and optimize the breeding environment.

[0003] Traditional water pollutant detection technology mainly relies on laboratory chemical analysis or single sensor on-site detection, and has defects such as slow response speed, weak multi-parameter collaborative analysis capability, and inability to perform real-time dynamic prediction. Summary of the invention

[0004] The present invention provides an intelligent detection system for water pollutant concentration and a detection method thereof, which solve the technical problems in the related art of slow data processing response speed, rapid diffusion of pollutants to water bodies during aquaculture, and low aquaculture efficiency caused by abnormal behavior of aquatic organisms.

[0005] The present invention provides an intelligent detection method for water pollutant concentration, comprising: Step 100, acquiring optical data, chemical data, electrical data, environmental data and biological data; Step 200, normalizing and standardizing the acquired data; Step 300, extracting corresponding optical features, chemical features, electrical features and biological features from the optical data, chemical data, electrical data and biological data respectively; Step 400, combining the optical features, chemical features, electrical features, and biological features to obtain a fusion feature; Step 500, inputting the fused features into a multi-layer perceptron to predict the concentration of water pollutants; Step 600, using the fusion features and the acquired environmental data in combination with fluid mechanics formulas to predict the diffusion of pollutants at different time points in the future; Step 700, inputting the fused features into a decision tree model, the decision tree divides the features, constructs decision rules, and determines the immediate impact of water pollutants on aquaculture organisms; Step 800, adjusting the aquaculture environment parameters based on the determined immediate impact of water pollutants on the aquaculture organisms, and optimizing the aquaculture environment parameters.

[0006] In a preferred embodiment, edge nodes are used to complete data preprocessing of optical data, chemical data, electrical data, environmental data, and biological data.

[0007] In a preferred embodiment, optical data of light transmittance and color of the water body are obtained from optical sensors, chemical data of dissolved oxygen content, ammonia nitrogen concentration, and total phosphorus content of the water body are obtained from chemical sensors, electrical data of electrical conductivity and redox potential of the water body are obtained from electrical sensors, environmental data of water flow velocity and direction of the water body are obtained from environmental sensors, and biological data of the swimming speed of fish are obtained from biological monitors.

[0008] In a preferred embodiment, the formula for extracting optical features from optical data using a convolutional neural network is as follows: ; in, represents the first convolution function, represents the normalized transmittance measurement value, is the normalized RGB color channel value, Represents optical characteristics.

[0009] The formula for chemical feature extraction of chemical data through recurrent neural network is as follows: ; ; ; ; ; Among them, LSTM represents long short-term memory network, represents the hidden state at time t, , , They represent input gate, forget gate, and output gate respectively. , , They represent the standardized dissolved oxygen content, ammonia nitrogen concentration, and total phosphorus content, respectively. represents the sigmoid activation function, , , Respectively represent the chemical data weight matrix of the input gate, forget gate, and output gate, , , Respectively represent the hidden state weight matrices of the input gate, forget gate, and output gate, , , Respectively represent the input gate, forget gate, and output gate bias terms, is the attention weight coefficient at the tth moment, represents the total step length of the time series, Indicates chemical characteristics.

[0010] The formula for extracting electrical features from electrical data through an autoencoder is as follows: ; ; ; in, represents the hyperbolic tangent activation function, represents the rectified linear unit activation function, , Represent the standardized conductivity and redox potential, respectively. represents the reconstructed value of electrical data, , are the first and second layer weight matrices of the encoder respectively, , are the first and second layer weight matrices of the decoder, respectively. and are the first and second bias items of the encoder, , are the first and second bias items of the decoder respectively, represents the mean squared error reconstruction loss of the autoencoder.

[0011] The formula for extracting biometric features from biological data through a convolutional neural network is as follows: ; in, represents the second convolution function, represents the normalized swimming speed of fish, Represents biological characteristics.

[0012] The calculation formula of feature fusion is as follows: ; in represents the fusion feature, , , , Respectively represent optical characteristics, chemical characteristics, electrical characteristics, and biological characteristics. Represents the concatenation function.

[0013] The calculation formula for predicting water pollutant concentration is as follows: ; in, is the output first weight matrix, To output the second weight matrix, and To output the first and second bias terms, represents the S-type function, is the predicted pollutant concentration value.

[0014] In a preferred embodiment, the calculation formula for diffusion trend prediction is as follows: ; in, represents the pollutant concentration field, represents the velocity vector field, is the diffusion coefficient tensor, is the pollution source term; Learning parameters through RNN: ; ; in, The diffusion coefficient matrix representing the spatiotemporal variation, represents the intensity distribution of pollution sources, is the grid resolution of the prediction area, , Represent the first and second recurrent neural network functions respectively.

[0015] In a preferred embodiment, the predicted water pollutant concentration value is the ammonia nitrogen concentration value, with the unit of mg / L.

[0016] In a preferred embodiment, the decision rule is as follows: ; in, is the safe concentration threshold of pollutants, is the normalized swimming speed of fish, ImpactLevel=0 indicates a safe state, 1 indicates a warning state, and 2 indicates a dangerous state.

[0017] In a preferred embodiment, the immediate impact of the water pollutants on the farmed organisms specifically includes three levels: no impact, slight impact, and severe impact, which correspond to the safe state, warning state, and dangerous state in the decision-making rules respectively.

[0018] In a preferred embodiment, the optimization of the aquaculture environment parameters includes: adjusting the water change frequency, the power of the aeration equipment, and the feed input amount.

[0019] A water body pollutant concentration intelligent detection system is used to execute any of the above-mentioned water body pollutant concentration intelligent detection methods.

[0020] The beneficial effects of the present invention are as follows: the present invention uses edge computing to quickly process data and predict the diffusion trend of water pollutants. It also combines the behavior of aquatic organisms to judge the impact of the diffusion of water pollutants on aquatic organisms, thereby adjusting the breeding environment parameters, achieving accurate water quality monitoring during aquaculture, and optimizing the breeding environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a method for intelligent detection of water pollutant concentration of the present invention; Figure 2 It is an example of data collection in a method for intelligent detection of water pollutant concentration of the present invention; Figure 3 It is an example of aquaculture environment parameters in the intelligent detection method of water pollutant concentration of the present invention. DETAILED DESCRIPTION

[0022] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0023] At least one embodiment of the present invention discloses a method for intelligently detecting the concentration of pollutants in water bodies, such as Figure 1 As shown, including: Step 100, acquiring optical data, chemical data, electrical data, environmental data and biological data; Obtain optical data of light transmittance and color of water from optical sensors, chemical data of dissolved oxygen content, ammonia nitrogen concentration, and total phosphorus content of water from chemical sensors, electrical data of conductivity and redox potential of water from electrical sensors, environmental data of water flow velocity and direction from environmental sensors, and biological data of fish swimming speed from biological monitors; In one embodiment of the present invention, the color is specifically represented by the color channel values ​​of red, green, and blue.

[0024] In one embodiment of the present invention, the swimming speed of the fish is obtained by capturing video using an underwater high-definition camera in a biological monitor, and the swimming speed of the fish is calculated using target detection and motion tracking.

[0025] Step 200, normalizing and standardizing the acquired data; In one embodiment of the present invention, edge nodes are used to complete data preprocessing of optical data, chemical data, electrical data, environmental data, and biological data.

[0026] Step 300, extracting corresponding optical features, chemical features, electrical features and biological features from the optical data, chemical data, electrical data and biological data respectively; In one embodiment of the present invention, the formula for extracting optical features from optical data using a convolutional neural network is as follows: ; in, represents the first convolution function, represents the normalized transmittance measurement value, is the normalized RGB color channel value, Represents optical characteristics.

[0027] In one embodiment of the present invention, the formula for extracting chemical features from chemical data using a recurrent neural network is as follows: ; ; ; ; ; Among them, LSTM represents long short-term memory network, represents the hidden state at time t, , , They represent input gate, forget gate, and output gate respectively. , , They represent the standardized dissolved oxygen content, ammonia nitrogen concentration, and total phosphorus content, respectively. represents the sigmoid activation function, , , Respectively represent the chemical data weight matrix of the input gate, forget gate, and output gate, , , Respectively represent the hidden state weight matrices of the input gate, forget gate, and output gate, , , Respectively represent the input gate, forget gate, and output gate bias terms, is the attention weight coefficient at the tth moment, represents the total step length of the time series, Indicates chemical characteristics.

[0028] In one embodiment of the present invention, the formula for extracting electrical features from electrical data using an autoencoder is as follows: ; ; ; in, represents the hyperbolic tangent activation function, represents the rectified linear unit activation function, , Represent the standardized conductivity and redox potential, respectively. represents the reconstructed value of electrical data, , are the first and second layer weight matrices of the encoder respectively, , are the first and second layer weight matrices of the decoder, respectively. and are the first and second bias items of the encoder, , are the first and second bias items of the decoder respectively, represents the mean squared error reconstruction loss of the autoencoder.

[0029] In one embodiment of the present invention, the formula for extracting biometric features from biological data using a convolutional neural network is as follows: ; in, represents the second convolution function, represents the normalized swimming speed of fish, Represents biological characteristics.

[0030] Step 400, combining the optical features, chemical features, electrical features, and biological features to obtain a fusion feature; In one embodiment of the present invention, the calculation formula of feature fusion is as follows: ; in represents the fusion feature, , , , Respectively represent optical characteristics, chemical characteristics, electrical characteristics, and biological characteristics. Represents the concatenation function.

[0031] Step 500, inputting the fused features into a multi-layer perceptron to predict the concentration of water pollutants; In one embodiment of the present invention, the calculation formula for predicting the concentration of water pollutants is as follows: ; in, is the output first weight matrix, To output the second weight matrix, and To output the first and second bias terms, represents the S-type function, is the predicted pollutant concentration value.

[0032] In one embodiment of the present invention, the predicted water pollutant concentration value is the ammonia nitrogen concentration value, with the unit of mg / L.

[0033] Step 600, using the fusion features and the acquired environmental data in combination with fluid mechanics formulas to predict the diffusion of pollutants at different time points in the future; In one embodiment of the present invention, the calculation formula for diffusion trend prediction is as follows: ; in, represents the pollutant concentration field, represents the velocity vector field, is the diffusion coefficient tensor, is the pollution source term.

[0034] Learning parameters through RNN: ; ; in, The diffusion coefficient matrix representing the spatiotemporal variation, represents the intensity distribution of pollution sources, is the grid resolution of the prediction area, , Represent the first and second recurrent neural network functions respectively.

[0035] Step 700, inputting the fused features into a decision tree model, the decision tree divides the features, constructs decision rules, and determines the immediate impact of water pollutants on aquaculture organisms; In one embodiment of the present invention, the decision rule is as follows: ; in, is the safe concentration threshold of pollutants, is the normalized swimming speed of fish, ImpactLevel=0 indicates a safe state, 1 indicates a warning state, and 2 indicates a dangerous state.

[0036] In one embodiment of the present invention, the immediate impact of water pollutants on farmed organisms specifically includes three levels: no impact, slight impact, and severe impact, which correspond to the safe state, warning state, and dangerous state in the decision-making rules respectively.

[0037] Step 800, adjusting the aquaculture environment parameters based on the determined immediate impact of water pollutants on the aquaculture organisms, and optimizing the aquaculture environment parameters.

[0038] In one embodiment of the present invention, the optimization of the aquaculture environment parameters includes: adjusting the water change frequency, the power of the aeration equipment, and the feed input amount.

[0039] In one embodiment of the present invention, a water pollutant concentration intelligent detection system, which is used to execute any of the above-mentioned water pollutant concentration intelligent detection methods, includes the following modules: The data acquisition module is used to obtain the optical data of the light transmittance and color of the water body, the chemical data of the dissolved oxygen content, ammonia nitrogen concentration and total phosphorus content of the water body, the electrical data of the conductivity and redox potential of the water body, the environmental data of the water flow speed and direction of the water body, and the biological data of the swimming speed of fish; A data preprocessing module is used to normalize and standardize the acquired data; A feature extraction module, for extracting optical features from optical data through a convolutional neural network; extracting chemical features from chemical data through a recurrent neural network, extracting electrical features from electrical data through an autoencoder, and extracting biological features from biological data through a convolutional neural network; Feature fusion module, used to combine optical features, chemical features, electrical features, and biological features to obtain fused features; The concentration prediction module is used to input the fusion features into the multi-layer perceptron to predict the concentration of water pollutants; The diffusion assessment module is used to use the fusion features and the acquired environmental data combined with fluid mechanics formulas to predict the diffusion of pollutants at different time points in the future; The decision tree judgment module is used to input the fusion features into the decision tree model. The decision tree divides the features, constructs decision rules, and judges the immediate impact of water pollutants on aquaculture organisms. The decision-making module is used to adjust and optimize the aquaculture environment parameters by judging the immediate impact of water pollutants on aquaculture organisms.

[0040] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A method for intelligent detection of water pollutant concentration, characterized in that: The following steps are involved: Acquire optical data, chemical data, electrical data, environmental data, and biological data; Normalize and standardize the acquired data; Extracting corresponding optical features, chemical features, electrical features and biological features from the optical data, chemical data, electrical data and biological data respectively; The optical features, chemical features, electrical features, and biological features are spliced ​​together to obtain fusion features; The fused features are input into the multi-layer perceptron to predict the concentration of water pollutants; Using fusion features and acquired environmental data combined with fluid mechanics formulas, we can predict the diffusion of pollutants at different time points in the future. The fused features are input into the decision tree model, which divides the features and constructs decision rules to determine the immediate impact of water pollutants on aquaculture organisms; The aquaculture environment parameters are adjusted and optimized by determining the immediate impact of water pollutants on the aquaculture organisms.

2. The method for intelligent detection of water pollutant concentration according to claim 1, characterized in that: Use edge nodes to complete data preprocessing of optical data, chemical data, electrical data, environmental data, and biological data.

3. The method for intelligent detection of water pollutant concentration according to claim 2, characterized in that: The optical data of the water's transmittance and color are obtained from optical sensors. The chemical data of the water's dissolved oxygen content, ammonia nitrogen concentration, and total phosphorus content are obtained from chemical sensors. The electrical data of the water's conductivity and redox potential are obtained from electrical sensors. The environmental data of the water's flow velocity and direction are obtained from environmental sensors. The biological data of the fish's swimming speed is obtained from biological monitors.

4. The method for intelligent detection of water pollutant concentration according to claim 3, characterized in that: The formula for extracting optical features from optical data using a convolutional neural network is as follows: ； in, represents the first convolution function, represents the normalized transmittance measurement value, is the normalized RGB color channel value, Represents optical characteristics; The formula for chemical feature extraction of chemical data through recurrent neural network is as follows: ； ； ； ； ； Among them, LSTM represents long short-term memory network, represents the hidden state at time t, , , They represent input gate, forget gate, and output gate respectively. , , They represent the standardized dissolved oxygen content, ammonia nitrogen concentration, and total phosphorus content, respectively. represents the sigmoid activation function, , , Respectively represent the chemical data weight matrix of the input gate, forget gate, and output gate, , , Respectively represent the hidden state weight matrices of the input gate, forget gate, and output gate, , , Respectively represent the input gate, forget gate, and output gate bias terms, is the attention weight coefficient at the tth moment, represents the total step length of the time series, Indicates chemical characteristics; The formula for extracting electrical features from electrical data through an autoencoder is as follows: ； ； ； in, represents the hyperbolic tangent activation function, represents the rectified linear unit activation function, , Represent the standardized conductivity and redox potential, respectively. represents the electrical data reconstruction value, , are the first and second layer weight matrices of the encoder respectively, , are the first and second layer weight matrices of the decoder, respectively. and are the first and second bias items of the encoder, , are the first and second bias items of the decoder respectively, represents the mean square error reconstruction loss of the autoencoder; The formula for extracting biometric features from biological data through a convolutional neural network is as follows: ； in, represents the second convolution function, represents the normalized swimming speed of fish, Indicates biological characteristics; The calculation formula of feature fusion is as follows: ； in represents the fusion feature, , , , Respectively represent optical characteristics, chemical characteristics, electrical characteristics, and biological characteristics. represents the concatenation function; The calculation formula for predicting water pollutant concentration is as follows: ； in, is the output first weight matrix, To output the second weight matrix, and To output the first and second bias terms, represents the S-type function, is the predicted pollutant concentration value.

5. The method for intelligent detection of water pollutant concentration according to claim 4, characterized in that: The calculation formula for diffusion trend prediction is as follows: ； in, represents the pollutant concentration field, represents the velocity vector field, is the diffusion coefficient tensor, is the pollution source term; Learning parameters through RNN: ； ； in, The diffusion coefficient matrix representing the spatiotemporal variation, represents the intensity distribution of pollution sources, is the grid resolution of the prediction area, , Represent the first and second recurrent neural network functions respectively.

6. The method for intelligent detection of water pollutant concentration according to claim 5, characterized in that: The predicted water pollutant concentration value is the ammonia nitrogen concentration value, in mg / L.

7. The method for intelligent detection of water pollutant concentration according to claim 6, characterized in that: The decision rules are as follows: ； in, is the safe concentration threshold of pollutants, is the normalized swimming speed of fish, ImpactLevel=0 indicates a safe state, 1 indicates a warning state, and 2 indicates a dangerous state.

8. The method for intelligent detection of water pollutant concentration according to claim 7, characterized in that: The immediate impact of water pollutants on farmed organisms specifically includes three levels: no impact, slight impact, and severe impact, which correspond to the safe state, warning state, and dangerous state in the decision-making rules respectively.

9. The method for intelligent detection of water pollutant concentration according to claim 8, characterized in that: Optimization of breeding environment parameters includes: adjustment of water change frequency, oxygenation equipment power, and feed input amount.

10. An intelligent detection system for water pollutant concentration, characterized in that: It is used to execute the intelligent detection method of water pollutant concentration described in any one of claims 1-9.

Images