Shellfish culture pollution reduction and carbon reduction effect prediction method based on gated neural network

Abstract

The invention provides a method for predicting pollution and carbon reduction effects of shellfish culture based on a gated neural network, and relates to the technical field of fishery ecological environment monitoring, and the method comprises the steps: obtaining a culture feature vector through normalization processing based on shellfish culture data; constructing a shellfish culture pollution prediction model based on the gated neural network; the shellfish culture consumption prediction model comprises a consumption prediction layer and a pollution prediction layer; a consumption prediction layer is used for processing the breeding feature vectors, key features in the breeding feature vectors are screened out, and a consumption prediction result is obtained through weighted calculation; and processing the consumption prediction result by using a pollution prediction layer, calculating the discharge amount of each breeding stage, obtaining a pollution prediction result, and completing pollution reduction and carbon reduction effect prediction of shellfish breeding. The method solves the problem that the pollution-reducing and carbon-reducing effect of shellfish culture is difficult to dynamically predict.

Description

Technical Field

[0001] The present invention relates to the technical field of fishery ecological environment monitoring, and in particular to a method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on a gated neural network. Background Art

[0002] The ocean is the largest carbon pool in the earth system, with a carbon pool that is 50 times that of the atmosphere and 20 times that of terrestrial ecosystems. The ocean carbon sink is stored in the deep ocean and occupies an important position in the earth's carbon cycle. Therefore, the ocean has the characteristics of large carbon fixation, high efficiency and long storage time. Exploring its huge carbon sink potential is an important way to implement the "dual carbon" commitment. The marine aquaculture industry is now huge in scale, but its extensive development has also brought many ecological and environmental problems. Shellfish are high-quality carbon sink organisms. On the one hand, during the growth and development of shellfish, they can increase the transparency of seawater and alleviate eutrophication of water bodies by filtering particulate matter and phytoplankton in seawater, thereby increasing primary productivity and carbon sinks. On the other hand, organic carbon not absorbed by shellfish can be deposited on the seabed in the form of feces and pseudo-feces, which can effectively accelerate the deposition and storage of organic carbon.

[0003] At present, there is a lack of a complete technical system for calculating carbon sinks in marine fisheries. Although some scholars have applied the life cycle assessment (LCA) method to the carbon sink accounting of marine shellfish farming throughout its life cycle in recent years, the calculation scope still does not cover the entire process and all elements of farming. For example, in the farming process, due to the difficulty in obtaining data on shellfish seedlings, most scholars only calculate the carbon storage of the harvested adult part, ignoring the carbon storage of seedlings. In addition, LCA is a typical static method and has a certain degree of subjectivity, ignoring the dynamic changes in time and space, which will affect the evaluation results. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on a gated neural network, which solves the problem that the pollution reduction and carbon reduction effectiveness of shellfish aquaculture is difficult to dynamically predict.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a method for predicting the pollution reduction and carbon reduction effect of shellfish aquaculture based on a gated neural network, comprising:

[0006] S1: Based on shellfish aquaculture data, the aquaculture feature vector is obtained through normalization processing;

[0007] S2: Based on the gated neural network, a shellfish aquaculture pollution prediction model is constructed; the shellfish aquaculture consumption prediction model includes a consumption prediction layer and a pollution prediction layer;

[0008] S3: Processing the breeding feature vector using the consumption prediction layer, screening out key features therein, and obtaining a consumption prediction result through weighted calculation;

[0009] S4: Use the pollution prediction layer to process the consumption prediction results, calculate the emissions in each breeding stage, obtain pollution prediction results, and complete the prediction of pollution reduction and carbon reduction effectiveness of shellfish breeding.

[0010] Furthermore, the S1 includes:

[0011] Based on the aquaculture stage, basic aquaculture data, input application data and tailwater quality data are collected as shellfish aquaculture data;

[0012] The shellfish farming data is normalized to obtain a farming feature vector.

[0013] Further, the S3 includes:

[0014] The aquaculture feature vector is screened by using a reset gate to obtain the feature that has the greatest impact on tailwater quality and carbon emissions; wherein the feature that has the greatest impact on tailwater quality and carbon emissions is a key feature;

[0015] The update gate is used to perform weighted calculation on the features that have the greatest impact on tailwater quality and carbon emissions to obtain a consumption prediction result; wherein the reset gate and the update gate belong to the consumption prediction layer.

[0016] Furthermore, the expression of the key feature is:

[0017]

[0018] r t =σ(ω r ·[h t-1 , x t ]);

[0019] in, represents the key feature, tanh represents the tanh activation function, ω represents the weight of each breeding feature vector in the reset gate, r t represents the reset gate, h t-1 represents the breeding feature vector at time t-1, x t represents the breeding feature vector at time t, σ represents the sigmoid activation function, ω r Represents the weight of each breeding feature vector.

[0020] Furthermore, the expression of the consumption prediction result is:

[0021]

[0022] z t =σ(ω z ·[h t-1 , xt ]);

[0023] Among them, h t represents the consumption forecast result at time t, z t represents the update gate, h t-1 represents the breeding feature vector at time t-1, represents the key feature, σ represents the sigmoid activation function, ω z Represents the weight of each breeding feature vector in the update gate.

[0024] Further, the S4 includes:

[0025] The consumption prediction results are processed using the pollution prediction layer to calculate the water quality index emissions and carbon emissions at each breeding stage;

[0026] The water quality index emissions and carbon emissions in each aquaculture stage are summed up according to the aquaculture stage to obtain the pollution prediction results, thus completing the prediction of the pollution reduction and carbon reduction effects of shellfish aquaculture.

[0027] Furthermore, the expressions of water quality index emission and carbon emission in each breeding stage are respectively:

[0028]

[0029] G i =E i ×Z i ;

[0030] in, represents the discharge of water quality index in the i-th breeding stage and the j-th breeding stage, represents the concentration of water quality index at the i-th breeding stage and the j-th breeding stage, Q i represents the amount of aquaculture tailwater discharged into the environment in the i-th aquaculture stage, G i represents the carbon emissions in the ith breeding stage, E i represents the carbon emission coefficient of the i-th breeding stage, Z i Represents the increase in dry weight of scallops in the i-th culture stage.

[0031] Furthermore, the expression of the pollution prediction result is:

[0032]

[0033] G = G1 + G2 + G3;

[0034] Among them, W j represents the total discharge of the jth water quality index in the whole cycle, represents the j-th water quality index discharge in the first breeding stage, represents the discharge of the jth water quality index in the second breeding stage, represents the emission of the jth water quality index in the third breeding stage, G represents the total carbon emission in the whole cycle, G1 represents the carbon emission in the first breeding stage, G2 represents the carbon emission in the second breeding stage, and G3 represents the carbon emission in the third breeding stage.

[0035] The beneficial effect of the present invention is that the processor processes the aquaculture feature vector using the shellfish aquaculture consumption prediction model to obtain a pollution prediction result. In this way, the impact of various factors on pollutant emissions and carbon sinks can be more comprehensively considered, and dynamic prediction of shellfish aquaculture pollutant emissions and carbon sink effects can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0037] Figure 1 It is an exemplary flow chart of a method for predicting the effectiveness of pollution reduction and carbon reduction in shellfish aquaculture based on a gated neural network according to some embodiments of this specification. DETAILED DESCRIPTION

[0038] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0039] Example

[0040] Figure 1 This is an exemplary flow chart of a method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on a gated neural network according to some embodiments of this specification. Figure 1 As shown, the process includes the following steps. In some embodiments, the process can be executed by a processor.

[0041] S1: Based on the shellfish farming data, the farming feature vector is obtained through normalization processing.

[0042] Shellfish farming data refers to data related to pollution emissions and carbon emissions in the shellfish farming process. For example, shellfish farming data may include basic data, energy consumption data, input data, and tailwater quality data of shellfish farming.

[0043] The aquaculture feature vector is the normalized features of the shellfish aquaculture data.

[0044] In some embodiments, the processor may collect basic aquaculture data, input application data and tailwater quality data based on the aquaculture stage as shellfish aquaculture data; and normalize the shellfish aquaculture data to obtain an aquaculture feature vector.

[0045] The culture phases are the different stages of shellfish farming. For example, the culture phases may include the nursery phase, the at-sea culture phase and the harvest phase.

[0046] S2: Based on the gated neural network, a shellfish aquaculture pollution prediction model is constructed; the shellfish aquaculture consumption prediction model includes a consumption prediction layer and a pollution prediction layer.

[0047] A gated neural network is a neural network that uses reset gates and update gates to make predictions on data.

[0048] The shellfish aquaculture pollution prediction model is a neural network model for predicting pollution results. For example, the shellfish aquaculture pollution prediction model can be a gated neural network model.

[0049] In some embodiments, the input of the shellfish aquaculture pollution prediction model may be an aquaculture feature vector, and the output of the shellfish aquaculture pollution prediction model may be a pollution prediction result.

[0050] In some embodiments, the shellfish aquaculture pollution prediction model has the following structure:

[0051] The shellfish aquaculture pollution prediction model includes a consumption prediction layer and a pollution prediction layer. The output of the consumption prediction layer is used as the input of the pollution prediction layer, and the output of the pollution prediction layer is used as the final output of the shellfish aquaculture pollution prediction model.

[0052] The consumption prediction layer is used to predict the consumption of shellfish farming at each stage, including reset gate and update gate. The input of the consumption prediction layer can include farming feature vectors, and the output can include consumption prediction results.

[0053] The pollution prediction layer is used to predict the discharge of shellfish farming at each stage. The input of the pollution prediction layer can include consumption prediction results, and the output can include pollution prediction results.

[0054] In some embodiments, the shellfish aquaculture pollution prediction model can be obtained by training multiple labeled training samples. For example, multiple labeled training samples can be input into the initial shellfish aquaculture pollution prediction model, and a loss function can be constructed by using the labels and the results of the initial shellfish aquaculture pollution prediction model. The parameters of the initial shellfish aquaculture pollution prediction model are iteratively updated based on the loss function by gradient descent or other methods. When the preset conditions are met, the model training is completed and a trained shellfish aquaculture pollution prediction model is obtained. The preset conditions can be that the loss function converges, the number of iterations reaches a threshold, etc.

[0055] In some embodiments, the training samples may include historical aquaculture feature vectors. The labels may be corresponding water quality indicator emissions and carbon emissions. The labels may be manually annotated.

[0056] S3: Using the consumption prediction layer to process the breeding feature vector, screen out key features therein, and obtain the consumption prediction result through weighted calculation.

[0057] Key characteristics are the aquaculture characteristics that have the greatest impact on pollution prediction. For example, key characteristics may include those that have the greatest impact on tailwater quality and carbon emissions.

[0058] The consumption forecast results reflect the consumption situation in each breeding stage. For example, the consumption forecast results may include the concentration of water quality indicators, the amount of breeding tail water, the carbon emission coefficient and the increase in dry weight of scallops in each breeding stage.

[0059] In some embodiments, the processor can use a reset gate to filter the aquaculture feature vector to obtain the features that have the greatest impact on tailwater quality and carbon emissions; use an update gate to perform weighted calculations on the features that have the greatest impact on tailwater quality and carbon emissions to obtain consumption prediction results.

[0060] In some embodiments, the expression of the key feature may be:

[0061]

[0062] r t =σ(ω r ·[h t-1 , x t ]);

[0063] in, represents the key feature, tanh represents the tanh activation function, ω represents the weight of each breeding feature vector in the reset gate, r t represents the reset gate, h t-1 represents the breeding feature vector at time t-1, x t represents the breeding feature vector at time t, σ represents the sigmoid activation function, ω r Represents the weight of each breeding feature vector.

[0064] In some embodiments, the expression of the consumption prediction result may be:

[0065]

[0066] z t =σ(ω z ·[h t-1 , x t ]);

[0067] Among them, h t represents the consumption forecast result at time t, z t represents the update gate, h t-1 represents the breeding feature vector at time t-1, represents the key feature, σ represents the sigmoid activation function, ω z Represents the weight of each breeding feature vector in the update gate.

[0068] S4: Use the pollution prediction layer to process the consumption prediction results, calculate the emissions in each breeding stage, obtain pollution prediction results, and complete the prediction of pollution reduction and carbon reduction effectiveness of shellfish breeding.

[0069] Emissions at each farming stage. For example, emissions at each farming stage may include water quality index emissions and carbon emissions at each farming stage.

[0070] The water quality index discharge is the discharge of each evaluated water quality parameter.

[0071] The pollution prediction results are the predicted data reflecting the emission of shellfish farming in the whole cycle. For example, they may include the total emission of water quality indicators and the total carbon emission in the whole cycle.

[0072] In some embodiments, the processor can use the pollution prediction layer to process the consumption prediction results, calculate the water quality index emissions and carbon emissions in each breeding stage; sum the water quality index emissions and carbon emissions in each breeding stage according to the breeding stage, obtain the pollution prediction results, and complete the prediction of the pollution reduction and carbon reduction effectiveness of shellfish breeding.

[0073] In some embodiments, the processor can analyze the changes in water quality indicator emissions and carbon emissions over a period of time based on the pollution prediction results, obtain the pollution reduction and carbon reduction effectiveness of shellfish aquaculture, and complete the prediction of the pollution reduction and carbon reduction effectiveness of shellfish aquaculture.

[0074] In some embodiments, the expressions for water quality index emission and carbon emission at each breeding stage can be:

[0075]

[0076] G i =E i ×Z i ;

[0077] in, represents the discharge of water quality index in the i-th breeding stage and the j-th breeding stage, represents the concentration of water quality index at the i-th breeding stage and the j-th breeding stage, Q irepresents the amount of aquaculture tailwater discharged into the environment in the i-th aquaculture stage, G i represents the carbon emissions in the ith breeding stage, E i represents the carbon emission coefficient of the i-th breeding stage, Z i Represents the increase in dry weight of scallops in the i-th culture stage.

[0078] In some embodiments, the expression of the pollution prediction result may be:

[0079]

[0080] G = G1 + G2 + G3;

[0081] Among them, W j represents the total discharge of the jth water quality index in the whole cycle, represents the j-th water quality index discharge in the first breeding stage, represents the discharge of the jth water quality index in the second breeding stage, represents the emission of the jth water quality index in the third breeding stage, G represents the total carbon emission in the whole cycle, G1 represents the carbon emission in the first breeding stage, G2 represents the carbon emission in the second breeding stage, and G3 represents the carbon emission in the third breeding stage.

[0082] In some embodiments, the first culture stage may be a nursery stage, the second culture stage may be an offshore culture stage, and the third culture stage may be a harvesting stage.

[0083] In some embodiments of this specification, the processor processes the aquaculture feature vector using the shellfish aquaculture consumption prediction model to obtain a pollution prediction result. In this way, the impact of various factors on pollutant emissions and carbon sinks can be more comprehensively considered, and dynamic prediction of shellfish aquaculture pollutant emissions and carbon sinks can be achieved.

Claims

1. A method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on a gated neural network, characterized in that: include: S1: Based on shellfish aquaculture data, the aquaculture feature vector is obtained through normalization processing; S2: Based on the gated neural network, a shellfish aquaculture pollution prediction model is constructed; the shellfish aquaculture consumption prediction model includes a consumption prediction layer and a pollution prediction layer; S3: Processing the breeding feature vector using the consumption prediction layer, screening out key features therein, and obtaining a consumption prediction result through weighted calculation; S4: Use the pollution prediction layer to process the consumption prediction results, calculate the emissions in each breeding stage, obtain pollution prediction results, and complete the prediction of pollution reduction and carbon reduction effectiveness of shellfish breeding.

2. The method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on gated neural network according to claim 1 is characterized in that: The S1 includes: Based on the aquaculture stage, basic aquaculture data, input application data and tailwater quality data are collected as shellfish aquaculture data; The shellfish farming data is normalized to obtain a farming feature vector.

3. The method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on gated neural network according to claim 1 is characterized in that: The S3 includes: The aquaculture feature vector is screened by using a reset gate to obtain the feature that has the greatest impact on tailwater quality and carbon emissions; wherein the feature that has the greatest impact on tailwater quality and carbon emissions is a key feature; The update gate is used to perform weighted calculation on the features that have the greatest impact on tailwater quality and carbon emissions to obtain a consumption prediction result; wherein the reset gate and the update gate belong to the consumption prediction layer.

4. The method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on gated neural network according to claim 3 is characterized in that: The expression of the key feature is: r t =σ(ω r ·[h t-1 ,x t ]); in, represents the key feature, tanh represents the tanh activation function, ω represents the weight of each breeding feature vector in the reset gate, r t represents the reset gate, h t-1 represents the breeding feature vector at time t-1, x t represents the breeding feature vector at time t, σ represents the sigmoid activation function, ω r Represents the weight of each breeding feature vector.

5. The method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on gated neural network according to claim 3 is characterized in that: The expression of the consumption prediction result is: z t =σ(ω z ·[h t-1 ,x t ]); Among them, h t represents the consumption forecast result at time t, z t represents the update gate, h t-1 represents the breeding feature vector at time t-1, represents the key feature, σ represents the sigmoid activation function, ω z Represents the weight of each breeding feature vector in the update gate.

6. The method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on gated neural network according to claim 1 is characterized in that: The S4 includes: The consumption prediction results are processed using the pollution prediction layer to calculate the water quality index emissions and carbon emissions at each breeding stage; The water quality index emissions and carbon emissions in each aquaculture stage are summed up according to the aquaculture stage to obtain the pollution prediction results, thus completing the prediction of the pollution reduction and carbon reduction effects of shellfish aquaculture.

7. The method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on gated neural network according to claim 6 is characterized in that: The expressions of water quality index emission and carbon emission in each breeding stage are: G i =E i ×Z i ; Among them, W i j represents the discharge of water quality index in the i-th breeding stage and the j-th breeding stage, represents the concentration of water quality index at the i-th breeding stage and the j-th breeding stage, Q i represents the amount of aquaculture tailwater discharged into the environment in the i-th aquaculture stage, G i represents the carbon emissions in the ith breeding stage, E i represents the carbon emission coefficient of the i-th breeding stage, Z i Represents the increase in dry weight of scallops in the i-th culture stage.

8. The method for predicting the pollution reduction and carbon reduction effectiveness of shellfish aquaculture based on gated neural network according to claim 6 is characterized in that: The expression of the pollution prediction result is: G = G1 + G2 + G3; Among them, W j W1 represents the total discharge of the jth water quality index in the whole cycle, j represents the j-th water quality index discharge in the first breeding stage, represents the discharge of the jth water quality index in the second breeding stage, represents the emission of the jth water quality index in the third breeding stage, G represents the total carbon emission in the whole cycle, G1 represents the carbon emission in the first breeding stage, G2 represents the carbon emission in the second breeding stage, and G3 represents the carbon emission in the third breeding stage.

Images