Harmful algal bloom growth prediction method based on Tent-GWO-GRU
Optimizing the GRU model parameters through the Tent-GWO-GRU model, the problem of insufficient prediction accuracy and stability of harmful algae burn growth in the existing technology is solved, and higher prediction accuracy and stability are achieved, fully considering the delay of algae burn growth Post-sex.
Patent Information
- Application Number
- CN202410939981.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-15
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-15
AI Technical Summary
The existing machine learning model has poor accuracy and stability when predicting the growth and evolution of harmful algae burns, especially the slow convergence speed of the particle swarm optimization algorithm and the random initial point selection, which affects the model's global search ability.
Adopting the harmful algae flower growth prediction method based on Tent-GWO-GRU, the gated recurrent unit neural network GRU model is constructed by obtaining historical water quality data, and the GRU model parameters are optimized using Tent chaotic mapping algorithm. To improve prediction accuracy.
The stability and prediction accuracy of the model are improved, and the delay in the growth of algae burns is considered through time delay, the initial wolf pack distribution is improved, the global search ability of the algorithm is enhanced, and the prediction accuracy is achieved.
Smart Images

Figure CN118917958A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the technical field of harmful algal bloom growth, and specifically relates to a harmful algal bloom growth prediction method based on Tent-GWO-GRU. Background Art
[0002] Harmful algal blooms (HABs) refer to ecological disasters caused by the rapid proliferation or large-scale aggregation of planktonic algae in the water. In recent decades, due to eutrophication of water bodies, global warming, the development of maritime transportation and marine aquaculture, the frequency, duration, scope of impact and degree of harm of harmful algal blooms have shown an increasing trend worldwide. In order to reduce the potential losses caused by harmful algal blooms, it is very necessary to establish an effective model to predict the growth and evolution of algae.
[0003] In recent years, machine learning models have been widely used in water quality prediction modeling, such as artificial neural network (ANN), recurrent neural network (GRU), extreme learning machine (ELM), random forest (RF), support vector machine (SVM), etc., which have been widely used for regression prediction. Machine learning models bypass the complex mechanism of algal bloom growth and evolution, focus on the mapping relationship between the output and input of the research system, and can accurately approximate highly nonlinear relationships by learning a large amount of historical data. Mamun et al. used support vector machines to predict algal chlorophyll a and water transparency in clean water ecosystems. The experimental results showed that the prediction accuracy and model stability of the SVM model were poor, and it could not accurately predict the growth and evolution process of harmful algal blooms. Lu Yuanbo et al. used the recurrent neural network (GRU) model to predict the water quality of cross-border sections in Guangdong Province with high prediction accuracy. In order to realize the migration and use of the GRU model, a PSO-GRU water quality prediction model was proposed, and the particle swarm optimization algorithm was introduced to optimize the hyperparameters of GRU, reducing the empirical and random nature of the hyperparameter settings and improving the prediction accuracy of the model. However, the convergence speed of the particle swarm optimization algorithm is slow. The gray wolf optimization algorithm can solve this problem well, but there is a problem of relatively random initial point selection. Therefore, Li Jin et al. used the randomness and ergodicity of the chaotic system to generate a more evenly distributed Tent chaotic sequence to initialize the gray wolf population, and obtained an initial solution with a more even distribution and higher diversity, which improved the global search ability of the algorithm. However, how to apply Tent-GWO and GRU to the field of harmful algal blooms still needs further exploration. Summary of the invention
[0004] The present application provides a harmful algal bloom growth prediction method based on Tent-GWO-GRU to solve the above-mentioned technical problems.
[0005] In order to solve the above technical problems, a technical solution adopted in this application is: a harmful algal bloom growth prediction method based on Tent-GWO-GRU, comprising the following steps:
[0006] Obtain historical water quality data of the target basin and perform preprocessing;
[0007] Based on historical water quality data, a gated recurrent unit neural network GRU model was constructed;
[0008] Based on the Tent chaotic mapping algorithm, the Grey Wolf Algorithm GWO is optimized;
[0009] Based on the gated recurrent unit neural network GRU model and the grey wolf algorithm GWO, the GRU model parameters are optimized and predictions are performed.
[0010] Furthermore, the method of obtaining the historical water quality data of the target watershed and performing preprocessing includes:
[0011] Collect historical water quality data of the target watershed and select chlorophyll concentration as an indicator of algae growth;
[0012] Based on historical water quality data and factors affecting algae growth, a first feature data set is obtained;
[0013] Based on the linear interpolation method, the missing value and outlier test are performed on the water quality characteristic data set to obtain the second characteristic data set;
[0014] Performing range normalization processing on the indicator data in the second feature data set to obtain a third feature data set;
[0015] Based on the time delay and the indicator data in the third feature data set, obtaining preprocessed data;
[0016] The preprocessed data is divided into training set and test set according to the preset ratio.
[0017] Furthermore, the method for constructing a gated recurrent unit neural network GRU includes:
[0018] Reorganize the index data in the preprocessed data to obtain a vector set;
[0019] Based on the vector set, the reset gate of the gated recurrent unit neural network GRU model is constructed;
[0020] Based on the reset gate of the gated recurrent unit neural network GRU model, the update gate of the gated recurrent unit neural network GRU model is obtained;
[0021] Based on the update gate of the gated recurrent unit neural network GRU model, a fully connected layer of the gated recurrent unit neural network GRU model is obtained;
[0022] Based on the fully connected layer of the gated recurrent unit neural network GRU model, the output value of the gated recurrent unit neural network GRU is obtained.
[0023] Furthermore, based on the Tent chaotic mapping algorithm, the method for optimizing the Grey Wolf Algorithm GWO includes:
[0024] Based on the Tent chaotic mapping algorithm, the wolf group in the grey wolf algorithm GWO is initialized to obtain the initialization parameters of the wolf group and the initial number of iterations;
[0025] Based on the initialized wolf pack parameters, the fitness value of each individual gray wolf is obtained;
[0026] Compare the fitness values of all gray wolf individuals to obtain the first optimal solution, the second optimal solution, and the third optimal solution;
[0027] Based on the optimal gray wolf individuals corresponding to the first optimal solution, the second optimal solution, and the third optimal solution, the distance parameters between the remaining gray wolf individuals and the optimal gray wolf individual are obtained;
[0028] Based on the distance parameter, update the parameter vector;
[0029] Repeat the above steps until the number of iterations reaches the preset value to obtain the optimal gray wolf parameters.
[0030] Furthermore, the method of optimizing the GRU model parameters and making predictions includes:
[0031] The data in the training set is input into the gated recurrent unit neural network GRU model to obtain the predicted value and prediction accuracy of chlorophyll concentration;
[0032] Based on the true value and predicted value of chlorophyll concentration, obtain the model loss function;
[0033] Optimize the model loss function based on the Adam optimization algorithm and obtain the pre-trained model;
[0034] The loss function is used as the fitness function, and the learning rate, number of hidden layer nodes and regularization coefficient in the pre-trained model are optimized according to the optimized Tent-GWO algorithm to obtain the pre-trained Tent-GWO-GRU model;
[0035] Based on the Tent-GWO-GRU model, the chlorophyll concentration in the test set is obtained.
[0036] The beneficial effects of the present application are as follows: the present application performs time delay on the index data, fully considers the delay of the growth evolution of algal blooms, and improves the feasibility of the method. The Tent chaos mapping algorithm is introduced to improve the method of generating the initial wolf pack by the GWO algorithm, so that the distribution of the initial wolf pack is more uniform, which significantly improves the fitness value of the algorithm and makes it easier to find the global optimal solution. In addition, the present application uses the Tent-GWO optimization algorithm to optimize the hyperparameters and network structure of the GRU model, which not only improves the stability of the model, but also achieves a higher prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 It is a flow chart of an embodiment of a harmful algal bloom growth prediction method based on Tent-GWO-GRU of the present application;
[0038] Figure 2 for Figure 1 A schematic flow chart of an embodiment of step S1 in FIG.
[0039] Figure 3 for Figure 1 A schematic flow chart of an embodiment of step S2 in FIG.
[0040] Figure 4 yes Figure 3 A schematic diagram of a unit structure of an embodiment of a gated recurrent unit neural network GRU model in FIG.
[0041] Figure 5 for Figure 1 A schematic flow chart of an embodiment of step S3 in FIG.
[0042] Figure 6 for Figure 1 A schematic flow chart of an embodiment of step S4 in FIG.
[0043] Figure 7 This is a fitness comparison diagram of the improved GWO and the original GWO in one embodiment of the harmful algal bloom growth prediction method based on Tent-GWO-GRU of the present application;
[0044] Figure 8 A comparison diagram of the goodness of fit between a harmful algal bloom growth prediction method and an SVM model in an embodiment of a harmful algal bloom growth prediction method based on Tent-GWO-GRU of the present application;
[0045] Fig. 9 This is a model prediction effect diagram of a harmful algal bloom growth prediction method according to an embodiment of the harmful algal bloom growth prediction method based on Tent-GWO-GRU of the present application;
[0046] Fig.10 This is a diagram showing the model effect of predicting the growth of harmful algal blooms based on the SVM model. DETAILED DESCRIPTION
[0047] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below in conjunction with specific embodiments.
[0048] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.
[0049] See also Figure 1 , Figure 1 : is a flow chart of an embodiment of a method for predicting harmful algal bloom growth based on Tent-GWO-GRU of the present application. The method comprises:
[0050] Step S1. Obtain historical water quality data of the target basin and perform preprocessing.
[0051] For details, see Figure 2 , Figure 2 for Figure 1 Schematic diagram of the process of step S1 in an embodiment. Step S1 includes:
[0052] Step S11: Collect historical water quality data of the target watershed and select the concentration of chlorophyll as an indicator to measure the degree of algae growth.
[0053] Specifically, water quality data from January 1, 2000 to December 31, 2023 were collected from the official website of the monitoring station in the target watershed. The concentration of chlorophyll-a (Chl-a) was selected as an indicator to measure the degree of algae growth, and eight indicators including total inorganic nitrogen (TIN) concentration, total phosphorus (PO4) concentration, chlorophyll-a (Chl-a) concentration, dissolved oxygen (DO) concentration, water temperature (Temp), Secchi disk depth (SSD), 5-day biological oxygen demand (BOD5) and pH value (PH) were selected as model inputs from the factors that directly affect algae growth.
[0054] Step S12: Based on historical water quality data and factors affecting algae growth, a first feature data set is obtained.
[0055] Specifically, the water quality data from January 1, 2000 to December 31, 2023 are integrated according to the selected indicators to obtain the first feature data set W1 = {TIN t ,PO4 i ,Chl-a t ,DO t ,Temp t ,SSD t ,BOD5 t,PH t}.
[0056] Step S13: Based on the linear interpolation method, the water quality characteristic data set is tested for missing values and outliers to obtain a second characteristic data set.
[0057] Specifically, the first feature data set W1 is tested for missing values and outliers, and is filled using a linear interpolation method to obtain a second feature data set W2.
[0058] Step S14: Perform range normalization processing on the indicator data in the second feature data set to obtain a third feature data set.
[0059] Specifically, the eight indicator data in the second feature data set W2 are subjected to range normalization processing, as shown in the following formula:
[0060]
[0061] Among them, x and x' represent the original value and normalized value of the variable respectively, x max and x min Represent the maximum and minimum values of the variable respectively.
[0062] Suppose the standardized third feature data set W3 is:
[0063] W3={TIN' t ,PO4' t ,Chl-a' t ,DO' t ,Temp' t ,SSD' t ,BOD5' t ,PH' t};
[0064] Among them, TIN' t ,PO4' t ,Chl-a' t ,DO' t ,Temp' t ,SSD' t ,BOD5' t and PH' t Respectively represent the normalized values of 8 indicator data.
[0065] Step S15: Based on the time delay and the indicator data in the third feature data set, obtain preprocessing data.
[0066] Specifically, in order to accurately predict the growth and evolution process of harmful algal blooms, a time delay of 7 to 13 days (t-7, t-8, t-9, t-10, t-11, t-12, t-13) was introduced into the 8 indicators in the third feature data set W3, and preprocessed data containing 56 input variables were constructed; among which, the preprocessed data is:
[0067]
[0068] Among them, the start and end time of t is from January 14, 2000 to December 31, 2023.
[0069] Step S16: Divide the preprocessed data into a training set and a test set according to a preset ratio.
[0070] Specifically, the preprocessed data is divided into training set and test set in a ratio of 7:3.
[0071] Step S2. Based on historical water quality data, a gated recurrent unit neural network GRU model is constructed.
[0072] For details, see Figure 3 , Figure 3 for Figure 1 Schematic diagram of the process of step S2 in an embodiment. Step S2 includes:
[0073] Step S21: Reorganize the index data in the preprocessed data to obtain a vector set.
[0074] Specifically, the 56 indicator data in the data set W3 are reorganized to obtain a vector set;
[0075] X t =(TIN′ t-7 ,PO4′ t-7 ,Chla′ t-7 ,DO′ t-7 ,…,Temp′ t-13 ,SSD′ t-13 ,BOD5′ t-13 ,PH′ t-13 );
[0076] Among them, X t For vector set.
[0077] Step S22. Based on the vector set, a reset gate of the gated recurrent unit neural network GRU model is constructed.
[0078] Specifically, based on formula (1), a reset gate of the gated recurrent unit neural network GRU model is constructed; wherein formula (1) is:
[0079] r t=Sigmoid(W r x t +U r h t-1 ,+b r ) (1);
[0080] Among them, the reset gate r t The smaller the value of W, the more information needs to be forgotten at time t; r and U r are weight matrices, b r is the bias vector of the reset gate, h t-1 Represents the output of the hidden layer at time t-1.
[0081] Step S23. Based on the reset gate of the gated recurrent unit neural network GRU model, obtain the update gate of the gated recurrent unit neural network GRU model.
[0082] For details, see Figure 4 , based on formulas (2)-(4), the update gate of the gated recurrent unit neural network GRU model is constructed; where formulas (2)-(4) are:
[0083] z t =Sigmoid(W z x t +U z h t-1 ,+b z ) (2);
[0084]
[0085] Among them, reset gate z t The smaller the value, the more historical information is retained. is the candidate hidden layer state, h t W represents the state vector output at time t. r ,W h ,U z and U h are weight matrices, b z and b h is the bias vector, and ⊙ represents the dot product at the corresponding position.
[0086] Step S24. Based on the update gate of the gated recurrent unit neural network GRU model, obtain the fully connected layer of the gated recurrent unit neural network GRU model.
[0087] Specifically, based on formula (5), a fully connected layer of the gated recurrent unit neural network GRU model is constructed; wherein formula (5) is:
[0088] y m =f(∑Wmt h t +b m ) (5);
[0089] Among them, y m is the output of the mth neuron in the fully connected layer; W mt represents the weight of the fully connected layer; b m represents the bias value of the fully connected layer; f(x) represents the ReLU activation function.
[0090] Step S25. Based on the fully connected layer of the gated recurrent unit neural network GRU model, obtain the output value of the gated recurrent unit neural network GRU.
[0091] Specifically, the output layer of the gated recurrent unit neural network GRU is a real-valued regression of a single neuron. The output value can be expressed as formula (6):
[0092] Y=Linear(∑α m y m +b) (6);
[0093] Among them, α m represents the normalized weight, and b represents the bias value.
[0094] Step S3: Optimize the Grey Wolf Algorithm GWO based on the Tent Chaotic Mapping Algorithm.
[0095] For details, see Figure 5 , Figure 5 for Figure 1 Schematic diagram of the process of step S3 of an embodiment. Step S3 includes:
[0096] Step S31. Based on the Tent chaotic mapping algorithm, the wolf pack in the grey wolf algorithm GWO is initialized to obtain the initialization wolf pack parameters and the initial number of iterations.
[0097] Specifically, the Tent chaotic mapping algorithm is used to initialize the wolf pack and generate n individual positions of gray wolves. The specific formula is as follows:
[0098]
[0099] Among them, L i represents the original position of the gray wolf individual, Represents the position after Tent chaos mapping, i=1,2,…,n. Initialize the wolf group parameters a, A, C, which are defined as follows:
[0100] C = 2r2, A = 2ar1-a;
[0101]
[0102] Among them, C is the swing factor, A is the convergence factor, r1 and r2 are random numbers between [0,1], a decreases linearly from 2 to 0 as the number of iterations increases, and max k is the maximum number of iterations, and k represents the kth iteration.
[0103] Step S32: Based on the initialized wolf pack parameters, obtain the fitness value of each individual gray wolf.
[0104] Step S33: Compare the fitness values of all gray wolf individuals to obtain the first optimal solution, the second optimal solution and the third optimal solution.
[0105] Specifically, compare the fitness values of individuals to determine the current first optimal solution The second best solution And the third optimal solution
[0106] Step S34. Based on the optimal gray wolf individuals corresponding to the first optimal solution, the second optimal solution, and the third optimal solution, obtain the distance parameters between the remaining gray wolf individuals and the optimal gray wolf individual.
[0107] Specifically, according to the swing factor C, the distance parameter D between other gray wolf individuals in the population and the optimal gray wolf individuals α, β, δ is calculated α , D β , D δ .
[0108] Step S35: Update the parameter vector based on the distance parameter.
[0109] Specifically, according to the distance parameter D α , D β , D δ And the convergence factor A, update the position of each gray wolf, and update the wolf pack parameters a, A, C.
[0110] Step S36. Repeat the above steps until the number of iterations reaches a preset value to obtain the optimal gray wolf parameters.
[0111] Specifically, determine whether the current number of iterations k reaches the maximum number of iterations max k , if k <max k , return to step S32, otherwise determine the optimal gray wolf parameters.
[0112] Step S4. Based on the gated recurrent unit neural network GRU model and the Grey Wolf Algorithm GWO, optimize the GRU model parameters and perform prediction.
[0113] For details, see Figure 6 , Figure 6 for Figure 1 Schematic diagram of the process of step S4 in an embodiment. Step S4 includes:
[0114] Step S41. Input the data in the training set into the gated recurrent unit neural network GRU model to obtain the predicted value and prediction accuracy of chlorophyll concentration.
[0115] Specifically, the data of the training set and the chlorophyll concentration at the corresponding time are input into the gated recurrent unit neural network GRU model, and the chlorophyll concentration prediction value and prediction accuracy are output.
[0116] Step S42: Obtain a model loss function based on the actual value and predicted value of chlorophyll concentration.
[0117] Specifically, according to the predicted value and true value of chlorophyll concentration, the model loss function RMSE of the GRU model is calculated; wherein, the model loss function RMS is as follows:
[0118]
[0119] in, represents the predicted value of chlorophyll concentration of the jth sample, y t represents the true value of the chlorophyll concentration of the jth sample, and N represents the total number of samples.
[0120] Step S43. Optimize the model loss function based on the Adam optimization algorithm to obtain a pre-trained model.
[0121] Specifically, the Adam optimization algorithm is used to optimize the model loss function to obtain the pre-trained model.
[0122] Step S44. Using the loss function as the fitness function, according to the optimized Tent-GWO algorithm, the learning rate, the number of hidden layer nodes and the regularization coefficient in the pre-trained model are optimized to obtain the pre-trained Tent-GWO-GRU model.
[0123] Step S45: Based on the Tent-GWO-GRU model, obtain the chlorophyll concentration in the test set.
[0124] See also Figure 7-10 In order to verify the effectiveness of the implementation of this application, MATLAB is used for numerical simulation verification. This example obtains Hong Kong water quality data from January 1, 1988 to December 31, 2018 from the Hong Kong Conservation Department's Weakest Flush Monitoring Station, uses linear interpolation for data preprocessing, and introduces a time delay of 7 to 13 days for the 8 indicator data. Based on the above data, in the Tent-GWO-GRU model proposed in the present invention, the gray wolf population is 5, the number of iterations is 20, Figure 7is the fitness curve of the improved GWO algorithm and the original GWO algorithm. The optimal fitness of the Tent-GWO algorithm is 0.0432, and the optimal fitness of the original GWO algorithm is 0.0475. Figure 7 It can be found that the convergence speed and convergence accuracy of the improved GWO algorithm are better than those of the original GWO algorithm, indicating that the improved GWO algorithm has better global search capabilities than the original GWO algorithm. Figure 8 As shown in the figure, it can be seen that the Tent-GWO-GRU model can better capture the growth trend and change amplitude of ChI-a concentration in the test set, with fewer outliers. Fig. 9 , Fig.10 The following are the fitting results of the prediction results of Tent-GWO-GRU and SVM and the true value, respectively. It can be seen from the figure that the model fitting effect of Tent-GWO-GRU is obviously better than that of SVM. The Tent-GWO-GRU model proposed in the present invention has good prediction performance.
[0125] This application performs time delay on the indicator data, fully considering the delay of algal bloom growth evolution, and improving the feasibility of the method. The Tent chaos mapping algorithm is introduced to improve the method of generating the initial wolf pack by the GWO algorithm, so that the distribution of the initial wolf pack is more uniform, which significantly improves the fitness value of the algorithm and makes it easier to find the global optimal solution. In addition, this application uses the Tent-GWO optimization algorithm to optimize the hyperparameters and network structure of the GRU model, which not only improves the stability of the model, but also achieves a higher prediction accuracy.
[0126] The above descriptions are merely embodiments of the present application and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.
Claims
1. A method for predicting the growth of harmful algal blooms based on Tent-GWO-GRU, characterized in that: The following steps are involved: Obtain historical water quality data of the target basin and perform preprocessing; Based on the historical water quality data, a gated recurrent unit neural network GRU model is constructed; Based on the Tent chaotic mapping algorithm, the Grey Wolf Algorithm GWO is optimized; Based on the gated recurrent unit neural network GRU model and the grey wolf algorithm GWO, the GRU model parameters are optimized and prediction is performed.
2. The method according to claim 1, characterized in that: The method for obtaining historical water quality data of the target watershed and performing preprocessing includes: Collect historical water quality data of the target watershed and select chlorophyll concentration as an indicator to measure the degree of algae growth; Based on the historical water quality data and factors affecting algae growth, obtaining a first feature data set; Based on the linear interpolation method, the water quality characteristic data set is tested for missing values and outliers to obtain a second characteristic data set; Performing range normalization processing on the indicator data in the second feature data set to obtain a third feature data set; Based on the time delay and the indicator data in the third feature data set, obtaining preprocessing data; The preprocessed data is divided into a training set and a test set according to a preset ratio.
3. The method according to claim 2, characterized in that The method for constructing a gated recurrent unit neural network GRU comprises: Reorganize the index data in the preprocessed data to obtain a vector set; Based on the vector set, a reset gate of a gated recurrent unit neural network GRU model is constructed; Based on the reset gate of the gated recurrent unit neural network GRU model, obtaining the update gate of the gated recurrent unit neural network GRU model; Based on the update gate of the gated recurrent unit neural network GRU model, obtaining a fully connected layer of the gated recurrent unit neural network GRU model; Based on the fully connected layer of the gated recurrent unit neural network GRU model, an output value of the gated recurrent unit neural network GRU is obtained.
4. The method according to claim 3, characterized in that The method for optimizing the grey wolf algorithm GWO based on the Tent chaotic mapping algorithm includes: Based on the Tent chaotic mapping algorithm, the wolf group in the grey wolf algorithm GWO is initialized to obtain the initialization wolf group parameters and the initial number of iterations; Based on the initialized wolf pack parameters, obtaining the fitness value of each individual gray wolf; Comparing the fitness values of all the gray wolf individuals to obtain a first optimal solution, a second optimal solution, and a third optimal solution; Based on the optimal gray wolf individuals corresponding to the first optimal solution, the second optimal solution, and the third optimal solution, obtaining distance parameters between the remaining gray wolf individuals and the optimal gray wolf individual; Based on the distance parameter, updating the parameter vector; Repeat the above steps until the number of iterations reaches a preset value to obtain the optimal gray wolf parameters.
5. The method according to claim 4, characterized in that The method for optimizing GRU model parameters and performing prediction includes: Inputting the data in the training set into the gated recurrent unit neural network GRU model to obtain the predicted value and prediction accuracy of the chlorophyll concentration; Obtaining a model loss function based on the true value and the predicted value of the chlorophyll concentration; Optimize the model loss function based on the Adam optimization algorithm to obtain a pre-trained model; The loss function is used as the fitness function, and the learning rate, number of hidden layer nodes and regularization coefficient in the pre-trained model are optimized according to the optimized Tent-GWO algorithm to obtain the pre-trained Tent-GWO-GRU model; Based on the Tent-GWO-GRU model, the chlorophyll concentration in the test set is obtained.
Citation Information
Patent Citations
Cyanobacterial bloom cross-water-area prediction method and device based on transfer learning
CN115330053A
Method for predicting putting amount of algistat for treating cyanobacterial bloom
CN116663723A
System for detection and prediction of water quality events
US20070233397A1
