A method, system, device and storage medium for predicting runoff sediment content

Abstract

The application provides a runoff sediment content prediction method, system, device and storage medium, belongs to the river sediment prediction technical field, and includes the following steps: introducing the position updating mechanism of the spotted hyena optimization algorithm and the position updating mechanism of the gradient optimization algorithm in the position updating mechanism of the gorilla troop optimization algorithm to obtain an improved gorilla troop optimization algorithm; the model parameters of a BiLSTM model are optimized by using the improved gorilla troop optimization algorithm, and a runoff sediment prediction model is obtained by training the BiLSTM model after parameter optimization. The runoff sediment content prediction result and the prediction accuracy are obtained based on the runoff sediment prediction model. Strong support is provided for ecological protection and engineering planning.

Description

Technical Field

[0001] This invention belongs to the field of river sediment prediction technology, specifically relating to a method, system, equipment, and storage medium for predicting runoff sediment content. Background Technology

[0002] The sediment content in runoff is a key parameter for measuring the degree of soil erosion, assessing reservoir siltation, providing early warning of geological disasters, and planning water conservancy projects. Under the influence of multiple factors such as global climate change, land use changes (e.g., urbanization, agricultural expansion), and the implementation of soil and water conservation projects, the runoff and sediment content of rivers and reservoirs exhibit dynamic and complex changes. This can lead to engineering problems such as river siltation and reservoir capacity reduction, as well as environmental risks such as eutrophication and ecological degradation. Therefore, predicting runoff sediment content is beneficial for protecting water resources and the ecological environment, effectively preventing soil erosion, maintaining river ecological balance, ensuring water quality safety, and preventing pollutants carried by sediment from damaging water bodies.

[0003] With the development of artificial intelligence technology, methods such as BP neural networks, LSTM, and BiLSTM are gradually being applied to runoff and sediment prediction, leveraging the nonlinear fitting capabilities of models to uncover hidden relationships between data. The predictive performance of a single deep learning model is highly dependent on hyperparameter configuration, and traditional hyperparameter optimization methods (grid search, random search) are inefficient for direct optimization. To address the challenge of hyperparameter optimization for single deep learning models, swarm intelligence optimization algorithms are combined with deep learning, utilizing the global optimization capabilities of swarm intelligence algorithms to optimize the hyperparameters of deep learning models. Swarm intelligence optimization algorithms simulate the foraging, migration, and competition behaviors of organisms, iteratively searching for optimal solutions in the solution space, making them suitable for solving complex nonlinear optimization problems such as hyperparameter optimization of deep learning models. However, swarm intelligence optimization algorithms are prone to getting trapped in local optima during the optimization process, which, when directly applied to runoff and sediment prediction, leads to insufficient model prediction accuracy and stability. Summary of the Invention

[0004] To address the problem that existing technologies are ineffective in predicting runoff sediment content, this invention provides a method, system, device, and storage medium for predicting runoff sediment content.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for predicting runoff sediment content includes the following steps: By introducing the position update mechanisms of the spotted hyena optimization algorithm and the gradient optimization algorithm into the original gorilla force optimization algorithm, an improved gorilla force optimization algorithm is obtained. The improved gorilla troop optimization algorithm is used to optimize the parameters of the BiLSTM model, resulting in a parameter-optimized BiLSTM model. Specifically, in the search phase, the convergence and control factors of the spotted hyenas are used to adjust individuals to move closer to the optimal individual, and gradient adaptive parameters are used to guide the search direction for global exploration and position updates. In the search sub-phase of the development phase, individuals form a variable search radius around the optimal individual, and gradient adaptive parameters are used to adjust the step size of individuals moving towards the optimal individual. The positions of other individuals in the population are introduced as correction terms to update the candidate positions of individuals. In the competition sub-phase of the development phase, the convergence and control factors of the spotted hyenas are used to control individuals to dynamically surround the optimal individual, and gradient adaptive parameters are used to dynamically adjust the gap between individuals and the optimal individual. The positions of other individuals in the population are randomly introduced as competition to update the candidate positions of individuals. The population refers to n sets of BiLSTM model parameters, with one set of BiLSTM model parameters corresponding to one individual. A dataset is constructed by acquiring geographical environment data, rainfall and meteorological data, and historical runoff sediment content measurements. Using the geographical environment data and rainfall and meteorological data in the dataset as inputs and the historical runoff sediment content measurements as output targets, the BiLSTM model with optimized parameters is trained to obtain a runoff sediment prediction model.

[0006] Preferably, the position update rule during the search phase is as follows: ; Where GX(t+1) represents the candidate position of the gorilla in the next iteration; t represents the current iteration number; X(t) represents the current position of each gorilla; r1, r2, and r3 are random values ​​in the range [0,1] updated in each iteration; X r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively; gbest is the position of the optimal individual, i.e., the position of the silverback gorilla. A controlling factor for spotted hyenas; and All are adaptive parameters of the gradient; E is the convergence factor of the spotted hyena; C, L, and H are algorithm control parameters.

[0007] Preferably, the position update rule in the search sub-stage is as follows: ; Where gbest is the location of the silverback gorilla, X r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively. and All are adaptive parameters of the gradient, t represents the current iteration number, E is the convergence factor of the spotted hyena, and h is the control factor of the spotted hyena.

[0008] Preferably, the position update rule in the competitive sub-stage is as follows: ; Where Q represents combat power, A is combat coefficient, gbest is the position of the silverback gorilla, and X... r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively. and All are adaptive parameters of the gradient, t represents the current iteration number, E is the convergence factor of the spotted hyena, and h is the control factor of the spotted hyena.

[0009] Preferably, the method further includes acquiring real-time geographic environmental data and rainfall and meteorological data and inputting them into the runoff and sediment prediction model to obtain the predicted value of runoff and sediment content.

[0010] Preferably, calculating the fitness value of an individual in the population specifically involves decoding the current position of the individual to obtain the hyperparameters of the BiLSTM model, training the BiLSTM model corresponding to the hyperparameters using the dataset parameters, constructing an error objective function to minimize the deviation between the predicted value and the actual runoff and sediment value, and calculating the error value based on the error objective function. This error value is the fitness value of the individual. The error objective function can be selected as mean absolute error (MAE), mean relative error (MRE), or root mean square error (RMSE).

[0011] Preferably, the method further includes setting constraints on the objective function, specifically on the number of hidden layer nodes k and the learning rate of the BiLSTM model. Batch size Dropout rate Set upper and lower limits.

[0012] This invention also provides a runoff sediment content prediction system, specifically comprising: The algorithm improvement module is used to introduce the position update mechanisms of the spotted hyena optimization algorithm and the gradient optimization algorithm into the position update mechanism of the original gorilla force optimization algorithm, so as to obtain the improved gorilla force optimization algorithm.

[0013] The model optimization module is used to optimize the BiLSTM model parameters using the improved gorilla army optimization algorithm, resulting in a parameter-optimized BiLSTM model. Specifically, in the search phase, the convergence and control factors of the spotted hyenas are used to adjust individuals to move closer to the optimal individual, and gradient adaptive parameters are used to guide the search direction for global exploration and position updates. In the search sub-phase of the development phase, individuals form a variable search radius around the optimal individual, and gradient adaptive parameters are used to adjust the step size of individuals moving towards the optimal individual, while the positions of other individuals in the population are introduced as correction terms to update the candidate positions of individuals. In the competition sub-phase of the development phase, the convergence and control factors of the spotted hyenas are used to control individuals to dynamically surround the optimal individual, and gradient adaptive parameters are used to dynamically adjust the gap between individuals and the optimal individual, while the positions of other individuals in the population are randomly introduced as competition to update the candidate positions of individuals. The population refers to n sets of BiLSTM model parameters, with one set of BiLSTM model parameters corresponding to one individual.

[0014] The sediment prediction module is used to acquire geographical environment data, rainfall and meteorological data, and historical runoff sediment content measured values ​​to construct a dataset. Using the geographical environment data and rainfall and meteorological data in the dataset as inputs and the historical runoff sediment content measured values ​​as output targets, the BiLSTM model with optimized parameters is trained to obtain the runoff sediment prediction model.

[0015] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps described in the method for predicting runoff sediment content.

[0016] The present invention also provides a computer-readable storage medium storing a computer program that, when loaded by a processor, can execute the steps described in the method for predicting runoff sediment content.

[0017] The method for predicting runoff sediment content provided by this invention has the following beneficial effects: This invention introduces the spotted hyena optimization algorithm and gradient optimization algorithm into the position update mechanism during the search phase, improving the individual gorilla position update method. It transforms the purely random, unguided migration into a guided global exploration, allowing individuals to maintain a certain degree of randomness to cover unknown areas while utilizing optimal solutions and gradient information to accelerate convergence and increase the diversity of individual positions. The improved gorilla troop optimization algorithm is used to optimize the parameters of the BiLSTM model, enabling individuals to explore more potential positions in competition, improving the algorithm's optimization efficiency and accuracy, reducing the influence of suboptimal solutions, and finding the parameter combination that minimizes the objective function more quickly, avoiding parameters getting trapped in local optima. Based on the optimal individual positions output by the algorithm, a BiLSTM model with optimized parameters is obtained. The higher the quality of the parameters output by the algorithm, the higher the prediction accuracy of the BiLSTM. This is beneficial for subsequent ecological environmental protection and engineering planning. Attached Figure Description

[0018] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a method for predicting runoff sediment content according to an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0021] Example The Gorilla Troops Optimization Algorithm (GTOA) is a swarm intelligence optimization algorithm inspired by the group behavior of gorillas. It primarily consists of two phases: exploration and development. The exploration phase uses three different operational methods: migrating to an unknown location, migrating to a known location, and migrating to other groups. The development phase uses two operational methods: following a silverback gorilla (the leader) and competing with female gorillas. These five operational methods simulate the behavioral patterns of gorillas.

[0022] This is achieved by simulating the lifestyle and social relationships of gorillas within a population. First, assume a region with multiple gorilla groups. Each group consists of one adult male gorilla (optimal solution), multiple adult female gorillas (other solutions), and their offspring, young male gorillas (candidate solutions). The adult male gorilla, also called a silverback gorilla, is the leader of the group, while the young males are called blackback gorillas. Typically, both male and female gorillas may migrate from their birth group to a new one. Even adult male gorillas may leave their current group to attract migrating females and form new groups. Meanwhile, some adult male gorillas may choose to remain in their current group, continuing to follow the silverback gorilla. When the silverback gorilla dies, these adult males may fight for leadership and mate with adult female gorillas.

[0023] However, this algorithm overemphasizes global search, which may cause it to linger too long near known promising areas, preventing in-depth exploration; or the search may be entirely dominated by the location of the silverback gorillas, with all gorillas closely surrounding them, lacking a mechanism for actively discovering new directions. If the silverback gorillas themselves get stuck in local optima, the entire population will be quickly pulled towards suboptimal solutions, leading to a sharp decline in diversity and making the algorithm prone to local optima. This results in inaccurate or insufficiently precise predictions of runoff and sediment, failing to achieve the desired predictive results.

[0024] Basic principles of the Gorilla Force optimization algorithm: (1) Initialization phase: During the initialization phase, the positions of the gorilla group are randomly initialized. Assuming there are nPop gorillas in a Dim-dimensional space, the i-th gorilla can be represented as... (1); Assuming the upper and lower boundaries of the target space are denoted by [UpB, LoB], then X i The initial position can be generated by equation (2).

[0025] (2); (2) Exploration phase: During the exploration phase, all gorillas were considered young males, i.e., candidate solutions. Silverback gorillas were the optimal solution in each iteration. To better simulate the natural migration behavior of gorilla groups, three different mathematical expressions were used to simulate gorilla migration to unknown locations, migration to known locations, and migration to other groups, as shown below.

[0026] (3); Among them, GX(t + 1) represents the candidate position of the gorilla individual in the next iteration, t represents the current iteration number, and X(t) represents the current position of each gorilla. r1, r2, r3, and rand are random values within the range of [0, 1] updated in each iteration. is the lower bound for optimization, is the upper bound for optimization. Xr(t) and GXr(t) represent the current position of a randomly selected gorilla from the entire population and the candidate position of a randomly selected gorilla, respectively. The algorithm process parameters C, F, L, and H are calculated through the following formulas respectively.

[0027] (4); (5); (6); (7); Among them, l is a random value within [-1, 1], Z is a random value within [-C, C], T is the maximum number of iterations of the gorilla group, is a random number. The random number in F is to introduce random perturbations during the attenuation process of controlling C, thereby increasing the diversity of exploration.

[0028] (3) Exploration stage: In the exploration stage of the gorilla troop optimization algorithm, in order to simulate the two behaviors of following the silverback gorilla and competing for adult female gorillas, the parameter w is introduced to control the switching between them, and its value is 0.8.

[0029] If C ≥ w, it enters the first sub-stage: the search sub-stage, and chooses to follow the silverback gorilla. Its mathematical expression is as follows.

[0030] (8); (9); (10); Among them, gbest is the position of the silverback gorilla, is the average position of the population.

[0031] If C < w, it enters the second sub-stage: the competition sub-stage. When young gorillas enter puberty, they will fight with other male gorillas to compete for adult female gorillas. This behavior is simulated by Equation (11).

[0032] (11); Among them, Q represents the combat power, A is the combat coefficient, and the calculation methods of Q and A are as follows.

[0033] (12); (13); Where β=3, E is used to simulate the influence of the dimension of the solution on combat power, as shown in equation (14).

[0034] (14); When rand ≥ 0.5, E will be equal to a D-dimensional array of random numbers that follow a normal distribution; when rand < 0.5, E will be equal to a random value that follows a normal distribution.

[0035] The optimization algorithm process for the "Gorilla Force" is described below.

[0036] Step 1: Initialize the population.

[0037] Step 2: Calculate the individual fitness value and select the optimal individual.

[0038] Step 3: Update C, F, L, H.

[0039] Step 4: Individuals begin to explore behaviors based on formula (3).

[0040] Step 5: If C≥w, proceed to the first stage, choose to follow the silverback gorilla, and update the position using formula (8); otherwise, proceed to the second stage, choose to fight other male gorillas, and update the position using formula (11).

[0041] Step 6: Calculate the individual fitness value and select the optimal individual.

[0042] Step 7: Determine if the termination condition is met. If it is, exit the loop; otherwise, return to step 3.

[0043] Step 8: Output the optimal position and optimal fitness value.

[0044] However, the Gorilla Squad Optimization Algorithm has several key shortcomings: In the search phase: the algorithm overemphasizes global search, potentially causing it to linger too long near known promising areas without timely in-depth exploration, thus slowing down convergence. In the development phase, the search sub-phase is entirely dominated by the silverback gorillas' positions. All gorillas are closely clustered around a single silverback, resulting in a very small perturbation range. Once a silverback falls into a local optimum, the population is almost unable to escape, forced to search only within a very small suboptimal region, lacking a mechanism for proactively discovering new directions. In the development phase, the competition sub-phase is overly sensitive to the silverback's position. If a silverback falls into a local optimum, the entire population is quickly pulled towards a suboptimal solution, leading to a sharp decline in diversity and making the algorithm prone to local optima from which it struggles to escape.

[0045] Therefore, an improved Gorilla Troops Optimization Algorithm (IGTOA) method for runoff and sediment prediction is proposed. Specifically, the position update mechanism of the Spotted Hyena Optimization Algorithm and the Gradient Optimization Algorithm is introduced into the position update mechanism. The Spotted Hyena Optimization Algorithm and the Gradient Optimization Algorithm are introduced into the search sub-stage and the competitive search sub-stage of the development phase.

[0046] In the GTOA search phase, the spotted hyena optimization algorithm and gradient optimization algorithm are introduced to improve the gorilla individual position update method. The improved algorithm comprehensively considers factors such as the single position update mode, the position of the silverback gorilla, the current position of the individual gorilla, the current positions of other gorillas in the population, the convergence factor of the spotted hyena, the control factor of the spotted hyena, and the adaptive parameters of the gradient to update the individual gorilla position, thereby improving the algorithm's convergence speed and saving computational resources. The improved formula for updating the gorilla individual position is: (15); (16); (17); (18); (19); Where GX(t+1) represents the candidate position of an individual gorilla in the next iteration, t represents the current iteration number, and X(t) represents the current position of each gorilla. r1, r2, and r3 are random values ​​in the range [0,1] that are updated in each iteration. X r (t) and GX r(t) represents the current position of a randomly selected gorilla from the entire population and the candidate position of a randomly selected gorilla, respectively; gbest is the position of the silverback gorilla; E is the convergence factor for the spotted hyena. A random number within the range [0,1] is a random number within [0,1]; h is the control factor of the spotted hyena, which decreases linearly with the number of iterations and takes values ​​in the range [0,5]. , For the adaptive parameters of the gradient, , The values ​​are 1.2 and 0.2 respectively.

[0047] In the search sub-phase of the GTOA development phase, the spotted hyena optimization algorithm and gradient optimization algorithm are introduced. The improved formula for updating the position of individual gorillas is as follows: (20); In the competitive sub-phase of the GTOA development process, the spotted hyena optimization algorithm and gradient optimization algorithm are introduced. The improved position update formula for individual gorillas is as follows: (twenty one).

[0048] Based on an improved gorilla force optimization algorithm, this invention provides a method for predicting runoff sediment content, such as... Figure 1 As shown, the specific steps include: Step 1: Use the raw data for runoff and sediment prediction (such as real-time and historical rainfall intensity, duration, flow, water level, etc.; vegetation index, land use type, soil moisture, digital elevation model (DEM) and derived topographic factors such as slope and aspect obtained by satellite or drone; measured values ​​of runoff and sediment content in historical periods; auxiliary water quality parameters such as turbidity and conductivity obtained by automatic monitoring stations) to construct a dataset, and divide the dataset into a training dataset and a test dataset.

[0049] Step 2: Establish the objective function for runoff and sediment prediction based on the improved gorilla army optimization algorithm. Options include: Mean Absolute Error (MAE), Mean Relative Error (MRE), and Root Mean Square Error (RMSE). Here, minimizing MAE is taken as an example, and the objective function is: Where M is the length of the samples in the test dataset, To test the actual runoff sediment of sample i in the dataset, To test the predicted runoff sediment for sample i in the dataset.

[0050] At the same time, set the corresponding constraints: the number of hidden layer nodes k in the BiLSTM, the learning rate. Batch size Dropout rate The upper and lower limits.

[0051] Step 3: Set the parameters, which mainly include: population size (i.e., the number of individual gorillas); maximum number of iterations (i.e., the condition for stopping the iteration) T; lower boundary LoB for problem optimization; and upper boundary UpB for problem optimization.

[0052] Step 4: Calculate the individual fitness value of all individuals in the population according to the objective function, and select the optimal individual.

[0053] Step 5: Update C, F, L, H.

[0054] Step 6: Individuals begin exploring behavior based on formula (15).

[0055] Step 7: If C≥w, proceed to the first stage (search sub-stage), choose to follow the silverback gorilla, and update the position using equation (20); otherwise, proceed to the second stage (competition sub-stage), choose to fight other male gorillas, and update the position using equation (21).

[0056] Step 8: Calculate the individual fitness value and select the optimal individual.

[0057] Step 9: Determine if the termination condition is met. If it is, exit the loop; otherwise, return to step 5.

[0058] Step 10: Output the optimal position, which is the number of hidden layer nodes for obtaining the optimal parameters of BiLSTM. Learning rate Batch size Dropout rate .

[0059] Step 11: Using the optimal parameters output as the model parameters for the BiLSTM, train the optimized BiLSTM model using the training dataset to obtain the runoff and sediment prediction model. Input the test dataset into the model to obtain the runoff and sediment content prediction results and prediction accuracy.

[0060] To verify the effectiveness of the above-mentioned sediment content prediction method, 2000 sets of runoff sediment prediction samples were selected, with 1600 sets randomly used as training samples and the remaining 400 sets as test samples. GTOA-BiLSTM and IGTOA-BiLSTM were used to predict runoff sediment content. MATLAB was used as the simulation platform. The parameters in the GTOA algorithm were: nPop=50, T=200, number of hidden layer nodes k, and learning rate. Batch size Dropout rate The search range is between -100 and 100; the parameters in the IGTOA algorithm are: nPop=50, T=200, number of hidden layer nodes k, and learning rate. Batch size Dropout rate The search range is between -100 and 100.

[0061] The evaluation metrics for the GTOA-BiLSTM and IGTOA-BiLSTM models can be selected as: Mean Absolute Error (MAE), Mean Relative Error (MRE), and Root Mean Square Error (RMSE). As shown in Table 1, compared with GTOA-BiLSTM, IGTOA-BiLSTM has a higher accuracy in predicting runoff and sediment. In other words, the BiLSTM parameters obtained by IGTOA search are better than those obtained by GTOA search. Simulation results show that the IGTOA algorithm has stronger search capabilities than the GTOA algorithm, and IGTOA-BiLSTM has a higher evaluation accuracy than GTOA-BiLSTM, verifying the effectiveness of the method.

[0062] Table 1. Comparison of Prediction Methods

[0063] The present invention also provides a runoff sediment content prediction system, comprising: The algorithm improvement module is used to introduce the position update mechanisms of the spotted hyena optimization algorithm and the gradient optimization algorithm into the position update mechanism of the original gorilla force optimization algorithm, so as to obtain the improved gorilla force optimization algorithm.

[0064] The model optimization module optimizes the BiLSTM model parameters using an improved gorilla army optimization algorithm, resulting in a parameter-optimized BiLSTM model. Specifically, during the search phase, the convergence and control factors of the spotted hyenas are used to adjust individuals to move closer to the optimal individual, and gradient adaptive parameters guide the search direction for global exploration and position updates. In the development phase's search sub-phase, individuals form a variable search radius around the optimal individual, and gradient adaptive parameters adjust the step size of individuals moving towards the optimal individual, while the positions of other individuals in the population are introduced as correction terms to update individual candidate positions. In the development phase's competition sub-phase, the convergence and control factors of the spotted hyenas are used to control individuals to dynamically surround the optimal individual, gradient adaptive parameters dynamically adjust the gap between individuals and the optimal individual, and the positions of other individuals in the population are randomly introduced as competition to update individual candidate positions. The population refers to n sets of BiLSTM model parameters, with one set of BiLSTM model parameters corresponding to one individual.

[0065] The sediment prediction module is used to acquire geographic environmental data, rainfall and meteorological data, and historical runoff sediment content measured values ​​to construct a dataset. Using the geographic environmental data and rainfall and meteorological data in the dataset as inputs and the historical runoff sediment content measured values ​​as output targets, the BiLSTM model with optimized parameters is trained to obtain the runoff sediment prediction model.

[0066] The modules in the aforementioned runoff sediment content prediction system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0067] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps in an embodiment of a runoff sediment content prediction method. Specific implementation methods can be found in the method embodiments, and will not be repeated here.

[0068] Furthermore, the present invention also provides a non-transitory computer-readable storage medium containing instructions on which a computer program is stored. For example, a memory containing instructions that can be executed by a processor of a computer device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. When the computer program is executed by the processor, it can implement the steps in an embodiment of a runoff sediment content prediction method. Specific implementation methods can be found in the method embodiments, which will not be repeated here.

[0069] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0073] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the present invention patent. No reference numerals in the claims should be construed as limiting the scope of the claims. Any simple variations or equivalent substitutions of technical solutions that can be readily obtained by those skilled in the art within the scope of the technology disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method of predicting runoff sediment concentration, characterized by, Includes the following steps: By introducing the position update mechanisms of the spotted hyena optimization algorithm and the gradient optimization algorithm into the original gorilla force optimization algorithm, an improved gorilla force optimization algorithm is obtained. The improved gorilla troop optimization algorithm is used to optimize the parameters of the BiLSTM model, resulting in a parameter-optimized BiLSTM model. During the search phase, the convergence factor and control factor of the spotted hyenas are used to adjust individuals towards the optimal individual, and gradient adaptive parameters are used to guide the search direction for global exploration and position updates. The position update rules are as follows: ； where GX(t+1) represents the candidate position of gorilla individual in the next iteration; t represents the current iteration number; X(t) represents the current position of each gorilla; r1, r2 and r3 are random values in the range of [0, 1] updated in each iteration; X r (t) and GX r (t) represent the current position of a randomly selected gorilla from the entire population and the candidate position of a randomly selected gorilla, respectively; gbest is the position of the optimal individual, i.e., the position of the silverback gorilla; is the control factor of spotted hyena; and are adaptive parameters of gradient; E is the convergence factor of spotted hyena; C, L and H are algorithm control parameters; In the search sub-phase of the development phase, individuals form a variable search radius around the optimal individual. The step size of an individual moving towards the optimal individual is adjusted using gradient adaptive parameters, and the positions of other individuals in the population are introduced as correction terms to update the candidate positions of individuals. The position update rules are as follows: Where gbest is the location of the silverback gorilla, X r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively. and All are adaptive parameters of the gradient, t represents the current iteration number, E is the convergence factor of the spotted hyena, and h is the control factor of the spotted hyena; In the competitive sub-phase of the development phase, the convergence factor and control factor of the spotted hyena are used to control the dynamic surrounding of the optimal individual by individuals. The gradient adaptive parameter is used to dynamically adjust the gap between individuals and the optimal individual, and the positions of other individuals in the population are randomly introduced as competition to update the candidate positions of individuals. The specific position update rules are as follows: ； Where Q represents combat power, A is combat coefficient, gbest is the position of the silverback gorilla, and X... r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively. and All are adaptive parameters of the gradient, t represents the current iteration number, E is the convergence factor of the spotted hyena, and h is the control factor of the spotted hyena; the population refers to n sets of BiLSTM model parameters, with one set of BiLSTM model parameters corresponding to one individual. A dataset is constructed by acquiring geographic environmental data, rainfall and meteorological data, and historical measured values ​​of runoff sediment content. Using the geographic environmental data and rainfall and meteorological data in the dataset as inputs and the historical measured values ​​of runoff sediment content as output targets, a BiLSTM model with optimized parameters is trained to obtain a runoff sediment prediction model. The dataset specifically includes real-time and historical rainfall intensity, duration, flow rate, water level, vegetation index, land use type, soil moisture, water turbidity, water conductivity, digital elevation model (DEM), and derived slope and aspect.

2. The method for predicting runoff sediment content according to claim 1, characterized in that, It also includes obtaining real-time geographic environmental data and rainfall and meteorological data and inputting them into the runoff and sediment prediction model to obtain the predicted value of runoff and sediment content.

3. The method for predicting runoff sediment content according to claim 1, characterized in that, Calculating the fitness value of an individual in the population involves decoding the current position of the individual to obtain the hyperparameters of the BiLSTM model. The BiLSTM model corresponding to these hyperparameters is then trained using the dataset parameters to minimize the deviation between the predicted value and the actual runoff and sediment value, thus constructing an error objective function. Based on this error objective function, the error value is calculated, and this error value is the fitness value of the individual. The error objective function can be selected from mean absolute error (MAE), mean relative error (MRE), or root mean square error (RMSE).

4. The method for predicting runoff sediment content according to claim 3, characterized in that, It also includes setting constraints on the objective function, specifically on the number of hidden layer nodes k and the learning rate of the BiLSTM model. Batch size Dropout rate Set upper and lower limits.

5. A runoff sediment content prediction system, characterized in that, include: The algorithm improvement module is used to introduce the position update mechanisms of the spotted hyena optimization algorithm and the gradient optimization algorithm into the position update mechanism of the original gorilla force optimization algorithm, so as to obtain the improved gorilla force optimization algorithm. The model optimization module is used to optimize the BiLSTM model parameters using the improved gorilla army optimization algorithm, resulting in a BiLSTM model with optimized parameters. Specifically, during the search phase, the convergence factor and control factor of the spotted hyenas are used to adjust individuals to move closer to the optimal individual, and gradient adaptive parameters are used to guide the search direction for global exploration and position updates. The position update rules are as follows: ； Where GX(t+1) represents the candidate position of the gorilla in the next iteration; t represents the current iteration number; X(t) represents the current position of each gorilla; r1, r2, and r3 are random values ​​in the range [0,1] updated in each iteration; X r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively; gbest is the position of the optimal individual, i.e., the position of the silverback gorilla. A controlling factor for spotted hyenas; and All are adaptive parameters of the gradient; E is the convergence factor of the spotted hyena algorithm; C, L, and H are algorithm control parameters. In the search sub-phase of the development phase, individuals form a variable search radius around the optimal individual. The step size of an individual moving towards the optimal individual is adjusted using gradient adaptive parameters, and the positions of other individuals in the population are introduced as correction terms to update the candidate positions of individuals. The position update rules are as follows: Where gbest is the location of the silverback gorilla, X r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively. and All are adaptive parameters of the gradient, t represents the current iteration number, E is the convergence factor of the spotted hyena, and h is the control factor of the spotted hyena; In the competitive sub-phase of the development phase, the convergence factor and control factor of the spotted hyena are used to control the dynamic surrounding of the optimal individual by individuals. The gradient adaptive parameter is used to dynamically adjust the gap between individuals and the optimal individual, and the positions of other individuals in the population are randomly introduced as competition to update the candidate positions of individuals. The specific position update rules are as follows: ； Where Q represents combat power, A is combat coefficient, gbest is the position of the silverback gorilla, and X... r (t) and GX r (t) represents the current position of a gorilla randomly selected from the entire population and the candidate position of a gorilla randomly selected, respectively. and All are adaptive parameters of the gradient, t represents the current iteration number, E is the convergence factor of the spotted hyena, and h is the control factor of the spotted hyena; the population refers to n sets of BiLSTM model parameters, with one set of BiLSTM model parameters corresponding to one individual. The sediment prediction module is used to construct a dataset by acquiring geographical environmental data, rainfall and meteorological data, and historical measured values ​​of runoff sediment content. Using the geographical environmental data and rainfall and meteorological data in the dataset as inputs and the historical measured values ​​of runoff sediment content as output targets, the BiLSTM model with optimized parameters is trained to obtain the runoff sediment prediction model. The dataset specifically includes real-time and historical rainfall intensity, duration, flow rate, water level, vegetation index, land use type, soil moisture, water turbidity, water conductivity, digital elevation model (DEM), and derived slope and aspect.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 4.

Images