NOA optimization-based island type fishway high-speed flow ratio prediction method

Abstract

The invention relates to an island type fishway high-speed flow ratio prediction method based on NOA optimization. According to the method, a BP neural network model is established according to the problem that optimal parameters are difficult to determine due to the fact that the flow characteristics in the island type fishway are different in change and trend influenced by geometric parameters, and the BP neural network is optimized by using a genetic algorithm. Compared with the prior art, the prediction average error of the optimized neural network is reduced, so that the BP neural network optimized by the genetic algorithm is higher in prediction accuracy and better in effect, and a foundation is laid for optimization of a subsequent structure. And performing parameter optimization on the model predicted by the GA-BP neural network by using an NOA algorithm, performing three-dimensional modeling on geometric parameters obtained by optimization, and importing ANSYS Fluent simulation software to obtain an optimized geometric parameter combination fishway high-speed flow ratio result value.

Description

Technical Field

[0001] The invention relates to a method for predicting high-speed flow ratio of an island fishway based on NOA optimization, and in particular to the technical field of water conservancy equipment. Technical Background

[0002] The island fishway is a new type of fishway that introduces the Tesla valve membrane structure and combines it with the classic fishway pool to form a barrier island-valve multi-stage energy dissipation. The unique island-valve structure in the fishway not only reduces the upstream flow velocity step by step, but also causes a large area of ​​high flow velocity and high turbulent kinetic energy in front of the valve and the front half of the pool. In the actual research process of the fishway, the selection and optimization of geometric parameters are important factors that directly affect the distribution of key influencing factors such as flow velocity and turbulent kinetic energy. By adjusting the geometric parameters such as island length, island width, pool pseudo-vertical slits and island distance, reducing the proportion of high-speed area, it is possible to optimize the selectivity of the migration route and thus increase the success rate of fish migration, help upstream and downstream fish species to communicate, and thus achieve the purpose of protecting ecological diversity. It has great practical significance for achieving the optimization of the fishway model structure. The traditional fishway optimization method uses manual adjustment to further optimize the internal flow characteristics. This method consumes a lot of energy and is difficult to achieve the optimization of parameter traversal, and cannot maximize the structural optimization effect. Therefore, a high-speed flow ratio prediction method for island fishway based on NOA optimization is proposed to solve this problem. Summary of the invention

[0003] The present invention provides a method for predicting the high-speed flow ratio of an island fishway based on NOA optimization. The method can accurately predict the flow parameters under different structures, and at the same time, optimize and adjust the geometric parameters based on the NOA algorithm. The method is of great significance for more conveniently obtaining the corresponding flow characteristic parameters directly with the help of hydraulic geometric parameters, and for the refined research of subsequent structural adjustment. In addition, the introduction of the method reduces the experimental cost and the amount of numerical simulation calculation, and reduces the waste of manpower and material resources.

[0004] The present invention provides a method for predicting high-speed flow ratio of an island fishway based on NOA optimization, and the specific steps are as follows:

[0005] Step S1: Collect initial data for numerical simulation model establishment. By collecting and recording physical model test data, provide boundary and initial conditions for the subsequent numerical model establishment, and ensure and evaluate the rationality and accuracy of simulation data.

[0006] Step S2: construct CFD models under different geometric parameters to obtain a training data set of high-speed flow ratio in the pool chamber.

[0007] Step S3: The data set obtained in step S2 is randomly divided into a training set and a test set as a training sample set, and data normalization is performed.

[0008] Step S4: construct a BP neural network, using the data set obtained in step S2 as input values, island length factor, island width factor, pool chamber pseudo-vertical seam factor and island distance factor, and the pool chamber high-speed flow ratio as an evaluation index to construct a BP neural network structure.

[0009] Step S5: The GA algorithm optimizes the initialization weights and thresholds of the BP neural network, substitutes the optimal initial weights and thresholds into the BP neural network for prediction, and obtains a GA-BP prediction model based on genetic-neural network.

[0009] Step S6, after obtaining the high-speed flow ratio prediction model of the pool chamber based on GA-BP, the trained pool chamber high-speed flow ratio prediction model is optimized using the Nudibranch algorithm to obtain the optimal geometric parameter combination of island length factor, island width factor, pool chamber pseudo-vertical seam factor and island distance factor.

[0010] Step S7, constructing a three-dimensional model of the optimal parameter combination obtained by optimization, and importing it into simulation software for numerical calculation to obtain the result of the high-speed flow ratio of the optimized structure fishway.

[0011] Furthermore, the initial data collected in step S1 mainly includes flow rate data and water level data.

[0012] Furthermore, the step S2 includes: S2.1, selecting geometric parameters of island length factor, island width factor, pool chamber pseudo-vertical seam factor and island distance factor; S2.2, taking the experimental conditions, flow velocity data and water level data of the above step S1 as initial data, taking the flow velocity at the measuring point in the pool chamber and the water level along the way as evaluation indicators, selecting the appropriate number of grids and turbulence model to construct a CFD simulation model of flow characteristics in the fishway; S2.3, using the single factor evolution analysis method to determine the value range of each parameter; S2.4, using the Latin hypercube sampling method to sample, obtain multiple sample points, each sample point is a set of geometric parameters, and obtain CFD model pool chamber high-speed flow ratio data with different parameters; S2.5, the sample point parameters obtained by sampling and the CFD model pool chamber high-speed flow ratio data are a set of training samples to obtain a sample set containing multiple training samples.

[0013] Furthermore, the step S5 includes: S5.1, initializing the population and selecting the population size as N; S5.2, using the inverse of the mean square error function MSE of the BP neural network as the fitness function of the genetic algorithm; S5.3, obtaining the neural network optimization weights and thresholds through selection operations, crossover operations and mutation operations; S5.4, cyclically performing selection, crossover and mutation operations and judging whether the fitness value reaches the target value or whether the number of iterations reaches the maximum number, to obtain a neural network model based on GA-BP.

[0014] Furthermore, the step S6 includes: S6.1, taking minimization of the high-speed flow ratio in the pool chamber as the optimization goal, establishing an optimization objective function between the optimization variable and the target optimization parameter; S6.2, setting the geometric parameter value range and the nutcracker search space dimension according to the data obtained in step S2; S6.3, setting the population size and number of iterations of the NOA algorithm; S6.4, population initialization; S6.5, generating random numbers σ and σ1 between 0 and 1, and judging whether to enter the first stage of exploration and development based on the random numbers, if the former is smaller, then enter, otherwise enter the second stage of exploration and development; S6.6, stopping the operation when the maximum number of iterations is reached and comparing the current fitness with the existing best fitness, updating the best fitness and the corresponding position (geometric parameter combination).

[0015] Due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0016] The BP network model optimized by GA can effectively overcome the nonlinear coupling relationship between geometric parameters and the high-speed flow ratio in the pool chamber, and the predicted output values ​​generated by the BP neural network model optimized by the GA algorithm are more consistent with the simulation values. Therefore, the optimized BP neural network has higher prediction accuracy.

[0017] By using NOA to optimize the geometric parameters of the model predicted by the GA-BP neural network, the best combination of geometric parameters can be determined, further improving the efficiency and quality of model optimization.

[0018] After optimization by the method described in the above invention, the high-speed flow ratio in the pool chamber was relatively reduced by 88.55%. In addition, the maximum flow rate in the pool chamber was also reduced by 7.15%, which to a certain extent achieved the effect of optimizing the selectivity of the migratory route, thereby increasing the success rate of fish migration and further improving the efficiency of the fishway operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] When considered in conjunction with the accompanying drawings, the advantages of the present invention can be better understood by referring to the following detailed description, but the drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention, wherein:

[0020] Figure 1 It is an overall flow chart of the method for predicting the high-speed flow ratio in a pool room proposed by the present invention;

[0021] Figure 2 It is a flow chart of the GA-BP neural network proposed by the present invention;

[0022] Figure 3 It is a schematic diagram of the island fishway structure;

[0023] Figure 4 This is a schematic diagram of the distribution test of flow velocity measurement points in the pool room;

[0024] Figure 5 This is a flow chart of the optimization of the Star Crow algorithm of the present invention; DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The following examples are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0026] like Figure 1 As shown, the present invention provides a method for predicting high-speed flow ratio of island fishway based on NOA optimization, and its specific steps are as follows:

[0027] Step S1: Collect initial data for numerical simulation model establishment. By collecting and recording physical model test data, provide boundary and initial conditions for the subsequent numerical model establishment, and ensure and evaluate the rationality and accuracy of simulation data.

[0028] Step S2: construct CFD models under different geometric parameters to obtain a training data set of high-speed flow ratio in the pool chamber.

[0029] Step S3: The data set obtained in step S2 is randomly divided into a training set and a test set as a training sample set, and data normalization is performed.

[0030] Step S4: construct a BP neural network, using the data set obtained in step S2 as input values, island length factor, island width factor, pool chamber pseudo-vertical seam factor and island distance factor, and the pool chamber high-speed flow ratio as an evaluation index to construct a BP neural network structure.

[0031] Step S5: The GA algorithm optimizes the initialization weights and thresholds of the BP neural network, substitutes the optimal initial weights and thresholds into the BP neural network for prediction, and obtains a GA-BP prediction model based on genetic-neural network.

[0032] Step S6: After obtaining the indoor high-speed flow ratio prediction model based on GA-BP, the trained indoor high-speed flow ratio prediction model is optimized using the Nudibranch algorithm to obtain the optimal parameter combination of island length factor, island width factor, chamber pseudo-vertical seam factor and island distance factor based on NOA optimization.

[0033] Step S7: construct a three-dimensional model of the optimal parameter combination obtained by optimization, and import it into simulation software for numerical calculation to obtain the high-speed flow ratio of the optimized structure fishway.

[0034] Example

[0035] In the specific implementation process, the specific implementation steps of the present invention are as follows:

[0036] The step S1 specifically includes the following steps:

[0037] S1.1. Use an open channel flowmeter to measure the radial velocity at the inlet and the flow velocity design measurement points in the pool chamber, and record the velocity data at each point. The pool chamber measurement points include the first-level pool chamber and the first half of the second-level pool chamber to better understand the flow velocity changes in the pool chamber and in front of the valve. The detailed pool chamber measurement point design is as follows: Figure 4 shown.

[0038] S1.2. The water level along the fishway changes rapidly and is difficult to capture. In order to minimize the reading error, a high-speed camera is used to capture the water level changes at the entrance and various water points along the way, and the average water level value of the frame data is read and calculated, and the data is recorded, laying the foundation for the subsequent construction and evaluation of the numerical model.

[0039] The step S2 specifically includes the following steps:

[0040] S2.1. Select the island length factor, island width factor, pseudo vertical seam factor and island distance factor geometric parameters as the target variables. The factors are shown as follows: Figure 3 As shown: (1): island length, (2): island width, (3): pseudo vertical seam, (4): island distance.

[0041] S2.2. Taking the experimental conditions, flow velocity data and water level data of step S1 as the initial data, the flow velocity at the measuring points in the pool and the water level along the way as the evaluation indicators, select the appropriate number of grids and turbulence model to construct a CFD simulation model of the flow characteristics in the fishway to ensure the validity of the simulation results.

[0042] S2.3. Use the single factor evolution analysis method for the geometric parameters selected in step S2.1, establish the CFD simulation model of the value range of each geometric parameter, analyze the simulation results to determine the value range of each geometric parameter, and obtain the value ranges of these four factors (unit: mm): island length value range [30,40], island width value range [15,30], pseudo vertical seam value range [0,2], island distance value range [30,40].

[0043] S2.4, in multi-objective optimization and design space exploration, in order to ensure the uniform distribution of samples in multi-dimensional space and the extraction efficiency, this embodiment adopts Latin hypercube sampling for sample sampling. The number of sample points is based on the BP neural network training sample requirements: ten to fifteen times the input variable, and the number of sample libraries is finally selected as 55 samples. The high-speed flow ratio in the pool room of each CFD model is obtained by numerical simulation calculation method, and finally a sample set containing multiple training samples is obtained.

[0044] The step S3 comprises:

[0045] S3.1. Randomly divide the data set obtained in step S2 into a training set and a test set.

[0046] S3.2. Perform data normalization on the data set divided in step S3.1.

[0047] The step S4 comprises:

[0048] S4.1. Set the number of neurons in the input layer m=4, corresponding to the four input features of island length factor, island width factor, chamber pseudo-vertical slit factor and island distance factor; set the number of neurons in the output layer n=1, corresponding to the high-speed flow ratio output feature.

[0049] S4.2. Determine the optimal number of nodes in the hidden layer of the BP neural network. The hidden layer is determined according to the following formula:

[0050]

[0051] Where m1 is the number of nodes in the hidden layer, n is the number of neurons at the input end, m is the number of neurons at the output end, and r is a constant between 1 and 10.

[0052] S4.3. Error function selects mean square error function MSE:

[0053]

[0054] Where N is the number of training samples, y i is the expected output, y i is the neural network prediction output.

[0055] S4.4. Set training parameters: maximum number of iterations: 1000; error threshold: 0.000001; learning rate: 0.01.

[0056] According to Figure 2 The flowchart of the GA-BP neural network shown in FIG. 1 , wherein the GA optimization BP neural network prediction model in step S5 specifically includes the following steps:

[0057] S5.1. Initialize the population and select the population size as N.

[0058] S5.2. Determine the fitness function. Take the inverse of the mean square error function MSE of the BP neural network as the fitness function of the genetic algorithm. The function formula is as follows:

[0059]

[0060] S5.3. Selection operation: Roulette method is used for selection operation to select sub-populations, thereby retaining excellent individuals; Crossover operation: two individuals are selected, and part of the genes of the individuals are selected for exchange according to the crossover probability to generate new individuals to expand the solution space; Mutation operation: New individuals are generated by randomly perturbing the genes of individuals to avoid falling into local optimal solutions. The generation of new individuals adopts adaptive mutation operators.

[0061] S5.4, loop step S5.3, stop the operation when the fitness value reaches the target value or reaches the maximum number of iterations, and assign the network initial weight and threshold corresponding to the optimal individual to the BP neural network to obtain a neural network model based on GA-BP.

[0062] The comparison table of evaluation indexes of indoor high-speed flow ratio in the pool predicted by BP neural network and GA-BP neural network model is shown in Table 1.

[0063] Table 1 Comparison of evaluation indexes of high-speed flow ratio in pool chamber

[0064] From the data comparison in Table 1, it can be seen that the BP neural network after GA optimization has a higher determination coefficient and a lower prediction error, which indicates that the prediction accuracy of the optimized GA-BP neural network is greater than that of the BP neural network, and it is more suitable for the prediction of high-speed flow ratio.

[0065] The optimal parameter combination optimization based on NOA-GA-BP in step S6 specifically includes the following steps:

[0066] Step S6.1, population initialization, the prediction model network obtained by the GA-BP algorithm is used as the problem search space solution set of the NOA algorithm, and the population initialization formula is as follows:

[0067]

[0068] Where, i = 1, 2, ..., N; j = 1, 2, ..., D; t represents the number of iterations; represents the j-dimensional vector of the i-th nutcracker at the current iteration number; U j and L j They represent the upper and lower bounds of the j-dimensional variable respectively; represents a random vector between [0,1].

[0069] S6.2, foraging strategy phase: the first phase of exploration (Exploration phase 1). In this phase, the nutcracker begins to take the initial position generated by the initialization formula in S3.1 in the search space, that is, to find the best position in the problem search space. The process position update formula is as follows:

[0070]

[0071] in, Indicates the latest position of the i-th starling at the current iteration number; represents the position of the i-th nutcracker at the t-th iteration; γ is a random number generated according to the Levy flight; A, B, and C are three different indices randomly selected from the population to promote the exploration of high-quality food sources; τ1, τ2, and r are random real numbers in the range [0,1]; Represents the mean of the jth dimension of all solutions in the current tth iteration; μ is a random number generated based on normal distribution (τ4), Levy flight (τ5) and between [0,1], as shown in the following formula:

[0072]

[0073] Where r1, r2 and r3 are random numbers in the range [0,1].

[0074] S6.3, Storage Strategy Phase: Exploitation Phase 1, in this phase, the Starlings begin to transport the better combinations obtained in the previous phase to temporary storage locations for storage. The formula is as follows:

[0075]

[0076] In the formula, is the new position in the Nutcracker storage area at the current iteration t, is the optimal position solution obtained so far; λ is a random number generated according to the Levy flight; τ3 is a random real number in the range [0,1]; l is a factor that linearly decreases from 1 to 0 in the NOA development behavior to diversify the NOA development behavior. and is the position of the randomly selected nutcracker individual in the tth iteration.

[0077] To maintain a balance between exploration and exploitation operators, Nutcrackers swap between the first foraging phase and storage using the following formula:

[0078]

[0079] in, is a random number in [0,1]; P a1 As the number of iterations decreases linearly from 1 to 0.

[0080] S6.4, cache search strategy phase: Exploration phase 2, in the process of solving the optimal parameter combination, the nutcracker uses the reference point as the signal of a single cache to determine the initial position. Here, it is assumed that each cache has only two objects, that is, two reference points (RPs). The RP generation formula is as follows:

[0081]

[0082] in, and They represent the two cached reference points of the i-th nutcracker in the current iteration t.

[0083] During the exploration process, the new position of the nutcracker is updated by the following formula:

[0084]

[0085]

[0086] The balance between the exploration behaviors of the first and second RPs is as follows:

[0087]

[0088] S6.5, Retrieval strategy phase: Exploitation phase 2. In the retrieval phase, there are two situations when the nutcracker searches for the cache: the first situation is that the nutcracker can remember the cache location through the first RP. At this time, there are two situations: the food exists and the food does not exist. The mathematical expression is as follows:

[0089]

[0090] The first state of formula (13) simulates the first possibility - food exists, which means that some ranges in this cache solution may also be applicable in the next iteration; the second state simulates the second possibility - food does not exist, and the solution area is not optimistic at this time, so the algorithm will use an escape mechanism to avoid falling into the local minimum. Among them, τ4 is a random number in the range of [0,1], and C is the index of a solution randomly selected from the population.

[0091] The second case is that the nutcracker does not remember the location of the food stored using the first RP, and then uses the second RP to find food. The update formula for the second RP is:

[0092]

[0093] Here, τ6 is a random number in the range [0,1].

[0094] Furthermore, the following formula is used to determine whether the nutcracker remembers the location where the first RP was used to store food:

[0095]

[0096] Among them, τ7 and τ8 are random numbers in the range of [0,1].

[0097] Furthermore, in order to maintain a balance between exploration and exploitation, the exchange between the cache search phase and the retrieval phase uses the following formula:

[0098]

[0099] Where φ is a random number between 0 and 1; P a2 Take 0.2; the above Star Crow algorithm optimization flow chart is as follows Figure 5 shown.

[0100] Table 2 Comparison of geometric parameters and results before and after optimization

[0101] The geometric parameter combination obtained by NOA optimization is: island length: 30mm, island width: 30mm, pseudo vertical seam: 0.2mm, island distance: 35mm. The optimal geometric parameter combination obtained by NOA optimization was used to build a three-dimensional model using SolidWorks and imported into ANSYS fluent simulation software for verification, and the optimization results based on NOA optimization with high-speed flow ratio as the optimization target were obtained. The comparison of geometric parameters and results before and after optimization is shown in Table 2. After optimization by the method described in the invention, the high-speed flow ratio in the pool chamber was relatively reduced by 88.55%. In addition, the maximum flow rate in the pool chamber was also reduced by 7.15%. The optimized structure can improve the operation efficiency of the fishway to a certain extent.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for predicting high-speed flow ratio of island fishway based on NOA optimization, characterized by , including the following steps: Step S1: Collect initial data for numerical simulation model establishment. By collecting and recording physical model test data, provide boundary and initial conditions for the subsequent numerical model establishment, and ensure and evaluate the rationality and accuracy of simulation data. Step S2: construct CFD models under different geometric parameters to obtain a training data set of high-speed flow ratio in the pool chamber. Step S3: The data set obtained in step S2 is randomly divided into a training set and a test set as a training sample set, and data normalization is performed. Step S4: construct a BP neural network, using the data set obtained in step 2 as input values ​​of island length factor, island width factor, chamber pseudo-vertical seam factor and island distance factor, and high-speed flow ratio as an evaluation index to construct a BP neural network structure. Step S5: The GA algorithm optimizes the initialization weights and thresholds of the BP neural network, substitutes the optimal initial weights and thresholds into the BP neural network for prediction, and obtains a GA-BP prediction model based on genetic-neural network. Step S6: After obtaining the high-speed flow ratio prediction model based on GA-BP, the trained high-speed flow ratio prediction model is optimized using the Nudibranch algorithm to obtain the optimal parameter combination of island length factor, island width factor, chamber pseudo-vertical slit factor and island distance factor based on NOA optimization. Step S7: construct a three-dimensional model of the optimal parameter combination obtained by optimization, and import it into simulation software for numerical calculation to obtain the high-speed flow ratio of the optimized structure fishway.

2. The method for predicting high-speed flow ratio of an island fishway based on NOA optimization according to claim 1, characterized in that: Step S1 specifically includes the following steps: S1.

1. Use an open channel flowmeter to measure the radial flow velocity at the inlet and the chamber flow velocity design measurement points, and record the flow velocity data at each point. The chamber measurement points include the first-level chamber and the first half of the second-level chamber to better understand the flow velocity changes in the chamber and in front of the valve. The detailed chamber measurement point design is shown in Figure 3. S1.

2. The water level along the fishway changes rapidly and is difficult to capture. In order to minimize the reading error, a high-speed camera is used to shoot the water level changes at the entrance and various water points along the way, and the average water level value of the frame data is read and calculated, and the data is recorded, laying the foundation for the subsequent construction and evaluation of the numerical model.

3. The method for predicting high-speed flow ratio of island fishway based on NOA optimization according to claim 1, characterized in that: Step S2 specifically includes the following steps: S2.1, select the geometric parameters of island length factor, island width factor, pseudo vertical seam factor and island distance factor as target variables; S2.

2. Taking the experimental conditions, flow velocity data and water level data of step S1 as the initial data, the flow velocity at the measuring points in the pool and the water level along the way as the evaluation indicators, select the appropriate number of grids and turbulence model to construct a CFD simulation model of the flow characteristics in the fishway to ensure the validity of the simulation results. S2.

3. Use the single factor evolution analysis method for the geometric parameters selected in step S2.1, establish the CFD simulation model of the value range of each geometric parameter, analyze the simulation results to determine the value range of each geometric parameter, and obtain the value ranges of these four factors (unit: mm): island length value range [30,40], island width value range [15,30], pseudo vertical seam value range [0,2], island distance value range [30,40]. S2.4, in multi-objective optimization and design space exploration, in order to ensure the uniform distribution of samples in multi-dimensional space and the extraction efficiency, the Latin hypercube sampling method is used for sample sampling. The number of sample points is based on the BP neural network training sample requirements: ten to fifteen times the input variable, and the number of sample libraries is finally selected as 55 samples. The high-speed flow ratio in the pool chamber of each CFD model is obtained by numerical simulation calculation method, and finally a sample set containing multiple training samples is obtained.

4. The method for predicting high-speed flow ratio of an island fishway based on NOA optimization according to claim 1 is characterized by: Step S4 The specific steps include: S4.

1. Set the number of neurons in the input layer m=4, corresponding to the four input features of island length factor, island width factor, chamber pseudo-vertical slit factor and island distance factor; set the number of neurons in the output layer n=1, corresponding to the high-speed flow ratio output feature. S4.

2. Determine the optimal number of nodes in the hidden layer of the BP neural network. S4.

3. The error function is the mean square error function MSE. S4.

4. Set training parameters: maximum number of iterations: 1000; error threshold: 0.000001; learning rate: 0.

01.

5. The method for predicting high-speed flow ratio of an island fishway based on NOA optimization according to claim 1 is characterized by: Step S5 specifically includes the following steps: S5.

1. Initialize the population and select the population size as N. S5.

2. Determine the fitness function and take the inverse of the mean square error function MSE of the BP neural network as the fitness function of the genetic algorithm. S5.

3. Selection operation: Roulette method is used for selection operation to select sub-populations, thereby retaining excellent individuals; Crossover operation: two individuals are selected, and part of the genes of the individuals are selected for exchange according to the crossover probability to generate new individuals to expand the solution space; Mutation operation: New individuals are generated by randomly perturbing the genes of individuals to avoid falling into local optimal solutions. The generation of new individuals adopts adaptive mutation operators. S5.4, loop step S5.3, stop the operation when the fitness value reaches the target value or reaches the maximum number of iterations, and assign the network initial weight and threshold corresponding to the optimal individual to the BP neural network to obtain a neural network model based on GA-BP.

6. The method for predicting high-speed flow ratio of an island fishway based on NOA optimization according to claim 1 is characterized by: Step S6 specifically includes the following steps: Step S6.1 Population initialization: The prediction model network obtained by the GA-BP algorithm is used as the problem search space solution set of the NOA algorithm. S6.2 Foraging strategy phase: Exploration phase 1, taking the initial position generated by the initialization formula in the search space. S6.3 Storage strategy phase: In the first phase of development (Exploitation phase 1), the Starling transports the better combination obtained in the previous phase to a temporary storage location for storage. S6.4 Cache search strategy phase: Exploration phase 2. In the process of solving the optimal parameter combination, the nutcracker uses the reference point as the signal of a single cache to determine the initial position. Here, it is assumed that each cache has only two objects, namely two reference points (RPs). S6.5 Retrieval strategy phase: Exploitation phase 2. In the retrieval phase, there are two situations when the nutcrackers search for caches: the first situation is that the nutcrackers can remember the cache location through the first RP, which is divided into two situations: the food exists and the food does not exist. The second situation is that the nutcrackers do not remember the location of the food stored using the first RP, and then use the second RP to find food.

Images