Axial-flow Kaplan turbine high-precision modeling method based on seagull optimization algorithm and BP neural network

Abstract

The invention discloses a high-precision modeling method for an axial-flow Kaplan turbine based on a seagull optimization algorithm and a BP (Back Propagation) neural network. The high-precision modeling method comprises the following steps: acquiring basic modeling data of the axial-flow Kaplan turbine; the opening degree characteristic and efficiency characteristic data of the axial-flow Kaplan turbine are supplemented; constructing a flow characteristic and torque characteristic data set of the axial-flow Kaplan turbine; correcting the guide vane opening degree in the flow characteristic and torque characteristic data set of the axial flow Kaplan turbine; constructing a guide vane and paddle joint characteristic model based on a seagull optimization algorithm and a BP neural network; and in combination with the corrected data set, the BP neural network and a seagull optimization algorithm, reconstructing a flow characteristic and torque characteristic neural network model of the axial-flow Kaplan turbine. According to the axial-flow movable propeller water turbine modeling method considering the modeling data quality and the guide vane and paddle joint characteristics, a high-precision and nonlinear axial-flow movable propeller water turbine model can be obtained, and a model basis is provided for axial-flow movable propeller water turbine unit control strategy research.

Description

Technical Field

[0001] This invention relates to the field of power engineering technology, and in particular to a high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network. Background Technology

[0002] Axial-flow propeller turbines play a crucial role in hydropower systems, and accurate modeling of their performance and operating status is essential for stable operation, optimized control, and performance evaluation of the entire hydropower system. However, traditional modeling methods for axial-flow propeller turbines have many limitations.

[0003] On the one hand, traditional modeling methods often rely on simple physical equations and empirical formulas. These models can describe the basic operating characteristics of water turbines to a certain extent, but they are difficult to accurately characterize complex nonlinear relationships and multivariate interactions. For example, under different operating conditions, the influence of guide vane opening and blade angle on water turbine flow and torque has complex nonlinear characteristics. Traditional linear or approximately linear models cannot fully capture these complex changing patterns, resulting in limited model accuracy.

[0004] On the other hand, with the continuous changes in turbine operating time and environment, the quality and completeness of modeling data also face challenges. In actual modeling processes, obtaining comprehensive and accurate basic data for axial-flow propeller turbines is difficult. Particularly regarding opening and efficiency characteristics, issues such as missing data, noise interference, and insufficient data coverage may exist, further affecting the reliability and accuracy of traditional modeling methods.

[0005] Furthermore, there is a close synergistic relationship between the guide vanes and blades of an axial-flow propeller turbine, which significantly affects the turbine's performance. However, previous studies have shown that many modeling methods have failed to fully consider this synergistic relationship or have only used simplistic approaches without delving into its complex dynamics, thus limiting the accuracy of the models in describing and predicting the actual operating conditions of the turbines.

[0006] In the application of optimization algorithms and neural networks, traditional methods for combining the Seagull Optimization Algorithm (SOP) and Backpropagation (BP) neural networks are not yet fully mature. While the Seagull Optimization Algorithm has certain advantages in global optimization, its combination with BP neural networks for modeling axial-flow propeller turbines still requires improvement in aspects such as parameter settings, network training strategies, and the collaborative mechanism between the two. Existing attempts often fail to fully leverage the potential of the Seagull Optimization Algorithm in optimizing neural network weights and structure, resulting in limited performance improvements and an inability to meet the requirements of high-precision modeling.

[0007] Therefore, current axial-flow propeller turbine modeling technology has many problems that need to be solved in terms of model accuracy, data processing, consideration of synergistic characteristics, and effective combination of optimization algorithms and neural networks. Therefore, a new high-precision modeling method that can overcome the above-mentioned shortcomings is needed. Summary of the Invention

[0008] The technical problem to be solved by this invention is to provide a high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network. By making reasonable use of the advantages of the Seagull optimization algorithm and BP neural network, and combining the optimization of modeling data quality and in-depth modeling of the cooperative characteristics of guide vanes and blades, the accuracy and applicability of the axial-flow propeller turbine model are improved, providing a more reliable model basis for the research of hydropower unit control strategies.

[0009] To achieve the above objectives, this application provides a high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network, comprising the following steps:

[0010] S1. Obtain basic data for modeling axial-flow propeller turbines; S2. Supplement the opening characteristics and efficiency characteristics data of the axial-flow propeller turbine, and calculate the unit flow rate of the turbine; S3. Construct a dataset of flow characteristics and torque characteristics for an axial-flow propeller turbine. S4. Based on the torque characteristic dataset, BP neural network and Seagull algorithm, correct the guide vane opening in the flow characteristic and torque characteristic dataset of axial flow propeller turbine. S5. Construct a cooperative characteristic model of guide vane and blade based on the Seagull optimization algorithm and BP neural network; S6. Combine the corrected dataset, BP neural network and Seagull optimization algorithm to reconstruct the neural network model of the flow characteristics and torque characteristics of the axial flow propeller turbine.

[0011] In S1, the basic data includes the basic structural parameters of the turbine, the comprehensive operating characteristic curve data, and the field dynamic test data.

[0012] The basic structural parameters of the turbine include: Runner geometry parameters: Runner diameter; Water guiding mechanism parameters: opening range; Water intake and discharge component parameters: Obtain the dimensions of the volute, tailpipe height and diffusion angle, and seat ring fixed guide vane dimensions based on the design drawings.

[0013] The comprehensive operational characteristic curve data includes: With head H as the abscissa and flow rate Q as the ordinate, the curves are: equal efficiency curve, equal guide vane opening curve, equal output curve, and equal blade angle curve. With output P as the abscissa and net head H as the ordinate: equal efficiency curve, equal guide vane opening curve, equal output curve, and equal blade rotation angle curve; A constant head curve with guide vane opening Y as the abscissa and blade rotation angle B as the ordinate.

[0014] The on-site dynamic test data includes: Dynamic test data from the hydropower station are acquired through high-precision sensors and data acquisition systems, including primary frequency regulation data and active power regulation data; the high-precision sensors include head pressure transmitters, flow meters, and torque meters.

[0015] In S2, a curve with output P as the abscissa and net head H as the ordinate is selected for modeling axial-flow propeller turbines. Data types on iso-efficiency curves include efficiency or Unit speed n 11 and unit torque M 11 ; The data on the constant opening curve includes the guide vane opening. Y Unit speed n 11 and unit torque M 11 ; A neural network model (ECNN) for the efficiency characteristics of a hydro turbine was established, with unit rotational speed and unit torque as inputs and efficiency as the output. or = or ( n 11 , M 11 );and A neural network model for turbine opening characteristics (OCNN) with unit rotational speed and unit torque as inputs and guide vane opening as output. Y = Y ( n 11 , M 11 ); Substitute the equal opening data into the turbine efficiency characteristic neural network model (ECNN) to calculate the corresponding guide vane opening Y; Substituting the equivalent efficiency data into the turbine opening characteristic neural network model (OCNN), the corresponding efficiency is calculated. or ; Then calculate the unit flow rate corresponding to the water turbine. Q 11 ;in, (1) (2) In the formula: , It is the density of water. D is the acceleration due to gravity; D is the diameter of the wheel; X is the rotational speed. It is a unit torque; It's about efficiency; It's a water purifier.

[0016] In S3, the following steps are taken to construct a dataset of flow characteristics and torque characteristics for an axial-flow propeller turbine: Merge the supplemented data on equal aperture and equal efficiency. Take unit speed n 11 Guide vane opening Y, unit flow rate Q 11 A sample dataset of flow characteristics of axial-flow propeller turbines was obtained. Take unit speed n 11 Guide vane opening Y, unit torque M 11 A sample dataset of torque characteristics of axial-flow propeller turbines was obtained.

[0017] In S4, based on the torque characteristic dataset, BP neural network, and Seagull algorithm, the guide vane opening Y in the flow characteristic and torque characteristic datasets of the axial-flow propeller turbine is corrected, including the following steps: Based on a BP neural network, a system is constructed with a unit rotational speed... n 11 And guide vane opening Y is the input, unit flow rate Q 11 and unit torque M 11 These are the output flow characteristic neural network model DCNN and torque characteristic neural network model TCNN, respectively. The initial weights and thresholds of the BP neural network, as well as the number of neurons in the hidden layer, are optimized using the Seagull optimization algorithm. Unit speed was obtained from the dynamic test data. n 1. Guide vane opening Y and unit torque M 11 Constructing a torque characteristic running dataset ( n 11 , Y , M 11 ); Running the dataset in conjunction with torque characteristics ( n 11 , Y , M 11The torque characteristic neural network model TCNN and the Seagull optimization algorithm were used to obtain the correction coefficient of guide vane opening Y in the sample dataset of flow torque and torque characteristics of axial flow propeller turbine in S3. The guide vane opening Y is corrected by incorporating a correction factor.

[0018] In S5, a cooperative characteristic model of guide vanes and propeller blades is constructed based on the Seagull optimization algorithm and BP neural network, including the following steps: Based on the guide vane and impeller coordination data and their boundary conditions from the field dynamic test data, BP neural network and Seagull optimization algorithm, a system based on the net water head is constructed. H and guide vane opening Y A neural network model of blade characteristics with input as the input and blade rotation angle B as the output; The initial weights and thresholds of the BP neural network, as well as the number of neurons in the hidden layer, are optimized using the Seagull optimization algorithm. The boundary conditions of the propeller data are considered in terms of synergy.

[0019] In S6, a dataset of modified axial-flow propeller turbine flow and torque characteristics, along with a BP neural network and the Seagull optimization algorithm, is constructed to measure the flow rate per unit rotational speed. n 11 Guide vane opening Y The blade rotation angle B is the input and the unit flow rate. Q 11 and unit torque M 11 The output of the axial-flow propeller turbine flow characteristic neural network model (RDCNN) is shown below. Q 11 = Q 11 ( n 11 , Y , B ()) and torque characteristic neural network model (RTCNN, M 11 = M 11 ( n 11 , Y , B )); Among them, the initial weights and threshold of the BP neural network, and the number of neurons in the hidden layer... N h The optimal value is obtained using the Seagull optimization algorithm; During parameter optimization, the number of hidden layer neurons N h The maximum value is limited by equation (7), and then the mean square error is expressed as shown in equation (8). MSETo obtain the optimal initial weights and thresholds, and the number of hidden layer neurons in the BP neural network for the target; (7) (8) In the formula, N i It is the number of neurons in the input layer; N o It is the number of neurons in the output layer; Z It is a constant between 0 and 10; k Indicates the number of sample data; This represents the predicted data; D i This represents sample data.

[0020] Compared with the prior art, the above-conceptual technical solution conceived in this application has the following beneficial effects: This invention effectively improves the accuracy of an axial-flow propeller turbine model by combining the Seagull optimization algorithm with a backpropagation (BP) neural network. During the modeling process, not only is the original data supplemented and improved, but key parameters such as guide vane opening are also corrected, thereby improving data quality and providing more reliable data support for model training. Simultaneously, the collaborative characteristics of the guide vanes and propeller blades are fully considered, enabling the model to more comprehensively reflect the actual operating state of the turbine. Furthermore, the combination of the Seagull optimization algorithm and the BP neural network ensures both global search capability and local optimization capability, improving the optimization efficiency of the model. The high-precision model constructed by this invention not only provides a solid theoretical foundation for the research of control strategies for axial-flow propeller turbines but also helps optimize the operating performance of hydropower units, improve the overall efficiency and stability of hydropower systems, and has significant practical application value. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0022] Figure 1 The method design steps of the present invention are described below.

[0023] Figure 2 This diagram illustrates the Seagull optimization algorithm of the present invention for optimizing the initial weights, thresholds, and number of hidden layer neurons in a BP neural network.

[0024] Figure 3 This is a schematic diagram of the axial-flow propeller turbine model of the present invention. Detailed Implementation

[0025] To more clearly illustrate the purpose, technical solution, and beneficial effects of this application, a further detailed description of this application is provided below in conjunction with illustrations and specific embodiments. It should be specifically noted that the specific embodiments described below are only for illustrating the technical content of this application and do not constitute a limitation on the scope of protection of this application.

[0026] Regarding the description of the embodiments: The terms "exemplary" and "for example" appearing in this application are only used to illustrate the technical solutions through specific examples. It should be particularly emphasized that any implementation method or design scheme marked as "exemplary" or "for example" should not be construed as having an advantage over other solutions. Such expressions are only used to present the technical concepts more intuitively.

[0027] Example 1: See Figure 1 to 3 This embodiment provides a high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network, including the following steps: S1. Obtain basic data for modeling axial-flow propeller turbines.

[0028] The basic data for modeling an axial-flow propeller turbine needs to cover multi-dimensional information such as structural parameters, dynamic characteristics, boundary conditions, and operating conditions. Specific steps include: 1) Obtain the basic structural parameters of the water turbine.

[0029] ①Gear geometric parameters: Gear diameter; ② Water guiding mechanism parameters: opening range (0~100%); ③ Parameters of water intake and discharge components: Obtain the dimensions of the volute, the height and diffusion angle of the tailpipe, and the dimensions of the guide vanes for the seat ring according to the design drawings.

[0030] 2) Obtain the comprehensive operational characteristic curve data.

[0031] The test reports or drawings provided by the manufacturer are digitized using a high-precision scanner, and then the data is extracted using image reading software. Data types include: ①Water purification head ( H ) is the horizontal axis, and the flow rate is ( Q () is the vertical axis: equal efficiency curve, equal guide vane opening curve, equal output curve, and equal blade rotation angle curve; ②Output ( P ) is the horizontal axis, and the net water head is the vertical axis. H () is the vertical axis: equal efficiency curve, equal guide vane opening curve, equal output curve, and equal blade rotation angle curve; ③ Guide vane opening ( Y ) is the x-axis and the blade rotation angle ( B () is the vertical axis: equal head curve.

[0032] 3) Obtain on-site dynamic test data.

[0033] High-precision sensors and data acquisition systems are used to obtain dynamic test data from the hydropower station, including primary frequency regulation data and active power regulation data.

[0034] High-precision sensors include head pressure transmitters, flow meters, and torque meters. The sampling frequency of the data acquisition system is ≥100 Hz.

[0035] S2. Supplement the opening characteristics and efficiency characteristics data of the axial-flow propeller turbine, and calculate the unit flow rate of the turbine.

[0036] Select output P The horizontal axis represents the net water head. H The curve with the vertical axis is used for modeling axial-flow propeller turbines.

[0037] For iso-efficiency curves, the data type includes efficiency. or Unit speed n 11 and unit torque M 11 The data on the constant opening curve does not include the guide vane opening Y; while the data on the constant opening curve includes the guide vane opening. Y Unit speed n 11 and unit torque M 11 Efficiency not included or That is, the unit flow rate of the water turbine cannot be calculated. Q 11 .

[0038] in, n 11 and M 11 The calculation methods are shown in equations (1) and (2).

[0039] (1) (2) In the formula: , It is the density of water. D is the acceleration due to gravity; D is the diameter of the wheel; X is the rotational speed. It is a unit torque; It's about efficiency; It's a water purifier.

[0040] To fully utilize the data from the comprehensive operating characteristic curves of axial-flow propeller turbines and ensure model accuracy, an ECNN (Efficiency Neural Network) model of turbine efficiency characteristics was established, with unit speed and unit torque as inputs and efficiency as output. or = or ( n 11 , M 11 ), and an Opening Characteristic Neural Network (OCNN) with unit rotational speed and unit torque as inputs and guide vane opening as output. Y = Y ( n 11 , M 11 The constant opening data is then substituted into the turbine efficiency characteristic neural network model (ECNN) to calculate the corresponding guide vane opening. Y The corresponding efficiency is calculated by substituting the equivalent efficiency data into the turbine opening characteristic neural network model (OCNN). or Then, the corresponding unit flow rate is calculated using equation (2). Q 11 .

[0041] S3. Construct a dataset of flow characteristics and torque characteristics for an axial-flow propeller turbine.

[0042] Merge the supplemented data on equal aperture and equal efficiency. Take unit speed n 11 Guide vane opening Y, unit flow rate Q 11 A sample dataset of flow characteristics of axial-flow propeller turbines was obtained. Take unit speed n 11 Guide vane opening Y, unit torque M 11 A sample dataset of torque characteristics of axial-flow propeller turbines was obtained.

[0043] S4. Based on the torque characteristic dataset, BP neural network and Seagull algorithm, the guide vane opening in the flow characteristic and torque characteristic dataset of axial flow propeller turbine is corrected.

[0044] Based on a BP neural network, a system is constructed with a unit rotational speed... n 11 And guide vane opening Y is the input, unit flow rate Q 11 and unit torque M 11 The output flow characteristic neural network model (DCNN) is respectively. Q 11 = Q 11 ( n 11 , Y (), and torque characteristic neural network model (TCNN, M11 = M 11 ( n 11 , Y )); The initial weights and thresholds of the BP neural network, as well as the number of neurons in the hidden layer, are optimized using the Seagull optimization algorithm. Unit speed was obtained from the dynamic test data. n 1. Guide vane opening Y and unit torque M 11 Constructing a torque characteristic running dataset ( n 11 , Y , M 11 ); Running the dataset in conjunction with torque characteristics ( n 11 , Y , M 11 The torque characteristic neural network model TCNN and the Seagull optimization algorithm were used to obtain the correction coefficient of guide vane opening Y in the sample dataset of flow torque and torque characteristics of axial flow propeller turbine in S3. By combining the correction coefficient of the guide vane opening Y, the correction of the guide vane opening Y is achieved.

[0045] S5. Construct a cooperative characteristic model of guide vane and propeller blade based on the Seagull optimization algorithm and BP neural network.

[0046] See Figure 2 Based on the guide vane and impeller coordination data and their boundary conditions from the field dynamic test data, BP neural network and Seagull optimization algorithm, a system based on the net water head is constructed. H and guide vane opening Y A neural network model of blade characteristics with input as the input and blade rotation angle B as the output; The initial weights and thresholds of the BP neural network, as well as the number of neurons in the hidden layer, are optimized using the Seagull optimization algorithm. The boundary conditions of the propeller data are considered in terms of synergy.

[0047] The Seagull Optimization Algorithm (SOA) is an intelligent optimization algorithm that simulates the flocking behavior of seagulls. Seagulls exhibit unique behavioral patterns in nature, such as flock migration, circling prey, and rapid swooping to capture prey. These behaviors contain efficient search and optimization strategies. The algorithm, based on these behavioral patterns, uses mathematical modeling and abstraction to transform them into a computational framework for solving complex optimization problems.

[0048] In SOA, each seagull represents a candidate solution, and all seagulls in the population constitute the solution space. At the start of the algorithm, a group of seagulls is randomly generated within the solution space, and relevant parameters, such as the population size, are initialized. N Maximum number of iterations T And so on. The position of each seagull is represented by X. i =( X i1 , X i2 , …, X in ),in n This is the dimension of the problem to be optimized.

[0049] 1) Seagull migratory behavior: Seagull migration is the core mechanism of the algorithm's global search. During migration, individual seagulls update their positions based on the current optimal seagull position and the distribution of other individuals. The position update formula is as follows: X i,j (t+1) = X i,j (t) + α ( X best,j (t) X i,j (t) )+ β rand() ( X avg,j (t) X i,j (t) (3) in: X i,j ( t ) is the first i The seagull was in the first t During the nth iteration j The position of the dimension; X best,j (t) It is the first t In the nth iteration, the optimal seagull population is at the ... j The position of the dimension; X avg,j (t) It is the first t In the next iteration, the population of seagulls was at the [number]th iteration. j The average position of the dimension;α and β It is a control parameter used to balance global search and local development; rand() is a random number in the range [0, 1].

[0050] 2) Flying around prey: When a seagull spots prey, it circles the prey and gradually approaches. In the algorithm, individual seagulls perform a local search around the current optimal solution, as shown in the following formula: X i,j (t+1) = X best,j (t) + c rand() ( X i,j (t) X best,j (t) (4) in: c It is a step size control parameter around flight, and its value is usually a small positive number.

[0051] 3) Dive-and-capture behavior; After locating prey, seagulls quickly swoop down to capture it. In the algorithm, this swoop-capture behavior is used to perform a fast local search of the region near the current optimal solution, as shown in the following formula: X i,j (t+1) = X best,j (t) + d randn();(5) in: d It is the step size control parameter for dive capture; randn() is a random number that follows a standard normal distribution.

[0052] Algorithm steps: ① Initialize the population: generate randomly N A seagull individual, initialized position X i and related parameters; ② Calculate fitness value: Calculate the fitness value of each seagull using the objective function. f (X i ); ③ Seagull migration: Update the seagull positions according to the migration behavior formula above; ④ Flying around prey: Further optimize the seagull's position based on the flying around formula; ⑤ Dive Capture: Perform a local search around the optimal solution based on the dive capture formula; ⑥ Update the optimal solution: Record the current optimal solution X of the population. best ; ⑦ Iteration Termination: Repeat steps ②-⑥ until the maximum number of iterations is reached. T Or it may satisfy the convergence condition.

[0053] With its unique search mechanism, the Seagull optimization algorithm achieves a good balance between global search and local exploitation, effectively avoiding getting trapped in local optima and demonstrating strong robustness and flexibility.

[0054] S6. Combine the corrected dataset, BP neural network and Seagull optimization algorithm to reconstruct the neural network model of the flow characteristics and torque characteristics of the axial flow propeller turbine.

[0055] See Figure 3 Combining the corrected axial-flow propeller turbine flow and torque characteristic datasets, BP neural networks, and the Seagull optimization algorithm, a system was constructed based on the flow and torque characteristics of the turbine per unit rotational speed. n 11 Guide vane opening Y The blade rotation angle B is the input and the unit flow rate. Q 11 and unit torque M 11 The output of the axial-flow propeller turbine flow characteristic neural network model (RDCNN) is shown below. Q 11 = Q 11 ( n 11 , Y , B ()) and torque characteristic neural network model (RTCNN, M 11 = M 11 ( n 11 , Y , B ), ; Among them, the initial weights and threshold of the BP neural network, and the number of neurons in the hidden layer... N h The optimal value is obtained through the Seagull optimization algorithm; data conversion module 1 is used to calculate ( n 11 , Y The dataset, data transformation module 2 is used to calculate ( H , Y ) dataset;X , H and Y These are the unit speed, head, and guide vane opening, respectively. x , h and y These are the relative deviations of the unit speed, head, and guide vane opening, respectively. X r and H r These are the unit's rated speed and rated head, respectively. X 0、 H 0 and Y 0 represents the initial values ​​for unit speed, head, and guide vane opening, respectively; During parameter optimization, the number of hidden layer neurons N h The maximum value is limited by equation (6), and then the mean square error is expressed as shown in equation (7). MSE To obtain the optimal initial weights and thresholds, and the number of hidden layer neurons in the BP neural network for the target; (6) (7) In the formula, N i It is the number of neurons in the input layer; N o It is the number of neurons in the output layer; Z It is a constant between 0 and 10; k Indicates the number of sample data; This represents the predicted data; D i This represents sample data.

[0056] To address the limitations of traditional axial-flow propeller turbine modeling methods that fail to adequately consider turbine nonlinearity and guide vane / blade coordination characteristics, this invention proposes a high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and a backpropagation neural network. This method comprehensively considers factors such as turbine nonlinearity and guide vane / blade coordination characteristics. This approach will provide a solid model foundation for the research of control strategies for axial-flow propeller turbines and contribute to optimizing the operational regulation performance of hydroelectric units.

[0057] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the invention. Modifications and variations made by those skilled in the art in accordance with the spirit of the invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network, characterized in that, Includes the following steps: S1. Obtain basic data for modeling axial-flow propeller turbines; S2. Supplement the opening characteristics and efficiency characteristics data of the axial-flow propeller turbine, and calculate the unit flow rate of the turbine; S3. Construct a dataset of flow characteristics and torque characteristics for an axial-flow propeller turbine. S4. Based on the torque characteristic dataset, BP neural network and Seagull algorithm, correct the guide vane opening in the flow characteristic and torque characteristic dataset of axial flow propeller turbine. S5. Construct a cooperative characteristic model of guide vane and blade based on the Seagull optimization algorithm and BP neural network; S6. Combine the corrected dataset, BP neural network and Seagull optimization algorithm to reconstruct the neural network model of the flow characteristics and torque characteristics of the axial flow propeller turbine.

2. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 1, characterized in that, In S1, the basic data includes the basic structural parameters of the turbine, the comprehensive operating characteristic curve data, and the field dynamic test data.

3. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 2, characterized in that, The basic structural parameters of the turbine include: Runner geometry parameters: Runner diameter; Water guiding mechanism parameters: opening range; Water intake and discharge component parameters: Obtain the dimensions of the volute, tailpipe height and diffusion angle, and seat ring fixed guide vane dimensions based on the design drawings.

4. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 2, characterized in that, The comprehensive operational characteristic curve data includes: With head H as the abscissa and flow rate Q as the ordinate, the curves are: equal efficiency curve, equal guide vane opening curve, equal output curve, and equal blade angle curve. With output P as the abscissa and net head H as the ordinate: equal efficiency curve, equal guide vane opening curve, equal output curve, and equal blade rotation angle curve; A constant head curve with guide vane opening Y as the abscissa and blade rotation angle B as the ordinate.

5. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 2, characterized in that, The on-site dynamic test data includes: Dynamic test data from the hydropower station are acquired through high-precision sensors and data acquisition systems, including primary frequency regulation data and active power regulation data; the high-precision sensors include head pressure transmitters, flow meters, and torque meters.

6. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 4, characterized in that, In S2, A curve with output P as the x-axis and head H as the y-axis is selected for modeling axial-flow propeller turbines. Data types on iso-efficiency curves include efficiency η Unit speed n 11 and unit torque M 11 ; The data on the constant opening curve includes the guide vane opening. Y Unit speed n 11 and unit torque M 11 ; Establish a neural network model of turbine efficiency characteristics with unit rotational speed and unit torque as inputs and efficiency as output. η = η ( n 11 , M 11 );and A neural network model of turbine opening characteristics with unit rotational speed and unit torque as inputs and guide vane opening as output. Y = Y ( n 11 , M 11 ); Substitute the constant opening data into the neural network model of turbine efficiency characteristics to calculate the corresponding guide vane opening Y; Substituting the equivalent efficiency data into the neural network model of the turbine opening characteristics, the corresponding efficiency is calculated. η ; Then calculate the unit flow rate corresponding to the water turbine. Q 11 ; in, ；(1) ； (2) In the formula: , It is the density of water. D is the acceleration due to gravity; D is the diameter of the wheel; X is the rotational speed. It is a unit torque; It's about efficiency; It's a water purifier.

7. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 6, characterized in that, In S3, the following steps are taken to construct a dataset of flow characteristics and torque characteristics for an axial-flow propeller turbine: Merge the supplemented data on equal aperture and equal efficiency. Take unit speed n 11 Guide vane opening Y, unit flow rate Q 11 A sample dataset of flow characteristics of axial-flow propeller turbines was obtained. Take unit speed n 11 Guide vane opening Y, unit torque M 11 A sample dataset of torque characteristics of axial-flow propeller turbines was obtained.

8. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 1 or 7, characterized in that, In S4, based on the torque characteristic dataset, BP neural network, and Seagull algorithm, the guide vane opening Y in the flow characteristic and torque characteristic datasets of the axial-flow propeller turbine is corrected, including the following steps: Based on a BP neural network, a system is constructed with a unit rotational speed... n 11 And guide vane opening Y is the input, unit flow rate Q 11 and unit torque M 11 These are the output flow characteristic neural network model DCNN and torque characteristic neural network model TCNN, respectively. The initial weights and thresholds of the BP neural network, as well as the number of neurons in the hidden layer, are optimized using the Seagull optimization algorithm. Unit speed was obtained from the dynamic test data. n 1. Guide vane opening Y and unit torque M 11 Constructing a torque characteristic running dataset ( n 11 , Y , M 11 ); Running the dataset in conjunction with torque characteristics ( n 11 , Y , M 11 The torque characteristic neural network model TCNN and the Seagull optimization algorithm were used to obtain the correction coefficient of guide vane opening Y in the sample dataset of flow torque and torque characteristics of axial flow propeller turbine in S3. The guide vane opening Y is corrected by incorporating a correction factor.

9. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 1, characterized in that, In S5, a cooperative characteristic model of guide vanes and propeller blades is constructed based on the Seagull optimization algorithm and BP neural network, including the following steps: Based on the guide vane and impeller coordination data and their boundary conditions from the field dynamic test data, BP neural network and Seagull optimization algorithm, a system based on the net water head is constructed. H and guide vane opening Y A neural network model of blade characteristics with input as the input and blade rotation angle B as the output; The initial weights and thresholds of the BP neural network, as well as the number of neurons in the hidden layer, are optimized using the Seagull optimization algorithm. The boundary conditions of the propeller data are considered in terms of synergy.

10. The high-precision modeling method for axial-flow propeller turbines based on the Seagull optimization algorithm and BP neural network according to claim 1, characterized in that, In S6, a dataset of modified axial-flow propeller turbine flow and torque characteristics, along with a BP neural network and the Seagull optimization algorithm, is constructed to measure the flow rate per unit rotational speed. n 11 Guide vane opening Y The blade rotation angle B is the input and the unit flow rate. Q 11 and unit torque M 11 These are the output neural network models of the flow characteristics and torque characteristics of the axial-flow propeller turbine, respectively. Among them, the initial weights and threshold of the BP neural network, and the number of neurons in the hidden layer... N h The optimal value is obtained using the Seagull optimization algorithm; During parameter optimization, the number of hidden layer neurons N h The maximum value is limited by equation (6), and then the mean square error is expressed as shown in equation (7). MSE To obtain the optimal initial weights and thresholds, and the number of hidden layer neurons in the BP neural network for the target; ；(6) ；(7) In the formula, N i It is the number of neurons in the input layer; N o It is the number of neurons in the output layer; Z It is a constant between 0 and 10; k Indicates the number of sample data; This represents the predicted data; D i This represents sample data.

Images