An optimization method and system for a jet mechanism of an impulse turbine

Through deep learning and multi-objective optimization algorithms, the geometric structure of the impact turbine jet mechanism is optimized, which solves the problem of serious wear under silt and sand conditions, and realizes the efficient and stable operation of the jet mechanism and the guarantee of hydraulic characteristics.

CN118569113BActive Publication Date: 2024-11-12XIAN UNIV OF TECH
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411063692.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-12
Estimated Expiration
2044-08-05

AI Technical Summary

Technical Problem

The existing impact turbine jet mechanism is severely worn under silt conditions, resulting in a decrease in jet quality and an increase in operating and maintenance costs, and the geometric structure optimization efficiency is low, so it is not possible to effectively consider the characteristics of silt particles and hydraulic characteristics constraints.

Method used

Using deep learning and multi-objective differential evolution optimization algorithm, the geometric structural parameters of the injection mechanism are optimized through the three-phase numerical simulation of water, gas and sand and the Transformer neural network model, reducing wear and ensuring hydraulic characteristics, including nozzle angle, needle angle, spraying Optimization of needle diameter and needle transition section diameter.

Benefits of technology

Significantly suppress the wear degree of the injection mechanism of the impact turbine, shorten the design cycle, reduce design costs, and improve operating stability and jet quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118569113B_ABST
    Figure CN118569113B_ABST
Patent Text Reader

Abstract

The invention discloses an optimization method and system for an impulse turbine jet mechanism, belonging to the field of impulse turbine jet mechanism design. The method generates a plurality of optimization parameter groups according to design variables of the jet mechanism, adopts each optimization parameter group to respectively update the initial three-dimensional model, and combines water, air and sand three-phase numerical simulation calculation to obtain the wear degree and hydraulic characteristics of each optimization parameter group; constructs a training data set according to each optimization parameter group, and the corresponding wear degree, hydraulic characteristics and sediment particle characteristics, and trains a deep neural network model to obtain a jet mechanism performance prediction model; takes reducing the wear degree as the optimization goal, combines a multi-objective differential evolution optimization algorithm to optimize the geometric structure parameters of the jet mechanism, and obtains the optimal solution with the minimum optimization goal; the method can efficiently optimize the geometric structure design of the jet mechanism, can significantly suppress the wear of the jet mechanism, and has strong practical value and promotion value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of design of an impulse turbine jet mechanism, and in particular to an optimization method and system for an impulse turbine jet mechanism. Background Art

[0002] The application head range of the impulse turbine is very wide, and it can be applied within 30m to 3000m. It is an important model suitable for large-head power stations. During the operation of the impulse turbine, the water flows through the jet mechanism and accelerates out. After contacting the air, it forms a high-speed jet, which is sprayed into the bucket surface with a sharp change in curvature and flows out of the bucket quickly. In the above flow process, there is a complex water-gas two-phase flow. The power station is generally built at a higher foothill, and the water flow will carry a certain amount of sediment through the impulse turbine. For some rivers with high sediment content, the jet mechanism and the bucket will be severely worn, resulting in the decrease of jet quality, power reduction and increase of operation and maintenance costs. In severe cases, it may also threaten the safe and stable operation of the impulse turbine. Therefore, it is of great significance to analyze the influence mechanism and degree of sediment particles on the jet mechanism and optimize the geometric structure of the impulse turbine jet mechanism under sandy conditions to improve the operation stability of the impulse turbine.

[0003] Nowadays, the geometric structure optimization of impulse turbines is mainly based on the empirical data of power stations. When optimizing the geometric structure of the injection mechanism, the influence of operating conditions (such as sediment particle characteristics) on the optimization parameters is not taken into account, and the hydraulic characteristics constraints of the injection mechanism are not carried out. Secondly, the geometric structure optimization design is manually performed based on calculated data, or the geometric structure optimization design is performed using traditional single optimization algorithms (such as genetic algorithms), which requires repeated simulation and verification of the geometric structure, resulting in low optimization efficiency and poor optimization effect.

[0004] The continuous development of deep learning has overcome the limitations of empirical data to a certain extent, and can effectively reduce the need for designer experience, thereby quickly and efficiently completing the geometric design of the ejector mechanism. Therefore, how to optimize the geometric design of the ejector mechanism with the help of intelligent computing is a technical problem that needs to be solved urgently. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides an optimization method and system for an impulse turbine jet mechanism. The optimized jet mechanism can effectively reduce the degree of wear thereof, so that the impulse turbine can operate efficiently and stably.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for optimizing an injection mechanism of an impulse turbine comprises the following steps:

[0008] Step 1: construct an initial three-dimensional model of the injection mechanism according to the initial geometric structure parameters of the injection mechanism;

[0009] Step 2: Generate multiple optimization parameter groups within the value range of each design variable of the injection mechanism, use each optimization parameter group to update the initial three-dimensional model respectively, obtain the three-dimensional model corresponding to each optimization parameter group, and then perform water-gas-sand three-phase numerical simulation calculation on each three-dimensional model respectively to obtain the wear degree and hydraulic characteristics of each optimization parameter group under the characteristics of sediment particles;

[0010] Step 3: construct a training data set according to each optimization parameter group and the corresponding wear degree, hydraulic characteristics and sediment particle characteristics, and use the training data set to train the deep neural network model to obtain a jet mechanism performance prediction model;

[0011] Step 4: Taking reducing the degree of wear of the injection mechanism as the optimization target and the outlet flow rate of the injection mechanism as the constraint condition of the optimization target, the optimization parameter group is optimized by combining the injection mechanism performance prediction model and the multi-objective optimization algorithm to obtain the optimal solution with the minimum degree of wear, and the optimal solution is used to design the geometric structure of the injection mechanism.

[0012] Preferably, the design variables of the injection mechanism in step 2 include nozzle angle, spray needle angle, spray needle diameter, nozzle diameter and spray needle transition section diameter.

[0013] Preferably, in step 2, a uniform sampling method is used to sample each design variable within a value range to obtain multiple optimization parameter groups.

[0014] Preferably, the method for numerical simulation calculation of water, air and sand three phases is as follows:

[0015] The VOF multiphase flow model is used to simulate the water-gas two-phase flow. After the unsteady calculation of the water-gas two-phase flow is stable, discrete phase particles are added to calculate the impact wear of the discrete phase particles on the injection mechanism, and the wear degree and hydraulic characteristics of the optimized parameter group under the characteristics of sediment particles are obtained.

[0016] Preferably, the wear degree includes a maximum wear rate, an average wear rate, and a wear range;

[0017] The hydraulic characteristics include the outlet flow rate of the injection mechanism;

[0018] The sediment particle characteristics include the particle size of the sediment particles and the concentration of the sediment particles.

[0019] Preferably, the deep neural network model is a Transformer neural network model;

[0020] The input of the Transformer neural network model is the particle size of the sediment particles, the concentration of the sediment particles and the optimization parameter group, and the output is the wear degree and hydraulic characteristics of different partitions of the injection mechanism under rated working conditions.

[0021] Preferably, in step 4, a multi-objective differential evolution optimization algorithm is used to optimize the optimization parameter group to obtain the optimal solution with the minimum wear degree.

[0022] Preferably, the method for determining the optimal solution is as follows:

[0023] According to the constrained optimization objectives, the optimization parameter group is optimized in the entire value range space. In the multi-objective optimization process, the injection mechanism performance prediction model is used to predict the wear degree and outlet flow of the current optimal solution, and the iterative optimization is used to obtain the optimization parameter group with the minimum wear degree.

[0024] Preferably, a multivariate polynomial is used to construct the optimization target, and a penalty function is used to constrain the deviation of the outlet flow.

[0025] An optimization system for an impulse turbine jet mechanism, comprising:

[0026] An initial module, used for constructing an initial three-dimensional model of the injection mechanism according to initial geometric structure parameters of the injection mechanism;

[0027] The model updating module is used to generate multiple optimization parameter groups within the value range of each design variable of the injection mechanism, and use each optimization parameter group to update the initial three-dimensional model respectively to obtain the three-dimensional model corresponding to each optimization parameter group, and then perform water-gas-sand three-phase numerical simulation calculation on each three-dimensional model to obtain the wear degree and hydraulic characteristics of each optimization parameter group under the characteristics of sediment particles;

[0028] A prediction module is used to construct a training data set according to each optimization parameter group and the corresponding wear degree, hydraulic characteristics and sediment particle characteristics, and use the training data set to train the deep neural network model to obtain a jet mechanism performance prediction model;

[0029] The optimization module is used to optimize the optimization parameter group by taking reducing the wear degree of the injection mechanism as the optimization target and the outlet flow rate of the injection mechanism as the constraint condition of the optimization target, combining the injection mechanism performance prediction model and the multi-objective optimization algorithm to obtain the optimal solution with the minimum wear degree, and use this optimal solution to design the geometric structure of the injection mechanism.

[0030] Compared with the prior art, the present invention has the following beneficial technical effects:

[0031] The present invention provides an optimization method for an impulse turbine jet mechanism, which samples within the value range of each design variable to form a plurality of optimization parameter groups, and combines water, air and sand three-phase numerical simulation calculations to obtain the wear degree and hydraulic characteristics of each optimization parameter group under the characteristics of sediment particles, and uses this as a training data set to train a deep neural network model to obtain a jet mechanism performance prediction model, and combines a multi-objective differential evolution optimization algorithm to determine the optimal solution of the optimization parameter group, thereby realizing the geometric structure optimization of the jet mechanism; the method couples deep learning with a multi-objective differential evolution optimization algorithm, takes the wear degree of the jet mechanism as the optimization target, comprehensively considers the maximum wear rate, the average wear rate and the wear range of the jet mechanism, takes the outlet flow of the jet mechanism as a constraint, seeks the best within the value range of each design variable, obtains the optimal solution, and optimizes the geometric structure of the jet mechanism according to the optimal solution. The method can significantly suppress the wear degree of the impulse turbine jet mechanism and ensure the hydraulic characteristics. The entire optimization method shortens the design cycle of the jet mechanism and reduces the design cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a structural schematic diagram of an impulse turbine;

[0033] Among them, 1. water guide mechanism; 2. injection mechanism; 3. water bucket;

[0034] Figure 2 for Figure 1 A cross-sectional view of the injection mechanism at center A;

[0035] Among them, α is the nozzle angle, β is the needle angle, d is the needle diameter, D is the nozzle diameter, and h is the needle transition section diameter;

[0036] Figure 3 A flow chart of an optimization method for a jet mechanism of an impulse turbine according to the present invention;

[0037] Figure 4 The figure is a schematic diagram of the Transformer neural network model of the present invention;

[0038] Figure 5 A comparison diagram of the geometric structures of the initial injection mechanism and the optimized injection mechanism;

[0039] Among them, the solid line represents the geometric structure of the initial injection mechanism, and the dotted line represents the geometric structure of the injection mechanism after optimization;

[0040] Figure 6 The nozzle throat wear comparison diagram of the initial injection mechanism and the optimized injection mechanism;

[0041] in, Figure 6 Figure a is the nozzle throat wear diagram of the initial injection mechanism, and Figure b is the nozzle throat wear diagram of the optimized injection mechanism;

[0042] Figure 7 The comparison diagram of the outer wear of the injection needle of the initial injection mechanism and the injection mechanism after optimization;

[0043] in, Figure 7 Figure a is the wear diagram of the outer side of the spray needle of the initial spray mechanism, and Figure b is the wear diagram of the outer side of the spray needle of the optimized spray mechanism;

[0044] Figure 8 The comparison diagram of the inner side wear of the injection needle of the initial injection mechanism and the injection mechanism after optimization;

[0045] in, Figure 8 Figure a is the wear diagram of the inner side of the nozzle needle of the initial injection mechanism, and Figure b is the wear diagram of the inner side of the nozzle needle of the optimized injection mechanism. DETAILED DESCRIPTION

[0046] The present invention will be further described in detail below in conjunction with the accompanying drawings, which are intended to explain the present invention rather than to limit it.

[0047] See also Figure 1 The impulse turbine includes a water guiding mechanism 1, an ejection mechanism 2 and a bucket 3. The water guiding mechanism 1 is responsible for guiding the water flow to the ejection mechanism 2. The ejection mechanism 2 generates a high-speed water flow through a nozzle and ejects it to the bucket 3. The bucket 3 captures the kinetic energy of the high-speed water flow and converts it into mechanical energy, thereby driving the generator to generate electrical energy.

[0048] See also Figure 2 The injection mechanism 2 includes a nozzle, a spray needle and a guide body. In the contraction section of the nozzle, the flow cross section decreases, the speed increases, the pressure energy of the water flow is converted into kinetic energy, and the water flow forms a high-speed jet in the air after it is ejected from the nozzle. The spray needle mainly controls the flow rate of the water flow by moving back and forth. The guide body is used to guide the water flow along the axis of the spray needle and supports the spray needle. During the flow of sand-containing water, the sediment particles impact the injection mechanism 2 with the flow of water. The structural changes of the injection mechanism 2 can change the flow characteristics of the water flow, and then change the movement characteristics of the sediment particles, which leads to different wear of the sediment particles on the wall of the injection mechanism 2.

[0049] The purpose of the present invention is to take the degree of wear of the injection mechanism 2 caused by mud and sand particles as the optimization target. The optimization target comprehensively considers the maximum wear rate, average wear rate and wear range of the injection mechanism 2, optimizes the geometric structure of the injection mechanism 2, and thus reduces the degree of wear of the injection mechanism 2 caused by mud and sand particles. The geometric structure parameters include the nozzle angle α, the needle angle β, the needle diameter d, the nozzle diameter D and the needle transition section diameter h. The injection mechanism 2 is modified by changing the geometric structure parameters to achieve the optimization of the geometric structure of the injection mechanism 2.

[0050] See also Figure 3 The object of the present invention is to provide a method for optimizing the injection mechanism of an impulse turbine, comprising the following steps:

[0051] Step 1: construct an initial three-dimensional model of the injection mechanism 2 according to the initial geometric structure parameters of the injection mechanism 2.

[0052] The initial three-dimensional model is a parameterized model, and the geometric structure parameters are used as design variables to be optimized. The initial three-dimensional model is optimized to reduce the wear degree of the injection mechanism 2. The design variables include the nozzle angle α, the needle angle β, the needle diameter d, the nozzle diameter D and the needle transition section diameter h.

[0053] It should be noted that the parametric model of the injection mechanism 2 is the basis for carrying out optimization design. Its function is to use fewer geometric structure parameters to accurately describe the injection mechanism 2, so that a new injection mechanism 2 can be easily obtained by changing the geometric structure parameters, thereby realizing the geometric structure optimization of the injection mechanism 2.

[0054] Step 2: Within the design range of each design variable, use the uniform sampling method to sample each design variable to obtain multiple optimization parameter groups, use each optimization parameter group to update the initial three-dimensional model respectively, and obtain the three-dimensional model corresponding to each optimization parameter group. Then, perform water-air-sand three-phase numerical simulation calculations on each three-dimensional model to obtain the wear degree and hydraulic characteristics of each optimization parameter group under the characteristics of sediment particles.

[0055] The wear degree includes a maximum wear rate, an average wear rate and a wear range.

[0056] The hydraulic characteristics include the outlet flow rate of the injection mechanism 2 .

[0057] The water-gas-sand three-phase numerical simulation calculation method of the three-dimensional model is as follows:

[0058] Within the design range of each design variable, multiple optimization parameter groups are generated based on the uniform sampling method, and the design variables in each optimization parameter group are used to update the initial three-dimensional model of the injection mechanism 2 to obtain the three-dimensional model corresponding to each optimization parameter group, determine the characteristics of the sediment particles, and perform water-gas-sand three-phase numerical simulation calculations on the three-dimensional model corresponding to each optimization parameter group. The volume of fluid (VOF) multiphase flow model is used to simulate the water-gas two-phase flow. After the unsteady calculation of the water-gas two-phase flow is stable, discrete phase particles are added, and the impact wear caused by the discrete phase particles on the injection mechanism 2 is calculated, so as to obtain the maximum wear rate, average wear rate and wear range of different areas of the three-dimensional model corresponding to each injection mechanism 2 under rated conditions, as well as the outlet flow rate of the injection mechanism 2.

[0059] Step 3: construct a training data set based on each optimization parameter group and the corresponding wear degree, hydraulic characteristics and sediment particle characteristics, take the wear degree of the injection mechanism 2 as the optimization target, use the training data set to train the deep neural network model, and obtain the injection mechanism performance prediction model.

[0060] The optimization target comprehensively considers the maximum wear rate, average wear rate and wear range of the injection mechanism 2. The expression of the optimization target is as follows:

[0061]

[0062] In the formula, represents the design variable; is the minimum value of the optimization objective; is the total area of ​​each region of the injection mechanism; The area where the wear rate of different areas of the flow-through parts is greater than the average wear rate; Design flow for reference; To optimize the outlet flow of the injection mechanism corresponding to different optimization parameter groups in the design, is a constraint condition.

[0063] The input of the injection mechanism performance prediction model is the characteristics of the sediment particles and the value of the optimization parameter group, and the output is the wear degree and hydraulic characteristics of different partitions of the injection mechanism 2 under rated working conditions.

[0064] In this embodiment, the actual need for optimizing the design variables of the injection mechanism 2 is taken into consideration, and the deep neural network model is a Transformer neural network model, which is a self-attention mechanism neural network model.

[0065] As a preferred implementation method of an embodiment of the present invention, the design variables to be optimized of the injection mechanism include the nozzle angle α, the needle angle β, the needle diameter d, the nozzle diameter D, and the needle transition section diameter h. The design variables to be optimized are determined after in-depth research on the injection mechanism 2, and can better cooperate with the injection mechanism performance prediction model to realize the prediction of the degree of wear of the injection mechanism; at the same time, in the prediction process, the characteristics of the sediment particles are used as the input of the injection mechanism performance prediction model to obtain the degree of wear of the injection mechanism 2 under the sediment particle characteristics. Since different sediment particle characteristics cause different wear on the injection mechanism 2, the injection mechanism performance prediction model needs to take the sediment particle characteristics as an input, and the sediment particle characteristics include the particle size and concentration of the sediment particles.

[0066] Step 4: Use the test data set to verify the prediction accuracy of the trained injection mechanism performance prediction model. When the prediction accuracy of the injection mechanism performance prediction model meets the requirements, execute step 5. When the prediction accuracy of the injection mechanism parameter performance prediction model does not meet the requirements, execute step 2 to expand the training data set, mainly expand the training set at the location where the prediction accuracy error is large, and use the expanded training data set to iteratively optimize the deep neural network model until the prediction accuracy of the deep neural network model meets the requirements.

[0067] In this embodiment, a test data set is used to verify the injection mechanism performance prediction model, and the degree of wear output by the injection mechanism performance prediction model, i.e., the maximum wear rate, the average wear rate and the wear range, are compared with the true values ​​corresponding to the test data set. When the errors of the maximum wear rate, the average wear rate and the wear range are all less than the corresponding set thresholds, the prediction accuracy of the injection mechanism performance prediction model meets the requirements.

[0068] When the error of the maximum wear rate, the average wear rate, and any value in the wear range is greater than the set threshold, execute step 2 to expand the training data set, and iterate the deep neural network model trained in step 3 again until the error of each value in the wear degree is less than the corresponding set threshold.

[0069] Step 5: Taking the wear degree of the injection mechanism 2 as the optimization target, the optimization parameter group is optimized in combination with the injection mechanism performance prediction model obtained in step 4 to obtain the optimal solution with the minimum wear degree, and the optimal solution is used to optimize the geometric structure design of the injection mechanism 2.

[0070] The optimization target is constructed by multivariate polynomials. The constraint condition of the optimization target is the outlet flow rate of a single injection mechanism 2. The outlet flow rate deviation is required to be within 5%. The deviation of the outlet flow rate is constrained in the form of a penalty function. In the multi-objective optimization process, the multi-objective differential evolution optimization algorithm is used as the optimization algorithm. The injection mechanism performance prediction model is used to predict the wear degree of different optimization parameter groups and iteratively optimize until the wear degree of the current optimal solution is less than the optimal solution obtained in the previous iteration, and the wear degree of the optimal solution does not change after iterating for multiple times. Then the optimal solution is the optimal parameter group of the injection mechanism 2. Finally, the three-dimensional model is updated according to the optimal parameter group to obtain the optimal three-dimensional model design, and the water-gas-sand three-phase numerical calculation is used for analysis and verification.

[0071] Example 1

[0072] A method for optimizing an injection mechanism of an impulse turbine comprises the following steps:

[0073] Step 1: Construct an initial three-dimensional model according to the set initial values ​​of the geometric structure parameters.

[0074] Step 2, determining the design variables to be optimized of the impulse turbine jet mechanism;

[0075] The design variables include nozzle angle α, needle angle β, needle diameter d, nozzle diameter D and needle transition section diameter h.

[0076] Step 3: Determine the design range of the design variables, and use the uniform sampling method to sample within the design range of the design variables to obtain multiple optimization parameter groups. Table 1 is the design range table of the design variables.

[0077] Table 1

[0078]

[0079] The upper and lower limits in Table 1 represent the upper and lower limits of the design range of each design variable, respectively.

[0080] The number of design variables to be optimized is n=5. The uniform sampling method is used to generate 101 optimization parameter groups in the five-dimensional design space. These optimization parameter groups are evenly distributed in the two-dimensional space formed by any two design variables, which can better reflect the characteristics of the design space.

[0081] Step 4: Update the initial three-dimensional model according to each optimization parameter group to obtain a three-dimensional model of the injection mechanism 2 corresponding to each optimization parameter group.

[0082] In this embodiment, a parameterized initial three-dimensional model is produced in three-dimensional drawing software, and then the design variables of the initial three-dimensional model are updated according to each optimization parameter group to obtain a three-dimensional model of the injection mechanism 2 corresponding to each optimization parameter group.

[0083] Step 5: Set the characteristics of the sediment particles in the CFD (Computational Fluid Dynamics) software, and then perform a water-gas-sand three-phase numerical simulation calculation on the three-dimensional model of each injection mechanism 2 to obtain the maximum wear rate, average wear rate, wear range and outlet flow rate of each three-dimensional model.

[0084] CFD software is used to perform unsteady calculations of the water-gas two-phase state on the three-dimensional model. After the calculation is stable, discrete phase particles are added to the inlet of the injection mechanism 2 to perform a water-gas-sand three-phase numerical simulation. The characteristics of the sediment particles in all three-dimensional models are consistent, and the maximum wear rate, average wear rate, wear range and outlet flow rate of the injection mechanism 2 in different areas are obtained.

[0085] Step 6: construct a training data set based on the optimized parameter group, and the corresponding maximum wear rate, average wear rate, wear range, outlet flow rate of the injection mechanism 2, and sediment particle characteristics.

[0086] Step 7: Use the training data set to train the Transformer neural network model, and obtain the injection mechanism performance prediction model after training.

[0087] See also Figure 4 The Transformer neural network model includes an encoder and a decoder. The encoder is composed of N=3 identical encoding modules stacked together. Each encoding module includes three parts: a self-attention sublayer, a feedforward neural network sublayer, and a normalization layer. In each encoding module, the input information first enters the self-attention sublayer, which helps the encoding module read the input information so as to better encode a specific information; the output of the self-attention sublayer will be passed to the feedforward neural network sublayer; each encoding module is superimposed on each other, and the output information is mapped to a specific range using a normalization layer to improve the training stability of the network.

[0088] The decoder is composed of M=2 identical decoding modules stacked together. Each decoding module includes a self-attention sublayer, a normalization layer, a decoding-encoding attention sublayer, and a feedforward neural network sublayer. The role of the decoding-encoding attention sublayer is to enable the decoder to pay attention to the relevant information between the input information and the target information.

[0089] In this embodiment, the number of encoding modules in the encoder (Encoder) is 3, and the number of decoding modules in the decoder (Decoder) is 2, which can also be defined according to actual engineering conditions.

[0090] In this embodiment, the Transformer neural network model is iteratively trained using an optimized parameter group and sediment particle characteristics to obtain a jet mechanism performance prediction model, and the wear degree output by the jet mechanism performance prediction model is verified and analyzed with the wear degree obtained by the water, air and sand three-phase numerical simulation calculation, proving that the jet mechanism performance prediction model has high prediction accuracy.

[0091] Step 8: Take the degree of wear as the optimization target and constrain it, and perform multi-objective optimization on the optimization parameter group of the injection mechanism 2 in combination with the injection mechanism performance prediction model to obtain the optimal solution of the optimization parameter group.

[0092] The expression of the optimization objective is as follows:

[0093]

[0094] in, represents the design variable; A represents the nozzle throat, B represents the needle transition section, C represents the tip of the spray needle; S A , SB and S C are the geometric areas of the nozzle throat, the needle transition section and the needle tip of the initial three-dimensional model; S thA The geometric area of ​​the nozzle throat corresponding to the optimized parameter group where the wear rate is greater than the average wear rate, S thB To optimize the geometric area of ​​the nozzle transition section where the wear rate is greater than the average wear rate corresponding to the parameter group, S thC The geometric area of ​​the nozzle tip corresponding to the optimized parameter group has a wear rate greater than the average wear rate. The optimization target constraint condition is the outlet flow rate of a single injection mechanism, the deviation is required to be within 5%, and the constraint processing is performed in the form of a penalty function.

[0095] In the multi-objective optimization process, a multi-objective differential evolution optimization algorithm is used in combination with an injection mechanism performance prediction model to carry out iterative optimization. The optimization stopping criterion is that the iterative optimization is terminated until the optimal solution of the optimization parameter group obtained after five iterations no longer changes or the set maximum number of iterations has been reached. The optimal solution of the current optimization parameter group is the optimal parameter group, and the wear degree of the optimal parameter group is the smallest. The optimal parameter group is used to optimize the geometric structure of the injection mechanism 2.

[0096] Figure 5 It is a comparison diagram of the geometric structures of the initial injection mechanism and the injection mechanism after optimization; wherein, the solid line represents the geometric structure of the initial injection mechanism, and the dotted line represents the geometric structure of the injection mechanism after optimization.

[0097] Figure 6 The nozzle throat wear comparison diagram of the initial injection mechanism and the optimized injection mechanism; Figure 7 The comparison diagram of the outer wear of the injection needle of the initial injection mechanism and the injection mechanism after optimization; Figure 8 The figure shows the comparison of the inner wear of the nozzle needle of the initial injection mechanism and the optimized injection mechanism.

[0098] according to Figure 6 It can be seen that the nozzle throat of the initial injection mechanism is mainly symmetrically worn up and down, while after optimization, two obvious strip-shaped wear appears on the inner side of the nozzle throat, but the overall wear degree of the nozzle throat is significantly reduced. Figure 7 and Figure 8 It can be seen that the initial injection mechanism has more serious wear in the area II of the needle transition section and the needle tip, which is point wear, while the wear of the area I of the needle transition section, the guide body, and the tail of the needle is relatively light; after optimization, the wear of the area I of the needle transition section, the guide body, and the tail of the needle of the injection mechanism almost disappears, the wear of the area II of the needle transition section and the outer side of the needle tip is significantly reduced, and the point wear on the inner side of the needle tip increases slightly. Figure 6 , Figure 7and Figure 8 After optimization, the wear degree of the nozzle throat and the transition section of the nozzle needle of the injection mechanism is significantly reduced, and the high wear area is also significantly reduced. Therefore, it can be proved that the optimized injection mechanism exhibits good anti-wear performance.

[0099] Another aspect of the present invention further provides an optimization system for an injection mechanism of an impulse turbine, comprising:

[0100] An initial module, used for constructing an initial three-dimensional model of the injection mechanism according to initial geometric structure parameters of the injection mechanism;

[0101] The model updating module is used to generate multiple optimization parameter groups within the value range of each design variable of the injection mechanism, and use each optimization parameter group to update the initial three-dimensional model respectively to obtain the three-dimensional model corresponding to each optimization parameter group, and then perform water-gas-sand three-phase numerical simulation calculation on each three-dimensional model to obtain the wear degree and hydraulic characteristics of each optimization parameter group under the characteristics of sediment particles;

[0102] A prediction module is used to construct a training data set according to each optimization parameter group and the corresponding wear degree, hydraulic characteristics and sediment particle characteristics, and use the training data set to train the deep neural network model to obtain a jet mechanism performance prediction model;

[0103] The optimization module is used to optimize the optimization parameter group by taking reducing the wear degree of the injection mechanism as the optimization target and the outlet flow rate of the injection mechanism as the constraint condition of the optimization target, combining the injection mechanism performance prediction model and the multi-objective optimization algorithm to obtain the optimal solution with the minimum wear degree, and use this optimal solution to design the geometric structure of the injection mechanism.

[0104] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for optimizing the injection mechanism of an impulse turbine, characterized in that: The following steps are involved: Step 1: construct an initial three-dimensional model of the injection mechanism according to the initial geometric structure parameters of the injection mechanism; Step 2: Generate multiple optimization parameter groups within the value range of each design variable of the injection mechanism, use each optimization parameter group to update the initial three-dimensional model respectively, obtain the three-dimensional model corresponding to each optimization parameter group, and then perform water-gas-sand three-phase numerical simulation calculation on each three-dimensional model respectively to obtain the wear degree and hydraulic characteristics of each optimization parameter group under the characteristics of sediment particles; The design variables of the injection mechanism include nozzle angle, needle angle, needle diameter, nozzle diameter and needle transition section diameter; The wear degree includes the maximum wear rate, the average wear rate and the wear range; The hydraulic characteristics include the outlet flow rate of the injection mechanism; the sediment particle characteristics include the particle size of the sediment particles and the concentration of the sediment particles; The method for numerical simulation calculation of water, gas and sand three phases is as follows: The VOF multiphase flow model is used to simulate the water-gas two-phase flow. After the unsteady calculation of the water-gas two-phase flow is stable, discrete phase particles are added to calculate the impact wear of the discrete phase particles on the injection mechanism, and the wear degree and hydraulic characteristics of the optimized parameter group under the characteristics of sediment particles are obtained. Step 3: construct a training data set according to each optimization parameter group and the corresponding wear degree, hydraulic characteristics and sediment particle characteristics, and use the training data set to train the deep neural network model to obtain a jet mechanism performance prediction model; Step 4: Taking reducing the wear degree of the injection mechanism as the optimization target and the outlet flow rate of the injection mechanism as the constraint condition of the optimization target, the optimization parameter group is optimized by combining the injection mechanism performance prediction model and the multi-objective optimization algorithm to obtain the optimal solution with the minimum wear degree, and the geometric structure of the injection mechanism is designed by using the optimal solution; The multi-objective differential evolution optimization algorithm is used to optimize the optimization parameter group and obtain the optimal solution with the minimum wear degree; According to the constrained optimization objectives, the optimization parameter group is optimized in the entire value range space. In the multi-objective optimization process, the injection mechanism performance prediction model is used to predict the wear degree and outlet flow of the current optimal solution, and the iterative optimization is used to obtain the optimization parameter group with the minimum wear degree.

2. The method for optimizing the injection mechanism of an impulse turbine according to claim 1, characterized in that: In step 2, the uniform sampling method is used to sample each design variable within the value range to obtain multiple optimization parameter groups.

3. The method for optimizing the injection mechanism of an impulse turbine according to claim 1, characterized in that: The deep neural network model is a Transformer neural network model; The input of the Transformer neural network model is the particle size of the sediment particles, the concentration of the sediment particles and the optimization parameter group, and the output is the maximum wear rate of different partitions of the machine injection mechanism, the average wear rate and the outlet flow rate of the injection mechanism.

4. The method for optimizing the injection mechanism of an impulse turbine according to claim 1, characterized in that: The optimization objective is constructed using a multivariate polynomial, and the deviation of the outlet flow is constrained in the form of a penalty function.

5. A system for executing the optimization method of the injection mechanism of an impulse turbine according to any one of claims 1 to 4, characterized in that: include: An initial module, used for constructing an initial three-dimensional model of the injection mechanism according to initial geometric structure parameters of the injection mechanism; The model updating module is used to generate multiple optimization parameter groups within the value range of each design variable of the injection mechanism, and use each optimization parameter group to update the initial three-dimensional model respectively to obtain the three-dimensional model corresponding to each optimization parameter group, and then perform water-gas-sand three-phase numerical simulation calculation on each three-dimensional model to obtain the wear degree and hydraulic characteristics of each optimization parameter group under the characteristics of sediment particles; A prediction module is used to construct a training data set according to each optimization parameter group and the corresponding wear degree, hydraulic characteristics and sediment particle characteristics, and use the training data set to train the deep neural network model to obtain a jet mechanism performance prediction model; The optimization module is used to optimize the optimization parameter group by taking reducing the wear degree of the injection mechanism as the optimization target and the outlet flow rate of the injection mechanism as the constraint condition of the optimization target, combining the injection mechanism performance prediction model and the multi-objective optimization algorithm to obtain the optimal solution with the minimum wear degree, and use this optimal solution to design the geometric structure of the injection mechanism.

Citation Information

Patent Citations

  • Control law optimization method and device for water turbine speed regulation system

    CN112464478A

  • Impact turbine nozzle structure parameter optimization method and system

    CN114997073A

  • Bucket blade profile optimization method for impulse turbine

    CN114997084A

  • Deep neural network hyperparameter optimization method, electronic device and storage medium

    WO2021007812A1