An optimization method for the braking fan of a turbine expander

Abstract

The present invention relates to the technical field of turbine blade profile optimization, and specifically relates to an optimization method for a braking fan of a turbine expander. The method comprises the following steps: generating a sample parameter group based on one-dimensional parameters and set fixed parameters; importing the sample parameter group into a one-dimensional performance prediction program to obtain one-dimensional performance prediction parameters corresponding to the sample parameter group; combining the one-dimensional parameters with the one-dimensional performance prediction parameters to form a set of samples, and combining multiple sets of samples to generate a sample library. Importing the sample library into a neural network interface for fitting to obtain a neural network; importing the well-fitted neural network into an optimization algorithm interface for optimization. The present invention can be well matched with the refrigerating capacity of the turbine expander, and quickly find a one-dimensional parameter design scheme of a braking fan with good performance under the condition of meeting the braking power; this method can match the refrigerating capacity of the turbine expander, and at the same time take into account technical problems such as too large a fan size and poor performance caused by a large refrigerating capacity.

Description

Technical Field

[0001] The present invention relates to the technical field of turbine machinery blade profile optimization, and more particularly, to an optimization method for a braking fan of a turbine expander. Background Art

[0002] With the continuous development of large scientific installations and aerospace technology, new requirements are constantly put forward in the field of cryogenic refrigeration. Refrigerating capacity is one of the important requirements. The refrigerating capacity of Stirling-type or G-M refrigerators is relatively small, and turbine expansion refrigeration can provide ideal refrigerating capacity. Especially in the temperature range below 20K, almost only turbine expansion refrigeration can achieve refrigerating capacity of hundreds or even tens of thousands of watts. As the core component of the turbine expander, the turbine affects the efficiency of the whole machine. When the turbine extracts the energy from the working medium, corresponding components are needed to dissipate the extracted energy to maintain energy balance, which is called the braking part. Common braking methods include oil braking, electric motor braking, turbine braking, etc. Among them, oil braking is not suitable for high speeds and low temperatures, and the control system of electric motor braking is relatively complex. Therefore, turbine braking is generally used in cryogenic turbines. Usually, turbine braking only considers whether it can meet the energy dissipation requirements, without considering its efficiency, pressure ratio, flow rate and other performance parameters. This is acceptable in general small-power cryogenic turbines. However, when the power of the cryogenic turbine is large, the corresponding turbine braking power is also large. If the performance of the braking turbine is still not considered at this time, the overall economy will decline.

[0003] The braking turbine is a centrifugal impeller. The function of the braking turbine is to dissipate the energy generated at the turbine end. Compared with traditional turbines, efficiently dissipating energy becomes the primary goal. At this time, whether the braking power meets the design requirements, whether the working range is sufficient, and the efficiency level become important considerations. The design goal is to achieve as high an efficiency as possible and as wide a working range as possible under the condition of meeting the designed braking power. The design process of the centrifugal impeller generally includes one-dimensional preliminary design, two-dimensional through-flow design, full three-dimensional blade modeling, and CFD calculation to verify the performance, and this process is repeated until the performance meets the design requirements. One-dimensional preliminary design: According to one-dimensional aerodynamic and thermodynamic relationships, determine the basic geometric dimensions of the inlet and outlet of the centrifugal compressor, and use one-dimensional performance prediction methods to predict the performance of the preliminary design scheme under design conditions and off-design conditions, and preliminarily judge whether it meets the design requirements. As an important link in the design of centrifugal compressors, one-dimensional preliminary design has a decisive influence on the performance of the impeller.

[0004] For a long time, the research focus of turboexpanders has been on the impeller, bearings, etc., and insufficient attention has been paid to the braking end. However, the braking end also affects the performance of the whole machine. Improving the efficiency of the braking fan, reducing the volume of the braking end, increasing the heat transfer efficiency of water cooling, etc. will all improve the performance of the whole machine. The core component of the braking end is the impeller of the braking fan. However, the design of the impeller of the braking fan is usually rather arbitrary and does not consider the impact on the overall performance. Therefore, there is an urgent need for an impeller design scheme that takes into account the unique properties during braking and has good performance at the same time;

[0005] The braking fan is an important part of the turboexpander. In general small-power turboexpanders, the braking fan only needs to meet the braking power. However, as the refrigeration capacity increases, the braking power increases accordingly. And the existing design methods of braking fans are too rough. For high-power turboexpansion refrigerators, not only the braking power needs to be met, but also requirements for the size, performance, etc. of the braking fan are put forward. Therefore, the present invention scheme is proposed. Summary of the Invention

[0006] The embodiment of the present invention provides an optimization method for a braking fan used in a turboexpander, which can match the refrigeration capacity of the turboexpander and at the same time take into account the technical problems of the too large size and poor performance of the fan caused by the large refrigeration capacity.

[0007] According to an embodiment of the present invention, an optimization method for a braking fan of a turboexpander is provided, including the following steps:

[0008] Generate a sample parameter group based on one-dimensional parameters and set fixed parameters;

[0009] Import the sample parameter group into a one-dimensional performance prediction program, and obtain one-dimensional performance prediction parameters corresponding to the sample parameter group based on the one-dimensional performance prediction program;

[0010] The one-dimensional parameters are combined with the one-dimensional performance prediction parameters to form a group of samples, and multiple samples are combined to generate a sample library.

[0011] Import the sample library into a neural network interface for fitting to obtain a neural network;

[0012] Import the fitted neural network into an optimization algorithm interface for optimization.

[0013] Further, after importing the fitted neural network into the optimization algorithm interface for optimization, it also includes:

[0014] Obtain the optimized one-dimensional performance parameters;

[0015] Import the optimized one-dimensional performance parameters into the one-dimensional performance prediction program to verify the optimization effect.

[0016] Further, in the step of generating a sample parameter group based on one-dimensional parameters and set fixed parameters, it includes:

[0017] Select one-dimensional parameters and fixed parameters;

[0018] Determine the range of values for each one-dimensional parameter;

[0019] Select a data point from within the range of values for each one-dimensional parameter;

[0020] Combine the selected data points with the set fixed parameters to form a sample parameter group.

[0021] Further, each one-dimensional parameter in the sample parameters can vary simultaneously, or only some or several of them can vary as needed.

[0022] Further, in the step of importing the sample parameter group into a one-dimensional performance prediction program to obtain the one-dimensional performance parameters corresponding to the sample parameter group, it includes:

[0023] Import the one-dimensional parameters and fixed parameters into the one-dimensional performance prediction program;

[0024] Determine the one-dimensional structure parameters according to the one-dimensional parameters and fixed parameters imported into the one-dimensional prediction program;

[0025] Calculate the inlet aerodynamic parameters and outlet aerodynamic parameters according to the one-dimensional structure parameters, the estimated initial flow rate and initial efficiency included in the fixed parameters;

[0026] Calculate the efficiency according to the one-dimensional structure parameters, inlet aerodynamic parameters and outlet aerodynamic parameters;

[0027] If the error between the calculated efficiency and the initial efficiency is less than the set residual value, proceed to the next step of calculating the braking power, otherwise, recalculate the inlet aerodynamic parameters and outlet aerodynamic parameters with the calculated efficiency as the initial efficiency;

[0028] Calculate the braking power according to the efficiency. If the error between the calculated braking power and the initial power is less than the set residual value, output the one-dimensional performance prediction parameters, otherwise, recalculate the initial flow rate with the efficiency and braking power, generate a new initial flow rate value and continue the calculation until the one-dimensional performance prediction parameters are output.

[0029] Further, the one-dimensional performance prediction program is written based on a centrifugal one-dimensional loss model.

[0030] Further, quickly generate a set number of samples based on the written sample library program, and combine the set number of samples to generate a sample library.

[0031] Further, in the step of importing the sample library into a neural network interface for fitting to obtain a neural network, it includes:

[0032] Screen the samples in the sample library;

[0033] Import the screened sample library into the neural network interface;

[0034] Start fitting after setting the basic parameters;

[0035] Repeat fitting until a neural network meeting the accuracy requirements appears.

[0036] Further, in the step of importing the fitted neural network into the optimization algorithm interface for optimization, it includes:

[0037] Import the fitted neural network into the optimization algorithm interface;

[0038] Select one or several objectives to be optimized as needed; the objectives include flow rate, efficiency, pressure ratio, and power;

[0039] Select one-dimensional parameters and the range values of the one-dimensional parameters;

[0040] Set the basic parameters of the optimization algorithm;

[0041] Optimize the objectives to obtain the optimized one-dimensional performance parameters.

[0042] Further, select the genetic algorithm and the particle swarm optimization algorithm respectively to optimize the optimization objectives to compare the optimization effects with each other.

[0043] The beneficial effects of the present invention are as follows: Generate a sample parameter group based on one-dimensional parameters and set fixed parameters; import the sample parameter group into the one-dimensional performance prediction program to obtain the one-dimensional performance prediction parameters corresponding to the sample parameter group; combine the one-dimensional parameters with the one-dimensional performance prediction parameters to form a group of samples, and quickly generate a set number of samples based on the sample library program and form a sample library; import the sample library into the neural network interface for fitting to obtain a neural network; import the fitted neural network into the optimization algorithm interface for optimization. By proposing the design of this optimization method for the fan applicable to the use of a turboexpander, it can be well matched with the refrigerating capacity of the turboexpander, quickly obtain the performance of the preliminary design of one-dimensional parameters, and quickly find a one-dimensional parameter design scheme of a braking fan with good performance under the condition of meeting the braking power; this method can match the refrigerating capacity of the turboexpander and at the same time take into account technical problems such as the too large size and poor performance of the fan caused by a large refrigerating capacity; taking high efficiency as an example, it can find the efficiency matching the refrigerating capacity within a short time, and significantly improve the one-dimensional parameter design scheme of the braking fan. Description of the Drawings

[0044] The accompanying drawings described herein are used to provide a further understanding of the present invention, and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute an improper limitation of the present invention. In the drawings:

[0045] Figure 1 is a flowchart of an optimization method for a braking fan of a turbine expander according to the present invention;

[0046] Figure 2 is a flowchart of the design of a one-dimensional performance prediction program according to the present invention;

[0047] Figure 3 is a schematic diagram of the one-dimensional parameter range values according to the present invention;

[0048] Figure 4 is a schematic diagram of generating a sample parameter group from one-dimensional parameters and set fixed parameters according to the present invention;

[0049] Figure 5 is a schematic diagram of selecting data points from the one-dimensional parameter range values according to the present invention;

[0050] Figure 6 is a schematic diagram of the one-dimensional performance prediction parameters according to the present invention;

[0051] Figure 7 is a schematic diagram of the efficiency in the one-dimensional parameter range values according to the present invention;

[0052] Figure 8 is a schematic diagram of setting basic parameters after importing the sample library into the neural network interface according to the present invention;

[0053] Figure 9 and Figure 10 is a schematic diagram of the fitting quality of fitting after importing the sample library into the neural network interface according to the present invention;

[0054] Figure 11 is a schematic diagram of the parameter settings for particle swarm optimization according to the present invention;

[0055] Figure 12 is a schematic diagram of the optimization effect according to the present invention. Detailed implementation manners

[0056] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0057] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0058] According to an embodiment of the present invention, an optimization method for a turbine expansion machine braking fan is provided. Refer to Figure 1 , including the following steps:

[0059] S101: Generate a sample parameter group based on one-dimensional parameters and set fixed parameters;

[0060] S102: Import the sample parameter group into a one-dimensional performance prediction program to obtain one-dimensional performance prediction parameters corresponding to the sample parameter group;

[0061] S103: Combine the one-dimensional parameters with the one-dimensional performance prediction parameters to form a group of samples, and combine multiple samples to generate a sample library

[0062] S104: Import the sample library into a neural network interface for fitting to obtain a neural network;

[0063] S105: Import the fitted neural network into an optimization algorithm interface for optimization.

[0064] In the embodiment, a sample parameter group is generated based on one-dimensional parameters and set fixed parameters; and the sample parameter group is imported into a one-dimensional performance prediction program, and then one-dimensional performance prediction parameters corresponding to the sample parameter group are obtained; the one-dimensional parameters are combined with the one-dimensional performance prediction parameters to form a group of samples, and a set number of samples are quickly generated based on the sample library program and form a sample library; the sample library is imported into a neural network interface for fitting to obtain a neural network; the fitted neural network is imported into an optimization algorithm interface for optimization. By proposing such a design of an optimization method for a fan applicable to a turbine expansion machine, it can be well matched with the refrigerating capacity of the turbine expansion machine, quickly obtain the performance of the preliminary design of one-dimensional parameters, and quickly find a one-dimensional parameter design scheme of a braking fan with good performance under the condition of meeting the braking power; this method can match the refrigerating capacity of the turbine expansion machine and at the same time take into account technical problems such as too large a fan size and poor performance caused by a large refrigerating capacity; taking high efficiency as an example, it can find the efficiency matching the refrigerating capacity in a short time, and significantly improve the one-dimensional parameter design scheme of the braking fan.

[0065] In the embodiment, after importing the fitted neural network into the optimization algorithm interface for optimization, the following steps are further included:

[0066] The first step: obtaining the optimized one-dimensional performance parameters;

[0067] The second step: importing the optimized one-dimensional performance parameters into the one-dimensional performance prediction program to verify the optimization effect.

[0068] To clearly understand the optimized result, the optimized one-dimensional performance parameters are imported into the one-dimensional performance prediction program for verification; as Figure 12 shown, the original efficiency is 82.92%, the optimized efficiency is 87.61%, and the efficiency is increased by 5.65%. Based on this, it can be known that the efficiency has been significantly improved.

[0069] In the embodiment, in the step of generating the sample parameter group based on the one-dimensional parameter and the set fixed parameters, the following steps are included:

[0070] The first step: selecting the one-dimensional parameter and the fixed parameters;

[0071] The second step: determining the change range values of each one-dimensional parameter;

[0072] The third step: selecting a data point from within the range values of each one-dimensional parameter;

[0073] The fourth step: combining the selected data points with the set fixed parameters to form the sample parameter group.

[0074] As Figures 2 to 5 shown, the one-dimensional parameters include but are not limited to the inlet size, outlet size, inlet blade installation angle, outlet blade installation angle, number of blades, hub diameter, hub blade angle, impeller outer diameter, outlet blade height, etc.; the fixed parameters include but are not limited to the set braking power, estimated initial flow rate, estimated initial efficiency, inlet gas thermodynamic parameters, inlet total temperature, inlet total pressure, and gas constant, etc.

[0075] Select the range of the one-dimensional parameter. As Figure 3 shown, when selecting the parameter range, it should be appropriate. When selecting the range, the mutual constraints between the one-dimensional parameters should be considered, such as the inlet size being smaller than the outlet size, etc.; select the data points from within the range of the one-dimensional parameter. If there are obvious irrationalities in the current one-dimensional parameter, they should be excluded. As Figure 4 shown; combine the selected one-dimensional parameter and the fixed parameters to produce the sample parameter group and import it into the table as Figure 5 shown, for example, import it into an Excel table.

[0076] In the embodiment, each one-dimensional parameter in the sample parameters can change simultaneously, or only some or several of them can change according to needs.

[0077] For example, only select the change in the number of blades while keeping other one-dimensional parameters unchanged; or select the inlet size, outlet size, inlet blade installation angle, and outlet blade installation angle to change together while keeping other one-dimensional parameters unchanged.

[0078] In the embodiment, in the step of importing the sample parameter group into the one-dimensional performance prediction program to obtain the one-dimensional performance prediction parameters corresponding to the sample parameter group, it includes:

[0079] The first step: Import the one-dimensional parameters and fixed parameters into the one-dimensional performance prediction program;

[0080] The second step: Determine the one-dimensional structure parameters according to the one-dimensional parameters and fixed parameters imported into the one-dimensional prediction program;

[0081] The third step: Calculate the inlet aerodynamic parameters and outlet aerodynamic parameters according to the one-dimensional structure parameters, the estimated initial flow rate and initial efficiency included in the fixed parameters;

[0082] The third step: Calculate the efficiency according to the one-dimensional structure parameters, inlet aerodynamic parameters and outlet aerodynamic parameters;

[0083] The fifth step: If the error between the calculated efficiency and the initial efficiency is less than the set residual value, proceed to the next step of calculating the braking power; otherwise, use the calculated efficiency as the initial efficiency to recalculate the inlet aerodynamic parameters and outlet aerodynamic parameters;

[0084] The sixth step: Calculate the braking power according to the efficiency. If the error between the calculated braking power and the estimated initial power is less than the set residual value, output the one-dimensional performance parameters; otherwise, recalculate the initial flow rate with the efficiency and braking power, generate a new initial flow rate value and continue the calculation until the one-dimensional performance prediction parameters are output.

[0085] As Figure 2 shown, the design flow steps of the above one-dimensional performance prediction program are as follows:

[0086] S201: Input one-dimensional parameters and fixed parameters;

[0087] S202: Set the braking power;

[0088] S203: Estimate the initial flow rate;

[0089] S204: Estimate the initial efficiency;

[0090] S205: Calculate the inlet aerodynamic parameters and outlet aerodynamic parameters;

[0091] S206: Calculate the efficiency, and determine whether the error between the calculated efficiency and the estimated initial efficiency is less than the set residual value. If it is less, proceed to the next step; otherwise, return to S204 with the calculated efficiency as the initial value;

[0092] S207: Calculate the braking power based on the current efficiency and initial flow rate, and determine whether the error between the braking power and the initial flow rate is less than the set residual value. If so, proceed to the next step. Otherwise, calculate the flow rate based on the current efficiency and the calculated braking power, generate a new initial flow rate value and return to S203.

[0093] S208: Output one-dimensional performance prediction parameters.

[0094] Among them, a one-dimensional performance prediction program for high-speed, small-flow brake fans is written based on the currently widely recognized one-dimensional loss model of centrifugal fans; the output one-dimensional performance prediction parameters mainly include but are not limited to flow, efficiency, power and pressure ratio, etc.; among them, power and efficiency are as follows Figure 6 shown.

[0095] In the embodiment, a set number of samples is quickly generated based on the written sample library program, and the set number of samples are combined to generate a sample library. A set of randomly selected one-dimensional parameters plus the corresponding one-dimensional performance prediction parameters constitute a set of samples. A set number of samples are generated through the written sample library generation program to form a sample library. For example, the number of samples is set to 10,000, the number of valid samples is 4,088, and samples are randomly generated and one-dimensional performance prediction is performed. The prediction efficiency is as follows: Figure 7 shown.

[0096] In an embodiment, the step of importing the sample library into the neural network interface for fitting to obtain the neural network includes:

[0097] Step 1: Screen the samples in the sample library;

[0098] Step 2: Import the screened sample library into the neural network interface;

[0099] Step 3: Start fitting after setting basic parameters;

[0100] Step 4: Repeat the fitting process until a neural network that meets the accuracy requirements emerges.

[0101] Specifically, before neural network fitting, the sample library is screened to remove samples with low efficiency and flow rates that are too far from the estimated flow rate. Removing extreme points can significantly improve the fitting quality. The screened samples are imported into the neural network interface; then, basic parameters are set, such as the number of iterations, learning rate, etc. Figure 8 As shown; then start fitting. Due to the uncertainty of the fitting process, the fitting should be repeated many times until a neural network that meets the accuracy is found. If not, return to the sample library generation program to increase the sample library capacity and the ability to filter out extreme points. The BP neural network is selected for fitting, and the fitting quality is shown in Figure 9 andFigure 10 as shown

[0102] In the embodiment, in the step of importing the fitted neural network into the optimization algorithm interface for optimization, it includes:

[0103] Import the fitted neural network into the optimization algorithm interface;

[0104] Select one or several objectives to be optimized as needed; the objectives include flow rate, efficiency, pressure ratio, and power;

[0105] Select one-dimensional parameters and the range values of the one-dimensional parameters;

[0106] Set the basic parameters of the optimization algorithm;

[0107] Optimize the objectives to obtain the optimized one-dimensional performance parameters.

[0108] Specifically, there are a variety of current intelligent optimization algorithms, and their optimization effects are different. In this solution, the genetic algorithm and the particle swarm algorithm are selected for separate optimization to verify the optimization effects with each other. Import the fitted neural network into the optimization algorithm interface, and the optimization objectives can be changed according to the actual situation; for example, the optimization objectives can be any one or several of the performance parameters such as flow rate, pressure ratio, and efficiency; parameter constraints need to be considered during the optimization process, and both the structural performance and the braking power are considered. That is, first, the braking power needs to meet the design requirements, and second, the structural performance is considered, such as the circumferential diameter of the impeller, etc. Select the one-dimensional parameters and the range of the one-dimensional parameters, which is basically the same as when the samples are randomly generated; set the basic parameters of the optimization algorithm, such as the initial population size, etc. After setting, perform optimization to obtain the one-dimensional performance parameters, compare the results of the two algorithms, use the particle swarm optimization, and the parameter settings are as Figure 11 , and the optimization effect is as Figure 12 . After optimization, obtain the one-dimensional performance parameters, and import the optimized one-dimensional performance parameters into the one-dimensional performance prediction program to verify the optimization effect.

[0109] Furthermore, during the initial optimization, adjust the range of change of the one-dimensional parameters to be as large as possible. Take the one-dimensional performance parameters obtained after optimization as the initial value, take equal-length ranges on both sides with this initial value as the center point, and the secondary change range is smaller than the initial change range, and perform optimization and iterate multiple times until the optimization result converges.

[0110] As Figure 12 shown, the optimized efficiency is 87.61%, the original efficiency is 82.92%, and the efficiency is increased by 5.65%, indicating a significant improvement in efficiency.

[0111] In this example, the optimization objective is efficiency, and the inlet hub diameter, inlet shroud diameter, outlet impeller diameter, outlet blade height, and outlet blade angle are one-dimensional parameters. However, this does not mean that the optimization objective is only efficiency. It can be other single or multiple performance parameters. At the same time, the one-dimensional parameters are not limited to the above parameters and can also be other single or multiple one-dimensional parameters.

[0112] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. An optimization method for a braking fan of a turbine expander, characterized in that, It includes the following steps: Generate a sample parameter group based on one-dimensional parameters and set fixed parameters; Import the sample parameter group into a one-dimensional performance prediction program, and obtain one-dimensional performance prediction parameters corresponding to the sample parameter group based on the one-dimensional performance prediction program; The one-dimensional parameters are combined with the one-dimensional performance prediction parameters to form a set of samples, and multiple sets of the samples are combined to generate a sample library; Import the sample library into a neural network interface for fitting to obtain a neural network; Import the fitted neural network into an optimization algorithm interface for optimization; where: In the step of importing the sample parameter group into the one-dimensional performance prediction program and obtaining the one-dimensional performance parameters corresponding to the sample parameter group, it includes: Import the one-dimensional parameters and the fixed parameters into the one-dimensional performance prediction program; Determine one-dimensional structure parameters according to the one-dimensional parameters and the fixed parameters imported into the one-dimensional prediction program; Calculate inlet aerodynamic parameters and outlet aerodynamic parameters according to the one-dimensional structure parameters, the estimated initial flow rate and initial efficiency included in the fixed parameters; Calculate the efficiency according to the one-dimensional structure parameters, the inlet aerodynamic parameters and the outlet aerodynamic parameters; If the error between the calculated efficiency and the initial efficiency is less than the set residual value, proceed to the next step of calculating the braking power, otherwise, use the calculated efficiency as the initial efficiency to recalculate the inlet aerodynamic parameters and the outlet aerodynamic parameters; Calculate the braking power according to the efficiency. If the error between the calculated braking power and the initial power is less than the set residual value, output the one-dimensional performance prediction parameters, otherwise, recalculate the initial flow rate with the efficiency and the braking power to generate a new initial flow rate value and continue the calculation until the one-dimensional performance prediction parameters are output; In the step of importing the fitted neural network into the optimization algorithm interface for optimization, it includes: Import the fitted neural network into the optimization algorithm interface; Select one or several objectives to be optimized as needed; the objectives include flow rate, efficiency, pressure ratio and power; Select one-dimensional parameters and range values of the one-dimensional parameters; Set the basic parameters of the optimization algorithm; Optimize the objectives to obtain the optimized one-dimensional performance parameters.

2. The optimization method of the turbine expander braking fan according to claim 1, characterized in that After importing the fitted neural network into the optimization algorithm interface for optimization, it further includes: Obtain the optimized one-dimensional performance parameters; Import the optimized one-dimensional performance parameters into the one-dimensional performance prediction program to verify the optimization effect.

3. The optimization method of the turbine expansion machine braking fan according to claim 1, characterized in that, In the step of generating a sample parameter group based on one-dimensional parameters and set fixed parameters, it includes: Select the one-dimensional parameters and the fixed parameters; Select the range values of the change of each one-dimensional parameter; Select a data point from within the range values of each one-dimensional parameter; Combine the selected data points with the set fixed parameters to form the sample parameter group.

4. The optimization method of the turbine expansion machine braking fan according to claim 3, characterized in that, Each one-dimensional parameter in the sample parameters can change simultaneously, or only some of them can change according to needs.

5. The optimization method of the braking fan of the turbine expander according to claim 1, characterized in that, The one-dimensional performance prediction program is written based on a centrifugal one-dimensional loss model.

6. The optimization method of the turbine expansion machine braking fan according to claim 1, wherein Based on the written sample library program, quickly generate the set number of the samples, and combine the set number of samples to generate the sample library.

7. The optimization method of the braking fan of the turbine expander according to claim 1, wherein In the step of importing the sample library into the neural network interface for fitting to obtain the neural network, it includes: Screen the samples in the sample library; Import the screened sample library into the neural network interface; Start fitting after setting the basic parameters; Repeat the fitting until the neural network that meets the accuracy requirements appears.

8. The optimization method of the turbine expansion machine braking fan according to claim 1, characterized in that, Respectively select the genetic algorithm and the particle swarm algorithm to optimize the optimization target respectively to compare the optimization effects with each other.

Images