Method for optimizing loss of external pipeline of reignition gas compressor

By constructing the basic loss optimization model and agent model, combining multiple point-adding criteria and optimization algorithms, the problems of low efficiency and difficult flow matching in the external pipeline design of heavy-duty gas turbines are solved, and the total pressure loss is significantly reduced and the design efficiency is improved. .

CN119067003APending Publication Date: 2024-12-03SHANGHAI JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410936687.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing technology has problems in the design of external pipelines of heavy-duty gas turbines with low efficiency, long time consumption and difficulty in matching the flow matching scheme of front, middle and rear stage pumping pipelines. The traditional design method is limited to a single pipeline design, making it difficult to optimize the total pressure loss. .

Method used

By constructing the basic loss optimization model of the reignitment compressor, using Latin hypercube sampling and proxy models (such as Kriging and BPNN) to establish an approximate relationship between input design parameters and total pressure loss, combined with a variety of pointing criterion and optimization algorithms, iterate over time Reconstruction and optimization process to realize parameterized modeling and flow matching of pipeline systems.

Benefits of technology

It significantly reduces the total pressure loss of the external pipeline of the reignitment compressor and improves the design efficiency. The optimized pipeline shows a larger flow range at different flow ranges, with a loss difference of 3% at low Mach numbers, high The speed of subsonic is 6%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067003A_ABST
    Figure CN119067003A_ABST
Patent Text Reader

Abstract

A loss optimization method for an external pipeline of a reignition gas compressor comprises the following steps: constructing a basic loss optimization model of the reignition gas compressor, determining constraint conditions of the basic loss optimization model, performing parametric modeling on the external pipeline of the reignition gas compressor, and generating an initial sample in a design space through a Latin hypercube sampling method; obtaining a total pressure loss coefficient corresponding to the initial sample through one-dimensional simulation; then, a Kirging agent model and a BPNN agent model are used in parallel to establish an approximate relation between the input design parameters of the external pipeline of the lifting gas compressor and the corresponding total pressure loss coefficients, and the input design parameters, enabling the total pressure loss to be minimum, on the agent models are searched based on multiple point adding criteria and an optimization algorithm; and comparing a proxy model result with a one-dimensional simulation result, and realizing optimization through iterative reconstruction. According to the method, an optimization scheme with the lowest total pressure loss can be accurately and stably converged after iteration, the optimization process is stable, and the optimization result is reliable.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a technology in the field of re-ignition compressors, in particular to a method for optimizing external pipeline losses of re-ignition compressors. Background Art

[0002] Heavy-duty gas turbines occupy an important strategic position in the energy and power industry. When starting a combined cycle and maintaining stable operation, heavy-duty gas turbines need to draw air from the compressor and distribute it to various hot end components through external pipelines to take into account the functions of anti-surge and cooling sealing. The design of external pipelines affects efficiency, compressor workmanship and exhaust emission levels. However, traditional pipeline design methods rely more on human experience, are time-consuming and inefficient. Summary of the invention

[0003] The present invention aims at the shortcomings of the prior art in optimizing the external piping design or only simulating the intercooled cycle gas turbine, and proposes a method for optimizing the external piping loss of the re-ignition compressor, which comprehensively solves the defects that the current piping loss allocation method in the optimization design of the re-ignition external piping is difficult to match the current heavy-duty gas turbine, and the conventional design method is limited to a single pipeline design, which is difficult to match the flow matching scheme of the front, middle and rear stage exhaust pipelines, and cannot be optimized using an optimization function. Compared with the prior art, the present invention can accurately and stably converge to an optimization scheme with the lowest total pressure loss after iteration, and the optimization process is stable and the optimization results are reliable.

[0004] The present invention is achieved through the following technical solutions:

[0005] The invention relates to a method for optimizing the loss of an external pipeline of a re-ignition compressor. The method comprises the following steps: constructing a basic loss optimization model of a re-ignition compressor and determining its constraint conditions, performing parameterized modeling on the external pipeline of the re-ignition compressor, generating initial samples in a design space by a Latin hypercube sampling method, and obtaining a total pressure loss coefficient corresponding to the initial samples by one-dimensional simulation; then using a Kirging proxy model and a BPNN proxy model in parallel to establish an approximate relationship between input design parameters of the external pipeline of the re-ignition compressor and the total pressure loss coefficient corresponding to the initial samples, finding input design parameters on the proxy model that minimize the total pressure loss based on a plurality of point addition criteria and an optimization algorithm, comparing the proxy model result with the one-dimensional simulation result, and realizing optimization by iterative reconstruction.

[0006] The basic loss optimization model is specifically: the total loss of the external pipeline Where: Δp i is the pressure loss along the pipeline at different sections and local pressure loss λ is the Darcy friction factor, l i is the length of the pipeline, D is the diameter of the pipeline, is the average flow rate, θ i is the turning angle of the elbow, is the local pressure loss coefficient, and R is the radius of curvature of the center line of the elbow.

[0007] The constraints include: total flow constraint of engine cooling, branch flow constraint, pipeline system flow constraint, pipeline diameter constraint, pipeline quantity constraint and elbow parameter constraint.

[0008] The parametric modeling refers to: parameterizing key design parameters of the pipeline, including the pipeline diameter, the number of bends, the number of outlets, and the number of pipelines.

[0009] The one-dimensional simulation refers to: using numerical simulation method to solve the pipeline computational fluid dynamics flow equation: in: , where: ρ is gas density; u is velocity; p is pressure; T is temperature; h is convective heat transfer coefficient, T w is the wall temperature, T ∞ Take the local mainstream average temperature and heat flux density Friction coefficient p in 、p out are the average pressures of the inlet and outlet, L is the channel length, and D is the pipe diameter; parameters related to the pipe diameter

[0010] The iterative reconstruction means that when the error between the proxy model result and the one-dimensional simulation result is less than 3%, the design parameter is output and the optimization is stopped; otherwise, the sample point is added to the sample library, a new proxy model is reconstructed and iterated.

[0011] The BPNN proxy model is a back propagation neural network proxy model, including: an input layer, a hidden layer and an output layer. The initial output data output by the BPNN proxy model is compared with the data in the training set, and the comparison result is back-propagated from the output layer to the hidden layer and the input layer to adjust the connection weights so that the error between the output data and the training set data is minimized.

[0012] The multiple point-adding criteria refer to: three parallel point-adding criteria of MP, EI, and RMSE, among which: the MP point-adding criterion is the minimum value criterion of the objective function, which assumes that the proxy model is accurate enough. After the proxy model is established, the minimum value of the objective function on the proxy model is directly found. The EI point-adding criterion is the expected improvement criterion, which finds the maximum value of the expected function of the difference between the currently known minimum value and the unknown point value on the proxy model. The RMSE point-adding criterion is the maximum root mean square error criterion, which adds sample points where the proxy model has the largest error, can improve the overall accuracy of the model, and can better find the global optimal solution.

[0013] The hybrid optimization algorithm is a hybrid optimization algorithm constructed based on Latin hypercube sampling, genetic algorithm (GA) and sequential quadratic programming algorithm (SQP) to find the optimal point on the proxy model. The present invention uses genetic algorithm and Latin hypercube sampling to generate multiple initial optimal points respectively, which are used as the initial value of the SQP algorithm, and then optimized again to obtain multiple optimal points. On this basis, comparison is made and the final optimal point is output.

[0014] The present invention relates to a system for realizing the above method, comprising: an initial sample generation unit, a one-dimensional simulation unit, a proxy model unit and an optimization algorithm unit, wherein: the initial sample generation unit uses a Latin hypercube sampling method to generate a plurality of initial samples according to constraint conditions; the one-dimensional simulation unit calculates the initial samples to obtain calculation results corresponding to the initial samples, that is, the weighted total pressure loss of a plurality of pipelines; the proxy model unit establishes a Kriging proxy model and a BPNN proxy model in parallel according to the design variables and calculation results of the initial samples to obtain the corresponding relationship between the input design variables and the output weighted total pressure loss; the optimization algorithm unit performs optimization work according to the established proxy model to obtain the input design variables that minimize the weighted total pressure loss. Technical Effects

[0015] The present invention uses a parameterized modeling method for irregular multi-pipelines with complex layout trends; based on the multi-point criterion and multi-optimization algorithm, the pipeline system after hybrid optimization not only improves the aerodynamic performance of its design point, but also improves the aerodynamic performance of its non-design point, so that the optimized pipeline has a larger flow range. When the Mach number of the pipeline increases from 0.2Ma to 0.9Ma, the loss gap gradually increases. After shortening the pipeline length and increasing the pipeline diameter, the total pressure loss difference of the pipeline before and after optimization at low Mach number is about 3%, and at high subsonic speed, the loss difference gradually increases to 6%. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a test result diagram of various dot-adding criteria of the present invention;

[0017] Figure: (a) MP plus point criterion test results, (b) EI plus point criterion test results, (c) test function three-dimensional graph, (d test function cloud graph, (e) MP+EI plus point criterion test results, (f) MP+EI+RMSE plus point criterion test results

[0018] Figure 2 is a schematic diagram of a reference member of the present invention;

[0019] Figure 3 It is a flow chart of the present invention;

[0020] Figure 4 This is a comparison chart of optimization results;

[0021] Figure 5 This is a comparison chart of the pipelines before and after optimization;

[0022] In the figure: (a) pipeline prototype, (b) optimized pipeline

[0023] Figure 6 This is a schematic diagram of the gas source test bench;

[0024] In the figure: A, Roots blower, B, inlet valve, C, diffusion section, D, voltage stabilizing section, E, damping net, F, honeycomb rectifier grid, G, contraction section;

[0025] Figure 7 This is a schematic diagram of the extended section of the surge tank;

[0026] In the figure: (a) straight section of the extended section, (b) curved section of the extended section;

[0027] Figure 8 This is a schematic diagram of a pressure stabilizing tank;

[0028] In the figure: (a) surge tank, (b) surge tank exhaust port;

[0029] Fig. 9 This is a schematic diagram of the external test pipeline;

[0030] In the figure: (a) before the external test pipeline is installed, (b) after the external test pipeline is installed;

[0031] Fig.10 This is a schematic diagram of the pressure collection hose;

[0032] In the figure: (a) pressure collection hose, (b) pressure collection hose after installation;

[0033] Fig.11 It is a schematic diagram of a pressure scanning valve;

[0034] In the figure: (a) pressure scanning valve, (b) pressure scanning valve after installation;

[0035] Fig.12 This is a schematic diagram of the temperature and atmospheric pressure acquisition equipment;

[0036] In the figure: (a) total temperature acquisition equipment, (b) digital ambient atmospheric pressure acquisition equipment;

[0037] Fig.13 It is a schematic diagram of the effect of the embodiment;

[0038] In the figure: (a) the loss of 1 pipe changes with the Mach number before and after optimization, (b) the loss of 2 pipes changes with the Mach number before and after optimization, (c) the loss of 4 pipes changes with the Mach number before and after optimization, (d) the loss of T-type pipes changes with the Mach number before and after optimization. DETAILED DESCRIPTION

[0039] like Figure 2 As shown, this embodiment is directed to the flow path distribution of the secondary air system of the F-class heavy-duty gas turbine, wherein: the compressor part is provided with three external air extraction points. The front stage air extraction point has one air extraction pipeline, and the middle stage and rear stage air extraction points each have two air extraction pipelines, a total of five air extraction pipelines, wherein: the first external air extraction pipeline of the front stage connects the front stage of the compressor to the low-pressure cooling chamber of the turbine, the second and third external air extraction pipelines of the middle stage connect the middle stage of the compressor to the medium-pressure cooling chamber of the turbine, and the fourth and fifth external air extraction pipelines of the rear stage connect the rear stage of the compressor to the high-pressure cooling chamber of the turbine.

[0040] The cooling flow rate of the intermediate stage and the rear stage exhaust pipelines is controlled by butterfly valves.

[0041] According to engineering practice, the three external pipelines are involved in the cooling of the hot end components, and the front stage and the middle stage also play a role in preventing surge. The unstable operating point of the gas turbine is generally at low speed and low power, and surge often occurs in the front stage. Therefore, when preventing surge, the front stage and the middle stage exhaust pipelines are used for exhaust.

[0042] like Figure 3 As shown, a method for optimizing the external pipeline loss of a re-ignition compressor involved in this embodiment includes the following steps:

[0043] Step 1. Determine the optimization benchmark solution: The optimization problem of the pipeline is to determine the minimum pressure loss under the conditions of pipeline system flow, pipeline diameter, bend parameters, length, and number of pipelines. The external exhaust pipeline of the heavy-duty gas turbine consists of the front stage, the middle stage, and the rear stage. Therefore, after determining a certain optimization analysis solution, a basic loss optimization model is formed. Optimizing pipeline loss can reduce the required flow rate without changing the cooling effect, and the total loss of the external pipeline Where: Δp i is the pressure loss along the pipeline at different sections and local pressure loss λ is the Darcy friction factor, li is the length of the pipeline, D is the diameter of the pipeline, is the average flow rate, θ i is the turning angle of the elbow, is the local pressure loss coefficient, and R is the radius of curvature of the centerline of the elbow. Finally, the total pressure loss of the pipeline is the sum of the local pressure loss and the overall pressure loss.

[0044] Step 2: Determine the constraints. Specifically, the total flow rate of the engine cooling is generally limited to a percentage of the compressor inlet flow rate. After the total pipeline flow rate is determined, the branch flow rate constraint is Pipeline system flow constraint q min <<q i <<q max , pipeline diameter constraint: l min <<l i <<l max , pipeline quantity constraint 3<<m<<12, bending parameter constraint 90°<<θ<<150°.

[0045] Step 3: parametrically model the pipeline, use the Latin hypercube sampling method, generate 200 initial samples, and then calculate the calculation results corresponding to the initial samples, that is, the weighted total pressure loss of multiple pipelines.

[0046] Step 4: Using the above initial samples, establish the Kriging proxy model and the BPNN proxy model to obtain the corresponding relationship between the input design variables and the output weighted total pressure loss.

[0047] Step 5: Based on the proxy model, find the input design parameters that minimize the total pressure loss on the self-built hybrid optimization platform. Compare the results obtained by the proxy model with the one-dimensional simulation results. If the error is less than three percent, stop the optimization and output the final result; otherwise, add the sample to the sample library, re-establish the Kirging proxy model and BPNN proxy model, and iterate.

[0048] According to the above steps, the optimization work is carried out to obtain the optimized design of the external piping of the re-ignition compressor. The piping parameters before and after optimization are shown in Table 1, and the piping geometry comparison before and after optimization is shown in Figure 5 . Comparison of pipeline losses before and after optimization is shown in Figure 4 It can be seen that the overall loss of the pipeline after optimization has decreased significantly, and significant optimization effects have been achieved under different flow ranges. The total pressure loss was reduced by 3% under the design conditions, and by 2% when the flow rate increased. When the flow rate decreased, the effect of pipeline optimization was more obvious.

[0049] Table 1 Pipeline parameters before and after optimization

[0050] After specific experiments, the re-ignition external pipeline test was conducted to verify the optimization results of the re-ignition external pipeline and analyze the aerodynamic layout of the external pipeline. The relevant test data of a certain F-class external pipeline was collected to accumulate design experience and provide technical support for the subsequent aerodynamic design of compressors and turbines. Before measurement, it is necessary to first determine the working state of the re-ignition external pipeline test piece and examine the uniformity of the flow. Select the inlet flow condition, adjust the corresponding re-ignition external pipeline installation angle, and adjust it to the specified incoming flow after the fan starts and the airflow is stable. Then adjust the direction of the external pipeline outlet to prevent the pipeline outlet from affecting the outlet relatively; when measuring, first determine the external pipeline working condition. After the fan starts and the airflow is stable, measure the total temperature, total pressure and static pressure at the inlet of the external pipeline, convert the inlet Mach number, and adjust it to the aerodynamic test condition of the external pipeline. Then, change the test condition until all the specified test work is completed; after measurement, determine whether there are obvious errors or omissions through data processing and analysis. If so, re-test.

[0051] like Figure 6 As shown, the experimental equipment includes:

[0052] 1) Roots blower: The air source is provided by Roots blower, with a design flow rate (inlet) of 343m3 / min and a pressure rise of 68.6kPa. Depending on the test conditions, the air source can be switched between the small blower and the large blower continuously or provided together. A pressure expansion connection section and a pressure stabilizing box with a diameter of Φ800mm are installed at the blower outlet. The pressure stabilizing box plays a role in stabilizing the air source of the blower. The built-in damping net has a total of 3 layers, 1 layer of honeycomb rectifier grid, and the outlet is connected to the extension section of the test pipeline.

[0053] 2) Surge tank extension, see Figure 7 , including 4 straight pipe extensions, a 90° elbow for diversion, and a flange adapter to connect the standard flange to the specific flange interface of the surge tank. The loss of the entire pipeline is the loss before the gas source. The total pressure probe is installed in the surge tank, so this section of the loss is not included in the test loss. In order to divert the gas source, sufficient space margin is left for pipeline testing.

[0054] 3) Pressure regulator tank, such as Figure 8 The figure shows the pressure-surge tank used in the test. The pressure-surge tank is about 1.7m×0.8m×2m in length, width and height, and has a layer of rectifier grid inside. The function of the pressure-surge tank is to stabilize the pressure and flow. There are different exhaust channels in and around the pressure-surge tank, including front-stage, middle-stage and rear-stage exhaust ports. In addition, there are 6 total pressure probes arranged in the circumference to explore the flow field inside the pressure-surge tube due to uneven exhaust, which plays the role of simulating the exhaust chamber.

[0055] 4) Test pipelines: There are 12 test pipelines in total, which are installed on the pressure-surge tank in 9 ways. The pipelines are divided into two sets, one set is the pipeline before optimization, and the other set is the pipeline after optimization. The pipelines before and after optimization are installed on the pressure-surge tank in the same way. The difference is some pipeline parameters, such as diameter, pipeline length, combination method, pneumatic layout, etc. Fig. 9 shown.

[0056] 5) Pressure collection hose: The pressure collection hose can collect the static pressure of the pipeline wall to the scanning valve. The collection pipeline used this time has an outer diameter of The longest is 3m and the shortest is 1.5m. Its function is to collect the average static pressure of the wall surface at the outlet of the test pipeline and collect pressure information at a distance of 1cm from the outlet of each pipeline. Fig.10 As shown in (b), the static pressure at the pipeline outlet is obtained by the average pressure of the four holes.

[0057] 6) Pressure scanning valve, the pressure acquisition part includes 2 PSI 9116 acquisition modules, a total of 32 pressure channels, and the acquisition signal is the time-averaged pipe wall static pressure after the pipeline pressure is stable, such as Fig.11 shown.

[0058] 7) Temperature and atmospheric pressure acquisition equipment, electronic total temperature acquisition device and digital barometer are connected to the acquisition computer to record the real-time temperature and pressure signals, such as Fig.12 shown.

[0059] In order to judge the eligibility of the test data, the present invention has passed the uncertainty analysis. In the re-ignition compressor external pipeline test, the uncertainty speed is measured to characterize the analytical error limit of various errors on the results during the test, or it is an interval in which the various errors have a comprehensive impact on the possible deviation of the measurement result from its true value. Measurement uncertainty can be divided into two categories, namely Class A and Class B uncertainty. Class A uncertainty is a statistical analysis method that characterizes a series of repeated observations, and Class B assessment is an assessment of other methods different from statistical analysis (based on probability distribution assessment recognized by experience or other information). The evaluation methods of Class A and Class B uncertainties are both expressed in digital characteristics of standard deviation. The smaller the standard deviation, the more accurate the data.

[0060] The factors that affect the uncertainty in this test include: the inadequacy of the test design, the incomplete definition of the pneumatic parameters of the measuring pipeline, the test sample pipeline cannot fully represent the optimized pipeline, the impact of the test environment on the data inclusion, the human error in the reading of the data acquisition instrument, the limitations of the data acquisition itself, and the rounding error caused by the initial data processing.

[0061] According to the one-dimensional and three-dimensional coupled optimization analysis, the optimized pipeline design parameters were obtained. However, the numerical simulation scheme is still quite different from the actual pipeline flow conditions. Therefore, experimental research is very necessary to verify the effects before and after optimization. The flow field observation and aerodynamic parameter measurement of this experiment were carried out in the aerodynamic anechoic chamber modified by the Aero Engine Research Institute of Shanghai Jiaotong University. The laboratory has a fully enclosed aerodynamic anechoic equipment, a blade test bench, and a transferable multifunctional precision flow test device. Its test equipment and testing technology can facilitate the observation of flow phenomena, display of flow fields, measurement of flow fields, etc. It has dynamic pressure sensors, 16-channel pressure scanning valve acquisition systems, etc., which are particularly suitable for basic research and analysis such as collecting flow field information, studying pressure fluctuations, loss coefficients, and total temperature changes. The steady-state data point acquisition of the flow field in the experiment helps to understand the total pressure recovery coefficient inside the pipeline and analyze the flow mechanism of the external pipeline of the re-ignition compressor. This experiment used 16-channel pressure scanning valves and temperature sensors to observe the flow field, which has been widely used in the acquisition of flow field pressure and temperature.

[0062] like Fig.13 As shown in the figure, the total pressure recovery coefficient changes with the Mach number before and after pipeline optimization (0-0.9Ma). It can be seen that the pipeline loss has a large difference before and after optimization in the four pipeline layouts. The overall loss of the optimized pipeline is better than the pipeline layout before optimization. When the pipeline Mach number increases from 0 to 0.9Ma, the loss gap is also gradually increasing. At low Mach numbers, the gap is about 3%. At high subsonic speeds, the loss gap gradually increases to 6%. Combined with the previous one-dimensional and three-dimensional comprehensive analysis results, as shown in Figure 4 As shown in the figure, the comprehensive optimization efficiency improvement level is about 3%. The test results also confirm the previous conclusions. The loss situation is similar under the distribution mode of 1 / 2 / 4 pipelines. One reason is that the gas source power is surplus. The test data is also the mean pressure operating point measured after the gas source is stable for 20s. Therefore, different outlet flow rates have no effect on the loss. The second is that the optimization of the pipelines before and after loss optimization under different distribution modes is very obvious, and there is no change in the optimization situation with the increase in the number of pipelines.

[0063] The Rastrigin test function is used to test and compare the existing single-point addition criterion optimization algorithm and the multiple-point addition criterion optimization algorithm of the present invention. The test function is: like Figure 1 As shown in (c) and (d), the test function has many peaks and valleys, which makes it easy for the optimization algorithm to converge to the local optimal point. The test function can simulate the complex objective function of the external pipeline loss optimization of the re-ignition compressor, and its global optimal point is the origin. After 300 rounds of adding points, the response surface cloud of the proxy model corresponding to different adding point criteria is shown in Figure 1The results show that, facing this complex objective function, only the optimization algorithm of multiple point addition criteria of the present invention can converge to the global optimal point, and its response surface cloud diagram can also better reflect the real objective function.

[0064] Compared with the prior art, the present invention uses a parameterized modeling method for irregular multi-pipelines with complex layout trends and a hybrid optimization platform including a Kriging proxy model, a back-propagation neural network proxy model, an objective function minimum point-adding criterion, an expected improvement point-adding criterion, a root mean square error maximum point-adding criterion and multiple optimization algorithms, which can more accurately simulate the flow matching state of the front, middle and rear stage pipelines. Compared with the traditional design method, the present invention can significantly reduce the total pressure loss of the external pipeline of the gas compressor, reduce time and manpower consumption, and improve the design efficiency. The above two technical means are used for optimization work. The final numerical and experimental results show that the optimized pipeline system not only improves the aerodynamic performance of its design point, but also improves the aerodynamic performance of its non-design point, so that the optimized pipeline has a larger flow range. When the Mach number of the pipeline increases from 0.2Ma to 0.9Ma, the loss gap is also gradually increasing. After shortening the pipeline length and increasing the pipeline diameter, the total pressure loss gap of the pipeline before and after optimization at low Mach numbers is about 3%, and at high subsonic speeds, the loss gap gradually increases to 6%. .

[0065] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A method for optimizing the external pipeline loss of a re-ignition compressor, characterized in that: By constructing a basic loss optimization model of the re-ignition compressor and determining its constraints, the external piping of the re-ignition compressor is parameterized and modeled. The initial samples are generated in the design space through the Latin hypercube sampling method, and the total pressure loss coefficient corresponding to the initial samples is obtained through one-dimensional simulation. The Kirging surrogate model and the BPNN surrogate model are used in parallel to establish an approximate relationship between the input design parameters of the external piping of the re-ignition compressor and its corresponding total pressure loss coefficient. The input design parameters that minimize the total pressure loss on the surrogate model are found based on a variety of point-adding criteria and optimization algorithms, and the surrogate model results are compared with the one-dimensional simulation results. Optimization is achieved through iterative reconstruction.

2. The method for optimizing the external pipeline loss of a re-ignition compressor according to claim 1, characterized in that: The basic loss optimization model is specifically: the total loss of the external pipeline Where: Δp i is the pressure loss along the pipeline at different sections and local pressure loss λ is the Darcy friction factor, l i is the length of the pipeline, D is the diameter of the pipeline, is the average flow rate, θ i is the turning angle of the elbow, is the local pressure loss coefficient, R is the radius of curvature of the centerline of the elbow; The constraints include: total flow constraint of engine cooling, branch flow constraint, pipeline system flow constraint, pipeline diameter constraint, pipeline quantity constraint and elbow parameter constraint.

3. The method for optimizing external pipeline loss of a re-ignition compressor according to claim 1, characterized in that: The parametric modeling refers to: parameterizing key design parameters of the pipeline, including the pipeline diameter, the number of bends, the number of outlets, and the number of pipelines.

4. The method for optimizing external pipeline loss of a re-ignition compressor according to claim 1, characterized in that: The one-dimensional simulation refers to: using numerical simulation method to solve the pipeline computational fluid dynamics flow equation: in: Where: ρ is gas density; u is velocity; p is pressure; T is temperature; h is convective heat transfer coefficient, T w is the wall temperature, T ∞ Take the local mainstream average temperature and heat flux density Friction coefficient p in 、p out are the average pressures of the inlet and outlet, L is the channel length, and D is the pipe diameter; parameters related to the pipe diameter 5. The method for optimizing external pipeline loss of a re-ignition compressor according to claim 1, characterized in that: The iterative reconstruction means that when the error between the proxy model result and the one-dimensional simulation result is less than 3%, the design parameter is output and the optimization is stopped; otherwise, the sample point is added to the sample library, a new proxy model is reconstructed and iterated.

6. The method for optimizing the external pipeline loss of a re-ignition compressor according to claim 1, characterized in that: The BPNN proxy model is a back propagation neural network proxy model, including: an input layer, a hidden layer and an output layer. The initial output data output by the BPNN proxy model is compared with the data in the training set, and the comparison result is back-propagated from the output layer to the hidden layer and the input layer to adjust the connection weights so that the error between the output data and the training set data is minimized.

7. The method for optimizing the external pipeline loss of a re-ignition compressor according to claim 1, characterized in that: The multiple point-adding criteria are: parallel MP, EI, and RMSE three point-adding criteria, among which: the MP point-adding criterion is the minimum value criterion of the objective function, which assumes that the proxy model is accurate enough. After the proxy model is established, the minimum value of the objective function on the proxy model is directly sought; the EI point-adding criterion is the expected improvement criterion, which seeks the maximum value of the expected function of the difference between the currently known minimum value and the unknown point value on the proxy model; the RMSE point-adding criterion is the root mean square error maximum value criterion, which adds sample points where the proxy model error is the largest, which can improve the overall accuracy of the model and better find the global optimal solution.

8. The method for optimizing the external pipeline loss of a re-ignition compressor according to claim 1, characterized in that: The hybrid optimization algorithm is a hybrid optimization algorithm constructed based on Latin hypercube sampling, genetic algorithm (GA) and sequential quadratic programming algorithm (SQP) to find the optimal point on the proxy model.

9. A system for optimizing external pipeline losses of a re-ignition compressor for implementing the method described in any one of claims 1 to 8, comprising: An initial sample generation unit, a one-dimensional simulation unit, a proxy model unit and an optimization algorithm unit, wherein: the initial sample generation unit uses a Latin hypercube sampling method to generate multiple initial samples according to constraints; the one-dimensional simulation unit calculates the initial samples to obtain the calculation results corresponding to the initial samples, that is, the weighted total pressure loss of multiple pipelines; the proxy model unit establishes a Kriging proxy model and a BPNN proxy model in parallel according to the design variables and calculation results of the initial samples to obtain the corresponding relationship between the input design variables and the output weighted total pressure loss; the optimization algorithm unit performs optimization work according to the established proxy model to obtain the input design variables that minimize the weighted total pressure loss.