Multi-objective optimization design method for combustion chamber of heavy-duty gas turbine

Abstract

The invention provides a multi-objective optimization design method for a combustion chamber of a heavy duty gas turbine, which comprises the following steps of: pre-sampling a multi-dimensional design variable, and calculating and determining a multi-dimensional design objective variable result of a pre-sampling point through a CFD (Computational Fluid Dynamics) model; establishing a sample set based on the pre-sampling points and a multi-dimensional design target variable result; on the basis of an outlet average temperature distribution interval in the sample set, performing supplementary sampling on the multi-dimensional design variables according to a special sampling rule to update the sample set and construct an agent model; in response to the fact that the generalization ability of the agent model does not meet the expectation, secondary supplementary sampling is carried out to update the sample set, and the agent model is constructed based on the sample set; repeating secondary supplementary sampling until the generalization ability of the agent model meets the expectation; performing multi-objective optimization by adopting a heuristic algorithm, and evaluating the candidate solutions by utilizing the proxy model to determine a Pareto solution set; and performing multi-objective decision according to the Pareto solution set to obtain an optimization design solution of the combustion chamber.

Description

Technical Field

[0001] The present invention relates to the field of gas turbines, and in particular to a multi-objective optimization design method for a heavy-duty gas turbine combustion chamber, a multi-objective optimization design system for a heavy-duty gas turbine combustion chamber, and a computer-readable storage medium. Background Art

[0002] The combustor is one of the three core components of a heavy-duty gas turbine. It receives high-pressure gas from the compressor and fuel from the fuel channel, converting chemical energy into heat energy to power subsequent components.

[0003] The combustion chamber directly determines the thermal efficiency of a heavy-duty gas turbine. Therefore, efficient operation of a heavy-duty gas turbine places high demands on the combustor's performance. During operation, the combustor's internal working environment is harsh, subject to high temperatures caused by fuel combustion. This increases the risk of cracking and failure under complex and variable operating conditions. Therefore, optimizing the combustor's design while balancing performance and lifespan is crucial.

[0004] Prior art methods for combustion chamber optimization design include experimental methods and numerical simulation methods. Among them, computational fluid dynamics (CFD) numerical simulation methods can establish high-fidelity models, provide intuitive and accurate information on the combustion chamber flow field distribution, and guide combustion chamber optimization design based on the flow field distribution characteristics under different operating conditions.

[0005] However, directly using numerical simulation methods for combustion chamber optimization often consumes significant computing resources and increases computational time. For example, while high-fidelity CFD models can accurately simulate the combustion chamber flow field, they typically require smaller mesh sizes and a larger number of cells, significantly impacting computational time. Furthermore, combustion chamber optimization requires a large amount of data, often resulting in an unacceptable time cost.

[0006] In order to overcome the above-mentioned defects of the existing technology, the field urgently needs a multi-objective optimization design technology for heavy-duty gas turbine combustion chambers, which can establish a high-precision proxy model based on a high-quality sample set, thereby greatly reducing the high time cost brought by numerical simulation while ensuring the accuracy of the combustion chamber optimization design, and at the same time reducing the combustion chamber optimization design cost while meeting the needs of optimization design of multiple types of combustion chamber indicators under different working conditions. Summary of the Invention

[0007] A brief overview of one or more aspects is given below to provide a basic understanding of these aspects. This overview is not an exhaustive survey of all contemplated aspects and is neither intended to identify key or decisive elements of all aspects nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to a more detailed description given later.

[0008] To overcome the above-mentioned deficiencies existing in the prior art, the present invention provides a multi-objective optimization design method for a heavy-duty gas turbine combustor, a multi-objective optimization design system for a heavy-duty gas turbine combustor, and a computer-readable storage medium, which can establish a high-precision surrogate model based on a high-quality sample set, thereby greatly reducing the high time cost brought by numerical simulation on the premise of ensuring the accuracy of the combustor optimization design, and meeting the requirements for the optimization design of multiple types of indexes of the combustor under different working conditions while reducing the cost of the combustor optimization design.

[0009] Specifically, the above-mentioned multi-objective optimization design method for a heavy-duty gas turbine combustor provided according to the first aspect of the present invention includes the steps of: S1, pre-sampling multi-dimensional design variables, and calculating the results of multi-dimensional design objective variables of the pre-sampling points through a CFD model, where the CFD model is a combustor CFD model, and the results of multi-dimensional design objective variables include the average outlet temperature; S2, establishing a sample set based on the pre-sampling points and the results of multi-dimensional design objective variables of the pre-sampling points; S3, based on the average outlet temperature distribution interval in the sample set, supplementing the sampling of the multi-dimensional design variables and performing CFD model calculations according to special sampling rules to update the sample set and constructing a surrogate model based on the sample set; S4, in response to the generalization ability of the surrogate model not meeting the expectation, based on the average outlet temperature distribution interval in the sample set, performing secondary supplementary sampling of the multi-dimensional design variables and CFD model calculations according to special sampling rules to update the sample set and constructing a surrogate model based on the sample set; S5, repeating S4 until the generalization ability of the surrogate model meets the expectation; S6, performing multi-objective optimization using a heuristic algorithm, and evaluating candidate solutions using the surrogate model to determine a Pareto solution set that takes into account both the performance and life indexes of the combustor; and S7, making a multi-objective decision according to the Pareto solution set to obtain an optimized design solution for the combustor.

[0010] Preferably, in an embodiment of the present invention, the steps of determining the multi-dimensional design variables and the multi-dimensional design target variables include: based on the optimization design requirements for the operating performance and reliability of a heavy-duty gas turbine combustor, determining the multi-dimensional design variables that take into account both the combustor structure and operating conditions, and the multi-dimensional design target variables that take into account both the combustor performance and life, wherein the multi-dimensional design variables include structural design variables and operating condition design variables, and the multi-dimensional design target variables include performance design target variables and life design target variables.

[0011] Preferably, in an embodiment of the present invention, S3 includes: based on the outlet average temperature distribution interval in the sample set, dividing the operating condition design variables into layers according to a special sampling rule; determining the layers corresponding to each sample in the sample set according to the sample set, and performing supplementary sampling on the operating condition design variables in the layers without samples; performing supplementary sampling on the structural design variables, where the number of supplementary sampling points of the structural design variables is equal to the number of supplementary sampling points of the operating condition design variables; combining the supplementary sampling points of the structural design variables and the supplementary sampling points of the operating condition design variables to determine the supplementary sampling points for supplementary sampling; calculating and determining the results of the multi-dimensional design target variables of the supplementary sampling points through a CFD model, updating the sample set based on the supplementary sampling points and the results of the multi-dimensional design target variables of the supplementary sampling points; and constructing a surrogate model based on the training set in the sample set and evaluating the generalization ability of the surrogate model based on the test set in the sample set.

[0012] Preferably, in an embodiment of the present invention, S4 includes: in response to the generalization ability of the surrogate model not meeting the expectation, based on the outlet average temperature distribution interval in the sample set, further dividing the operating condition design variables into layers according to a special sampling rule; determining the layers corresponding to each sample in the sample set according to the sample set, and performing secondary supplementary sampling on the operating condition design variables in the layers without samples; performing secondary supplementary sampling on the structural design variables, where the number of secondary supplementary sampling points of the structural design variables is equal to the number of secondary supplementary sampling points of the operating condition design variables; combining the secondary supplementary sampling points of the structural design variables and the secondary supplementary sampling points of the operating condition design variables to determine the secondary supplementary sampling points for secondary supplementary sampling; calculating and determining the results of the multi-dimensional design target variables of the secondary supplementary sampling points through a CFD model, updating the sample set based on the secondary supplementary sampling points and the results of the multi-dimensional design target variables of the secondary supplementary sampling points; and constructing a surrogate model based on the training set in the sample set and evaluating the generalization ability of the surrogate model based on the test set in the sample set.

[0013] Preferably, in an embodiment of the present invention, the structural design variables include the diameter of the combustion chamber metering hole, the diameter of the combustion chamber mixing hole, the length of the combustion chamber mixing section, and the outlet area of the combustion chamber transition section, and the operating condition design variables include the fuel quantity of the primary premixed nozzle, the fuel quantity of the secondary premixed nozzle, the fuel quantity of the secondary diffusion nozzle, and the air intake of the combustion chamber; the performance design target variables include the mass flow rate of outlet nitrogen oxides, the average outlet temperature, and the pressure drop of the combustion chamber, and the life design target variable includes the average temperature gradient of the combustion chamber flame tube.

[0014] Preferably, in an embodiment of the present invention, the combustion chamber CFD model is established based on the turbulent combustion model and the kinetic model of combustion chamber simulation.

[0015] Preferably, in an embodiment of the present invention, the decision variables of the multi-objective optimization problem of the multi-objective optimization are the variables of each dimension of the multi-dimensional design variables, the objective function is a function of the multi-dimensional design variables, and the constraint conditions include load number constraint, equivalence ratio constraint, fuel flow distribution ratio constraint, and average outlet temperature constraint.

[0016] Preferably, in an embodiment of the present invention, the heuristic algorithm includes genetic algorithm, multi-objective particle swarm optimization algorithm or simulated annealing algorithm, and the multi-objective decision-making includes technique for order preference by similarity to an ideal solution (TOPSIS) or analytic hierarchy process (AHP).

[0017] In addition, the multi-objective optimization design system for a heavy-duty gas turbine combustion chamber provided according to the second aspect of the present invention includes a memory and a processor. Computer instructions are stored on the memory. The processor is connected to the memory and is configured to execute the computer instructions stored on the memory to implement the multi-objective optimization design method for a heavy-duty gas turbine combustion chamber provided in any of the above embodiments.

[0018] In addition, computer instructions are stored on the computer-readable storage medium provided according to the third aspect of the present invention. When the computer instructions are executed by a processor, the multi-objective optimization design method for a heavy-duty gas turbine combustion chamber provided in any of the above embodiments is implemented. Description of the Drawings

[0019] After reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings, the above features and advantages of the present invention can be better understood. In the drawings, the components are not necessarily drawn to scale, and components with similar relevant characteristics or features may have the same or similar reference numerals.

[0020] Figure 1 Shows a schematic diagram of a multi-objective optimization design system for a heavy-duty gas turbine combustion chamber provided according to some embodiments of the present invention;

[0021] Figure 2 A flowchart of a multi-objective optimization design method for a heavy-duty gas turbine combustor provided according to some embodiments of the present invention is shown;

[0022] Figure 3 A flowchart of a multi-objective optimization design method for a heavy-duty gas turbine combustion chamber according to a preferred embodiment of the present invention is shown;

[0023] Figure 4 A table showing reaction kinetic parameters of a kinetic model provided according to a preferred embodiment of the present invention is shown;

[0024] Figure 5 A sampling range table of pre-sampling provided according to a preferred embodiment of the present invention is shown;

[0025] Figure 6 A table showing the number of layers for supplementary sampling according to a preferred embodiment of the present invention is shown;

[0026] Figure 7 A table showing the number of layers for secondary supplementary sampling according to a preferred embodiment of the present invention is shown;

[0027] Figure 8 shows an evaluation result table of the agent model provided according to a preferred embodiment of the present invention; and

[0028] Figure 9 A comparison diagram of the normalized values of the compromise solution and different preference solutions in the Pareto solution set provided by the preferred embodiment of the present invention and the training set is shown.

[0029] Reference numerals:

[0030] 100: Multi-objective optimization design system for heavy-duty gas turbine combustor;

[0031] 110: memory;

[0032] 111: Computer readable storage medium;

[0033] 120: processor;

[0034] 901: outlet nitrogen oxide mass flow rate;

[0035] 902: Average temperature at the combustion chamber outlet;

[0036] 903: Combustion chamber pressure drop;

[0037] 904: Average temperature gradient of combustion chamber flame tube;

[0038] 910, 920, 930, 940, 950: curves; and

[0039] S1 to S7: Steps. Detailed implementation manners

[0040] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Note that the aspects described below in conjunction with the accompanying drawings and specific embodiments are merely exemplary and should not be construed as imposing any limitation on the protection scope of the present invention.

[0041] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0042] In addition, the "upper", "lower", "left", "right", "top", "bottom", "horizontal", and "vertical" used in the following description should be understood as the orientations shown in this section and the related drawings. This relative term is only for convenience of description and does not represent that the device described needs to be manufactured or operated in a specific orientation, so it should not be construed as a limitation on the present invention.

[0043] It can be understood that although terms such as "first", "second", and "third" can be used herein to describe various components, regions, layers, and / or parts, these components, regions, layers, and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers, and / or parts. Therefore, the first component, region, layer, and / or part discussed below can be referred to as the second component, region, layer, and / or part without departing from some embodiments of the present invention.

[0044] As described above, directly using numerical simulation methods for the optimization design of the combustion chamber often consumes a large amount of computing resources and increases the time cost. For example, although a high-fidelity CFD model can accurately simulate the combustion chamber flow field, it usually requires a smaller grid size and a larger number of grids, significantly affecting the computing time. Moreover, the optimization design of the combustion chamber requires a large amount of data, and the time cost spent is often unacceptable.

[0045] To overcome the above-mentioned defects existing in the prior art, the present invention provides a multi-objective optimization design method for a heavy-duty gas turbine combustor, a multi-objective optimization design system for a heavy-duty gas turbine combustor, and a computer-readable storage medium, which can establish a high-precision surrogate model based on a high-quality sample set, thereby greatly reducing the high time cost brought by numerical simulation on the premise of ensuring the accuracy of the combustor optimization design, and meeting the requirements for the optimization design of multiple types of indicators of the combustor under different working conditions while reducing the combustor optimization design cost.

[0046] In some non-limiting embodiments, the above-mentioned multi-objective optimization design method for a heavy-duty gas turbine combustor provided by the first aspect of the present invention can be implemented via the above-mentioned multi-objective optimization design system for a heavy-duty gas turbine combustor provided by the second aspect of the present invention.

[0047] Please refer to Figure 1 , Figure 1 which shows a schematic diagram of a multi-objective optimization design system for a heavy-duty gas turbine combustor provided by some embodiments of the present invention.

[0048] As Figure 1 shown, the multi-objective optimization design system 100 for a heavy-duty gas turbine combustor may be configured with a memory 110 and a processor 120. The memory 110 includes but is not limited to the above-mentioned computer-readable storage medium 111 provided by the third aspect of the present invention, on which computer instructions are stored. The processor 120 is connected to the memory 110 and is configured to execute the computer instructions stored on the memory 110 to implement the multi-objective optimization design method for a heavy-duty gas turbine combustor provided by the first aspect of the present invention.

[0049] The working principle of the above-mentioned multi-objective optimization design system for a heavy-duty gas turbine combustor will be described below in combination with some embodiments of the multi-objective optimization design method for a heavy-duty gas turbine combustor. Those skilled in the art can understand that these embodiments of the multi-objective optimization design method for a heavy-duty gas turbine combustor are only some non-limiting implementation manners provided by the present invention, aiming to clearly show the main concept of the present invention and provide some specific solutions convenient for the public to implement, rather than limiting all functions or all working modes of the multi-objective optimization design system for a heavy-duty gas turbine combustor. Similarly, the multi-objective optimization design system for a heavy-duty gas turbine combustor is also a non-limiting implementation manner provided by the present invention, and does not limit the execution subject and execution order of each step in these multi-objective optimization design methods for a heavy-duty gas turbine combustor.

[0050] Before implementing the multi-objective optimization design method for a heavy-duty gas turbine combustor provided by the present invention, the multi-objective optimization design system for a heavy-duty gas turbine combustor can first establish a combustor CFD model based on the turbulent combustion model and kinetic model of the heavy-duty gas turbine combustor simulation, as well as the real three-dimensional geometry of the combustor.

[0051] The turbulent combustion model is a mathematical model that describes the coupled action of turbulent flow and combustion reaction, with the bidirectional coupling characteristics of turbulent flow and combustion reaction, and can realize the organic integration of the turbulent model and the combustion model.

[0052] Specifically, the turbulent model can be the Standard k-epsilon model (standard k-ε model), the Realizable k-epsilon model (realizable k-ε model), or the RNG k-epsilon model (Renormalization Group k-epsilon, renormalization group k-ε model), etc. The combustion model can be the Finite-Rate / Eddy-Dissipation model (finite rate / vortex dissipation model), the equilibrium PDF model (Equilibrium Probability Density Function Model, equilibrium probability density function model), the The partially premixed combustion model (partially premixed combustion model), etc. in the component transport model. The kinetic model can be a detailed chemical reaction mechanism and a simplified chemical reaction mechanism. For example, GRI Mech1.2, GRI Mech3.0; the kinetic model can also be a global chemical reaction mechanism for methane combustion, such as a one-step global reaction mechanism and a two-step global reaction mechanism.

[0053] The multi-objective optimization design system can determine multi-dimensional design variables and multi-dimensional design objective variables according to the optimization design requirements of the operating performance and reliability of the heavy-duty gas turbine combustor.

[0054] In some embodiments, the goal of the multi-objective optimization design system is to achieve multi-objective optimization of the combustor performance and life while taking into account the combustor structure and operating conditions. Thus, the multi-objective optimization design system determines the variables that can take into account the combustor structure and operating conditions as multi-dimensional design variables, and determines the target variables that can take into account the combustor performance and life as multi-dimensional design objective variables.

[0055] For example, the multi-dimensional design variables that take into account the combustion chamber structure and operating conditions may include the combustion chamber metering hole diameter, the combustion chamber mixing hole diameter, the combustion chamber mixing section length, the combustion chamber transition section outlet area, the fuel quantity of the primary premixed nozzle, the fuel quantity of the secondary premixed nozzle, the fuel quantity of the secondary diffusion nozzle, and the combustion chamber inlet air quantity. Among them, the first four dimensions (the combustion chamber metering hole diameter, the combustion chamber mixing hole diameter, the combustion chamber mixing section length, and the combustion chamber transition section outlet area) can be classified as structural design variables, and the last four dimensions (the fuel quantity of the primary premixed nozzle, the fuel quantity of the secondary premixed nozzle, the fuel quantity of the secondary diffusion nozzle, and the combustion chamber inlet air quantity) can be classified as operating condition design variables.

[0056] The multi-dimensional design target variables that take into account the combustion chamber performance and life can include the outlet nitrogen oxide mass flow rate of the combustion chamber, the outlet average temperature of the combustion chamber, the combustion chamber pressure drop, and the average temperature gradient of the combustion chamber flame tube. Among them, the first three dimensions (the outlet nitrogen oxide mass flow rate, the outlet average temperature, and the combustion chamber pressure drop) can be classified as performance design target variables, and the average temperature gradient of the combustion chamber flame tube can be classified as a life design target variable.

[0057] Those skilled in the art should understand that the selection of the multi-dimensional design variables and the multi-dimensional design target variables can be adjusted accordingly according to the actual situation and the optimization design objectives of the combustion chamber, and is not limited to the design variables and design target variables listed above.

[0058] Please refer to Figure 2 , Figure 2 which shows a flowchart of a multi-objective optimization design method for a heavy-duty gas turbine combustion chamber provided according to some embodiments of the present invention.

[0059] As Figure 2 shown, the multi-objective optimization design method 200 for a heavy-duty gas turbine combustion chamber may include step S1: pre-sampling the multi-dimensional design variables, and calculating and determining the results of the multi-dimensional design target variables of the pre-sampling points through a CFD model.

[0060] The multi-objective optimization design system pre-samples the multi-dimensional design variables to obtain a certain number of pre-sampling points. The pre-sampling can be achieved by means of stratified sampling, and the methods of stratified sampling may include Latin hypercube sampling, equal-proportion sampling, single-variable factor sampling, etc.

[0061] Based on the obtained pre-sampling points, the multi-objective optimization design system can calculate the results of the multi-dimensional design target variables of the pre-sampling points through the previously established combustion chamber CFD model. Here, the results of the multi-dimensional design target variables include the outlet average temperature. The outlet average temperature distribution interval can be determined from the outlet average temperatures among the results of the multi-dimensional design target variables of multiple pre-sampling points calculated by the CFD model.

[0062] After that, the multi-objective optimization design system can establish a sample set based on the pre-sampled points and the multi-dimensional design objective variable results of the pre-sampled points. Each sample in the sample set can be composed of the pre-sampled points and the multi-dimensional design objective variable results of the pre-sampled points. According to the outlet average temperature in the multi-dimensional design objective variable results of each sample, the outlet average temperature distribution interval of the sample set can be determined.

[0063] Here, through pre-sampling and CFD model calculation, the multi-objective optimization design system can identify which multi-dimensional design variables the outlet average temperature of the combustion chamber is mainly affected by, so as to formulate special sampling rules for the process of supplementary sampling and secondary supplementary sampling of multi-dimensional design variables, so as to ensure that the outlet average temperature distribution of the finally formed sample set can be more uniform.

[0064] In some embodiments, according to pre-sampling and CFD model calculation, the multi-objective optimization design system can identify that the influence degree of the structural design variables in the multi-dimensional design variables on the outlet average temperature is less than that of the operating condition design variables. Therefore, the multi-objective optimization design system does not design special sampling rules for the structural design variables with less influence, and sets special sampling rules for the operating condition design variables with greater influence. The set special sampling rules can be used for subsequent supplementary sampling and secondary sampling of the operating condition design variables.

[0065] Furthermore, in the above embodiments, the outlet average temperature is mainly affected by the load number and the equivalence ratio mapped by the operating condition design variables. Therefore, the multi-objective optimization design system can use the load number and the equivalence ratio as the characteristics of the special sampling rules.

[0066] Specifically, the relationship between the operating condition design variable and the load number can be described as:

[0067] qM total =qM pp +qM sp +qM sd (1)

[0068]

[0069] Wherein, qM pp is the fuel quantity of the primary premixed nozzle, qM sp is the fuel quantity of the secondary premixed nozzle, qM sd is the fuel quantity of the secondary diffusion nozzle, qM total is the current total fuel quantity, qM 100％ is the total fuel quantity at 100% load, and k is the load number.

[0070] The relationship between the operating condition design variable and the equivalence ratio can be described as:

[0071] qMair1 = qM total ×4÷0.233÷0.69966199 (3)

[0072] eq = qM air1 ÷qM air (4)

[0073] Here, the chemical reaction during the operation of the combustion chamber is the combustion reaction of methane and oxygen. In formula (3), qM total multiplied by 4 is the mass flow rate of oxygen required to completely consume qM total and then divided by 0.233 is the mass flow rate of air required to completely consume qM total Finally, divided by the air flow distribution coefficient 0.69966199 for distribution reaction and cooling to obtain the total air volume qM participating in cooling and reaction air1 , and the equivalence ratio eq is qM air1 divided by the air intake volume qM of the combustion chamber air .

[0074] As Figure 2 shown, the multi-objective optimization design system can continue to execute step S3: Based on the outlet average temperature distribution interval in the sample set, perform supplementary sampling on the multi-dimensional design variables according to special sampling rules and conduct CFD model calculations to update the sample set and construct a surrogate model based on the sample set.

[0075] Under the outlet average temperature distribution interval of the sample set, the multi-objective optimization design system can continue to perform supplementary sampling on the multi-dimensional design variables according to special sampling rules. Perform CFD model calculations based on the supplementary sampling points to determine the results of the multi-dimensional design objective variables of the supplementary sampling points. Then supplement the samples formed by the supplementary sampling points and the results of the multi-dimensional design objective variables of the supplementary sampling points to the sample set formed in step S2 to update the sample set.

[0076] As described above, in some embodiments, the multi-objective optimization design system takes the load number and the equivalence ratio as the characteristics of the special sampling rules for the operating condition design variables.

[0077] During the process of the multi-objective optimization design system performing supplementary sampling on the multi-dimensional design variables, first, based on the outlet average temperature distribution interval in the sample set, according to the special sampling rules, that is, according to the load number and the equivalence ratio, divide the operating condition design variables into layers and limit the sampling range of each layer. Here, since the outlet average temperature is mainly affected by the load number and the equivalence ratio, there is a certain corresponding relationship between different layers and different outlet average temperature values.

[0078] After that, according to the sample set, determine the layers corresponding to each sample in the sample set, and perform supplementary sampling on the operation condition design variables in the layers without samples to ensure that there are sample distributions in each layer. Here, the supplementary sampling can be equal-number sampling.

[0079] For the structural design variables with less influence, supplementary sampling can be performed in the same sampling manner as the pre-sampling. The number of supplementary sampling points of the structural design variables is equal to the number of supplementary sampling points of the operation condition design variables.

[0080] The multi-objective optimization design system then combines the supplementary sampling points of the structural design variables and the supplementary sampling points of the operation condition design variables with equal numbers to determine the supplementary sampling points of the supplementary sampling. The combination of the supplementary sampling points of the structural design variables and the supplementary sampling points of the operation condition design variables can be a random combination.

[0081] After determining the supplementary sampling points of the supplementary sampling, the multi-objective optimization design system can perform CFD model calculations based on the supplementary sampling points to determine the results of the multi-dimensional design objective variables of the supplementary sampling points and update the sample set. Through supplementary sampling, the outlet average temperature distribution in the updated sample set will be more uniform.

[0082] After that, based on the updated sample set, the multi-objective optimization design system constructs a surrogate model. First, the multi-objective optimization design system can divide the updated sample set into a training set and a test set. In some embodiments, the method of dividing the sample set may include, but is not limited to, the Sample Set Partitioning Based On Joint X-Y Distances (SPXY) method, the Sample Set Partitioning Based on X distances (SPX), or the Kennard–Stone (KS) algorithm.

[0083] After that, the multi-objective optimization design system can use the training set to construct and train the surrogate model, and use the test set to test the generalization ability of the surrogate model. In some embodiments, the constructed surrogate model can be an Epsilon-Support Vector Regression (Epsilon-SVR), a random forest, a partial least squares method, a Lasso (Least Absolute Shrinkage and Selection Operator) regression method. The multi-objective optimization design system adopts the Coefficient of Determination, R 2) to evaluate the generalization ability of the surrogate model.

[0084] Please continue to refer to Figure 2 , the multi-objective optimization design system can continue to execute step S4: in response to the generalization ability of the surrogate model not meeting the expectation, based on the outlet average temperature distribution interval in the sample set, perform secondary supplementary sampling on the multi-dimensional design variables according to the special sampling rule and CFD model calculation to update the sample set and construct a surrogate model based on the sample set.

[0085] After the multi-objective optimization design system constructs the surrogate model and evaluates its generalization ability according to the updated sample set, if the generalization ability of the constructed surrogate model meets the expectation, the multi-objective optimization design system can directly use this surrogate model for the multi-objective optimization task of the combustor.

[0086] If the generalization ability of the surrogate model does not meet the expectation, the multi-objective optimization design system can continue to perform secondary supplementary sampling on the multi-dimensional design variables according to the special sampling rule to form a final sample set with a uniform outlet average temperature distribution, and form a surrogate model with the expected generalization ability based on the final sample set.

[0087] Specifically, in some embodiments, among the multi-dimensional design variables, the operation condition design variables have a greater impact on the outlet average temperature, and the structure design variables have a smaller impact. When the generalization ability of the surrogate model constructed in step S3 does not meet the expectation, the multi-objective optimization design system based on the outlet average temperature distribution interval in the sample set divides the operation condition design variables into layers according to the special sampling rule, that is, further divides the layers according to the load number and equivalence ratio, and re-limits the sampling range of each layer.

[0088] After that, determine the layer corresponding to each sample according to the sample set, and perform secondary supplementary sampling on the operation condition design variables in the layers without samples. The secondary supplementary sampling can be equal-number sampling. Thus, ensure that there are sample distributions in each layer.

[0089] Then, perform secondary supplementary sampling on the structure design variables in the pre-sampling manner. The number of secondary supplementary sampling points of the structure design variables is equal to the number of secondary supplementary sampling points of the operation condition design variables.

[0090] The multi-objective optimization design system then combines the equal-number secondary supplementary sampling points of the structure design variables and the secondary supplementary sampling points of the operation condition design variables to determine the secondary supplementary sampling points of the secondary supplementary sampling. The combination of the secondary supplementary sampling points of the structure design variables and the secondary supplementary sampling points of the operation condition design variables can be a random combination.

[0091] After that, the multi-objective optimization design system determines the results of the multi-dimensional design objective variables of the secondary supplementary sampling points through CFD model calculation, and updates the sample set based on the secondary supplementary sampling points and the results of the multi-dimensional design objective variables of the secondary supplementary sampling points, so as to further improve the uniformity of the outlet average temperature distribution in the sample set.

[0092] Divide the updated sample set into a training set and a test set, construct and train a surrogate model based on the training set, and evaluate the generalization ability of the surrogate model based on the test set.

[0093] As Figure 2 shown, the multi-objective optimization design method includes step S5: repeat S4 until the generalization ability of the surrogate model meets the expectation.

[0094] When the generalization ability of the surrogate model does not meet the expectation, the multi-objective optimization design system can repeat the process of secondary supplementary sampling, and evaluate the generalization ability of the re-constructed and trained surrogate model until the generalization ability of the constructed surrogate model meets the expectation.

[0095] Through the above steps S1-S5, the multi-objective optimization design method for the heavy-duty gas turbine combustor provided by the present invention uses a stratified sampling method to divide the sample space and sample according to special sampling rules, so as to construct a high-quality sample set, and then construct a high-precision surrogate model.

[0096] In the prior art, the construction of a high-precision surrogate model requires a large number of CFD calculation samples as the data basis for establishing the surrogate model. The multi-objective optimization design method provided by the present invention uses a stratified sampling method and the set special sampling rules when constructing the surrogate model, and samples are extracted from each layer of the sample space with a certain probability, which has a significant statistical effect. Therefore, while ensuring the accuracy of the surrogate model, the calculation resources and calculation time are greatly reduced.

[0097] Please continue to refer to Figure 2 , the multi-objective optimization design system can continue to execute step S6: perform multi-objective optimization using a heuristic algorithm, and use the surrogate model to evaluate candidate solutions to determine the Pareto solution set that takes into account both the combustor performance and life index.

[0098] Before performing multi-objective optimization using a heuristic algorithm, the multi-objective optimization design system can first determine the decision variables, objective functions, and constraint conditions of the multi-objective optimization problem.

[0099] In some embodiments, the decision variables of the multi-objective optimization problem can be the design variables of each dimension in the multi-dimensional design variables, and the objective functions can be functions of the multi-dimensional design variables. In some embodiments, the multi-objective optimization design system can input the design variables of each dimension into the objective function to output the design objective variables of each dimension. In some embodiments, the multi-objective optimization design system can use the surrogate model constructed above that can obtain the results of multi-dimensional design objective variables as the objective function.

[0100] The constraint conditions of the multi-objective optimization problem can further include load number constraint, equivalence ratio constraint, fuel flow distribution ratio constraint, and outlet average temperature constraint. Specifically, the constraint conditions of the multi-objective optimization problem can be described as:

[0101]

[0102] eq min ≤eq≤eq max (6)

[0103] q min ≤qM pp / qM total ≤q max (7)

[0104] T ave_min ≤T ave ≤T ave_max (8)

[0105] Among them, formula (5) is the load number constraint, k min is the minimum load number, k max is the maximum load number; formula (6) is the equivalence ratio eq constraint, eq min is the minimum equivalence ratio, eq max is the maximum equivalence ratio; formula (7) is the fuel flow distribution ratio constraint, q min is the minimum flow distribution ratio, q max is the maximum flow distribution ratio; formula (8) is the outlet average temperature T ave constraint, T ave_min is the minimum outlet average temperature, T ave_max is the maximum outlet average temperature.

[0106] Afterwards, based on the decision variables, objective functions, and constraint conditions determined by the multi-objective optimization problem, the multi-objective optimization design system can use heuristic algorithms for multi-objective optimization. Here, the heuristic algorithms can include Genetic Algorithm (GA), Multi-Objective Particle Swarm Optimization (MOPSO), or Simulated Annealing (SA).

[0107] Through multi-objective optimization, a large number of candidate solutions that take into account the key structures and operating conditions of the combustion chamber can be obtained. Then, the multi-objective optimization design system can use the surrogate model to evaluate a large number of candidate solutions that take into account the key structures and operating conditions of the combustion chamber, thereby obtaining the Pareto solution set that takes into account the performance and life indexes of the combustion chamber.

[0108] As Figure 2 shown, the multi-objective optimization design system can continue to execute step S7: perform multi-objective decision-making based on the Pareto solution set to obtain the optimized design solution of the combustion chamber.

[0109] The multi-objective optimization design system makes multi-objective decisions based on the obtained Pareto solution set. Here, the multi-objective decision-making methods can include but are not limited to Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) or Analytic Hierarchy Process (AHP). Thus, the multi-objective optimization design system can obtain the optimized design solutions of the combustion chamber with better performance and life than the original samples under different combinations of structures, loads, and equivalence ratios.

[0110] The following is a specific non-limiting preferred embodiment, based on which the multi-objective optimization design method and system for a heavy-duty gas turbine combustion chamber proposed by the present invention are described in detail.

[0111] Please refer to Figure 3 , Figure 3 which shows the flowchart of the multi-objective optimization design method for a heavy-duty gas turbine combustion chamber provided according to the preferred embodiment of the present invention.

[0112] As Figure 3 shown, the multi-objective optimization design system for a heavy-duty gas turbine combustion chamber can first establish a high-fidelity CFD model of the combustion chamber.

[0113] In this preferred embodiment, the combustion chamber CFD model can be constructed based on the turbulent combustion model and kinetic model for heavy-duty gas turbine combustion chamber simulation. Since the internal fluid in the combustion chamber involves fluid phenomena such as inlet vortex of the flame tube, swirl, and jet from the nozzle holes, the turbulent model can select the Realizable k-epsilon model that can be applied to solve strong streamline bending, vortex, and rotating fluid phenomena. Through this turbulent model, the fluid phenomena in the combustion chamber can be accurately solved.

[0114] The combustion model can be the Finite-Rate / Eddy-Dissipation model in the species transport model. The apparent reaction rate of this model is the smaller value of the finite-rate reaction rate and the eddy dissipation rate. Specifically, the molar rate of generation / destruction of substance i in reaction r can be described as:

[0115]

[0116] Equation (9) represents the calculation method of the finite-rate reaction rate. Wherein, Γ represents the net influence of the third body on the reaction rate, v′ ,r can be the stoichiometric fraction of the reactant in substance i in reaction r, v″ i,r can be the stoichiometric fraction of the product in substance i in reaction r, k f,r and k b,r are respectively the forward rate constant and the reverse rate constant in reaction r, C j,r is the molar concentration of substance j in reaction r, η′ j,r can be the rate exponent of the reactant in substance j in reaction r, η″ j,r can be the rate exponent of the product in substance j in reaction r.

[0117] In addition, the net production of substance i generated by reaction r can be given by the minimum value of Equation (10) and Equation (11):

[0118]

[0119] Equation (10) and Equation (11) represent the calculation method of the eddy dissipation rate. Wherein, M w,i is the molecular weight of substance i, A and B are empirical constants, Y R is the mass fraction of a specific reactant R, Y P is the mass fraction of any product, ε is the turbulent dissipation rate, and k is the turbulent kinetic energy.

[0120] In the unignited region of the combustion chamber, the low temperature conditions make the finite rate reaction rate dominate the apparent reaction rate. After the flame is established, the enhanced turbulent mixing makes the eddy dissipation rate become the dominant factor, realizing the dynamic conversion of the combustion process from chemical control to turbulent control. Thus, the combustion model synchronously calculates the finite rate reaction rate and the eddy dissipation rate through formulas (9), (10) and (11), and uses the competition mechanism between the two to control the combustion process. When the finite rate reaction rate is significantly lower than the eddy dissipation rate, the phenomenon of reaction occurring before the ignition of the flame in the simulation can be avoided, and the non-physical pre-reaction can be automatically suppressed.

[0121] In this preferred embodiment, the kinetic model can adopt a two-step global reaction mechanism for methane combustion, and the specific equations are as follows:

[0122] CH4 + 1.5O2 → CO + 2H2O (12)

[0123] CO + 0.5O2 → CO2 (13)

[0124] Please refer to Figure 4 , Figure 4 , which shows the reaction kinetic parameter table of the kinetic model provided according to the preferred embodiment of the present invention.

[0125] As Figure 4 shown, when the reaction of the kinetic model is formula (12), the pre-exponential factor is 1.5×10 13 , the activation energy is 1.256×10 8 , and the rate exponents are CH4: 1, O2: 1; when the reaction of the kinetic model is formula (13), the pre-exponential factor is 1×10 13 , the activation energy is 1.6747×10 8 , and the rate exponents are CO: 1, O2: 1.

[0126] According to the selected Realizable k-epsilon model as the turbulence model, the Finite-Rate / Eddy-Dissipation model as the combustion model, and the two-step global reaction mechanism of methane combustion as the kinetic model, the multi-objective optimization design system can construct a CFD model of the combustion chamber.

[0127] After that, according to the optimization design requirements of the operating performance and reliability of the heavy gas turbine combustion chamber, the multi-objective optimization design system can determine the multi-dimensional design variables and multi-dimensional design objective variables.

[0128] In this preferred embodiment, the multi-dimensional design variables may include the diameter of the metering hole of the combustion chamber (D meter ), the diameter of the mixing hole of the combustion chamber (D dilu ), the length of the mixing section of the combustion chamber (Ldilu-zone ) The outlet area of the combustion chamber transition section (S transi ), the fuel quantity of the primary premixed nozzle (qM pp ), the fuel quantity of the secondary premixed nozzle (qM sp ), the fuel quantity of the secondary diffusion nozzle (qM sd ) and the air intake quantity of the combustion chamber (qM air ). The first four dimensions can be classified as structural design variables, and the last four dimensions can be classified as operating condition design variables.

[0129] The multi-dimensional design objective variables can include four dimensions, namely the mass flow rate of nitrogen oxides at the outlet of the combustion chamber The average outlet temperature of the combustion chamber (T ave ), the pressure drop of the combustion chamber (p), and the average temperature gradient of the combustion chamber liner (gradT).

[0130] Specifically, the calculation method of the average temperature gradient gradT of the combustion chamber liner can first use CFD post-processing means to obtain the temperature gradient component data of all positions of the liner in the three-dimensional coordinate system for each sample in the sample set. Then, the magnitude of the temperature gradient value at all positions of the liner in the three-dimensional coordinate system is calculated through formula (14), and formula (14) is as follows:

[0131]

[0132] Wherein, is the temperature gradient component value at all positions of the liner in the x-axis direction, is the temperature gradient component value at all positions of the liner in the y-axis direction, is the temperature gradient component value at all positions of the liner in the z-axis direction.

[0133] The average value of the magnitudes of the temperature gradient values at all positions calculated by formula (14) is taken as the average temperature gradient value gradT of the combustion chamber liner for each sample.

[0134] Please continue to refer to Figure 3 , after constructing the CFD model and selecting and determining the multi-dimensional design variables and multi-dimensional design objective variables, the multi-objective optimization design system can perform pre-sampling on the multi-dimensional design variables and calculate the CFD model to determine the outlet average temperature distribution in the results of the multi-dimensional design objective variables, and establish an initial sample set.

[0135] Please refer to Figure 5 , Figure 5 shows a pre-sampling range table provided according to a preferred embodiment of the present invention.

[0136] The multi-objective optimization design system can be based on the settings such as Figure 5For the sampling range shown, the Latin Hypercube Sampling method is used to pre-sample the multi-dimensional design variables. Among them, the unit of the combustion chamber metering hole diameter (D meter ) is mm, the minimum value is 15 mm, and the maximum value is 45 mm, that is, the pre-sampling range is between 15 and 45 mm; the unit of the combustion chamber mixing hole diameter (D dilu ) is mm, the minimum value is 43 mm, and the maximum value is 73 mm, that is, the pre-sampling range is between 43 and 73 mm; the unit of the combustion chamber mixing section length (L dilu-zone ) is mm, the minimum value is 412 mm, and the maximum value is 442 mm, that is, the pre-sampling range is between 412 and 442 mm; the unit of the combustion chamber transition section outlet area (S transi ) is mm 2 , the minimum value is 32266 mm 2 , and the maximum value is 44675 mm 2 , that is, the pre-sampling range is between 32266 and 44675 mm 2 ; the unit of the fuel quantity of the primary premixed nozzle (qM pp ) is kg / s, the minimum value is 0.09324 kg / s, and the maximum value is 0.26936 kg / s, that is, the pre-sampling range is between 0.09324 and 0.26936 kg / s; the unit of the fuel quantity of the secondary premixed nozzle (qM sp ) is kg / s, the minimum value is 0 kg / s, and the maximum value is 0.06630 kg / s, that is, the pre-sampling range is between 0 and 0.06630 kg / s; the unit of the fuel quantity of the secondary diffusion nozzle (qM sd ) is kg / s, the minimum value is 0 kg / s, and the maximum value is 0.09946 kg / s, that is, the pre-sampling range is between 0 and 0.09946 kg / s; and the unit of the combustion chamber inlet air quantity (qM air ) is kg / s, the minimum value is 4.77 kg / s, and the maximum value is 33.89 kg / s, that is, the pre-sampling range is between 4.77 and 33.89 kg / s.

[0137] Based on Figure 5 the pre-sampled points obtained after pre-sampling within the shown sampling range, CFD model calculations are performed to determine the results of the multi-dimensional design target variables of the pre-sampled points. A sample set is established based on the pre-sampled points and the results of the multi-dimensional design target variables of the pre-sampled points. Here, the results of the multi-dimensional design target variables include the outlet average temperature, and based on the results of the multi-dimensional design target variables of each pre-sampled point, the outlet average temperature distribution of the sample set can be determined.

[0138] Through pre-sampling and CFD model calculation, the multi-objective optimization design system can identify that the average outlet temperature is mainly affected by the load number and equivalence ratio mapped by the operating condition design variables. Therefore, the multi-objective optimization design system can use the load number and equivalence ratio as the characteristics of the special sampling rule.

[0139] After that, as Figure 3 shown, based on the distribution interval of the average outlet temperature in the sample set, the multi-objective optimization design system can perform supplementary sampling and CFD model calculation on the multi-dimensional design variables according to the special sampling rule to update the sample set and construct a surrogate model based on the sample set.

[0140] Please refer to Figure 6 , Figure 6 which shows the layer division table of supplementary sampling provided according to the preferred embodiment of the present invention.

[0141] In this preferred embodiment, the multi-objective optimization design system can, according to the established sample set, based on the Figure 6 shown layer division table, divide the operating condition design variables into layers according to the load number and equivalence ratio values, and limit the sampling range of each layer after division according to the Figure 5 shown sampling range table. Here, the load number range to be studied is 30% - 80%, and the equivalence ratio range is 0.3 - 0.8. As Figure 6 shown, when the load number is in the ranges of 30% - 47%, 47% - 64%, and 64% - 80%, the equivalence ratio in the range of 0.30 - 0.47 can be divided into one layer, the equivalence ratio in the range of 0.47 - 0.67 can be divided into one layer, and the equivalence ratio in the range of 0.64 - 0.80 can be divided into one layer. Correspondingly, when the equivalence ratio is in the ranges of 0.30 - 0.47, 0.47 - 0.67, and 0.64 - 0.80, the load number in the range of 30% - 47% can be divided into one layer, the load number in the range of 47% - 64% can be divided into one layer, and the load number in the range of 64% - 80% can be divided into one layer.

[0142] After that, according to the Figure 6 shown divided layers, supplementary sampling is performed on the operating condition design variables. By performing equal-number sampling on the operating condition design variables in other layers without samples, it is ensured that there are sample distributions in each layer.

[0143] Specifically, based on the primary premixed nozzle fuel quantity qM pp of each sample in the sample set, the secondary premixed nozzle fuel quantity qM sp , the secondary diffusion nozzle fuel quantity qM sd and the combustion chamber inlet air quantity qM air, calculate the load number k and equivalence ratio eq of each sample through formulas (1), (2), (3) and (4). According to the load number k and equivalence ratio eq of each sample, determine the layer corresponding to each sample, and then determine the layer without samples. Then, perform equal - quantity sampling in the layer without samples according to the Figure 5 sampling range table, and the samples supplemented by sampling meet the limiting conditions of this layer, that is, meet the load number and equivalence ratio of this layer. Thus, after the supplementary sampling, there are sample distributions in each layer divided.

[0144] For the structural design variables, the multi - objective optimization design system can perform supplementary sampling according to the Latin hypercube sampling method. The number of supplementary sampling points of the structural design variables is equal to the number of supplementary sampling points of the operating condition design variables, and in this way, randomly combine the supplementary sampling points of the structural design variables and the supplementary sampling points of the operating condition design variables to construct the supplementary sampling points for supplementary sampling.

[0145] The multi - objective optimization design system performs CFD model calculations based on the supplementary sampling points to determine the multi - dimensional design objective variable results of the supplementary sampling points. Supplement the supplementary sampling points and the multi - dimensional design objective variable results of the supplementary sampling points to the sample set to update the sample set.

[0146] As Figure 3 shown, the multi - objective optimization design system can use the SPXY method to divide the sample set into a training set and a test set, and use the training set to construct and train the surrogate model and use the test set to test the generalization ability of the surrogate model. In this preferred embodiment, the type of the surrogate model constructed by the multi - objective optimization design system is a support vector machine (Epsilon - SVR), and use R 2 to evaluate the generalization effect of the current surrogate model.

[0147] Specifically, the steps of constructing the surrogate model and testing the generalization ability of the surrogate model can include hyperparameter determination, hyperparameter optimization, and surrogate model training and testing.

[0148] First, the multi - objective optimization design system can determine the hyperparameters of the surrogate model. In this preferred embodiment, Epsilon - SVR can have four hyperparameters that affect the fitting and generalization effects of the surrogate model, namely C, Kernel, gamma, and epsilon. Among them, C is the penalty coefficient. The larger the penalty coefficient C, the better the surrogate model fits the training set data, but it will reduce the generalization ability of the test set data; Kernel is the kernel function type of Epsilon - SVR. In one example, the multi - objective optimization design system can select the RBF kernel to provide the ability to map to a high - dimensional space; gamma is the selected kernel function coefficient, which is used to control the influence range of the kernel function; epsilon is the slack variable, which is used to control the error degree allowed by the surrogate model.

[0149] After that, the multi-objective optimization design system can perform hyperparameter optimization for each dimension of the design objective variables. In this preferred embodiment, the multi-objective optimization design system can first determine the ranges of the penalty coefficient C, kernel function coefficient gamma, and relaxation variable epsilon in the hyperparameters for each dimension of the design objective variables. Then, it uses the Optuna library based on the Bayesian optimization algorithm as the optimization method, and uses K-fold cross-validation as the method to evaluate the generalization ability of the model and prevent overfitting for hyperparameter optimization.

[0150] Then, the multi-objective optimization design system can perform surrogate model training and testing for each dimension of the design objective variables. In this preferred embodiment, the multi-objective optimization design system can, based on the results of K-fold cross-validation, select a set of hyperparameters with the highest cross-validation score to train and test the surrogate model. Then use R 2 to evaluate the test effect of the test set, that is, to evaluate the generalization ability of the surrogate model. When the R 2 value is greater than the preset threshold, it indicates that the generalization ability of the surrogate model is acceptable and meets the expectations; if it is less than the preset threshold, it indicates that the generalization ability of the surrogate model does not meet the expectations. In one embodiment, the preset threshold can be 0.8.

[0151] In the above embodiment, after the multi-objective optimization design system performs surrogate model testing for each dimension of the design objective variables, if the R 2 value of some design objective variables of the surrogate model is less than 0.8, it indicates that the generalization ability of the corresponding surrogate model does not meet the expectations. Only when the R 2 value of each dimension of the design objective variables is greater than 0.8, does it indicate that the generalization ability of the surrogate model meets the expectations.

[0152] Furthermore, please continue to refer to Figure 3 , when the generalization ability of the surrogate model does not meet the expectations, the multi-objective optimization design system divides the operating condition design variables into layers according to the load number and equivalent ratio based on the outlet average temperature distribution in the existing sample set, and re-limits the sampling range of each layer to perform secondary supplementary sampling. Then, based on the secondary supplementary sampling points, CFD model calculations are performed to supplement the sample set.

[0153] Please refer to Figure 7 , Figure 7 which shows the layer division table of the secondary supplementary sampling provided according to the preferred embodiment of the present invention.

[0154] The average outlet temperature among the multi-dimensional design target variables exhibits strong sensitivity to the equivalence ratio, and different equivalence ratio ranges correspond to a specific distribution range of the average outlet temperature. In this preferred embodiment, the multi-objective optimization design system can further divide the operating condition design variables into layers according to the equivalence ratio value based on the sample set established after supplementary sampling, and limit the sampling range of each layer according to the Figure 5 sampling range table.

[0155] For example, as Figure 7 shown, it is possible to further re-divide the layers based on the equivalence ratio under the sampling range shown in Figure 6 so as to refine the distribution range of the average outlet temperature. When the load numbers are in the ranges of 30% - 47%, 47% - 64%, and 64% - 80%, the equivalence ratio in the range of 0.30 - 0.47 can be further divided into 9 layers: 0.30 - 0.33 as one layer, 0.33 - 0.40 as one layer, 0.40 - 0.45 as one layer, 0.45 - 0.51 as one layer, 0.51 - 0.57 as one layer, 0.57 - 0.64 as one layer, 0.64 - 0.70 as one layer, 0.70 - 0.78 as one layer, and 0.78 - 0.80 as one layer.

[0156] In this way, the multi-objective optimization design system realizes active sampling of the average outlet temperature by dividing the number of layers of the equivalence ratio. Thus, when sampling the multi-dimensional design variables, sampling of the multi-dimensional design target variables is taken into account, thereby simultaneously considering the data distributions of the input and output of the CFD model, further improving the problem of uneven data distribution, and enhancing the quality of the sample set.

[0157] Then, similar to the above supplementary sampling, the multi-objective optimization design system can perform secondary supplementary sampling on the operating condition design variables according to the Figure 7 number of divided layers shown. By performing equal-number sampling on the operating condition design variables in other layers without samples, it is ensured that there are sample distributions in each layer.

[0158] For the structural design variables, the multi-objective optimization design system can perform secondary supplementary sampling according to the Latin hypercube sampling method. The number of secondary supplementary sampling points for the structural design variables is equal to the number of secondary supplementary sampling points for the operating condition design variables, and in this way, the secondary supplementary sampling points for the structural design variables and the secondary supplementary sampling points for the operating condition design variables are randomly combined to construct the secondary supplementary sampling points for the secondary supplementary sampling.

[0159] The multi-objective optimization design system performs CFD model calculations based on the secondary supplementary sampling points to determine the results of the multi-dimensional design target variables for the secondary supplementary sampling points. The secondary supplementary sampling points and the results of the multi-dimensional design target variables for the secondary supplementary sampling points are supplemented into the sample set to update the sample set.

[0160] After that, in this preferred embodiment, the multi-objective optimization design system can continue to use the same training and testing steps in the supplementary sampling stage to construct the surrogate model and evaluate the generalization ability. Specifically, the multi-objective optimization design system can use the SPXY method to divide the sample set into a training set and a testing set, and use the training set to construct and train the surrogate model and use the testing set to test the generalization ability of the surrogate model. The type of the surrogate model constructed by the multi-objective optimization design system is a support vector machine (Epsilon-SVR), and use R 2 to evaluate the generalization effect of the current surrogate model.

[0161] In the case where the generalization ability of the surrogate model does not meet the expectation, the multi-objective optimization design system can continue to repeat the above-mentioned secondary supplementary sampling process and update the sample set to construct the surrogate model until after testing the surrogate model for each dimensional design objective variable, the R 2 values of all design objective variables are greater than a preset threshold (for example, 0.8), indicating that the generalization ability of the surrogate model meets the expectation.

[0162] Please refer to Figure 8 , Figure 8 which shows the evaluation result table of the surrogate model provided according to the preferred embodiment of the present invention.

[0163] As Figure 8 shown, in this preferred embodiment, after the multi-objective optimization design system performs secondary supplementary sampling, it constructs, trains and evaluates the surrogate model based on the training set divided from the updated sample set, and performs surrogate model evaluation on the mass flow rate of nitrogen oxides at the outlet of the combustion chamber to obtain an R 2 value of 0.83, performs surrogate model evaluation on the average temperature (T ave ) at the outlet of the combustion chamber to obtain an R 2 value of 0.98, performs surrogate model evaluation on the pressure drop (p) of the combustion chamber to obtain an R 2 value of 1, and performs surrogate model evaluation on the average temperature gradient (gradT) of the combustion chamber flame tube to obtain an R 2 value of 0.98. Correspondingly, the surrogate model is tested and evaluated based on the testing set divided from the updated sample set. For the mass flow rate of nitrogen oxides at the outlet of the combustion chamber surrogate model evaluation is performed to obtain an R 2 value of 0.89, for the average temperature (T ave ) at the outlet of the combustion chamber surrogate model evaluation is performed to obtain an R 2 value of 0.98, for the pressure drop (p) of the combustion chamber surrogate model evaluation is performed to obtain an R 2The value is 0.99, and the surrogate model is evaluated for the average temperature gradient (gradT) of the combustion chamber flame tube, obtaining an R 2 value of 0.96. Thus, the R 2 values of all design objective variables are greater than the preset threshold of 0.8, and the generalization ability of this surrogate model meets the expectations. Preferably, the multi-objective optimization design system can construct a final surrogate model based on the training set.

[0164] As Figure 3 shown, after the multi-objective optimization design system constructs a surrogate model with satisfactory generalization ability, it can determine the decision variables, objective functions, and constraint conditions of the multi-objective optimization problem.

[0165] In this preferred embodiment, the decision variables of the multi-objective optimization problem can be the variables of each dimension in the multi-dimensional design variables, that is, the diameter of the metering hole of the combustion chamber (D meter ), the diameter of the mixing hole of the combustion chamber (D dilu ), the length of the mixing section of the combustion chamber (L dilu-zone ), the outlet area of the transition section of the combustion chamber (S transi ), the fuel quantity of the primary premixed nozzle (qM pp ), the fuel quantity of the secondary premixed nozzle (qM sp ), the fuel quantity of the secondary diffusion nozzle (qM sd ), and the air intake quantity of the combustion chamber (qM air ).

[0166] The objective function can be a function of the multi-dimensional design variables, that is, a function of the diameter of the metering hole of the combustion chamber (D meter ), the diameter of the mixing hole of the combustion chamber (D dilu ), the length of the mixing section of the combustion chamber (L dilu-zone ), the outlet area of the transition section of the combustion chamber (S transi ), the fuel quantity of the primary premixed nozzle (qM pp ), the fuel quantity of the secondary premixed nozzle (qM sp ), the fuel quantity of the secondary diffusion nozzle (qM sd ), and the air intake quantity of the combustion chamber (qM air ). In some embodiments, the multi-objective optimization design system can use the surrogate model constructed above that can obtain the results of multi-dimensional design objective variables as the objective function. [[ID=4८]]

[0167] The constraint conditions of the multi-objective optimization problem can be as shown in formulas (5) to (8):

[0168]

[0169] eq min ≤eq≤eq max (6)

[0170] q min ≤qM pp / qM total ≤q max (7)

[0171] T ave_min ≤T ave ≤T ave_max (8)

[0172] Among them, formula (5) is the load number constraint, the minimum load number k min It can be taken as 30%, the maximum load number k max It can be taken as 80%; Formula (6) is the equivalence ratio eq constraint, the minimum equivalence ratio eq min It can be taken as 0.3, the maximum equivalent ratio eq max It can be taken as 0.8. The equivalence ratio constraint range is the equivalence ratio range commonly used in combustion chamber operation. Formula (7) is the fuel flow distribution ratio constraint. The minimum flow distribution ratio q min It can be taken as 0.6, the maximum flow distribution ratio q max It can be taken as 0.65; Formula (8) is the average outlet temperature T ave Constraint, minimum average outlet temperature T ave_min The maximum outlet average temperature T can be 1397.15K. ave_max You can get 1623.15K.

[0173] like Figure 3 As shown, the multi-objective optimization design system can employ a heuristic algorithm for multi-objective optimization, then utilize a proxy model to evaluate a large number of candidate solutions that take into account the key structures and operating conditions of the combustion chamber, thereby obtaining a Pareto solution set that takes into account both the performance and life indicators of the combustion chamber. In this preferred embodiment, the heuristic algorithm can be a non-dominated sorting genetic algorithm (NSGA-II). The non-dominated sorting genetic algorithm is a heuristic algorithm that can obtain a non-dominated solution set and can solve multi-objective optimization problems.

[0174] In addition, before using NSGA-II to solve the optimization problem based on the agent model, it can also include a process of determining the optimization direction of the optimization target. In this preferred embodiment, the optimization target can include the outlet nitrogen oxide mass flow rate The average temperature at the combustion chamber outlet (T ave ), combustion chamber pressure drop (p) and combustion chamber flame tube average temperature gradient (gradT). When determining the optimization direction, it also involves minimizing and maximizing different optimization objectives. The multi-objective optimization design system can be used to calculate the outlet nitrogen oxide mass flow rate. Optimization is carried out to minimize the pressure drop (p) in the combustion chamber and the average temperature gradient (gradT) in the combustion chamber liner, and to maximize the average temperature (T ave ) at the outlet of the combustion chamber. To unify the optimization direction, the multi-objective optimization design system can reverse the optimization direction of the average temperature (T ave ) at the outlet of the combustion chamber.

[0175] After determining the optimization direction of the optimization objectives, the multi-objective optimization design system can use the surrogate model constructed above that can predict the values of all optimization objectives as the objective function. Then, NSGA-II can solve the multi-objective optimization problem based on the determined optimization direction and the objective function.

[0176] Next, the multi-objective optimization design system can set the NSGA-II parameters. The NSGA-II parameters mainly include the number of individuals pop_size in the initial population, the crossover probability prob, the mutation amplitude eta, the sampling method sampling, and the number of population iterations n_gen. In this preferred embodiment, the number of individuals pop_size in the initial population can be set to 200, the crossover probability prob can be set to 0.9, the mutation amplitude eta can be set to 20, the sampling method sampling can be Latin hypercube sampling, and the number of population iterations n_gen is set to 2000 times.

[0177] Finally, the multi-objective optimization design system can use NSGA-II for population iteration. First, perform non-dominated sorting on the initial population, and generate the first-generation subgroup through selection, crossover, and mutation. Then, merge the parent population and the first-generation subgroup, and then perform non-dominated sorting and crowding degree calculation to generate the Pareto front solutions. Then, select the elite individuals as the new parent population, and repeat the selection, crossover, and mutation processes. Continuously perform population replacement, and each replacement solves the objective function value and constraint conditions according to the surrogate model until the number of population iterations n_gen reaches 2000 times, and finally output the Pareto front solutions as the Pareto solution set that takes into account both the combustion chamber performance and life indicators.

[0178] In the process of multi-objective optimization based on the Pareto dominance relationship adopted by the multi-objective optimization design system, the Pareto solution set is screened through non-dominated sorting and crowding degree calculation. Among the finally obtained Pareto fronts, any two non-dominated solutions satisfy: if solution A is superior to solution B in a certain objective, then solution B must be superior to solution A in another objective. Through this characteristic of the finally obtained Pareto solution set, the competitive trade-off between different optimization objectives is reflected.

[0179] Such as Figure 3As shown, after obtaining the Pareto solution set, the multi-objective optimization design system can make multi-objective decisions based on the Pareto solution set to obtain the optimized design solution for the combustion chamber. In this preferred embodiment, the multi-objective optimization design system can use the TOPSIS decision-making method to obtain the compromise solution of the Pareto solution set.

[0180] Specifically, the multi-objective optimization design system can first normalize each non-dominated solution in the Pareto solution set to eliminate the influence of dimensions. Then, obtain the positive ideal solution and the negative ideal solution for each dimension among all non-dominated solutions, that is, the optimal value and the worst value, where d is the d-th dimension among all non-dominated solutions. Calculate the distances and between the positive ideal solution and respectively. The calculation formulas are as follows:

[0181]

[0182] where rl d为 is the l-th non-dominated solution in the d-th dimension.

[0183] Then, calculate the proximity S l to the ideal solution through formula (17):

[0184]

[0185] where, the closer to the positive ideal solution the smaller it is, and S l is closer to 1.

[0186] Based on the proximity S l calculated by formula (17), rank the superiority and inferiority of the non-dominated solutions. Select the non-dominated solution with the proximity S l closest to 1 as the compromise solution selected by TOPSIS decision-making, and use the optimal values of the non-dominated solutions of different optimization objectives in the Pareto solution set as the preference solutions.

[0187] Then, based on TOPSIS, extract the decision variable values (i.e., each dimension of design variables), substitute them into the CFD model for calculation, and obtain the verification result values. Then, compare the above verification result values with the multi-dimensional design objective variable results determined by the CFD model in the surrogate model training set to evaluate the improvement amplitude of the optimization algorithm.

[0188] Please refer to Figure 9 , Figure 9 which shows a comparison graph of the normalized values of the compromise solution and different preference solutions in the Pareto solution set provided according to the preferred embodiment of the present invention with the training set.

[0189] After the compromise solutions and preferred solutions obtained throughout the process are verified by the CFD model, the comparison with the target performance parameters of the original training set is as follows Figure 9 shown. The vertical axis is the normalized value (Normalized Indicator Value) of each dimension design objective variable, and the horizontal axis is each dimension design objective variable of the combustor, namely the outlet nitrogen oxide mass flow rate 901, the average outlet temperature (T ave ) 902 of the combustor, the pressure drop (p) 903 of the combustor, and the average temperature gradient (gradT) 904 of the combustor flame tube.

[0190] Curve 910 is the normalized value of the results calculated by the CFD model for each decision variable corresponding to the compromise solution of the Pareto solution set obtained by using the TOPSIS decision-making method; Curve 920 is the normalized value of the results calculated by the CFD model for each decision variable corresponding to the selected preferred solution when the outlet nitrogen oxide mass flow rate 901 is minimized; Curve 930 is the normalized value of the results calculated by the CFD model for each decision variable corresponding to the selected preferred solution when the average temperature gradient (gradT) 904 of the combustor flame tube is minimized; Curve 940 is the normalized value of the results calculated by the CFD model for each decision variable corresponding to the selected preferred solution when the average outlet temperature (T ave ) 902 of the combustor is maximized and the pressure drop (p) of the combustor is minimized; and Curve 950 is the normalized value of the results of the multi-dimensional design objective variables obtained by CFD calculation corresponding to multiple sampling points in the training set divided by the final sample set.

[0191] Among the verification result values calculated by the CFD model for the decision variables corresponding to each preferred solution, values better than the values of the multi-dimensional design objective variables obtained by CFD calculation in the original training set can be found.

[0192] For example, based on multiple curves 950, it can be determined that the normalized value of the outlet nitrogen oxide mass flow rate 901 of the optimal multi-dimensional design target variable result obtained by CFD calculation in the training set is 0.17145, the normalized value of the average outlet temperature 902 of the combustion chamber is 0.89624, the normalized value of the combustion chamber pressure drop 903 is 0.05638, and the normalized value of the average temperature gradient of the combustion chamber flame tube 904 is 0.18036. Based on the curves 920, 930, and 940 corresponding to each preference solution, it can be determined that the normalized value of the outlet nitrogen oxide mass flow rate 901 of the optimal result of each decision variable corresponding to each preference solution after being calculated by the CFD model is 0.14453, the normalized value of the average outlet temperature 902 of the combustion chamber is 0.92854, the normalized value of the combustion chamber pressure drop 903 is 0.0580303, and the normalized value of the average temperature gradient of the combustion chamber flame tube 904 is 0.11435. It can be seen that the normalized values of the latter are all better than those of the former.

[0193] In addition, based on curve 910, it can be determined that the normalized value of the outlet nitrogen oxide mass flow rate 901 of the result of the decision variable corresponding to the compromise solution after being calculated by CFD is 0.15992, the normalized value of the average outlet temperature 902 of the combustion chamber is 0.53699, the normalized value of the combustion chamber pressure drop 903 is 0.04029, and the normalized value of the average temperature gradient of the combustion chamber flame tube 904 is 0.29337. The normalized values of the outlet nitrogen oxide mass flow rate 901 and the combustion chamber pressure drop 903 are better than the normalized values of the optimal multi-dimensional design target variable results obtained by CFD calculation in the training set. Thus, it can be seen that after the decision variables corresponding to the above-mentioned compromise solution in the non-dominated solution set are calculated by CFD, the normalized values of nitrogen oxide emissions and combustion chamber pressure drop are both better than the normalized values of the multi-dimensional design target variable results obtained by CFD calculation in the original training set.

[0194] The following are the comparative examples of the above preferred embodiments, based on which the multi-objective optimization design system for a heavy-duty gas turbine combustion chamber proposed by the present invention is compared and described.

[0195] Different from the multi-objective optimization design method of the heavy-duty gas turbine combustion chamber based on the surrogate model in the above preferred embodiment, the comparative example adopts a traditional sensitivity-based design method.

[0196] First, determine the range of the design variables for sensitivity analysis. In the comparative example, the design variables are the same as the dimensions of the multi-dimensional design variables in the preferred embodiment, and the determined range of the design variables is the same as the sampling range determined when pre-sampling each dimension of the design variables in the preferred embodiment. The sensitivity-based design method of the design system in the comparative example sets the variation range of the combustion chamber metering hole diameter (D meter ) to 15 - 45 mm, and the combustion chamber mixing hole diameter (Ddilu ) is set to vary from 43 to 73 mm, and the length of the combustion chamber mixing section (L dilu-zone ) is set to vary from 412 to 442 mm, and the outlet area of the combustion chamber transition section (S transi ) is set to vary from 32266 to 44675 mm 2 , the fuel quantity of the primary premixed nozzle (qM pp ) is set to vary from 0.09324 to 0.26936 kg / s, the fuel quantity of the secondary premixed nozzle (qM sp ) is set to vary from 0 to 0.06630 kg / s, the fuel quantity of the secondary diffusion nozzle (qM sd ) is set to vary from 0 to 0.09946 kg / s, and the air intake quantity of the combustion chamber (qM air ) is set to vary from 4.77 to 33.89 kg / s.

[0197] After that, the design system in the comparative example sets the initial values and increment values of each design variable. According to the determined ranges of the design variables, the design system sets the initial values of the combustion chamber metering hole diameter (D meter ), the combustion chamber mixing hole diameter (D dilu ), the length of the combustion chamber mixing section (L dilu-zone ), the outlet area of the combustion chamber transition section (S transi ), the fuel quantity of the primary premixed nozzle (qM pp ), the fuel quantity of the secondary premixed nozzle (qM sp ), the fuel quantity of the secondary diffusion nozzle (qM sd ) and the air intake quantity of the combustion chamber (qM air ) to 15 mm, 43 mm, 412 mm, 3,2266 mm 2 , 0.09324 kg / s, 0 kg / s, 0 kg / s and 4.77 kg / s respectively, and sets appropriate increment values within the ranges of each design variable.

[0198] Based on the ranges of the design variables and the set initial values and increment values of each design variable, the sensitivity is calculated according to the single factor method. That is, according to the increment value, the value of one design variable is changed while the others remain unchanged, and then it is calculated using the combustion chamber CFD model, and then the sensitivity value of the design target variable is determined according to formula (18). Here, the design target variables include the mass flow rate of nitrogen oxides at the outlet of the combustion chamber the average temperature at the outlet of the combustion chamber (T ave ), the pressure drop of the combustion chamber (p), and the average temperature gradient of the combustion chamber flame tube (gradT). Formula (18) can be shown as follows:

[0199]

[0200] Among them, S hg is the sensitivity value of the h-th optimization objective (i.e., the above-mentioned design objective variables, a total of 4) with respect to the g-th decision variable (i.e., the above-mentioned design variables, a total of 8), Δf h / f h is the relative change of the h-th optimization objective when the g-th decision variable increases by Δz g , Δz g is the increment value of the g-th decision variable, and Δz g / z g is the relative change when the g-th decision variable increases by Δz g .

[0201] Then, the design system of the comparative example can separately find the intervals with high sensitivity values of the mass flow rate of nitrogen oxides at the outlet of the combustion chamber the average temperature (T ave ) at the outlet of the combustion chamber, the pressure drop (p) of the combustion chamber, and the average temperature gradient (gradT) of the combustion chamber flame tube. Based on the optimization directions of the optimization objectives, for minimizing the design objective variables, i.e., the mass flow rate of nitrogen oxides at the outlet the pressure drop (p) of the combustion chamber, and the average temperature gradient (gradT) of the combustion chamber flame tube, the values of different design variables are determined according to the negative sensitivity value intervals respectively; for maximizing the design objective variables, i.e., the average temperature (T ave ) at the outlet of the combustion chamber, the values of different design variables are determined according to the positive sensitivity value intervals respectively. Finally, the design system can determine 4 sets of combinations of design variable values for different design objective variables, and then perform CFD model calculations based on the combinations of design variable values to determine the values of the design objective variables corresponding to each combination of design variable values.

[0202] Comparing the sensitivity-based design method provided by the comparative example with the multi-objective optimization design method based on the surrogate model provided by the preferred embodiment, the comparative example calculates the sensitivity according to the single-factor method, adopts single-objective gradient optimization based on local sensitivity analysis, and only adjusts the parameters within the sensitive interval of the design variables, lacking the ability of global exploration. The comparative example considers different optimization objectives separately and does not consider the competition relationship between multiple objectives. Therefore, the comparative example cannot find the optimal combination of values that can balance the improvement of the combustion chamber performance and life.

[0203] In the multi-objective optimization design method based on the surrogate model provided by the preferred embodiment, the surrogate model is a global approximation model that comprehensively considers different optimization objectives, can avoid relying on local gradients, and explores the non-convex solution space through multi-objective optimization based on the Pareto dominance relationship to find a solution set closer to the true Pareto front. In the process of multi-objective optimization based on the Pareto dominance relationship, the Pareto solution set is screened through non-dominated sorting and crowding degree calculation to obtain a Pareto solution set that can reflect the competitive trade-off between different optimization objectives.

[0204] In summary, the multi-objective optimization design method for a heavy-duty gas turbine combustor provided by the present invention can establish a high-precision surrogate model based on a high-quality sample set, thereby greatly reducing the high time cost brought by numerical simulation on the premise of ensuring the accuracy of the combustor optimization design, and at the same time considering multiple optimization objectives while reducing the combustor optimization design cost to meet the requirements for optimizing the design of various indicators of the combustor under different working conditions.

[0205] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions not illustrated and described herein but understood by those skilled in the art.

[0206] Those skilled in the art will understand that information, signals, and data can be represented using any of a variety of different technologies and techniques. For example, the data, instructions, commands, information, signals, bits, symbols, and chips described throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

[0207] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability of hardware and software, the various illustrative components, blocks, modules, circuits, and steps are described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and the design constraints imposed on the overall system. Skilled artisans may implement the described functionality in different ways for each particular application, but such implementation decisions should not be construed as causing a departure from the scope of the present invention.

[0208] The various illustrative logical modules and circuits described in connection with the embodiments disclosed herein can be implemented or performed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

[0209] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such that the processor can read from, and write to, the storage medium. In the alternative, the storage medium may be integrated into the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In the alternative, the processor and the storage medium may reside as discrete components in a user terminal.

[0210] The foregoing description of the disclosure has been provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-objective optimization design method for a heavy-duty gas turbine combustor, characterized in that Including steps: S1. Pre-sampling multidimensional design variables and determining multidimensional design target variable results at pre-sampling points through CFD model calculation, wherein the CFD model is a combustion chamber CFD model, and the multidimensional design target variable results include an average outlet temperature; S2. establishing a sample set based on the pre-sampling points and the multi-dimensional design target variable results of the pre-sampling points; S3. Based on the outlet average temperature distribution interval in the sample set, perform supplementary sampling and CFD model calculation on the multidimensional design variables according to special sampling rules to update the sample set and construct a proxy model based on the sample set; S4. In response to the generalization ability of the proxy model not meeting expectations, based on the outlet average temperature distribution interval in the sample set, performing secondary supplementary sampling and CFD model calculation on the multidimensional design variables according to special sampling rules to update the sample set and construct a proxy model based on the sample set; S5. Repeat S4 until the generalization ability of the proxy model meets expectations; S6. performing multi-objective optimization using a heuristic algorithm and evaluating candidate solutions using the surrogate model to determine a Pareto solution set that takes into account both combustion chamber performance and life indicators; as well as S7. Perform multi-objective decision making based on the Pareto solution set to obtain an optimal design solution for the combustion chamber.

2. The multi-objective optimization design method according to claim 1, characterized in that, The step of determining the multidimensional design variables and the multidimensional design target variables includes: Based on the optimization design requirements of the operating performance and reliability of the heavy-duty gas turbine combustor, the multidimensional design variables that take into account the combustor structure and operating conditions, as well as the multidimensional design target variables that take into account the combustor performance and life are determined, wherein the multidimensional design variables include structural design variables and operating condition design variables, and the multidimensional design target variables include performance design target variables and life design target variables.

3. The multi-objective optimization design method according to claim 2, characterized in that The S3 includes: Based on the outlet average temperature distribution interval in the sample set, the operating condition design variables are divided into layers according to special sampling rules; Determining, based on the sample set, a layer corresponding to each sample in the sample set, and performing supplementary sampling on the operating condition design variables in the layer without samples; Performing supplementary sampling on the structural design variables, wherein the number of supplementary sampling points of the structural design variables is equal to the number of supplementary sampling points of the operating condition design variables; combining the supplementary sampling points of the structural design variables and the supplementary sampling points of the operating condition design variables to determine supplementary sampling points for supplementary sampling; Determine the multi-dimensional design target variable results of the supplementary sampling points by CFD model calculation, and update the sample set based on the supplementary sampling points and the multi-dimensional design target variable results of the supplementary sampling points; and A proxy model is constructed based on a training set in the sample set, and a generalization ability of the proxy model is evaluated based on a test set in the sample set.

4. The multi-objective optimization design method according to claim 2, wherein The S4 includes: In response to the generalization ability of the surrogate model not meeting expectations, based on the outlet average temperature distribution interval in the sample set, the operating condition design variable is subdivided into layers according to a special sampling rule; Determine the layer corresponding to each sample in the sample set according to the sample set, and perform secondary supplementary sampling on the operating condition design variables in the layers without samples; Perform secondary supplementary sampling on the structural design variables, and the number of secondary supplementary sampling points of the structural design variables is equal to the number of secondary supplementary sampling points of the operating condition design variables; Combine the secondary supplementary sampling points of the structural design variables and the secondary supplementary sampling points of the operating condition design variables to determine the secondary supplementary sampling points of the secondary supplementary sampling; Calculate and determine the multi-dimensional design target variable results of the secondary supplementary sampling points through the CFD model, and update the sample set based on the secondary supplementary sampling points and the multi-dimensional design target variable results of the secondary supplementary sampling points; and Construct a surrogate model based on the training set in the sample set, and evaluate the generalization ability of the surrogate model based on the test set in the sample set.

5. The multi-objective optimization design method according to claim 2, wherein The structural design variables include the combustion chamber metering hole diameter, the combustion chamber mixing hole diameter, the combustion chamber mixing section length, and the combustion chamber transition section outlet area, and the operating condition design variables include the fuel quantity of the primary premixed nozzle, the fuel quantity of the secondary premixed nozzle, the fuel quantity of the secondary diffusion nozzle, and the combustion chamber inlet air quantity; The performance design target variables include the outlet nitrogen oxide mass flow rate, the outlet average temperature, and the combustion chamber pressure drop, and the life design target variable includes the average temperature gradient of the combustion chamber flame tube.

6. The multi-objective optimization design method according to claim 1, wherein The combustion chamber CFD model is established based on the turbulent combustion model and the kinetic model of the combustion chamber simulation.

7. The multi-objective optimization design method according to claim 1, wherein: The decision variables of the multi-objective optimization problem of the multi-objective optimization are the design variables of each dimension of the multi-dimensional design variables, the objective function is a function of the multi-dimensional design variables, and the constraint conditions include the load number constraint, the equivalence ratio constraint, the fuel flow distribution ratio constraint, and the outlet average temperature constraint.

8. The multi-objective optimization design method according to claim 1, characterized in that The heuristic algorithm includes a genetic algorithm, a multi-objective particle swarm optimization algorithm, or a simulated annealing algorithm, and the multi-objective decision-making includes the technique for order preference by similarity to an ideal solution (TOPSIS) or the analytic hierarchy process (AHP).

9. A multi-objective optimization design system for a heavy-duty gas turbine combustor, characterized in that, Comprising: A memory on which computer instructions are stored; And A processor connected to the memory and configured to execute the computer instructions stored on the memory to implement the multi-objective optimization design method for a heavy-duty gas turbine combustion chamber according to any one of claims 1 to 8.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that, When the computer instructions are executed by the processor, the multi-objective optimization design method for a heavy-duty gas turbine combustion chamber according to any one of claims 1 to 8 is implemented.

Images