Gas turbine overall performance simulation method based on pod and neural network reduced order model

Abstract

This application proposes a method for simulating the overall performance of a gas turbine based on a reduced-order model using Point of View (POD) and neural networks. The method includes: first, sampling multiple results from a sample space formed based on the gas turbine's design requirements; then, using these multiple results as boundary conditions to obtain physical field data of the gas turbine under multiple operating conditions, dividing this data into training and test datasets. Next, based on the training dataset, an initial reduced-order model is constructed using POD and neural networks. Then, based on the accuracy and time taken by the model to calculate the test dataset, the initial reduced-order model is updated to obtain a target reduced-order model whose accuracy and time meet preset conditions. Finally, based on the target reduced-order model, a zero-dimensional performance simulation model is optimized, and the overall performance of the gas turbine is simulated to obtain simulation results. This method ensures the efficiency of the zero-dimensional overall performance simulation of the gas turbine while enabling the simulation results to display the flow details of components, thus improving the reliability of the performance simulation.

Description

Technical Field

[0001] This application relates to the field of gas turbine overall performance simulation technology, and in particular to a gas turbine overall performance simulation method based on POD and neural network reduced-order model. Background Technology

[0002] The operating status and performance of heavy-duty gas turbines are determined by the matching of various components. Overall performance simulation analysis plays an important role in the development of gas turbines, which can shorten the development cycle, reduce development costs, and achieve a comprehensive improvement in product performance, economy and reliability.

[0003] The most commonly used simulation method at present is zero-dimensional overall performance simulation based on component characteristic maps. This method uses high-dimensional component simulation models to perform large-scale calculations to obtain the power and efficiency under different inlet guide vanes (IGVs), equivalent speeds, pressure ratios, or flow rates, forming the component characteristic maps. Then, zero-dimensional simulation of the whole machine is carried out based on these maps. However, the simulation results of this method cannot reflect the flow details inside the components, which may lead to distorted performance predictions and affect the efficiency of gas turbine optimization and development. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in the related art.

[0005] Therefore, the first objective of this application is to propose a gas turbine overall performance simulation method based on POD and neural network reduced-order models. This method involves sampling results from a sample space formed by the boundary condition ranges of gas turbine components, merging and supplementing other CFD calculation results generated during the design iteration process, generating and iteratively updating the component reduced-order models, thereby optimizing the zero-dimensional performance simulation model and completing the overall performance simulation of the gas turbine. This aims to ensure the efficiency of the zero-dimensional overall performance simulation of the gas turbine while enabling the simulation results to display the flow details of the components, thus improving the reliability of the performance simulation.

[0006] The second objective of this application is to propose a gas turbine overall performance simulation device based on POD and a reduced-order neural network model.

[0007] The third objective of this application is to propose an electronic device.

[0008] The fourth objective of this application is to provide a computer-readable storage medium.

[0009] The fifth objective of this application is to provide a computer program product.

[0010] To achieve the above objectives, the first aspect of this application proposes a method for simulating the overall performance of a gas turbine based on a reduced-order model using a POD and a neural network, comprising:

[0011] Based on the component boundary condition value range determined by the gas turbine design requirements, the Latin hypercube sampling method is used to sample in the sample space formed by the boundary condition value range to obtain multiple sampling results, wherein the number of the sampling results is less than the number threshold and is uniformly distributed in the sample space. Using the multiple sampling results as boundary conditions, physical field data of the gas turbine under multiple operating conditions are obtained, wherein the physical field data includes at least temperature field data, pressure field data, and velocity field data; The physical field data under the multiple working conditions are divided into training datasets and test datasets according to the working conditions. Based on the training dataset, an initial reduced-order model is constructed using snapshot intrinsic orthogonal decomposition (POD) and neural network methods. Based on the accuracy and time of the initial order reduction model in calculating each physical field data in the test dataset, the initial order reduction model is updated to obtain a target order reduction model whose accuracy and time meet the preset conditions. Based on the target reduced-order model, the zero-dimensional performance simulation model corresponding to the gas turbine is optimized to obtain the target performance simulation model; Based on the target performance simulation model, the overall performance of the gas turbine is simulated, and the simulation results are obtained.

[0012] To achieve the above objectives, a second aspect of this application proposes a gas turbine overall performance simulation device based on a POD and a neural network reduced-order model, comprising: The sampling module is used to sample from the sample space formed by the boundary condition value range determined by the gas turbine design requirements using the Latin hypercube sampling method, obtaining multiple sampling results. The number of sampling results is less than a threshold and is uniformly distributed within the sample space. The determination module is used to obtain physical field data of the gas turbine under multiple operating conditions by using the multiple sampling results as boundary conditions respectively, wherein the physical field data includes at least temperature field data, pressure field data and velocity field data; The processing module is used to divide the physical field data under the multiple working conditions into training datasets and test datasets according to the working conditions. The generation module is used to construct an initial reduced-order model based on the training dataset using snapshot intrinsic orthogonal decomposition (POD) and neural network methods. The update module is used to update the initial order reduction model based on the accuracy and time calculated for each physical field data in the test dataset by the initial order reduction model, so as to obtain a target order reduction model whose accuracy and time meet the preset conditions. The optimization module is used to optimize the zero-dimensional performance simulation model of the gas turbine based on the target reduced-order model, so as to obtain the target performance simulation model. The simulation module is used to simulate the overall performance of the gas turbine based on the target performance simulation model and obtain simulation results.

[0013] To achieve the above objectives, a third aspect of this application provides an electronic device, including: a processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the gas turbine overall performance simulation method based on POD and neural network reduced-order model as described in the first aspect embodiment.

[0014] To achieve the above objectives, a fourth aspect of this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the gas turbine overall performance simulation method based on POD and a neural network reduced-order model as described in the first aspect embodiment.

[0015] To achieve the above objectives, the fifth aspect of this application proposes a computer program product, including a computer program that, when executed by a processor, implements the gas turbine overall performance simulation method based on POD and neural network reduced-order model as described in the first aspect embodiment.

[0016] The gas turbine overall performance simulation method based on POD and neural network reduced-order model provided in this application constructs a reduced-order model of the component using physical field data under different operating conditions determined by sampling the range of component boundary conditions. This reduced-order model is then used to optimize the corresponding zero-dimensional performance simulation model of the gas turbine. The optimized performance simulation model has the ability to quickly obtain details of the component's flow field, effectively integrating and fully utilizing a large number of high-dimensional simulation calculation results from the component design iteration process. While ensuring the efficiency of the zero-dimensional overall performance simulation of the gas turbine, the simulation results can display the flow details of the component, improving the reliability of the performance simulation and providing engineers with more information support in design decisions.

[0017] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0018] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a gas turbine overall performance simulation method based on POD and a neural network reduced-order model, provided for an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a gas turbine overall performance simulation device based on POD and a neural network reduced-order model, provided in an embodiment of this application. Detailed Implementation

[0019] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0020] The following describes, with reference to the accompanying drawings, a method and apparatus for simulating the overall performance of a gas turbine based on a POD and a neural network reduced-order model, according to embodiments of this application.

[0021] Figure 1 This is a flowchart illustrating a gas turbine overall performance simulation method based on a reduced-order model using POD and neural networks, provided as an embodiment of this application. Figure 1 As shown, the gas turbine overall performance simulation method based on POD and neural network reduced-order model includes the following steps: Step 101: Based on the range of component boundary conditions determined by the gas turbine design requirements, the Latin hypercube sampling method is used to sample within the sample space formed by the range of boundary conditions to obtain multiple sampling results.

[0022] Among them, the number of sampling results is less than the number threshold and is uniformly distributed in the sample space.

[0023] The components in a gas turbine can be a compressor, combustion chamber, turbine, etc.

[0024] In this embodiment, the boundary conditions corresponding to different components can be composed of different parameters, such as inlet pressure, inlet temperature, outlet back pressure, rotational speed, and fuel flow rate. These parameters can jointly define the operating conditions. The range of boundary condition values ​​is determined by the design requirements of the gas turbine. For example, the range of turbine inlet temperature can be from 1500K to 1800K, and the range of rotational speed can be from 0rpm to 3000rpm.

[0025] In this embodiment, the range of values ​​for the computational fluid dynamics (CFD) calculation boundary conditions of each component can be determined according to the design requirements of the gas turbine to be simulated, thus forming a sample space.

[0026] In this embodiment, each parameter combination in the sample space can represent a potential operating condition. Sampling can be performed in the sample space using any sampling method (such as Latin hypercube) to determine whether the parameter combinations corresponding to the extracted operating conditions are uniformly distributed or distributed according to probability density within the sample space, thus avoiding the omission of key operating conditions.

[0027] In this embodiment, Latin hypercube sampling (LHS) is used to perform stratified sampling from a sample space with a multivariate parameter distribution. By dividing each parameter dimension into equally probable intervals and randomly selecting sample points, uniform coverage of the sample in the multidimensional space is ensured. Furthermore, a maximum number of samples, i.e., a quantity threshold, can be set to ensure that the sample space can be uniformly covered with a smaller number of samples.

[0028] Step 102: Using multiple sampling results as boundary conditions, obtain the physical field data of the gas turbine under multiple operating conditions.

[0029] The physical field includes at least the temperature field, pressure field, and velocity field.

[0030] In some embodiments, each sampling result can be used as a boundary condition for computational fluid dynamics (CFD) simulation. The working condition corresponding to the sampling result can be determined based on the simulation result, and the physical field data under the working condition can be extracted from the simulation result.

[0031] In this embodiment, each sampling result corresponds to a specific combination of boundary conditions. By changing these parameters, the physical field changes under different operating conditions (such as full load, partial load, start-up / stop, variable speed, etc.) can be simulated. The sampling results can be used as boundary conditions to construct CFD simulation models of components one by one, perform numerical simulations, and extract the temperature field, pressure field, and velocity field from the calculation results to form a database of component temperature field, pressure field, and velocity field under different operating conditions.

[0032] It should be noted that, in the embodiments of this application, the temperature field, pressure field and velocity field obtained from CFD simulation of components under other working conditions formed during the design iteration process can be added to the corresponding database, so as to make full use of the large amount of high-dimensional field data generated by the components during the design iteration process, which is beneficial to expand the data dimension and improve the model generalization ability.

[0033] Step 103: Divide the physical field data under multiple working conditions into training datasets and test datasets according to the working conditions.

[0034] In this embodiment, the physical field data under multiple operating conditions can be divided into training and testing datasets according to a ratio (e.g., 70% training, 30% testing) using a simple random partitioning method. Alternatively, the physical field data can be layered according to the values ​​of key parameters (e.g., inlet temperature, rotational speed) under multiple operating conditions, and then the training and testing sets can be randomly partitioned proportionally within each layer, etc. This application does not limit the scope of this method.

[0035] It should be noted that when the physical field data contains multiple types of physical fields, the training dataset and test dataset can be divided separately for each type of physical field data, or the dataset can be divided according to the working conditions, and then all types of physical field data under the same working conditions can be put into the same dataset.

[0036] Step 104: Based on the training dataset, construct an initial reduced-order model using snapshot intrinsic orthogonal decomposition (POD) and neural network methods.

[0037] In this embodiment, the Snapshot Proper Orthogonal Decomposition (POD) method can be used to reduce the order of the component's temperature field, pressure field, and velocity field in the training dataset, respectively, to obtain the corresponding POD spatial modes and modal coefficients of each order. Then, a neural network method is used to construct a modal coefficient prediction model under arbitrary operating conditions. During the training process of the modal coefficient prediction model, the input data is the boundary conditions used in the CFD calculation of the component under a certain operating condition in the training database, and the output data is the modal coefficients under that operating condition. Subsequently, the obtained first N POD spatial modes and modal coefficient prediction models of each order can be used to form an initial order reduction model.

[0038] In some embodiments, the temperature field data, pressure field data, and velocity field data of each component under all operating conditions in the training dataset can be reduced in order using snapshot intrinsic orthogonal decomposition (POD) to obtain the first N spatial modes corresponding to each physical field type of the component.

[0039] Where N is a positive integer, that is, the order to be retained, and its value can be determined according to the energy percentage threshold (e.g., if the cumulative energy percentage of the first 10 modes reaches 90% of the threshold, then N is 10).

[0040] The physical field types are temperature field, pressure field, or velocity field.

[0041] In this embodiment of the application, the POD method first arranges the temperature field data (pressure field data or velocity field data) of a single component under all working conditions in the training dataset into a matrix in a certain order, and then calculates the covariance, decomposes the eigenvalues ​​and modulus stages, and retains the first N spatial modes.

[0042] Then, based on the Nth-order spatial modes and physical field data corresponding to any physical field type, the modal coefficients of each order of spatial modes under the corresponding working conditions of each physical field data can be determined.

[0043] In this embodiment, after obtaining the Nth-order spatial modes, since the Nth-order spatial modes are orthogonal, the physical field data under any working condition in the training dataset can be projected onto the Nth-order spatial modes to obtain the modal coefficients corresponding to each spatial mode under that working condition. That is, the physical field data can be reconstructed using the modal coefficients and the Nth-order spatial modes.

[0044] Then, based on all working conditions in the training dataset and the Nth-order modal coefficients for each working condition, a modal coefficient prediction model can be constructed using a neural network method.

[0045] In this embodiment of the application, a modal coefficient prediction model under any working condition can be constructed using a neural network method. The input data for training the modal coefficient prediction model is the boundary condition used in the CFD calculation of a component under a certain working condition in the training database, and the output data is the modal coefficient under that working condition calculated in the above steps.

[0046] Finally, the Nth-order spatial mode and the modal coefficient prediction model can be combined to obtain the initial reduced-order model corresponding to the component and any physical field type.

[0047] In this embodiment, different types of physical fields have different physical characteristics, distribution patterns and sensitivity to operating parameters, resulting in different mapping relationships between their POD mode spaces and mode coefficients and operating parameters. Therefore, different types of physical fields can determine different initial reduced-order models, and then the corresponding reduced-order models can be updated using physical field data of the same type.

[0048] In some embodiments, the mean values ​​for any physical field type in the training dataset can be determined first. Then, based on the mean values, the Nth-order spatial modes, and each physical field data, physical field reconstruction is performed to determine the modal coefficients of each order of spatial modes under the corresponding operating conditions of the physical field data.

[0049] In this embodiment of the application, the physical field reconstruction can be represented by the following formula (1).

[0050] (1) in, Here is the physical field matrix under a certain working condition, where x represents the position and t represents the working condition. To obtain the physical field mean values ​​for all operating conditions in the training dataset, Indicates the first Step, These are the modal coefficients for each order under a certain working condition. These are POD spatial modes (also known as bases).

[0051] Step 105: Based on the accuracy and time of the calculation of each physical field data in the test dataset by the initial order reduction model, update the initial order reduction model to obtain the target order reduction model whose accuracy and time meet the preset conditions.

[0052] In this embodiment, after obtaining the initial reduced-order model, the physical fields in the test dataset can be predicted. If the accuracy is lower than the preset accuracy value, the value of N is increased. If, as the value of N increases, the required computation time exceeds the preset computation time value, and the accuracy still does not meet the requirements, the process returns to step 101 to resample the sample space and obtain a new reduced-order model. This process continues until both the accuracy and computation time reach the preset values. The reduced-order model at this point is then determined as the target reduced-order model for optimizing the performance simulation model.

[0053] It should be noted that when the dataset contains multiple types of physical field data, the initial reduction model corresponding to each type can be used to predict the same type of physical field data and update the reduction model accordingly.

[0054] In some embodiments, an initial reduced-order model corresponding to any component and any physical field type can be used to predict each working condition in the test dataset to obtain the predicted physical field data and prediction time corresponding to that working condition.

[0055] Then, based on the predicted physics data and the corresponding reference physics data in the test dataset, the prediction accuracy of the initial price reduction model is determined.

[0056] In this embodiment, the predicted physical field data and the reference physical field data correspond to the same operating conditions, components, and physical field types. The prediction accuracy of the initial price reduction model can be obtained by calculating the error between the predicted physical field data and the reference physical field data.

[0057] Then, if the prediction accuracy is less than the accuracy threshold and the prediction time is less than the time threshold, the value of N can be increased, and the step of using POD to perform order reduction calculation to obtain the Nth order spatial mode can be returned to obtain the updated order reduction model.

[0058] The accuracy threshold and time threshold can both be set based on experiments or experience, and are used to determine the minimum value that the reduced-order model meets the requirements.

[0059] In this embodiment, if the prediction accuracy is less than the accuracy threshold, it indicates that the performance of the reduced-order model is insufficient and the prediction error is large. This may be due to insufficient model complexity. Therefore, the prediction accuracy of the physical field can be improved by increasing the number of spatial modes N, refining the spatial decomposition, reducing truncation error, and enhancing nonlinear approximation capability.

[0060] In this embodiment, the prediction time reflects the model efficiency. The longer the prediction time, the lower the efficiency. When the prediction time exceeds a certain time threshold, it may be because the sampling results of the current initial model reduction introduce redundant data or do not cover key areas, resulting in excessive computation during the model training or inference stage. In this case, the prediction time can be reduced by resampling to optimize the sampling results.

[0061] In this embodiment, when the prediction accuracy is less than the accuracy threshold and the prediction time is less than the time threshold, there is no need to optimize the sampling results. It is only necessary to increase the value of N and re-execute the calculation of the first N spatial modes corresponding to each physical field type of each component in step 104 to construct a new reduced-order model.

[0062] Then, based on the updated downgraded model, the prediction accuracy and prediction time can be recalculated, and so on, until the target downgraded model that satisfies the condition that the prediction accuracy is greater than the accuracy threshold and the prediction time is less than the time threshold is obtained.

[0063] In some embodiments, if the prediction accuracy is less than the accuracy threshold and the prediction time is greater than the time threshold, the sample space can be resampled to obtain new sampling results.

[0064] In this embodiment, when the prediction accuracy is less than the accuracy threshold and the prediction time is greater than the time threshold, simply increasing the value of N to improve the prediction accuracy may lead to an increase in the time period of the model prediction, resulting in low efficiency. Therefore, resampling can be performed in the sample space to obtain new sampling results. Then, based on the new sampling results, steps 102 to 105 are executed sequentially to obtain a new reduced-order model.

[0065] In this embodiment, the reduced-order model is updated by increasing the order of the spatial modes when the computation time does not exceed a threshold, thereby improving prediction accuracy. Alternatively, when the computation time exceeds the threshold, the sample space formed by the boundary condition values ​​of the gas turbine components is resampled to obtain new training and test datasets, and the reduced-order model is reconstructed. This ensures both the computational efficiency and the update efficiency of the reduced-order model.

[0066] Step 106: Based on the target reduced-order model, optimize the zero-dimensional performance simulation model corresponding to the gas turbine to obtain the target performance simulation model.

[0067] In this embodiment of the application, the component characteristic diagram in the zero-dimensional overall performance model of the gas turbine can be replaced by the verified target reduced-order model to complete the optimization of the zero-dimensional performance simulation model and obtain the overall performance simulation model based on the POD and neural network reduced-order model, i.e., the target performance simulation model.

[0068] Step 107: Based on the target performance simulation model, simulate the overall performance of the gas turbine and obtain the simulation results.

[0069] In this embodiment of the application, after obtaining the target performance simulation model, the overall performance of the gas turbine can be simulated using the target performance simulation model to obtain simulation results, providing engineers with more information support in design decisions.

[0070] In this embodiment, a reduced-order model of the component is constructed using POD and neural network methods based on physical field data under different operating conditions determined by sampling the range of component boundary conditions. This reduced-order model is then used to optimize the corresponding zero-dimensional performance simulation model of the gas turbine. The optimized performance simulation model has the ability to quickly obtain the details of the component flow field, effectively integrating and making full use of a large number of high-dimensional simulation calculation results performed during the component design iteration process. While ensuring the efficiency of the overall zero-dimensional performance simulation of the gas turbine, the simulation results can display the flow details of the component, improve the reliability of the performance simulation, and provide engineers with more information support in design decisions.

[0071] Figure 2 This is a schematic diagram of a gas turbine overall performance simulation device based on POD and a neural network reduced-order model, provided as an embodiment of this application.

[0072] like Figure 2 As shown, the gas turbine overall performance simulation device 20 based on POD and neural network reduced-order model includes: The sampling module 201 is used to sample the components in the sample space formed by the boundary condition value range determined by the gas turbine design requirements using the Latin hypercube sampling method to obtain multiple sampling results. The number of sampling results is less than the number threshold and is evenly distributed in the sample space. The determination module 202 is used to obtain physical field data of the gas turbine under multiple operating conditions by using multiple sampling results as boundary conditions respectively. The physical field data includes at least temperature field data, pressure field data and velocity field data. Processing module 203 is used to divide the physical field data under multiple working conditions into training dataset and test dataset according to the working conditions. The generation module 204 is used to construct an initial reduced-order model based on the training dataset using snapshot intrinsic orthogonal decomposition (POD) and neural network methods. The update module 205 is used to update the initial order reduction model based on the accuracy and time of the calculation of each physical field data in the test dataset according to the initial order reduction model, so as to obtain the target order reduction model whose accuracy and time meet the preset conditions. Optimization module 206 is used to optimize the zero-dimensional performance simulation model of the gas turbine based on the target reduced-order model, so as to obtain the target performance simulation model; Simulation module 207 is used to simulate the overall performance of the gas turbine based on the target performance simulation model and obtain simulation results.

[0073] Furthermore, in one possible implementation of this application embodiment, the determining module 202 may specifically be used for: Each sampling result is used as a boundary condition for computational fluid dynamics (CFD) simulation. The working conditions corresponding to the sampling results are determined based on the simulation results, and the physical field data under the working conditions are extracted from the simulation results.

[0074] Furthermore, in one possible implementation of this application embodiment, the generation module 204 may specifically be used for: For the temperature field data, pressure field data and velocity field data of each component under all working conditions in the training dataset, the order reduction calculation is performed by using snapshot intrinsic orthogonal decomposition (POD) to obtain the first N spatial modes corresponding to each physical field type of the component, where N is a positive integer and the physical field type is temperature field, pressure field or velocity field. Based on the Nth-order spatial modes and physical field data corresponding to any physical field type, determine the modal coefficients of each order of spatial modes under the working conditions corresponding to each physical field data. Based on all working conditions and the Nth-order modal coefficients for each working condition in the training dataset, a modal coefficient prediction model is constructed using a neural network method. By combining the Nth-order spatial mode with the modal coefficient prediction model, an initial reduced-order model corresponding to the component and any physical field type is obtained.

[0075] Furthermore, in one possible implementation of this application embodiment, the generation module 204 may specifically be used for: Determine the mean of all physical field data for any physical field type in the training dataset; Based on the mean, Nth-order spatial modes, and each physical field data, physical field reconstruction is performed to determine the modal coefficients of each order of spatial modes under the corresponding operating conditions of the physical field data.

[0076] Furthermore, in one possible implementation of this application embodiment, the update module 205 may specifically be used for: Using the initial reduced-order model corresponding to any component and any physical field type, predictions are made for each working condition in the test dataset to obtain the predicted physical field data and prediction time corresponding to the working condition. The prediction accuracy of the initial price reduction model is determined based on the predicted physical field data and the corresponding reference physical field data in the test dataset. The predicted physical field data and the reference physical field data have the same operating conditions, components and physical field types. If the prediction accuracy is less than the accuracy threshold and the prediction time is less than the time threshold, increase the value of N and return to the step of using POD to perform order reduction calculation to obtain the Nth order spatial mode, so as to obtain the updated order reduction model. Based on the updated downgraded model, the prediction accuracy and prediction time are recalculated, and so on, until the target downgraded model that satisfies the condition that the prediction accuracy is greater than the accuracy threshold and the prediction time is less than the time threshold is obtained.

[0077] Furthermore, in one possible implementation of this application embodiment, the update module 205 can also be used for: If the prediction accuracy is less than the accuracy threshold and the prediction time is greater than the time threshold, the sample space is resampled to obtain new sampling results.

[0078] It should be noted that the explanation of the above-mentioned embodiment of the gas turbine performance simulation model optimization method also applies to the gas turbine performance simulation model optimization device of this embodiment, and will not be repeated here.

[0079] In this embodiment, a reduced-order model of the component is constructed using POD and neural network methods based on physical field data under different operating conditions determined by sampling the range of component boundary conditions. This reduced-order model is then used to optimize the corresponding zero-dimensional performance simulation model of the gas turbine. The optimized performance simulation model has the ability to quickly obtain the details of the component flow field, effectively integrating and fully utilizing the large number of high-dimensional simulation calculation results performed during the component design iteration process. While ensuring the efficiency of the overall zero-dimensional performance simulation of the gas turbine, the simulation results can display the details of component flow, improve the reliability of performance simulation, and provide engineers with more information support in design decisions.

[0080] To implement the above embodiments, this application also proposes an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided in the foregoing embodiments.

[0081] To implement the above embodiments, this application also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.

[0082] To implement the above embodiments, this application also proposes a computer program product, including a computer program that, when executed by a processor, implements the methods provided in the foregoing embodiments.

[0083] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0084] It should be noted that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. Furthermore, such collection / sharing should only be conducted after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes authorization of relevant user information before the user uses the function. In addition, any necessary steps must be taken to protect and safeguard access to such personal information data and ensure that others with access to personal information data comply with their privacy policies and procedures.

[0085] This application is intended to provide an implementation scheme for users to selectively prevent the use or access to their personal information data. Specifically, this disclosure is intended to provide hardware and / or software to prevent or block access to such personal information data. Once personal information data is no longer needed, risks can be minimized by restricting data collection and deleting data. Furthermore, where applicable, such personal information is de-identified to protect user privacy.

[0086] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0087] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0088] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0089] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0090] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0091] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0092] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0093] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for simulating the overall performance of a gas turbine based on a reduced-order model using POD and neural networks, characterized in that, Includes the following steps: Based on the component boundary condition value range determined by the gas turbine design requirements, the Latin hypercube sampling method is used to sample in the sample space formed by the boundary condition value range to obtain multiple sampling results, wherein the number of the sampling results is less than the number threshold and is uniformly distributed in the sample space. Using the multiple sampling results as boundary conditions, physical field data of the gas turbine under multiple operating conditions are obtained, wherein the physical field data includes at least temperature field data, pressure field data, and velocity field data; The physical field data under the multiple working conditions are divided into training datasets and test datasets according to the working conditions. Based on the training dataset, an initial reduced-order model is constructed using snapshot intrinsic orthogonal decomposition (POD) and neural network methods. Based on the accuracy and time of the initial order reduction model in calculating each physical field data in the test dataset, the initial order reduction model is updated to obtain a target order reduction model whose accuracy and time meet the preset conditions. Based on the target reduced-order model, the zero-dimensional performance simulation model corresponding to the gas turbine is optimized to obtain the target performance simulation model; Based on the target performance simulation model, the overall performance of the gas turbine is simulated, and the simulation results are obtained.

2. The method according to claim 1, characterized in that, The step of using the multiple sampling results as boundary conditions to obtain the physical field data of the gas turbine under multiple operating conditions includes: Each sampling result is used as a boundary condition for computational fluid dynamics (CFD) simulation. The working condition corresponding to the sampling result is determined based on the simulation result, and the physical field data under the working condition is extracted from the simulation result.

3. The method according to claim 1, characterized in that, The initial order reduction model is constructed based on the training dataset using snapshot intrinsic orthogonal decomposition (POD) and neural network methods, including: For the temperature field data, pressure field data and velocity field data of each component under all working conditions in the training dataset, the order reduction calculation is performed by using snapshot intrinsic orthogonal decomposition (POD) to obtain the first N spatial modes corresponding to each physical field type of the component, where N is a positive integer and the physical field type is temperature field, pressure field or velocity field. Based on the Nth-order spatial modes and physical field data corresponding to any physical field type, determine the modal coefficients of each order of spatial modes under the working conditions corresponding to each physical field data. Based on all working conditions and the Nth-order modal coefficients for each working condition in the training dataset, a modal coefficient prediction model is constructed using a neural network method. By combining the Nth-order spatial mode with the modal coefficient prediction model, an initial reduced-order model corresponding to the component and any physical field type is obtained.

4. The method according to claim 3, characterized in that, The determination of the modal coefficients of each spatial mode under each physical field data condition, based on the Nth-order spatial mode and physical field data corresponding to any physical field type, includes: Determine the mean of all physical field data corresponding to any physical field type in the training dataset; Based on the mean, the Nth-order spatial mode, and each physical field data, physical field reconstruction is performed to determine the modal coefficients of each order spatial mode under the corresponding operating conditions of the physical field data.

5. The method according to claim 3, characterized in that, The step of updating the initial order reduction model based on the accuracy and time calculated for each physics field data in the test dataset using the initial order reduction model, to obtain a target order reduction model whose accuracy and time meet preset conditions, includes: Using an initial reduced-order model corresponding to any component and any physical field type, prediction is made for each working condition in the test dataset to obtain the predicted physical field data and prediction time corresponding to the working condition. The prediction accuracy of the initial price reduction model is determined based on the predicted physical field data and the corresponding reference physical field data in the test dataset, wherein the predicted physical field data and the reference physical field data correspond to the same working conditions, components and physical field types; If the prediction accuracy is less than the accuracy threshold and the prediction time is less than the time threshold, increase the value of N, and return to the step of using POD to perform order reduction calculation to obtain the Nth order spatial mode, so as to obtain the updated order reduction model. Based on the updated downgraded model, the prediction accuracy and prediction time are recalculated, and so on, until a target downgraded model is obtained that satisfies the condition that the prediction accuracy is greater than the accuracy threshold and the prediction time is less than the time threshold.

6. The method according to claim 5, characterized in that, Also includes: If the prediction accuracy is less than the accuracy threshold and the prediction time is greater than the time threshold, the sample space is resampled to obtain new sampling results.

7. A gas turbine overall performance simulation device based on POD and a reduced-order neural network model, characterized in that, include: The sampling module is used to sample multiple sampling results in the sample space formed by the boundary condition value range determined by the gas turbine design requirements, using the Latin hypercube sampling method. The number of the sampling results is less than a number threshold and is uniformly distributed in the sample space. The determination module is used to obtain physical field data of the gas turbine under multiple operating conditions by using the multiple sampling results as boundary conditions respectively, wherein the physical field data includes at least temperature field data, pressure field data and velocity field data; The processing module is used to divide the physical field data under the multiple working conditions into training datasets and test datasets according to the working conditions. The generation module is used to construct an initial reduced-order model based on the training dataset using snapshot intrinsic orthogonal decomposition (POD) and neural network methods. The update module is used to update the initial order reduction model based on the accuracy and time calculated for each physical field data in the test dataset by the initial order reduction model, so as to obtain a target order reduction model whose accuracy and time meet the preset conditions. The optimization module is used to optimize the zero-dimensional performance simulation model of the gas turbine based on the target reduced-order model, so as to obtain the target performance simulation model. The simulation module is used to simulate the overall performance of the gas turbine based on the target performance simulation model and obtain simulation results.

8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1-7.

Images