Gas turbine control simulation test method, test system, equipment and storage medium

Through deep learning algorithms and FPGA simulation environment, the problem of insufficient reliance on physical tests and software simulation in the design of gas turbine control system is solved, and efficient combustion model step reduction and system-level simulation verification is achieved, which improves the design and optimization capabilities of gas turbine control system. .

CN118938709APending Publication Date: 2024-11-12ZHEJIANG ZHENENG TECHN RES INST CO LTD
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202411142293.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The design of traditional gas turbine control system relies on physical tests to be time-consuming and labor-intensive, cannot simulate extreme working conditions, and software simulation is difficult to verify the reliability and matching of hardware components, and the optimization effect is not ideal.

Method used

Deep learning algorithm is used to establish a downgrade network of combustion models, combine the whole machine thermodynamic model to build a low-order theoretical model, and build a prototype simulation environment through FPGA, and conduct semi-physical simulation experiments to verify the matching of each component of the control system and optimize parameters.

Benefits of technology

High-precision combustion process simulation is realized, computing efficiency is improved, the hardware matching of the control system can be more realistically verified, and the impact of optimized design parameters on performance is systematically analyzed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure HDA0005001206430000011
    Figure HDA0005001206430000011
  • Figure HDA0005001206430000021
    Figure HDA0005001206430000021
  • Figure HDA0005001206430000031
    Figure HDA0005001206430000031
Patent Text Reader

Abstract

The invention discloses a gas turbine control simulation test method, test system, equipment and storage medium, and relates to the technical field of gas turbine control systems.The method comprises the steps that a hydrogen-mixed gas turbine theoretical model is established; the theoretical model of the hydrogen-mixed gas turbine is digitalized into a computer executable model, and operation of a gas turbine control system is simulated on a computer; based on a field programmable gate array (FPGA), a prototype simulation environment equivalent to actual control hardware is built to verify the matching performance of joint work of all parts of the gas turbine control system; connecting part of actual physical components to a prototype simulation environment, performing a semi-physical simulation test, and reproducing a real control flow; and analyzing and identifying key parameters influencing the control performance according to a simulation result, and correcting and optimizing the key parameters. According to the method, a combustion model order reduction network based on deep learning is established, multi-scale features of a complex combustion process can be captured with high precision, and lossless order reduction from a high-precision large-scale CFD model to a high-efficiency low-order agent model is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of gas turbine control systems, in particular to a gas turbine control simulation test method, a test system, equipment and a storage medium. Background Art

[0002] Traditional gas turbine control system design usually relies on a large number of physical tests, which is not only time-consuming and labor-intensive, but also poses certain safety risks due to the inability to simulate various extreme working conditions. On the other hand, it is also difficult to truly verify the reliability of the control system by relying solely on software simulation, especially the matching and collaborative performance between various hardware components. In addition, most of the existing control system optimization methods are based on experience or simple mathematical models, and it is difficult to systematically analyze the root cause parameters that affect key performance indicators, resulting in unsatisfactory optimization results. Summary of the invention

[0003] In view of the drawbacks of existing gas turbine control system design methods that they rely on a large number of physical tests, cannot simulate various extreme working conditions and have potential safety hazards, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is how to integrate and apply multiple technical means to avoid the drawbacks of relying on a large number of physical tests, truly verify the matching and reliability of various components of the control system, and systematically analyze and optimize the main parameters affecting key performance.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] The present invention provides a gas turbine control simulation test method, which comprises the following steps:

[0007] S1. Obtain the design parameters of the gas turbine and the characteristics of the mixed hydrogen combustion, and establish a theoretical model of the mixed hydrogen gas turbine;

[0008] S2. Digitize the theoretical model of the hydrogen-mixed gas turbine into a computer executable model, and simulate the operation of the gas turbine control system on the computer;

[0009] S3. Based on the field programmable gate array FPGA, build a prototype simulation environment equivalent to the actual control hardware to verify the compatibility of the various components of the gas turbine control system;

[0010] S4, connect the actual physical components to the prototype simulation environment, conduct semi-physical simulation tests, and reproduce the real control process;

[0011] S5. Analyze and identify parameters that affect control performance based on the simulation results, and modify and optimize the parameters;

[0012] S6. Connect the final control hardware prototype to the semi-physical simulation environment to carry out hardware-in-the-loop simulation;

[0013] S7. Feedback the problems and optimization requirements found in the simulation to the gas turbine control logic design link of steps S1-S6 to improve the gas turbine control logic framework.

[0014] The gas turbine control simulation test method according to claim 1 is characterized in that: the specific steps of establishing the hydrogen-mixed gas turbine theoretical model are as follows:

[0015] 1) Design and execute mixed hydrogen combustion experiments to measure flow field data, concentration distribution data, and temperature field data under different working conditions inside the combustion chamber;

[0016] 2) Based on the experimental measurement data, a three-dimensional numerical simulation model of the combustion chamber was established using CFD software to simulate the mixed hydrogen combustion process;

[0017] 3) A combustion model reduction network is trained using a deep learning algorithm to compress the three-dimensional numerical simulation model of the combustion chamber into a low-order proxy combustion model for order reduction calculation;

[0018] 4) Based on the low-order proxy combustion model after compression, a low-order theoretical model of the gas turbine system is constructed by introducing the thermodynamic and fluid dynamics control equations at the whole machine level;

[0019] 5) Based on the low-order theoretical model of the whole machine, analyze the influence of design parameters on performance indicators through numerical calculation to determine the optimal design parameter combination;

[0020] 6) Use other independent experimental data to verify the low-order theoretical model of the whole machine, calculate the prediction deviation, and perform secondary correction and optimization on the deviation part;

[0021] 7) After multiple rounds of data correction and optimization iterations, a high-precision theoretical model of the hydrogen-mixed gas turbine is obtained to simulate the thermal performance under various operating conditions.

[0022] 1. The gas turbine control simulation test method according to claim 2, characterized in that: the specific steps of compressing the three-dimensional numerical simulation model of the combustion chamber into a low-order proxy combustion model are as follows:

[0023] 1) Under different working conditions, a large number of sample data sets including model input and output are generated using the three-dimensional numerical simulation model of the combustion chamber, and the generated data sets are preprocessed in a standardized manner;

[0024] 2) Use deep learning algorithm to train a combustion model reduction network CMRN;

[0025] 3) The dataset is divided into training set, validation set and test set in proportion, the total loss function TLoss is defined, and the adaptive moment estimation optimization algorithm is used to train CMRN end-to-end;

[0026] 4) Deploy the trained CMRN network as an online inference engine to generate low-order proxy combustion model outputs in real time;

[0027] 5) Evaluate the generalization performance of CMRN on the test set and calculate the root mean square error of each prediction head;

[0028] 6) If the root mean square error RMSE of any prediction head of CMRN exceeds the preset threshold, the control logic is triggered to perform adaptive adjustment on CMRN;

[0029] 7) After the adjustment, redeploy the CMRN online service to form a new round of online reasoning;

[0030] The specific formula of the total loss function TLoss is as follows:

[0031] TLoss=α·RLoss+β·CLoss

[0032] RLoss=Σ(yy hat ) 2

[0033] CLoss=λ·θ1

[0034] Among them, RLoss represents the mean square error reconstruction loss between the predicted output and the true output, CLoss represents the compression loss, α and β represent the weight coefficients of the reconstruction loss RLoss and the compression loss CLoss in the total loss function, y represents the true combustion model output, y hat represents the combustion model output predicted by the CMRN network, θ represents the set of all trainable parameters in the CMRN network, θ1 represents the L1 norm of θ, that is, the sum of the absolute values ​​of each parameter, and λ represents the weight coefficient of CLoss.

[0035] 2. The gas turbine control simulation test method according to claim 3, characterized in that: the specific steps of the trigger control logic to adaptively fine-tune the CMRN are as follows:

[0036] 1) Analyze the sources of errors,

[0037] If the error under the corresponding working condition exceeds the standard, increase the training data under the corresponding working condition to improve the data distribution coverage;

[0038] If the error comes from temperature field prediction, increase the loss function weight of the temperature field prediction head;

[0039] If the error comes from high pressure / low pressure conditions, increase the data sampling rate of the corresponding conditions;

[0040] 2) Adjust the compression constraints,

[0041] If the root mean square error RMSE>1.5ε, relax the compression constraint, reduce λ, and increase the model capacity;

[0042] If the root mean square error RMSE>ε and RMSE≤1.5ε, then strengthen the compression constraint, increase λ, and reduce the model complexity;

[0043] 3) If the training is terminated and RMSE>ε, the model is refined based on the current parameters of CMRN through a rule-based pruning strategy, and the model is fine-tuned on the adjusted training set based on the refined model;

[0044] If the training is terminated and RMSE ≤ ε, the loss function weights α, β, and λ are adjusted based on the current model through the Monte Carlo search method, and the adjusted optimal model parameters and loss function parameters are saved;

[0045] 4) Evaluate the generalization performance of the adjusted CMRN. If RMSE > ε, repeat the adaptive fine-tuning process.

[0046] 3. The gas turbine control simulation test method according to claim 4, characterized in that: the specific steps of digitizing the hydrogen-mixed gas turbine theoretical model into a computer executable model are as follows:

[0047] 1) Select programming language and development environment, as well as corresponding numerical computing libraries and tools;

[0048] 2) Various equations, algorithms and data structures in the theoretical model of the hybrid hydrogen gas turbine are coded and implemented using the selected programming language to form a theoretical model computer program;

[0049] 3) Perform unit testing and integration testing on the coded computer programs to ensure the correctness and reliability of the code;

[0050] 4) Design and implement digital models of gas turbine control systems, including various controllers, actuators and sensor models;

[0051] 5) Integrate the theoretical model of the hydrogen-mixed gas turbine and the digital model of the gas turbine control system to construct a comprehensive digital model of the gas turbine;

[0052] 6) Design and implement the user interface to input various operating parameters, design parameters and control parameters;

[0053] 7) Run the gas turbine integrated digital model on the computer to simulate the dynamic process under various working conditions and output performance indicators and state quantity curves;

[0054] 8) Analyze and visualize simulation results to assist in design optimization;

[0055] 9) Adjust and improve the comprehensive digital model of gas turbine based on simulation results and actual data.

[0056] 4. The gas turbine control simulation test method according to claim 1, characterized in that: the specific steps of building a prototype simulation environment equivalent to the actual control hardware are as follows:

[0057] 1) Select FPGA development board and supporting design tools;

[0058] 2) According to the structure and function of the actual control hardware, use the hardware description language to perform logical design, simulate the digital circuits of each component, and synthesize the designed digital circuits into a hardware mapping file;

[0059] 3) Instantiate the logic circuits of each component in parallel or sequentially on the FPGA development board to build an FPGA prototype system;

[0060] 4) Write test platform programs to simulate the communication protocols and timing signals between components of the FPGA prototype system;

[0061] 5) Connect the gas turbine integrated digital model with the FPGA prototype system as the environmental input to build the FPGA prototype simulation environment;

[0062] 6) In the FPGA prototype simulation environment, simulate the interactive response of the FPGA prototype system under different working conditions;

[0063] 7) Collect and analyze the running data of the FPGA prototype system in the simulation environment to verify the normal coordination and matching between the components;

[0064] 8) Based on the verification results, feedback is provided to optimize the FPGA prototype design or the gas turbine integrated digital model design;

[0065] 9) If there is a mismatch between the FPGA prototype system and the integrated digital model of the gas turbine, feedback is provided to optimize the FPGA prototype system design;

[0066] If the deviation between the gas turbine integrated digital model and the actual situation exceeds a first preset threshold, the integrated digital model design of the gas turbine is optimized.

[0067] 5. The gas turbine control simulation test method according to claim 1, characterized in that: the specific steps of analyzing and identifying key parameters affecting control performance according to simulation results are as follows:

[0068] 1) Extracting original simulation data from the semi-physical simulation environment and calculating key performance indicators based on the original simulation data;

[0069] 2) For relevant performance indicators, determine their theoretical distribution range and confidence interval through mathematical and statistical analysis methods, correspond the actual simulation results to the theoretical distribution, and calculate their probability scores as scoring indicators;

[0070] 3) Setting the grading threshold of the probability score and grading the key performance indicators according to the probability score;

[0071] 4) For performance indicators with unqualified scores, trace back and analyze the root cause parameters or parameter combinations that lead to poor performance of this indicator;

[0072] 5) Construct a parameter optimization mathematical model for the identified parameters that affect performance;

[0073] 6) Use particle swarm and sequential quadratic programming algorithms to solve the parameter optimization mathematical model and obtain the optimal parameter combination;

[0074] 7) Update the optimal parameter combination to the semi-physical simulation environment, re-conduct the semi-physical simulation experiment, and obtain the optimized key indicator performance data;

[0075] 8) Based on the optimized semi-physical simulation results, re-evaluate the performance level of each key indicator.

[0076] 6. A gas turbine control simulation test system, based on the gas turbine control simulation test method according to any one of claims 1 to 7, characterized in that it also includes:

[0077] Model building module, used to obtain gas turbine design parameters and hydrogen-mixed combustion characteristics, and establish a hydrogen-mixed gas turbine theoretical model;

[0078] A digital module is used to digitize the theoretical model of the hydrogen-mixed gas turbine into a computer executable model and simulate the operation of the gas turbine control system on a computer;

[0079] The prototype simulation module is used to build a prototype simulation environment equivalent to the actual control hardware based on the field programmable gate array FPGA to verify the compatibility of the various components of the gas turbine control system;

[0080] Semi-physical simulation module, used to connect actual physical components to the prototype simulation environment, conduct semi-physical simulation experiments, and reproduce the real control process;

[0081] A parameter correction module is used to analyze and identify parameters that affect control performance based on simulation results, and to correct and optimize the parameters;

[0082] The hardware-in-the-loop simulation module is used to connect the final control hardware prototype with the semi-physical simulation environment to carry out hardware-in-the-loop simulation;

[0083] The feedback improvement module is used to feed back the problems and optimization requirements found in the simulation to the control logic design link to improve the control logic framework.

[0084] 7. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein the processor implements the steps of the gas turbine control simulation test method according to any one of claims 1 to 7 when executing the computer program.

[0085] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the gas turbine control simulation test method according to any one of claims 1 to 7 are implemented.

[0086] The beneficial effects of the present invention are as follows: the present invention establishes a combustion model reduction network based on deep learning, which can capture the multi-scale characteristics of complex combustion processes with high precision, and realize lossless reduction from high-precision large-scale CFD models to efficient low-order proxy models, providing key support for establishing a high-fidelity gas turbine whole machine model. The reduced-order combustion model is coupled with the whole machine thermodynamic model to construct a system-level low-order theoretical model of the hybrid hydrogen gas turbine. Compared with simple CFD simulation, the model has higher computational efficiency and can better analyze the impact of design parameters on system performance. A prototype system equivalent to the actual control hardware is built based on FPGA, and a simulation environment is constructed by connecting with the digital model. Compared with software simulation, it can more realistically verify the hardware matching between the various components of the control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0088] Figure 1 The framework flow chart of the gas turbine control simulation test method.

[0089] Figure 2 A simulation test method for gas turbine control.

[0090] Figure 3 Computer equipment diagram for gas turbine control simulation test method. DETAILED DESCRIPTION

[0091] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0092] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0093] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0094] Example 1

[0095] Reference Figure 1 to Figure 3 , which is the first embodiment of the present invention, provides a gas turbine control simulation test method, comprising:

[0096] S1: Obtain the gas turbine design parameters and hydrogen-mixed combustion characteristics, and establish a theoretical model of hydrogen-mixed gas turbine.

[0097] Specifically, the following steps are included:

[0098] S1.1: Design and execute mixed hydrogen combustion experiments to measure flow field data, concentration distribution data, and temperature field data under different working conditions inside the combustion chamber.

[0099] S1.2: Based on the experimental measurement data, a three-dimensional numerical simulation model of the combustion chamber is established using CFD software to simulate the mixed hydrogen combustion process.

[0100] S1.3: A combustion model reduction network is trained using a deep learning algorithm to compress the high-precision but computationally intensive three-dimensional numerical simulation model of the combustion chamber into a low-order proxy combustion model for order reduction calculations.

[0101] Specifically, the following steps are included:

[0102] S1.3.1: Generate a large number of sample data sets including model input and output using a three-dimensional numerical simulation model of the combustion chamber under different operating conditions, and perform standardized preprocessing on the generated data sets.

[0103] Specifically, the model input includes intake temperature, pressure, flow rate, fuel composition, etc., and the output includes temperature field, concentration field, flow field, etc.

[0104] S1.3.2: Use deep learning algorithm to train a combustion model reduction network CMRN.

[0105] Specifically, CMRN adopts self-attention mechanism and gated recurrent unit to capture the long-term dependency between model input and output; CMRN introduces wavelet transform encoder-decoder to learn multi-scale feature representation of model input and output; CMRN designs lossy compression module to compress intermediate feature representation through sparse constraints to reduce model complexity; CMRN's output layer contains several prediction heads, which predict different physical quantities such as temperature, concentration, flow field, etc.

[0106] S1.3.3: Split the dataset into training set, validation set and test set in proportion, define the loss function, and use the adaptive moment estimation optimization algorithm to train CMRN end-to-end.

[0107] Specifically, the end-to-end training of CMRN using an adaptive moment estimation optimization algorithm includes the following steps: randomly initializing all trainable parameters in the CMRN network, and inputting training data into the CMRN network in batches; for each batch of training data, calculating the reconstruction loss and the compression loss through forward propagation, and weighting the two to obtain the total loss function value; calculating the gradient of the total loss function value with respect to all trainable parameters; according to the Adam optimization algorithm rules, using the calculated gradient to update the value of each trainable parameter in the CMRN network; repeating the processing of each batch of data until the training data set has been seen in its entirety, or the maximum number of training rounds is reached, and outputting the CMRN model parameters that minimize the total loss function value on this training data.

[0108] Specifically, the specific formula of the loss function TLoss is as follows:

[0109] TLoss=α·RLoss+β·CLoss

[0110] RLoss=∑(yy hat ) 2

[0111] CLoss=λ·θ1

[0112] Among them, RLoss represents the mean square error reconstruction loss between the predicted output and the true output, CLoss represents the compression loss, α and β represent the weight coefficients of the reconstruction loss RLoss and the compression loss CLoss in the total loss function, y represents the true combustion model output, y hat represents the combustion model output predicted by the CMRN network, θ represents the set of all trainable parameters in the CMRN network, |θ1 represents the L1 norm of θ, that is, the sum of the absolute values ​​of each parameter, and λ represents the weight coefficient of CLoss.

[0113] S1.3.4: Deploy the trained CMRN network as an online inference engine to generate low-order proxy combustion model outputs in real time.

[0114] Specifically, the CMRN model is deployed as an online service to receive new model input data; the input data is passed to the encoder module of CMRN, and the encoder compresses the high-dimensional input data into a low-dimensional feature representation; the compressed low-dimensional feature representation is passed to the decoder module of CMRN, and the decoder generates a low-order proxy output of the combustion model; according to actual needs, the prediction results of different physical fields such as temperature field and concentration field are selectively obtained from the multiple output heads of the decoder.

[0115] S1.3.5: Evaluate the generalization performance of CMRN on the test set and calculate the root mean square error of each prediction head.

[0116] S1.3.6: If the root mean square error RMSE of any prediction head of CMRN exceeds the preset threshold ε, the control logic is triggered to perform adaptive fine-tuning on CMRN.

[0117] Specifically, analyze the source of the error. If the error is significantly exceeded under some working conditions, increase the training data under these working conditions to improve the data distribution coverage; if the error mainly comes from the temperature field prediction, increase the loss function weight of the temperature field prediction head; if the error mainly comes from specific working conditions such as high pressure / low pressure, increase the data sampling rate of the corresponding working conditions. Adjust the compression constraint. If the root mean square error RMSE>1.5ε, relax the compression constraint, reduce λ, and increase the model capacity; if the root mean square error RMSE>ε and RMSE≤1.5ε, strengthen the compression constraint, increase λ, and reduce the model complexity. If the training is terminated and RMSE>ε, based on the current parameters of CMRN, the model is refined through a rule-based pruning strategy to remove parameters with low impact on the error and retain parameters with large impact on the key prediction output. Based on the refined model, the model is fine-tuned on the fine-tuned training set. If the training is terminated and RMSE≤ε, the loss function weights α, β, and λ are fine-tuned by the Monte Carlo search method based on the current model, and the adjusted optimal model parameters and loss function parameters are saved. Evaluate the generalization performance of the fine-tuned CMRN. If RMSE > ε, repeat the above adaptive fine-tuning process.

[0118] S1.3.7: After fine-tuning, redeploy the CMRN online service to form a new round of online reasoning.

[0119] S1.4: Based on the low-order proxy combustion model after compression, a low-order theoretical model of the gas turbine system is constructed by introducing the thermodynamic and fluid dynamics control equations at the whole machine level.

[0120] S1.5: Based on the low-order theoretical model of the whole machine, analyze the impact of design parameters on performance indicators through numerical calculation to determine the optimal design parameter combination.

[0121] S1.6: Use other independent experimental data to verify the low-order theoretical model of the whole machine, calculate the prediction deviation, and perform secondary correction and optimization on the part with deviation.

[0122] S1.7: After multiple rounds of data correction and optimization iterations, a high-precision theoretical model of the hydrogen-mixed gas turbine is obtained to simulate the thermal performance under various operating conditions.

[0123] S2: Digitize the theoretical model of the hydrogen-mixed gas turbine into a computer executable model and simulate the operation of the gas turbine control system on the computer.

[0124] Specifically, select a suitable programming language and development environment, as well as corresponding numerical calculation libraries and tools; use the selected programming language to encode and implement various equations, algorithms and data structures in the theoretical model of the hybrid gas turbine to form a theoretical model computer program; perform unit testing and integration testing on the coded computer program to ensure the correctness and reliability of the code; design and implement a digital model of the gas turbine control system, including various controllers, actuators and sensor models; integrate the theoretical model of the hybrid gas turbine and the digital model of the gas turbine control system to construct a comprehensive digital model of the gas turbine; design and implement a user interface to input various operating parameters, design parameters and control parameters; run the comprehensive digital model of the gas turbine on a computer to simulate the dynamic process under various operating conditions and output performance indicators and state quantity curves; analyze and visualize the simulation results to assist in design optimization; adjust and improve the comprehensive digital model of the gas turbine based on the simulation results and actual data.

[0125] S3: Based on the field programmable gate array FPGA, build a prototype simulation environment equivalent to the actual control hardware to verify the compatibility of the various components of the gas turbine control system.

[0126] Specifically, select a suitable FPGA development board and supporting design tools; use hardware description language to perform logic design on the actual control hardware according to its structure and function, simulate the digital circuit of each component, and synthesize the designed digital circuit into a hardware mapping file that can be implemented on the FPGA; instantiate the logic circuit of each component in parallel or sequentially on the FPGA development board to build an FPGA prototype system; write a test platform program to simulate the communication protocol and timing signals between the components of the FPGA prototype system; connect the comprehensive digital model of the gas turbine obtained by S2 with the FPGA prototype system as the environment input to build an FPGA prototype simulation environment; in the FPGA prototype simulation environment, simulate the interactive response of the FPGA prototype system under different working conditions; collect and analyze the operating data of the FPGA prototype system in the simulation environment to verify the normal coordination and matching degree between the components; according to the verification results, feedback optimize the FPGA prototype design or the gas turbine comprehensive digital model design; if there is a mismatch between the FPGA prototype system and the gas turbine comprehensive digital model, feedback optimize the FPGA prototype system design; if the deviation between the gas turbine comprehensive digital model and the actual situation exceeds the first preset threshold, feedback optimize the gas turbine comprehensive digital model design.

[0127] It should be noted that the first preset threshold refers to a critical value of allowable deviation.

[0128] S4: Connect some actual physical components to the prototype simulation environment, conduct semi-physical simulation experiments, and reproduce the real control process.

[0129] It should be noted that when selecting actual physical components, the following points should be considered: priority should be given to key components that are critical to the operation of the system and have complex functions to verify their compatibility and collaborative working performance with other components; priority should be given to high-risk components with high failure rates and harsh working environments to verify the reliable operation performance of these components under extreme working conditions; priority should be given to components that may have mutual influences such as electromagnetic compatibility and thermal coupling in the actual state to verify the impact of these coupling effects on the system; priority should be given to actual components that are mature and ready to use to verify their compatibility with the FPGA prototype system; on the basis of the above principles, the cost and availability of the corresponding physical components must also be considered. For components that are costly or difficult to obtain in the short term, digital models can be used temporarily as substitutes.

[0130] S5: Analyze and identify key parameters that affect control performance based on the simulation results, and modify and optimize the key parameters.

[0131] Specifically, the following steps are included:

[0132] S5.1: Extract the original simulation data from the semi-physical simulation environment and calculate the key performance indicators based on the original simulation data.

[0133] Specifically, the original simulation data includes controller input, control quantity, state quantity of the controlled object, etc., and the key performance indicators include adjustment time, overshoot, energy consumption, fault recovery time, etc.

[0134] S5.2: For each key performance indicator, determine its theoretical distribution range and confidence interval through mathematical and statistical analysis methods, correspond the actual simulation results to the theoretical distribution, and calculate its probability score as the scoring indicator.

[0135] S5.3: Set the probability score classification threshold and classify the key performance indicators according to the probability score.

[0136] Specifically, [0.9, 1] is excellent, [0.7, 0.9) is good, [0.5, 0.7) is fair, [0.3, 0.5) is poor, and [0, 0.3) is unqualified.

[0137] S5.4: For performance indicators that are scored as unqualified or poor, trace back and analyze the root cause parameters or parameter combinations that led to the poor performance of this indicator.

[0138] S5.5: Construct a parameter optimization mathematical model for the identified parameters that affect key performance.

[0139] S5.6: Use particle swarm and sequential quadratic programming algorithms to solve the parameter optimization mathematical model and obtain the optimal parameter combination.

[0140] S5.7: Update the optimal parameter combination to the semi-physical simulation environment, re-conduct the semi-physical simulation experiment, and obtain the optimized key indicator performance data.

[0141] S5.8: Re-evaluate the performance level of each key indicator based on the optimized semi-physical simulation results.

[0142] Specifically, if all key performance indicator levels reach good or above, the current optimal parameter combination is saved as the output of this optimization; otherwise, determine whether to allow continued iterative optimization. If allowed, re-analyze the parameters that affect the poor performance based on the new simulation results, build a new parameter optimization mathematical model, solve and obtain a new optimal parameter combination, update it to a semi-physical simulation environment for verification simulation, and repeat the evaluation of the key performance indicator levels; if further iteration is not allowed, terminate the optimization and use the current optimal parameter combination as the final output result.

[0143] S6: Connect the final control hardware prototype to the semi-physical simulation environment to carry out hardware-in-the-loop simulation.

[0144] S7: Feedback the problems and optimization requirements found in the simulation to the control logic design phase to improve the control logic framework.

[0145] Specifically, the present embodiment further provides a gas turbine control simulation test system, including a model building module, which is used to obtain the design parameters of the gas turbine and the mixed hydrogen combustion characteristics, and establish a theoretical model of the mixed hydrogen gas turbine; a digitization module, which is used to digitize the theoretical model of the mixed hydrogen gas turbine into a computer executable model, and simulate the operation of the gas turbine control system on the computer; a prototype simulation module, which is used to build a prototype simulation environment equivalent to the actual control hardware based on the field programmable gate array FPGA, so as to verify the compatibility of the various components of the gas turbine control system working together; a semi-physical simulation module, which is used to connect some actual physical components to the prototype simulation environment, conduct semi-physical simulation experiments, and reproduce the real control process; a parameter correction module, which is used to analyze and identify key parameters that affect the control performance based on the simulation results, and correct and optimize the key parameters; an in-loop simulation module, which is used to connect the final control hardware prototype with the semi-physical simulation environment to carry out hardware-in-the-loop simulation; a feedback improvement module, which is used to feed back the problems and optimization requirements found in the simulation to the control logic design link to improve the control logic framework.

[0146] This embodiment also provides a computer device suitable for the case of a gas turbine control simulation test method, comprising a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the gas turbine control simulation test method proposed in the above embodiment.

[0147] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0148] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the gas turbine control simulation test method proposed in the above embodiment is implemented.

[0149] In summary, the present invention has established a combustion model reduction network based on deep learning, which can capture the multi-scale characteristics of complex combustion processes with high precision, and realize lossless reduction from high-precision large-scale CFD models to efficient low-order proxy models, providing key support for the establishment of a high-fidelity gas turbine whole machine model. The reduced-order combustion model is coupled with the whole machine thermodynamic model to construct a system-level low-order theoretical model of the hybrid hydrogen gas turbine. Compared with simple CFD simulation, the model has higher computational efficiency and can better analyze the impact of design parameters on system performance. A prototype system equivalent to the actual control hardware was built based on FPGA, and a simulation environment was built by connecting with the digital model. Compared with software simulation, it can more realistically verify the hardware matching between the various components of the control system.

[0150] Example 2

[0151] Reference Figure 1 to Figure 3 , which is the second embodiment of the present invention, provides a gas turbine control simulation test method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0152] Specifically, this embodiment aims to simulate and analyze the thermal performance and optimize the control system of a new type of hydrogen-mixed gas turbine. The experimental assumptions are as follows: the hydrogen-mixed fuel composition is 70% methane + 30% hydrogen (volume fraction); the combustion chamber inlet temperature is 550K and the pressure is 1.5MPa; the design point speed is 10000rpm, and the generator output power is 20MW.

[0153] Firstly, a mixed hydrogen combustion experiment was designed and executed to obtain temperature field, concentration field, flow field and other data under the above-mentioned working conditions. Based on the experimental data, a three-dimensional CFD model of the combustion chamber was established, and the CMRN network was trained to compress and reduce its order to obtain a low-order proxy combustion model with a computational efficiency improved by 2 orders of magnitude and a temperature field prediction error less than 5%. On this basis, a low-order theoretical model of the whole mixed hydrogen gas turbine was constructed, and data correction and optimization iteration were carried out.

[0154] Furthermore, after encoding the low-order theoretical model into a comprehensive digital model, the startup process of the gas turbine under given working conditions was simulated. The simulation results show that the gas turbine takes 6 minutes to start and reach a stable state, there is an overshoot of 13%, and the adjustment time is 120 seconds. Key indicators were extracted and scored from the simulation data, with the adjustment time score of 0.63 (average level), the overshoot score of 0.32 (poor level), and the startup time score of 0.87 (good level).

[0155] The analysis showed that the overshoot was mainly caused by improper PID parameters of the fuel controller, and the long adjustment time was due to the delayed response of the intake system. Therefore, an optimization model was constructed, and the intake valve opening curve was corrected after three iterations of optimization. After optimization, the start-up time was shortened to 4.5 minutes without overshoot, the adjustment time was shortened to 72 seconds, and all key indicators reached a good level.

[0156] The comparison indexes of the present invention and the traditional method are shown in Table 1.

[0157] Table 1 Comparison indexes of the present invention and the traditional method

[0158] index Traditional methods Method of the present invention Combustion model accuracy Low high Computational efficiency Large amount of calculation CMRN network is reduced in order, and the efficiency is improved by 2 orders of magnitude Model Interpretability Explainable Unexplainable Design space coverage Only covers a few typical working conditions Covering a large number of data points across the entire operating range Verify the environment Simple digital simulation High-fidelity semi-physical simulation environment Optimizing iteration efficiency Low (needs a lot of experimental feedback) High (fast iteration of computer models)

[0159] Specifically, as can be seen from Table 1, the method of the present invention has significant advantages and breakthroughs in many aspects such as model accuracy, computational efficiency, design space coverage, verification environment authenticity and optimization iteration efficiency.

[0160] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A gas turbine control simulation test method, characterized in that: The following steps are involved: S1. Obtain the design parameters of the gas turbine and the characteristics of the mixed hydrogen combustion, and establish a theoretical model of the mixed hydrogen gas turbine; S2. Digitize the theoretical model of the hydrogen-mixed gas turbine into a computer executable model, and simulate the operation of the gas turbine control system on the computer; S3. Based on the field programmable gate array FPGA, build a prototype simulation environment equivalent to the actual control hardware to verify the compatibility of the various components of the gas turbine control system; S4, connect the actual physical components to the prototype simulation environment, conduct semi-physical simulation tests, and reproduce the real control process; S5. Analyze and identify parameters that affect control performance based on the simulation results, and modify and optimize the parameters; S6. Connect the final control hardware prototype to the semi-physical simulation environment to carry out hardware-in-the-loop simulation; S7. Feedback the problems and optimization requirements found in the simulation to the gas turbine control logic design link of steps S1-S6 to improve the gas turbine control logic framework.

2. The gas turbine control simulation test method according to claim 1, characterized in that: The specific steps of establishing the theoretical model of the hydrogen-mixed gas turbine are as follows: 1) Design and execute mixed hydrogen combustion experiments to measure flow field data, concentration distribution data, and temperature field data under different working conditions inside the combustion chamber; 2) Based on the experimental measurement data, a three-dimensional numerical simulation model of the combustion chamber was established using CFD software to simulate the mixed hydrogen combustion process; 3) A combustion model reduction network is trained using a deep learning algorithm to compress the three-dimensional numerical simulation model of the combustion chamber into a low-order proxy combustion model for order reduction calculation; 4) Based on the low-order proxy combustion model after compression, a low-order theoretical model of the gas turbine system is constructed by introducing the thermodynamic and fluid dynamics control equations at the whole machine level; 5) Based on the low-order theoretical model of the whole machine, analyze the influence of design parameters on performance indicators through numerical calculation to determine the optimal design parameter combination; 6) Use other independent experimental data to verify the low-order theoretical model of the whole machine, calculate the prediction deviation, and perform secondary correction and optimization on the deviation part; 7) After multiple rounds of data correction and optimization iterations, a high-precision theoretical model of the hydrogen-mixed gas turbine is obtained to simulate the thermal performance under various operating conditions.

3. The gas turbine control simulation test method according to claim 2, characterized in that: The specific steps of compressing the three-dimensional numerical simulation model of the combustion chamber into a low-order proxy combustion model are as follows: 1) Under different working conditions, a large number of sample data sets including model input and output are generated using the three-dimensional numerical simulation model of the combustion chamber, and the generated data sets are preprocessed in a standardized manner; 2) Use deep learning algorithm to train a combustion model reduction network CMRN; 3) The dataset is divided into training set, validation set and test set in proportion, the total loss function TLoss is defined, and the adaptive moment estimation optimization algorithm is used to train CMRN end-to-end; 4) Deploy the trained CMRN network as an online inference engine to generate low-order proxy combustion model outputs in real time; 5) Evaluate the generalization performance of CMRN on the test set and calculate the root mean square error of each prediction head; 6) If the root mean square error RMSE of any prediction head of CMRN exceeds the preset threshold, the control logic is triggered to perform adaptive adjustment on CMRN; 7) After the adjustment, redeploy the CMRN online service to form a new round of online reasoning; The specific formula of the total loss function TLoss is as follows: TLoss=α·RLoss+β·CLoss RLoss=∑(yy hat ) 2 CLoss=λ·θ1 Among them, RLoss represents the mean square error reconstruction loss between the predicted output and the true output, CLoss represents the compression loss, α and β represent the weight coefficients of the reconstruction loss RLoss and the compression loss CLoss in the total loss function, y represents the true combustion model output, y hat represents the combustion model output predicted by the CMRN network, θ represents the set of all trainable parameters in the CMRN network, |θ1 represents the L1 norm of θ, that is, the sum of the absolute values ​​of each parameter, and λ represents the weight coefficient of CLoss.

4. The gas turbine control simulation test method according to claim 3, characterized in that: The specific steps of the trigger control logic to perform adaptive fine-tuning on CMRN are as follows: 1) Analyze the sources of errors, If the error under the corresponding working condition exceeds the standard, increase the training data under the corresponding working condition to improve the data distribution coverage; If the error comes from temperature field prediction, increase the loss function weight of the temperature field prediction head; If the error comes from high pressure / low pressure conditions, increase the data sampling rate of the corresponding conditions; 2) Adjust the compression constraints, If the root mean square error RMSE>1.5ε, relax the compression constraint, reduce λ, and increase the model capacity; If the root mean square error RMSE>ε and RMSE≤1.5ε, then strengthen the compression constraint, increase λ, and reduce the model complexity; 3) If the training is terminated and RMSE>ε, the model is refined based on the current parameters of CMRN through a rule-based pruning strategy, and the model is fine-tuned on the adjusted training set based on the refined model; If the training is terminated and RMSE ≤ ε, the loss function weights α, β, and λ are adjusted based on the current model through the Monte Carlo search method, and the adjusted optimal model parameters and loss function parameters are saved; 4) Evaluate the generalization performance of the adjusted CMRN. If RMSE > ε, repeat the adaptive fine-tuning process.

5. The gas turbine control simulation test method according to claim 4, characterized in that: The specific steps of digitizing the hydrogen-mixed gas turbine theoretical model into a computer executable model are as follows: 1) Select programming language and development environment, as well as corresponding numerical computing libraries and tools; 2) Various equations, algorithms and data structures in the theoretical model of the hybrid hydrogen gas turbine are coded and implemented using the selected programming language to form a theoretical model computer program; 3) Perform unit testing and integration testing on the coded computer programs to ensure the correctness and reliability of the code; 4) Design and implement digital models of gas turbine control systems, including various controllers, actuators and sensor models; 5) Integrate the theoretical model of the hydrogen-mixed gas turbine and the digital model of the gas turbine control system to construct a comprehensive digital model of the gas turbine; 6) Design and implement the user interface to input various operating parameters, design parameters and control parameters; 7) Run the gas turbine integrated digital model on the computer to simulate the dynamic process under various working conditions and output performance indicators and state quantity curves; 8) Analyze and visualize simulation results to assist in design optimization; 9) Adjust and improve the comprehensive digital model of gas turbine based on simulation results and actual data.

6. The gas turbine control simulation test method according to claim 1, characterized in that: The specific steps of building a prototype simulation environment equivalent to the actual control hardware are as follows: 1) Select FPGA development board and supporting design tools; 2) According to the structure and function of the actual control hardware, use the hardware description language to perform logical design, simulate the digital circuits of each component, and synthesize the designed digital circuits into a hardware mapping file; 3) Instantiate the logic circuits of each component in parallel or sequentially on the FPGA development board to build an FPGA prototype system; 4) Write test platform programs to simulate the communication protocols and timing signals between components of the FPGA prototype system; 5) Connect the gas turbine integrated digital model with the FPGA prototype system as the environmental input to build the FPGA prototype simulation environment; 6) In the FPGA prototype simulation environment, simulate the interactive response of the FPGA prototype system under different working conditions; 7) Collect and analyze the running data of the FPGA prototype system in the simulation environment to verify the normal coordination and matching between the components; 8) Based on the verification results, feedback is provided to optimize the FPGA prototype design or the gas turbine integrated digital model design; 9) If there is a mismatch between the FPGA prototype system and the integrated digital model of the gas turbine, feedback is provided to optimize the FPGA prototype system design; If the deviation between the gas turbine integrated digital model and the actual situation exceeds a first preset threshold, the integrated digital model design of the gas turbine is optimized.

7. The gas turbine control simulation test method according to claim 1, characterized in that: The specific steps of analyzing and identifying the key parameters that affect the control performance according to the simulation results are as follows: 1) Extracting original simulation data from the semi-physical simulation environment and calculating key performance indicators based on the original simulation data; 2) For relevant performance indicators, determine their theoretical distribution range and confidence interval through mathematical and statistical analysis methods, correspond the actual simulation results to the theoretical distribution, and calculate their probability scores as scoring indicators; 3) Setting the grading threshold of the probability score and grading the key performance indicators according to the probability score; 4) For performance indicators with unqualified scores, trace back and analyze the root cause parameters or parameter combinations that lead to poor performance of this indicator; 5) Construct a parameter optimization mathematical model for the identified parameters that affect performance; 6) Use particle swarm and sequential quadratic programming algorithms to solve the parameter optimization mathematical model and obtain the optimal parameter combination; 7) Update the optimal parameter combination to the semi-physical simulation environment, re-conduct the semi-physical simulation experiment, and obtain the optimized key indicator performance data; 8) Based on the optimized semi-physical simulation results, re-evaluate the performance level of each key indicator.

8. A gas turbine control simulation test system, based on the gas turbine control simulation test method according to any one of claims 1 to 7, characterized in that: Also includes, Model building module, used to obtain gas turbine design parameters and hydrogen-mixed combustion characteristics, and establish a hydrogen-mixed gas turbine theoretical model; A digital module is used to digitize the theoretical model of the hydrogen-mixed gas turbine into a computer executable model and simulate the operation of the gas turbine control system on a computer; The prototype simulation module is used to build a prototype simulation environment equivalent to the actual control hardware based on the field programmable gate array FPGA to verify the compatibility of the various components of the gas turbine control system; Semi-physical simulation module, used to connect actual physical components to the prototype simulation environment, conduct semi-physical simulation experiments, and reproduce the real control process; A parameter correction module is used to analyze and identify parameters that affect control performance based on simulation results, and to correct and optimize the parameters; The hardware-in-the-loop simulation module is used to connect the final control hardware prototype with the semi-physical simulation environment to carry out hardware-in-the-loop simulation; The feedback improvement module is used to feed back the problems and optimization requirements found in the simulation to the control logic design link to improve the control logic framework.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the gas turbine control simulation test method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the gas turbine control simulation test method according to any one of claims 1 to 7 are implemented.

Citation Information

Cited By

  • Semi-physical test system, method and program for gas turbine capable of reversing and storage medium

    CN119739041A

  • Agent-driven intelligent AI depth prediction method and system for power system

    CN120257826A

  • Complex system-oriented simulation model dynamic optimization and real-time correction system

    CN120449531A

  • Dynamic optimization and real-time correction system for simulation models of complex systems

    CN120449531B