Bayesian optimization-based combustion optimization control method and system

Abstract

The invention discloses a combustion optimization control method and system based on Bayesian optimization, and the method comprises the steps: obtaining the pollutant concentration of waste gas under an initial experiment working condition, and taking the pollutant concentration as initial sample data; training an agent model based on the initial sample data and a multi-target Bayesian optimization algorithm; based on the acquisition function, on the basis of proxy model prediction, selecting the next optimal group of experiment working conditions from the discretized experiment working condition parameter combination as the current experiment working condition; executing the current experiment working condition, detecting the pollutant concentration, judging whether a preset iteration termination condition is met or not, and if yes, ending optimization; and if not, continuing to select the optimal next group of experiment working conditions based on the acquisition function. According to the method, the rapid optimization of the experiment working condition can be realized by taking pollutant minimization as a target.

Description

Technical Field

[0001] The present invention belongs to the technical field of gas turbine combustion control, and in particular relates to a combustion optimization control method and system based on Bayesian optimization. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Traditional gas turbines usually use natural gas as fuel, but the combustion of natural gas produces a lot of carbon dioxide. Hydrogen, as a clean fuel, has been increasingly introduced into the combustion process of gas turbines in recent years.

[0004] Hydrogen blending technology mixes hydrogen with traditional fuels (such as natural gas) to reduce emissions of greenhouse gases and other pollutants. Optimization of the blending ratio is crucial, as it directly affects combustion characteristics, thermal efficiency, and emission levels. However, the combustion of hydrogen-rich fuels will promote the formation of nitrogen oxides, and will produce more carbon monoxide when the equivalence ratio is relatively low. Therefore, it is particularly important to conduct in-depth and comprehensive research on hydrogen blending combustion. Experimental research and analysis of the changes in pollutant concentrations produced by the combustion of hydrogen-blended fuels under different operating conditions will help improve combustion efficiency and reduce emissions. The traversal method is very labor- and time-intensive to find the operating conditions with the lowest pollutant concentration. Therefore, efficient optimization of the parameters of the combustion conditions based on the optimization algorithm can replace the traversal experimental method, which helps to improve the optimization efficiency.

[0005] Current optimization algorithms mostly use the number of iterations or the performance difference between two results as the iteration termination condition, ignoring the deviation in the model optimization direction caused by incomplete or inappropriate initial sample data settings. In addition, although an adjustable space can be set for each parameter, it does not mean that any combination of parameters belongs to a safe condition. Therefore, if the existing optimization algorithm is directly used, it is difficult to ensure that the system always operates stably under safe conditions during the optimization process. Summary of the invention

[0006] In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a combustion optimization control method and system based on Bayesian optimization, which can achieve rapid optimization of experimental conditions with the goal of minimizing pollutants.

[0007] To achieve the above object, a first aspect of the present invention provides a combustion optimization control method based on Bayesian optimization, comprising the following steps:

[0008] Obtain the pollutant concentration of the exhaust gas under the initial experimental conditions as initial sample data;

[0009] Based on the initial sample data, the agent model is trained using a multi-objective Bayesian optimization algorithm;

[0010] Based on the acquisition function and the prediction of the proxy model, the optimal next set of experimental conditions is selected from the discretized experimental condition parameter combinations as the current experimental conditions;

[0011] Execute the current experimental conditions and detect the pollutant concentration to determine whether the preset iteration termination condition is met. If so, the optimization ends; if not, further determine whether the current pollutant concentration is close to the set threshold:

[0012] If not, add this set of experimental conditions and pollutant concentrations to the sample data, update the proxy model, and continue to select the optimal next set of experimental conditions based on the acquisition function; if so, use the preset low-pollution conditions as the current experimental conditions, execute the current experimental conditions and detect the pollutant concentration.

[0013] In some embodiments, the pollutant concentration is expressed as a weighted sum of concentrations of a plurality of specified compounds under given experimental conditions.

[0014] In some embodiments, for each experimental condition, multiple cycle experiments are performed and an average value of the pollutant concentration is taken.

[0015] In some embodiments, the iteration termination condition is set as: when the pollutant concentration obtained in this experiment is reduced by less than a specific value compared with the pollutant concentration obtained in the previous experiment, or the operating conditions of this experiment are the same as those of the previous experiment.

[0016] In some embodiments, after obtaining the next set of experimental conditions, the values ​​of the parameters in the experimental conditions are rounded according to the adjustment steps set therefor.

[0017] A second aspect of the present invention provides a combustion optimization control device based on Bayesian optimization, comprising:

[0018] The initial sample acquisition module is configured to acquire the pollutant concentration of the exhaust gas under the initial experimental conditions as initial sample data;

[0019] An initial model training module is configured to train a proxy model based on the initial sample data and a multi-objective Bayesian optimization algorithm;

[0020] The experimental condition optimization module is configured to select the optimal next set of experimental conditions from the discretized experimental condition parameter combinations based on the acquisition function and the prediction of the surrogate model as the current experimental conditions;

[0021] The experimental condition iterative optimization module is configured to execute the current experimental condition and detect the pollutant concentration to determine whether the preset iteration termination condition is reached. If so, the optimization ends; if not, it further determines whether the current pollutant concentration is close to the set threshold:

[0022] If not, add this set of experimental conditions and pollutant concentrations to the sample data, update the proxy model, and continue to select the optimal next set of experimental conditions based on the acquisition function; if so, use the preset low-pollution conditions as the current experimental conditions, execute the current experimental conditions and detect the pollutant concentration.

[0023] A third aspect of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the electronic device executes the method described.

[0024] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method described when executed by a processor.

[0025] A fifth aspect of the present invention provides a computer program product, which implements the method when running on one or more processors.

[0026] The sixth aspect of the present invention provides a combustion optimization control system based on Bayesian optimization, including an electronic control system, and a constant volume combustion bomb, an air supply system and an exhaust system connected to the electronic control system; the constant volume combustion bomb is used to control the temperature and pressure of the combustion chamber, the air supply system is used to control the hydrogen blending ratio and the fuel gas temperature, and the exhaust system is used to recover the exhaust gas generated in the combustion chamber and perform pollutant analysis; the electronic control system is connected to an ECU controller, and the ECU controller is configured to execute the combustion optimization control method based on Bayesian optimization.

[0027] One or more of the above technical solutions can achieve rapid optimization of experimental conditions with the goal of minimizing pollutants. Moreover, during the entire optimization process, when it is detected that the pollutant concentration is close to the set threshold, the current experimental conditions are abandoned and the currently known low-pollution conditions are used as the next experimental conditions to continue optimization, ensuring that all sample data used to update the proxy model are under safe conditions. In addition, the safety of the optimization process and the stability of the combustion system are also guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0029] Figure 1 It is an overall architecture diagram of the combustion optimization control method and experimental system based on Bayesian optimization in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] Embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not used to limit the protection scope of the present application.

[0031] In the description of the embodiments of the present application, the term "including" and its similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "at least partially based on".

[0032] In order to find the experimental conditions with the lowest pollutant emissions, one or more embodiments of the present invention provide a combustion optimization control method based on Bayesian optimization, aiming at minimizing pollutant emissions, and capable of finding the optimal experimental conditions in the experimental conditions that ensure the safe and stable operation of the system.

[0033] Specifically, the method includes the following steps:

[0034] Step 1: Obtain the pollutant concentration of the exhaust gas under the initial experimental conditions as the initial sample data; the initial experimental conditions include the combustion chamber temperature and pressure, hydrogen addition ratio, and gas temperature.

[0035] Initial experimental design: Design 5 to 10 initial experimental conditions (including temperature, pressure, hydrogen addition ratio, and gas temperature) evenly distributed within the parameter range, and perform these experiments to record the pollutant concentration. Use these 5 to 10 groups of data as the initial samples of the Bayesian optimization algorithm to construct a preliminary approximation model, and the approximation model uses a probabilistic surrogate model.

[0036] The sampling points combine with the combustion characteristics to cover the differences in the generation rates of NOx and CO under different temperatures, pressures, and hydrogen addition ratios. To ensure that the initial data can effectively represent the pollutant emissions under different combustion conditions and facilitate the construction of an accurate model.

[0037] The pollutant concentration includes the concentration of a specified compound, which can be NO X , CO, CO 2 etc., and "minimization of pollutant concentration" is defined as the minimum of the combined function of the concentrations of a specified number of compounds, that is, the weighted sum of the concentrations of a specified number of compounds under the given experimental conditions:

[0038]

[0039] Where x represents the experimental condition, which is represented by an array of combustion chamber temperature, combustion chamber pressure, hydrogen blending ratio, and fuel gas temperature. n is the type of pollutant, g(x) is the concentration of a certain pollutant under experimental condition x, and a is a coefficient, whose value is a constant.

[0040] Step 2: Based on the initial sample data, train the proxy model using a multi-objective Bayesian optimization algorithm.

[0041] Build a surrogate model: Based on the initial experimental data, a Gaussian process regression model is constructed to approximate the relationship between pollutant concentrations and experimental conditions.

[0042] Step 3: Using the Bayesian optimization acquisition function, based on the proxy model prediction, select the optimal next set of experimental conditions from the discretized test condition parameter combinations as the current experimental conditions.

[0043] In step 3, the input parameters of the Bayesian optimization algorithm are limited to a specified set of discrete values, such as the adjustable range of the combustion chamber temperature is 500K to 1000K, the adjustable range of the combustion chamber pressure is 1MPa to 5MPa, the adjustable range of the fuel gas temperature is 300K to 1000K, and the adjustable range of the hydrogen blending ratio is 0% to 100%.

[0044] And for the next set of experimental conditions obtained each time, the rounding operation is performed according to the limited step size. For example, when the optimization result of the combustion chamber temperature is 723.4K, it is converted to the closest discrete value, that is, 720K. As an example, the adjustment step size of the combustion chamber temperature is limited to 10K, the adjustment step size of the combustion chamber pressure is limited to 0.5MPa, the adjustment step size of the gas temperature is limited to 10K, and the adjustment step size of the hydrogen blending ratio is limited to 5%. This restriction takes into account the accuracy of the experimental equipment, and can accelerate the convergence speed of the algorithm and reduce the number of experiments.

[0045] Step 4: Execute the current experimental conditions and detect the pollutant concentration to determine whether the preset iteration termination condition is met. If so, the optimization ends; if not, proceed to step 5.

[0046] In step 4, each time the current experimental condition is updated, multiple cycles of experiments are performed on the current experimental condition, and the average value of the pollutant concentration is taken for executing subsequent condition judgments to reduce experimental errors and ensure the repeatability of the experiment.

[0047] The iteration termination condition is set as follows: when the pollutant concentration obtained in this experiment is less than a certain value compared with the pollutant concentration obtained in the previous experiment, or the operating conditions of this experiment are the same as those of the previous experiment, there is no room for optimization. This can effectively improve the optimization efficiency and ensure that the combustion system operates at an acceptable pollution emission level.

[0048] Based on this, the optimal next set of experimental conditions can be continuously selected through the acquisition function. The system can automatically find the next optimal experimental condition. The model is updated according to each new experimental condition sample data to ensure accurate prediction of the pollutant generation law. Continue to increase the degree of exploration to obtain more experimental conditions as sample data, and then optimize and update the proxy model.

[0049] However, if the initial sample data is too small, or the initial experimental conditions are inappropriately selected, the acquisition model may select a test condition where the pollutant emissions do not meet the standards. Therefore, when the preset termination iteration condition is not met, it is further determined whether the pollutant concentration is close to the set threshold, and step 5 is further executed.

[0050] Step 5: Determine whether the current pollutant concentration is close to the set threshold. If not, add the set of experimental conditions and pollutant concentrations to the sample data, update the proxy model, and return to step 3; if so, use the preset low pollution condition as the current experimental condition and return to step 4.

[0051] Those skilled in the art will appreciate that an experimental condition with a low concentration of pollutants emitted may be selected in the initial test condition and recorded as a low-pollution condition.

[0052] Based on this, the system can be guaranteed to operate stably under safe conditions while continuously increasing the degree of exploration and obtaining more sample data; and when the pollutant concentration is detected to be close to the set threshold, the current experimental condition is abandoned, and the currently known low-pollution condition is used as the next experimental condition to continue optimization. It ensures that all sample data used to update the proxy model are under safe conditions, gradually enriching the sample data to improve the accuracy of the proxy model, and also ensures the safety of the optimization process and the stability of the combustion system.

[0053] Updating the proxy model with newly added test conditions and pollutant concentration data can improve the accuracy of the approximation to the target.

[0054] Based on the above method, one or more embodiments of the present invention also provide a combustion optimization control system based on the Bayesian optimization algorithm, including: a constant volume combustion bomb, an air supply system, an exhaust system and an electronic control system. The air supply system is connected to the burner input end; the exhaust system is connected to the burner output end; the combustion chamber is provided with a thermocouple, a pressure sensor, a heating device and a pressure regulating device, all of which are connected to the constant volume combustion bomb, and the constant volume combustion bomb is connected to the electronic control system.

[0055] The constant volume combustion bomb controls the temperature and pressure in the combustion chamber through a heating device and a pressure regulating device, and measures the temperature and pressure through a thermocouple and a pressure sensor.

[0056] The gas supply system is used to deliver fuel gas of different mixing ratios to the combustion chamber. The gas supply system also includes a mixer for mixing the fuel gas in proportion. The mixer is provided with an electric control valve, a heater, an agitator, a flow meter, a thermocouple and a gas analyzer, all of which are connected to the electric control system. Among them, the electric control valve is used to adjust the hydrogen blending ratio based on the fuel gas flow in the pipeline fed back by the flow meter; the heater is used to adjust the temperature of the mixed gas based on the temperature fed back by the thermocouple; the agitator is used to fully mix different types of gases; the gas analyzer is used to ensure the composition of the mixed gas;

[0057] The exhaust system includes a gas recovery device and an exhaust gas analyzer, both of which are connected to the electronic control system. The gas recovery device cooperates with the gas supply system to generate high-pressure airflow to clean the combustion chamber between each experiment or when the combustion conditions change. The exhaust gas analyzer can analyze the NO in the exhaust gas. X ,CO,CO 2 The concentration of pollutants.

[0058] The electronic control system is connected to an ECU controller, and the ECU controller is configured to: execute the combustion optimization control method based on Bayesian optimization. During the execution of the method, the constant volume combustion bomb is used to control the temperature and pressure of the combustion chamber, the gas supply system is used to control the hydrogen blending ratio and the gas temperature, and the pollutant composition and concentration data are obtained from the exhaust system.

[0059] Specifically, read the temperature and pressure of the constant volume combustion bomb, the composition and temperature of the mixed gas, and the composition of the exhaust gas;

[0060] The ECU controller can control the heating device and pressure regulator of the constant volume combustion bomb, as well as the electric control valve for controlling the mixed gas ratio and the heater for adjusting the mixed gas temperature.

[0061] Based on the above optimization control method, the ECU controller can adjust the temperature, pressure, hydrogen blending ratio and temperature of the combustion chamber in real time to optimize the combustion process and reduce pollutant emissions to obtain the optimal experimental operating conditions.

[0062] The system can control the temperature, pressure, hydrogen blending ratio and temperature of the combustion chamber through a computer, and optimize the control strategy based on exhaust gas detection feedback to minimize pollutant emissions, and infer the operating conditions and hydrogen blending ratio when pollutant emissions are minimum.

[0063] One or more embodiments of the present invention further provide an electronic device that can be used to implement the combustion optimization control method based on Bayesian optimization in the above embodiments. The electronic device includes one or more processors, one or more memories coupled to the processors, and a communication module coupled to the processors.

[0064] The memory may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD) or other magnetic storage and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: random access memory (RAM), or other volatile memories that do not last during the power outage duration. The computer program may be stored in the ROM. When the processor executes the computer program, the above-mentioned combustion optimization control method based on Bayesian optimization is implemented.

[0065] In some embodiments, the program may be tangibly contained in a computer-readable medium, which may be included in a device (such as in a memory) or other storage device accessible by the device. The program may be loaded from the computer-readable medium to the RAM for execution. The computer-readable medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, and the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned combustion optimization control method based on Bayesian optimization.

[0066] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a server or terminal, the process or function described in the embodiment of the present application is generated in whole or in part. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a server or terminal or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (such as a floppy disk, a hard disk and a tape, etc.), an optical medium (such as a digital video disk (digital video disk, DVD), etc.), or a semiconductor medium (such as a solid-state hard disk, etc.).

[0067] In addition, although each operation is described in a specific order, this should be understood as requiring such operation to be performed in the specific order shown or in a sequential order, or requiring that all illustrated operations should be performed to obtain desired results. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination mode.

[0068] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A combustion optimization control method based on Bayesian optimization, characterized in that: The following steps are involved: Obtain the pollutant concentration of the exhaust gas under the initial experimental conditions as initial sample data; Based on the initial sample data, the agent model is trained using a multi-objective Bayesian optimization algorithm; Based on the acquisition function and the prediction of the proxy model, the optimal next set of experimental conditions is selected from the discretized experimental condition parameter combinations as the current experimental conditions; Execute the current experimental conditions and detect the pollutant concentration to determine whether the preset iteration termination condition is met. If so, the optimization ends; if not, further determine whether the current pollutant concentration is close to the set threshold: If not, add this set of experimental conditions and pollutant concentrations to the sample data, update the proxy model, and continue to select the optimal next set of experimental conditions based on the acquisition function; if so, use the preset low-pollution conditions as the current experimental conditions, execute the current experimental conditions and detect the pollutant concentration.

2. The combustion optimization control method based on Bayesian optimization according to claim 1, characterized in that: The pollutant concentration is expressed as a weighted sum of the concentrations of multiple compounds under given experimental conditions.

3. The combustion optimization control method based on Bayesian optimization according to claim 1, characterized in that: For each experimental condition, multiple cycle experiments were carried out and the average value of pollutant concentration was taken.

4. The combustion optimization control method based on Bayesian optimization according to claim 1, characterized in that: The iteration termination condition is set as: when the pollutant concentration obtained in this experiment is reduced by less than a specific value compared with the pollutant concentration obtained in the previous experiment, or the operating conditions of this experiment are the same as those of the previous experiment.

5. The combustion optimization control method based on Bayesian optimization according to claim 1, characterized in that: After the next set of experimental conditions is obtained, the values ​​of the parameters in the experimental conditions are rounded according to the set adjustment steps.

6. A combustion optimization control device based on Bayesian optimization, characterized in that: include: The initial sample acquisition module is configured to acquire the pollutant concentration of the exhaust gas under the initial experimental conditions as initial sample data; An initial model training module is configured to train a proxy model based on the initial sample data and a multi-objective Bayesian optimization algorithm; The experimental condition optimization module is configured to select the optimal next set of experimental conditions from the discretized experimental condition parameter combinations based on the acquisition function and the prediction of the surrogate model as the current experimental conditions; The experimental condition iterative optimization module is configured to execute the current experimental condition and detect the pollutant concentration to determine whether the preset iteration termination condition is reached. If so, the optimization ends; if not, it further determines whether the current pollutant concentration is close to the set threshold: If not, add this set of experimental conditions and pollutant concentrations to the sample data, update the proxy model, and continue to select the optimal next set of experimental conditions based on the acquisition function; if so, use the preset low-pollution conditions as the current experimental conditions, execute the current experimental conditions and detect the pollutant concentration.

7. An electronic device, characterized in that: The electronic device comprises a processor and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the electronic device executes the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

9. A computer program product, characterized in that When the computer program product is run on one or more processors, the method according to any one of claims 1 to 5 is implemented.

10. A combustion optimization control system based on Bayesian optimization, characterized in that: It includes an electronic control system, and a constant volume combustion bomb, an air supply system and an exhaust system connected to the electronic control system; the constant volume combustion bomb is used to control the temperature and pressure of the combustion chamber, the air supply system is used to control the hydrogen blending ratio and temperature, and the exhaust system is used to recover the exhaust gas generated by the combustion chamber and perform pollutant analysis; the electronic control system is connected to an ECU controller, and the ECU controller is configured to: execute the combustion optimization control method based on Bayesian optimization as described in any one of claims 1-5.

Images