A combustion optimization method based on a combined cycle unit

Abstract

The application provides a combustion optimization method based on a combined cycle unit, and belongs to the technical field of combustion engineering. The method comprises the following steps: monitoring the operation parameters of the gas turbine through a sensor, evaluating the operation parameters of the gas turbine in combination with a first combustion law, and extracting important features from the combustion evaluation results based on a combustion optimization target; obtaining real-time heat value and density data of the combustion source, integrating the data with the operation parameters of the gas turbine, obtaining a first data set, performing feature construction to obtain a derived feature set, combining a second combustion law, and establishing a combustion model; combining the combustion model with historical operation data of the combined cycle unit, constructing a prediction model of the optimal opening degree of the premixing valve, and obtaining the optimal opening degree of the premixing valve; adjusting the premixing valve according to the optimal opening degree, collecting real-time combustion conditions after the premixing valve is adjusted, comparing the real-time combustion conditions with predicted combustion conditions, forming a feedback mechanism based on the comparison results, updating the combustion model, significantly improving the energy efficiency utilization rate of the combined cycle unit, and effectively reducing harmful emissions such as NOx.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of combustion engineering, and in particular to a combustion optimization method based on a combined cycle unit. BACKGROUND

[0002] Under the current background of increasing attention to energy utilization and environmental protection, as a highly efficient and relatively clean power generation method, the combined cycle unit, although relatively mature in technology, still faces the problem of how to further optimize the combustion process to improve efficiency and reduce emissions. The research on combustion optimization is not only related to the effective use of energy, but also related to reducing the negative impact on the environment. Traditional methods usually rely on fixed parameter settings and cannot adapt to dynamic changes in the environment. When dealing with historical data, these data are often not effectively extracted and applied.

[0003] Therefore, the present application provides a combustion optimization method based on a combined cycle unit. SUMMARY

[0004] The present application provides a combustion optimization method based on a combined cycle unit, which obtains real-time heat value and density data of the combustion source, integrates to obtain a first data set, and constructs a feature set. In combination with historical operation data and a combustion model, the optimal opening of the premixing valve is predicted. In actual operation, the valve opening is adjusted according to the predicted combustion condition, real-time data is collected for feedback, the energy efficiency utilization rate of the combined cycle unit is significantly improved, harmful emissions such as NOx are effectively reduced, dynamic response to environmental changes and fluctuations in operating conditions is achieved, and the system stability is improved.

[0005] The present application provides a combustion optimization method based on a combined cycle unit, which includes:

[0006] Step 1: monitor the operating parameters of the gas turbine through a sensor, evaluate the operating parameters of the gas turbine in combination with a first combustion law, and extract important features from the combustion evaluation results based on a combustion optimization target;

[0007] Step 2: obtain real-time heat value and density data of the combustion source, integrate with the operating parameters of the gas turbine, obtain a first data set, and construct a derived feature set from the first data set. In combination with a second combustion law, a combustion model is established;

[0008] Step 3: combine the combustion model with the historical operation data of the combined cycle unit to construct a prediction model of the optimal opening of the premixing valve, and obtain the optimal opening of the premixing valve;

[0009] Step 4: adjust the premixing valve according to the optimal opening, collect the real-time combustion condition after the premixing valve is adjusted, and compare it with the predicted combustion condition. Based on the comparison result, a feedback mechanism is formed to update the combustion model.

[0010] The application provides a combustion optimization method based on a combined cycle unit, the operation parameters of a gas turbine are monitored through a sensor, the operation parameters of the gas turbine are evaluated in combination with a first combustion law, important features are extracted from the combustion evaluation results based on a combustion optimization target, including:

[0011] Equipment basic information of the combined cycle unit is acquired, and the operation parameters of the gas turbine are acquired based on the equipment basic information;

[0012] The combustion relationship between the operation parameters of different gas turbines and the combustion condition is established according to the combustion theory, and the first combustion law of the combustion relationship under different working conditions is evaluated in combination with the combustion characteristics, and the combustion evaluation results are obtained;

[0013] The combustion optimization target is determined, and the important features related to the combustion optimization are extracted from the combustion evaluation results based on the combustion optimization target.

[0014] The application provides a combustion optimization method based on a combined cycle unit, real-time heat value and density data of a combustion source are acquired, and are integrated with the operation parameters of the gas turbine, first data sets are obtained, feature construction is performed on the first data sets to obtain a derived feature set, a combustion model is established in combination with a second combustion law, including:

[0015] Based on the important features and the first data sets, feature construction is performed by applying corresponding domain knowledge and the combustion theory to obtain a derived feature set;

[0016] According to the combustion optimization target, target variables are obtained, the derived feature set and the target variables are discretized, and the discretized derived feature set and the target variables are randomly combined to obtain category combinations, and a frequency table is created based on the category combinations;

[0017] According to the frequency table, the category combinations are tested, and corresponding category combinations with a correlation degree between the target variables and the derived features higher than a preset correlation degree are selected from the test results to generate a significant derived feature set;

[0018] The combustion influence of each significant derived feature set on each stage of the combustion process is determined in combination with the second combustion law, and a combustion model is constructed based on the combustion influence and the combustion evaluation results.

[0019] The application provides a combustion optimization method based on a combined cycle unit, based on the important features and the first data sets, feature construction is performed by applying corresponding domain knowledge and the combustion theory to obtain a derived feature set, including:

[0020] The important features and the first data sets are subjected to domain analysis based on the combustion optimization target, and the involved domain of the important features and the first data sets is determined;

[0021] The corresponding domain knowledge is acquired from the combustion knowledge base in combination with the domain, and a combustion theory consistent with the combustion optimization target is acquired, hierarchical feature construction of the domain is performed based on the important features and the first data set, and a derived feature set is obtained by comprehensively constructing features of all domains.

[0022] The application provides a combustion optimization method based on a combined cycle unit, which combines a second combustion law to determine the combustion influence of each significant derived feature set on each stage of the combustion process, constructs a combustion model based on the combustion influence and combustion evaluation results, including:

[0023] The second combustion law is used to establish a connection between the significant derived feature set and each stage of the combustion process, and the combustion influence of each significant derived feature at each stage is quantified;

[0024] Meanwhile, the contribution of the significant derived feature to the combustion optimization target is quantified;

[0025] The combustion model is constructed according to the combustion influence and the contribution in combination with the combustion evaluation results.

[0026] The application provides the combustion optimization method based on the combined cycle unit, which combines the combustion model and the historical operation data of the combined cycle unit to construct a prediction model of the optimal opening degree of the premixing valve, and obtains the optimal opening degree of the premixing valve, including:

[0027] The combustion model and the historical operation data of the combined cycle unit are combined to determine each environmental condition affecting the opening degree of the premixing valve, obtain baseline data of the opening degree of the premixing valve under stable operation conditions of the combined cycle unit, and determine the expected value range of the opening degree of the premixing valve under each environmental condition based on the baseline data;

[0028] Each environmental condition is identified according to the expected value range, and the change identification result is analyzed to obtain the change reason;

[0029] Each environmental condition and the corresponding change identification result and change reason are one-to-one corresponding, and the opening degree of the premixing valve is set according to the corresponding result, and then the prediction model of the optimal opening degree of the premixing valve is obtained, and the optimal opening degree of the premixing valve under different environmental conditions is predicted.

[0030] The application provides a combustion optimization method based on a combined cycle unit, which combines each environmental condition and the corresponding change identification result and change reason one-to-one, and sets the opening degree of the premixing valve according to the corresponding result, and then obtains a prediction model, and predicts the optimal opening degree of the premixing valve under different environmental conditions, including:

[0031] The first rule for adjusting the opening degree of the premixing valve is set according to the change identification result and the change reason;

[0032] A second rule for adjusting the opening of the premixing valve is set based on the environmental conditions and changes identified.

[0033] A third rule is set to adjust the opening of the premixing valve based on environmental conditions and the reasons for changes;

[0034] Based on the first, second, and third rules, the characteristics of fuel chamber temperature, combustion source flow rate, and air-to-combustion source ratio during the combustion process are determined, and a prediction model is constructed, setting the calculation formula for the optimal opening of the premixing valve:

[0035]

[0036] Where A represents the weighted sum of deviations for each environmental condition; m represents the total number of environmental conditions; k represents the k-th environmental condition; ω k (t) The weighting coefficient of the k-th environmental condition; X k X represents the actual variable of the k-th environmental condition; r,k The reference variable represents the k-th environmental condition; O new This indicates the optimal opening degree of the premixed valve under the current conditions; Indicates the reference opening degree of the premixing valve; O i This represents the i-th premixing valve opening value in the historical operating data; n represents the number of premixing valve opening values ​​in the historical operating data; F s (t) represents the current performance index of the combustion process; F q (t) represents the ideal performance index output by the combustion model; τ represents the dynamic adjustment factor; HV represents the real-time calorific value of the combustion source; D represents the density of the combustion source; E s β1, β2, and β3 represent the coefficients of the interaction term, the coefficients of the nonlinear term, and the coefficients of the exponential correlation, respectively; X1, X2, and X3 represent the characteristics of the fuel chamber temperature, the flow rate of the combustion source, and the ratio of air to combustion source during the combustion process, respectively.

[0037] This invention provides a combustion optimization method for combined cycle units, comprising adjusting a premixing valve according to the optimal opening degree, collecting real-time combustion data after premixing valve adjustment, comparing it with predicted combustion data, forming a feedback mechanism based on the comparison results, and updating the combustion model, including:

[0038] The predicted combustion situation is based on the output of the prediction model after the optimal opening degree is executed;

[0039] The real-time combustion situation is compared with the predicted combustion situation in multiple dimensions to obtain the multidimensional deviation. The deviation is judged. If the absolute value of the deviation is lower than the deviation threshold, it is marked as normal, the optimization effect is obtained, and positive feedback is given to the combustion model. If the absolute value of the deviation is higher than the deviation threshold, it is marked as abnormal, the negative adjustment result is obtained, and negative feedback is given to the combustion model.

[0040] The combustion model is updated based on the feedback mechanism.

[0041] Compared with existing technologies, the beneficial effects of this application are as follows: by acquiring real-time calorific value and density data of the combustion source, integrating them to obtain a first dataset, constructing a feature set, and then combining historical operating data with a combustion model, the optimal opening degree of the premixed valve is predicted. In actual operation, the valve opening degree is adjusted according to the predicted combustion situation, and data is collected in real time for feedback, which significantly improves the energy efficiency of the combined cycle unit, effectively reduces harmful emissions such as NOx, dynamically responds to environmental changes and fluctuations in operating conditions, and improves system stability.

[0042] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0043] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0044] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0045] Figure 1 This is a schematic flowchart of a combustion optimization method based on a combined cycle unit provided in an embodiment of the present invention. Detailed Implementation

[0046] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0047] Example 1:

[0048] This invention provides a combustion optimization method for combined cycle power units, such as... Figure 1 As shown, it includes:

[0049] Step 1: Monitor the operating parameters of the gas turbine through sensors, evaluate the operating parameters of the gas turbine in combination with the first combustion law, and extract important features from the combustion evaluation results based on the combustion optimization target;

[0050] Step 2: Obtain real-time calorific value and density data of the combustion source, integrate them with the operating parameters of the gas turbine to obtain the first dataset, construct features from the first dataset to obtain a derived feature set, and establish a combustion model by combining the second combustion law;

[0051] Step 3: Combining the combustion model with historical operating data of the combined cycle unit, construct a predictive model for the optimal opening of the premixed valve, and obtain the optimal opening of the premixed valve;

[0052] Step 4: Adjust the premixing valve according to the optimal opening, collect the real-time combustion data after the premixing valve adjustment, compare it with the predicted combustion data, form a feedback mechanism based on the comparison results, and update the combustion model.

[0053] In this embodiment, operating parameters are key indicators describing the operating status of combined cycle units, especially gas turbines, reflecting the current operating performance and environmental conditions of the equipment, including: gas temperature, pressure, air flow, gas composition, etc. For example, temperature: the temperature of the combustion chamber (e.g., 900°C), pressure: the pressure inside the combustion chamber (e.g., 100 kPa).

[0054] In this embodiment, the first combustion law is a summary of combustion behavior under specific conditions based on combustion theory and analysis results. It is usually an overview of the performance under different operating conditions, including: optimal combustion zone, combustion rate, etc. For example, combustion stability under stable temperature conditions: when the temperature is in the range of 900℃ to 950℃, NOx generation remains at an acceptable level.

[0055] In this embodiment, the combustion optimization objective is to achieve specific performance indicators by adjusting operating parameters, aiming to improve the safety, efficiency, and environmental protection level of the combustion process, including: improving combustion efficiency, reducing emissions, increasing energy output, reducing fuel density, optimizing combustion stability, reducing noise and vibration, and extending the lifespan of the combustion equipment. For example, maximizing combustion efficiency: the desired combustion efficiency is at least 95%.

[0056] In this embodiment, the density data is the density of the combustion source (such as natural gas, heavy oil, coal, etc.) under specific conditions. For example, liquid fuels such as heavy oil and gasoline are generally denser than gaseous fuels (such as natural gas).

[0057] In this embodiment, the first dataset contains measured data related to the operation of the gas turbine, such as operating parameters (temperature, pressure, fuel flow rate, air flow rate, etc.) and real-time calorific value and density data, for example, the fuel flow rate at a certain moment is 100 kg / h, the inlet air temperature is 600°C, and the pressure is 15 bar.

[0058] In this embodiment, the process of determining the derived feature set is to obtain relevant domain knowledge and combustion theory from the combustion knowledge base, combine important features with the first dataset, implement hierarchical feature construction, comprehensively analyze the feature construction results of different involved domains, and finally obtain the derived feature set.

[0059] In this embodiment, the combustion model is constructed by establishing a quantitative relationship between the significant derived feature set and each stage of the combustion process through the second combustion law, evaluating the impact of each feature on the combustion process, quantifying the contribution of the significant derived features to the combustion optimization objective, and constructing a combustion model by combining the combustion impact and contribution with the combustion evaluation results.

[0060] In this embodiment, by combining the combustion model with the historical operating data of the combined cycle unit, the environmental conditions affecting the opening of the premixed valve are identified, and baseline data under stable operating conditions are obtained. Based on these baseline data, the expected range of the premixed valve opening under various environmental conditions is determined, and the changes are identified and the causes are analyzed. By corresponding environmental conditions and causes of changes, the premixed valve opening is set, thereby establishing a predictive model for the optimal opening of the premixed valve.

[0061] In this embodiment, the feedback mechanism compares the real-time combustion status with the optimal opening degree output by the prediction model, calculates the multidimensional deviation, and judges whether the operation is normal by setting a deviation threshold. If the deviation is less than the threshold, it is marked as normal and positive feedback is given to the combustion model to improve the prediction accuracy; if the deviation is greater than the threshold, it is marked as abnormal and negative feedback is given to prompt the combustion model to adjust and update.

[0062] The working principle and beneficial effects of the above technical solution are as follows: by acquiring real-time calorific value and density data of the combustion source, integrating them to obtain the first dataset, constructing a feature set, and then combining historical operating data with the combustion model, the optimal opening degree of the premixed valve is predicted. In actual operation, the valve opening degree is adjusted according to the predicted combustion situation, and data is collected in real time for feedback, which significantly improves the energy efficiency of the combined cycle unit, effectively reduces harmful emissions such as NOx, dynamically responds to environmental changes and fluctuations in operating conditions, and improves system stability.

[0063] Example 2:

[0064] This invention provides a combustion optimization method for combined cycle units. The method monitors the gas turbine's operating parameters using sensors, evaluates these parameters in conjunction with a first combustion law, and extracts key features from the combustion evaluation results based on combustion optimization objectives. These features include:

[0065] Obtain basic equipment information of the combined cycle unit, and obtain the operating parameters of the gas turbine based on the basic equipment information;

[0066] Based on combustion theory, the combustion relationship between the operating parameters and combustion conditions of different gas engines is established, and the first combustion law of the combustion relationship under different operating conditions is evaluated in combination with combustion characteristics to obtain combustion evaluation results;

[0067] Define combustion optimization objectives, and extract key features related to combustion optimization from the combustion evaluation results based on these objectives.

[0068] In this embodiment, the basic equipment information describes the technical parameters and specifications of each component in the combined cycle unit, including the equipment's performance indicators and characteristics, such as model and specifications: a certain model of gas turbine, such as "GE LM6000", and rated power: the rated output power of the equipment (e.g., 40MW).

[0069] In this embodiment, the combustion relationship describes the interaction between different operating parameters and combustion results, including combustion products and temperature and pressure, such as temperature and NOx formation: increased temperature generally leads to increased NOx emissions.

[0070] In this embodiment, combustion performance refers to the combustion characteristics of a combined cycle unit under specific conditions, which is typically measured by emissions data and efficiency, including combustion mode, combustion stability, flame shape and characteristics, and unburned fuel monitoring. For example, combustion efficiency is the ratio of the energy produced by actual combustion to the theoretical energy (e.g., combustion efficiency = 95%).

[0071] In this embodiment, the combustion assessment results are data feedback obtained by analyzing the combustion process, describing the system's performance under specific conditions, including combustion efficiency, emission levels, etc. The emission assessment report describes the results of NOx, CO, and other emissions under different operating conditions (e.g., NOx = 30 ppm at 75% load and NOx = 50 ppm at 100% load).

[0072] The working principle and beneficial effects of the above technical solution are as follows: Based on the basic equipment information of the combined cycle unit, by real-time monitoring of the gas turbine's operating parameters and combining combustion theory to evaluate combustion relationships and characteristics, the first combustion law is determined, combustion optimization targets are clarified and important features are extracted, and finally, dynamic optimization of the combustion process is achieved, improving power generation efficiency and reducing emissions, adapting to changes in the operating environment, and enhancing the system's flexibility and responsiveness.

[0073] Example 3:

[0074] This invention provides a combustion optimization method for combined cycle units, which acquires real-time calorific value and density data of combustion sources, integrates them with the operating parameters of the gas turbine to obtain a first dataset, constructs features from the first dataset to obtain a derived feature set, and establishes a combustion model by combining a second combustion law, including:

[0075] Based on the aforementioned key features and the first dataset, corresponding domain knowledge and combustion theory are applied to construct features and derive a derived feature set.

[0076] Based on the combustion optimization objective, a target variable is derived. The derived feature set and the target variable are discretized, and the discretized derived feature set and the target variable are randomly combined to obtain a category combination. A frequency table is created based on the category combination.

[0077] The category combinations are tested according to the frequency table, and the corresponding category combinations with a higher correlation between the target variable and the derived features than the preset correlation are selected from the test results to generate a significant derived feature set.

[0078] The combustion impact of each significant derived feature set on each stage of the combustion process is determined by combining the second combustion law, and a combustion model is constructed based on the combustion impact and combustion evaluation results.

[0079] In this embodiment, domain knowledge includes the basic principles of combustion, chemical reaction characteristics, and fluid dynamics; combustion theory provides a theoretical framework for explaining and predicting combustion behavior. For example, based on Archimedes' constant and reaction rate equation, the influence of different fuel-air mixing ratios on combustion efficiency can be studied.

[0080] In this embodiment, the derived feature set is a new feature set extracted through mathematical calculations or physical models based on the first dataset and domain knowledge, such as the ratio of temperature to pressure, the ratio of oxygen to fuel, etc.

[0081] In this embodiment, the target variable is a key indicator determined based on the combustion optimization objective. It is usually an output result that needs to be optimized, such as combustion efficiency (expressed as a percentage) or emission concentration (e.g., NOx in g / kWh).

[0082] In this embodiment, the category combination is a pair of different categories (combinations) generated by randomly combining the discretized derived feature set and the target variable. For example, the category of the combination formed by the temperature, pressure and mixing ratio of the incoming air, such as (high temperature, high pressure, low mixing ratio).

[0083] In this embodiment, the frequency table is a table that records the frequency of occurrence of category combinations. It is used to detect which combinations are common in the dataset. For example, the table lists feature combinations, such as the "high temperature-high pressure" combination, which appeared 80 times in all test data.

[0084] In this embodiment, the significant derived feature set is a tested derived feature set whose correlation with the target variable is higher than the preset correlation, indicating features that have an important impact on the combustion process. For example, a significant correlation between oxygen concentration and NOx emissions was found, and the combination of oxygen concentration = 20% was considered a significant feature.

[0085] In this embodiment, the preset relevance is a threshold that defines the strength of the association between a feature and a target variable. Combinations below this threshold will be eliminated. For example, if the preset relevance is set to 0.7, only feature combinations with a relevance greater than 0.7 will be considered significant.

[0086] In this embodiment, combustion effects are the specific impacts of significant derivative characteristics on various stages of the combustion process, including the degree of impact on efficiency, stability, and emission characteristics. For example, a specific mixture ratio (such as an oxygen to fuel ratio of 1:14) may lead to optimal thermal efficiency and the lowest NOx emissions.

[0087] The working principle and beneficial effects of the above technical solution are as follows: Based on important features and the first dataset, feature construction is carried out using combustion theory to generate a derivative feature set. The target variable is determined according to the combustion optimization objective. The derivative feature set is discretized and randomly combined to create a category combination and frequency table. The correlation of the combination is tested, and significant derivative feature sets are screened. Combustion model is constructed in combination with the second combustion law to optimize and adjust each stage of the combustion process, improve power generation efficiency, achieve more effective energy utilization, optimize combustion characteristics, and reduce exhaust emissions.

[0088] Example 4:

[0089] This invention provides a combustion optimization method for combined cycle units. Based on the aforementioned key features and a first dataset, corresponding domain knowledge and combustion theory are applied to construct features, resulting in a derived feature set, including:

[0090] Based on the combustion optimization objective, a domain analysis is performed on the important features and the first dataset to determine the domains involved in the important features and the first dataset.

[0091] By combining the relevant domain knowledge from the combustion knowledge base, and obtaining the combustion theory consistent with the combustion optimization goal, and based on the important features and the first dataset, hierarchical feature construction of the relevant domain is carried out. By combining the feature construction results of all relevant domains, a derived feature set is obtained.

[0092] In this embodiment, the integration process refers to combining knowledge from different fields with combustion theory, and using a systematic feature construction method to combine multiple features to form an integrated derivative feature set. For example, suppose there is the following knowledge and theory from the combustion knowledge base: the relationship between combustion efficiency and fuel flow rate and air flow rate, the influence of temperature and pressure on chemical reaction rate, and more specifically, by combining the changes in temperature, pressure and flow rate, the feature "Feature 1 = (fuel flow rate / air flow rate) * temperature" is constructed.

[0093] In this embodiment, hierarchical feature construction refers to the process of layering features according to their characteristics, sources, or corresponding physical meanings during feature construction. For example, in combustion optimization research, the following layers of features may be constructed: First layer features (original features): features directly obtained from the first dataset, such as fuel temperature, pressure, and flow rate; Second layer features (derived features): features obtained through simple formulas, such as "thermal efficiency = output power / fuel flow rate"; Third layer features (composite features): combining multiple secondary features to represent more complex relationships, such as "feature 3 = interaction feature of temperature and flow rate".

[0094] The working principle and beneficial effects of the above technical solution are as follows: by acquiring relevant domain knowledge and combustion theory from the combustion knowledge base, combining important features with the first dataset, implementing hierarchical feature construction, comprehensively analyzing the feature construction results of different involved fields, and finally deriving a derived feature set, which helps to reveal the complexity of the combustion process, provides data support for optimizing combustion, helps to reduce harmful gas emissions, meets environmental protection standards, and promotes sustainable development.

[0095] Example 5:

[0096] This invention provides a combustion optimization method for combined cycle units, which determines the combustion impact of each significant derived feature set on each stage of the combustion process by combining a second combustion law, and constructs a combustion model based on the combustion impact and combustion evaluation results, including:

[0097] By establishing a connection between the set of significant derived features and each stage of the combustion process through the second combustion law, the combustion impact of each significant derived feature at each stage is quantified.

[0098] At the same time, the contribution of significant derived features to combustion optimization objectives was quantified;

[0099] A combustion model is constructed based on the combustion impact and contribution combined with the combustion assessment results.

[0100] In this embodiment, the process of establishing the connection is to analyze the relationship between the significant derivative feature set and each stage of the combustion process through the second combustion law analysis. This includes: preparing the significant derivative feature set and combustion data for each stage (such as initial combustion, stable combustion, and extinction stage), and using statistical methods (such as linear regression, correlation analysis, and causal inference) to study how the significant derivative features affect the specific performance of each combustion stage. For example, the analysis shows that the increase in temperature is significantly associated with the increase in NOx emissions, and this effect is particularly obvious in the stable combustion stage.

[0101] In this embodiment, the process of quantifying the impact of combustion involves calculating the specific effects of each significant derived feature on combustion efficiency, emission levels, stability, etc., and quantifying the impact of different features to a standard range for easy comparison. For example, standard deviation, percentage change, etc., can be used to express the degree of impact.

[0102] In this embodiment, the contribution calculation quantifies the specific contribution of each significant derived feature relative to the combustion optimization objective (such as reducing NOx emissions) to assess its importance. For example, through regression analysis, a quantitative relationship is found where a 1% increase in oxygen concentration leads to a 5% reduction in NOx emissions, while a 10°C increase in temperature leads to an 8% increase in NOx emissions.

[0103] In this embodiment, the process of constructing a combustion model involves selecting an appropriate model type (such as a linear model, a nonlinear model, a machine learning algorithm, etc.) to adapt to the data characteristics, using historical operating data and the influence derived from the quantification process for model training, adjusting the model parameters based on the training results to optimize its predictive power and accuracy, and using an independent validation set to test the model's performance to ensure its accuracy and reliability under different conditions. For example, a multivariate regression model is established through data fitting, which uses significant derived features (such as fuel flow rate, oxygen concentration, and temperature) to predict NOx emissions.

[0104] The working principle and beneficial effects of the above technical solution are as follows: by establishing a quantitative relationship between the significant derived feature set and each stage of the combustion process through the second combustion law, the influence of each feature on the combustion process is evaluated, the contribution of the significant derived features to the combustion optimization target is quantified, and a combustion model is constructed by combining the combustion influence and contribution with the combustion evaluation results to achieve dynamic optimization, improve combustion efficiency and reduce emissions.

[0105] Example 6:

[0106] This invention provides a combustion optimization method for combined cycle units. By combining a combustion model with historical operating data of the combined cycle unit, a predictive model for the optimal opening of the premixing valve is constructed to derive the optimal opening of the premixing valve, including:

[0107] By combining combustion models with historical operating data of combined cycle units, various environmental conditions affecting the opening degree of the premixed valve are determined, baseline data of the premixed valve opening degree of the combined cycle unit under stable operating conditions are obtained, and the expected value range of the premixed valve opening degree under various environmental conditions is determined based on the baseline data.

[0108] Based on the expected value range, changes in various environmental conditions are identified, and the causes of these changes are analyzed to determine the reasons for the changes.

[0109] By matching the identification results of various environmental conditions and their corresponding changes with the causes of the changes, and setting the opening degree of the premixed valve according to the corresponding results, a prediction model for the optimal opening degree of the premixed valve is obtained, which predicts the optimal opening degree of the premixed valve under different environmental conditions.

[0110] In this embodiment, environmental conditions include temperature, humidity, air pressure, fuel characteristics, etc.

[0111] In this embodiment, the baseline data is the premixed valve opening data collected under stable operating conditions. By collecting historical operating data of the combined cycle unit under various environmental conditions, the standard for stable operation is determined, such as data that has been maintained under the same load and environmental conditions for a long time. Statistical analysis is performed to extract the premixed valve opening data under these conditions. For example, under the environmental conditions of 30°C, 50% relative humidity, 1013hPa air pressure, and 250MW load, the average opening of the premixed valve is 30%, and the fluctuation range under these environmental conditions is 28% to 32%.

[0112] In this embodiment, the expected value range is determined by using statistical methods to calculate the expected value range of the premixed valve opening under different environmental conditions based on baseline data. For example, under the same environmental conditions, the calculation yields: mean: 30%, standard deviation: 2%, expected value range: 28% to 32%.

[0113] In this embodiment, change identification detects the deviation between actual operating data and expected value range. By monitoring data in real time, it identifies opening values ​​that are far from the expected value range, observes the opening trend over time, and determines whether an anomaly has occurred. Change cause analysis involves further analyzing the cause after the change is identified, comparing the current environmental conditions with historical operating conditions, and determining the differences. For example, high temperature impact: when the external ambient temperature rises to 40°C, the combustion efficiency decreases, and the gas flow needs to be increased to ensure output. Humidity impact: when the relative humidity is found to soar to 80%, the fuel-air ratio needs to be adjusted to maintain normal combustion.

[0114] In this embodiment, the one-to-one correspondence process involves matching each environmental condition, change identification result, and change cause with a table or matrix format. For example, if the environmental condition is high temperature (40°C), the change identification result is a decrease in opening degree, and the change cause is a decrease in combustion efficiency and an effect of gas density.

[0115] In this embodiment, the optimal opening prediction model is constructed by using machine learning algorithms or statistical regression analysis to build a prediction model for the opening of the premixed valve. This model can predict the optimal opening setting under different environmental conditions in real time, so as to achieve the best operating efficiency and emission control.

[0116] The working principle and beneficial effects of the above technical solution are as follows: By combining the combustion model with the historical operating data of the combined cycle unit, the environmental conditions affecting the opening of the premixed valve are identified, and baseline data under stable operating conditions are obtained. Based on these baseline data, the expected range of the premixed valve opening under various environmental conditions is determined, and the changes are identified and their causes are analyzed. By corresponding to the environmental conditions and the causes of the changes, the premixed valve opening is set, thereby establishing a predictive model for the optimal opening of the premixed valve to adapt to the operating requirements under different environmental conditions, optimize the fuel-air mixing ratio, thereby improving combustion efficiency and reducing fuel consumption.

[0117] Example 7:

[0118] This invention provides a combustion optimization method for combined cycle units, which identifies various environmental conditions and their corresponding changes, maps these changes to their causes, and sets the opening degree of the premixing valve based on these results. This leads to a prediction model that predicts the optimal opening degree of the premixing valve under different environmental conditions. The method includes:

[0119] Based on the change identification results and the cause of the change, a first rule for adjusting the opening of the premixing valve is set;

[0120] A second rule for adjusting the opening of the premixing valve is set based on the environmental conditions and changes identified.

[0121] A third rule is set to adjust the opening of the premixing valve based on environmental conditions and the reasons for changes;

[0122] Based on the first, second, and third rules, the characteristics of fuel chamber temperature, combustion source flow rate, and air-to-combustion source ratio during the combustion process are determined, and a prediction model is constructed, setting the calculation formula for the optimal opening of the premixing valve:

[0123]

[0124] Where A represents the weighted sum of deviations for each environmental condition; m represents the total number of environmental conditions; k represents the k-th environmental condition; ω k (t) The weighting coefficient of the k-th environmental condition; X k X represents the actual variable of the k-th environmental condition; r,k The reference variable represents the k-th environmental condition; O new This indicates the optimal opening degree of the premixed valve under the current conditions; Indicates the reference opening degree of the premixing valve; O i This represents the i-th premixing valve opening value in the historical operating data; n represents the number of premixing valve opening values ​​in the historical operating data; F s (t) represents the current performance index of the combustion process; F q (t) represents the ideal performance index output by the combustion model; τ represents the dynamic adjustment factor; HV represents the real-time calorific value of the combustion source; D represents the density of the combustion source; E s β1, β2, and β3 represent the coefficients of the interaction term, the coefficients of the nonlinear term, and the coefficients of the exponential correlation, respectively; X1, X2, and X3 represent the characteristics of the fuel chamber temperature, the flow rate of the combustion source, and the ratio of air to combustion source during the combustion process, respectively.

[0125] In this embodiment, the first rule is to adjust the opening of the premix valve based on changes (such as combustion efficiency, temperature, etc.) and their causes (such as environmental factors, high humidity, etc.) monitored by the system. For example, if it is detected that the temperature is too high and incomplete combustion is caused, the valve is adjusted, and the situation may be improved by reducing the opening of the premix valve by 5%.

[0126] In this embodiment, the second rule is to adjust the opening of the premix valve based on the current environmental conditions (such as air pressure and humidity) and the results of change identification. For example, when the humidity increases to 80%, the opening of the premix valve needs to be increased by 3% to ensure complete combustion.

[0127] In this embodiment, the third rule is to combine environmental conditions and the reasons for the change, and optimize the opening degree through dynamic adjustment. For example, if the oxygen is thin due to the air pressure dropping to 950 hPa, the premixing valve may need to be reduced by 2% to adapt to the lower oxygen environment.

[0128] In this embodiment, the interaction term coefficient is used to reflect the influence of the interaction between two or more variables on the model output. The value range is real number. For example, the interaction between fuel chamber temperature and combustion source flow may have a specific influence on the opening degree of the premixing valve.

[0129] In this embodiment, the nonlinear term coefficients allow certain variables in the model to affect the output in a nonlinear manner.

[0130] In this embodiment, the exponential correlation coefficient is typically used to describe the exponential relationship between a variable and its output. In a combustion process model, it is used to describe how variables such as temperature and flow rate affect the opening of the premixed valve in an exponential manner.

[0131] In this embodiment, β1, β2, β3 ∈ R.

[0132] In this embodiment, the fuel chamber temperature characteristic refers to the temperature of the fuel before or during combustion, which affects the volatility and combustion efficiency of the fuel. The general range is 20°C to 400°C. For example, if the temperature of the fuel chamber is 50°C, the premix valve may be set to 30%, while if the temperature increases to 70°C, it may need to be increased to 35%.

[0133] In this embodiment, the combustion source flow rate characteristic refers to the flow rate of the combustion source (gas, liquid, or solid) flowing into the combustion chamber, which directly affects the intensity and stability of combustion. The combustion source is gas, with a range of 20m. 3 / h--1000m 3 / h, with a liquid fuel source ranging from 10L / h to 2000L / h, and a solid fuel source ranging from 10kg / h to 1000kg / h, for example, a flow rate set to 100m³ / h. 3 / h, if the measured flow rate drops to 80m 3 / h, the premix valve needs to be adjusted to increase the flow rate and ensure efficiency.

[0134] In this embodiment, the air-to-fuel ratio is the ratio between air and fuel, which affects the completeness of combustion and emissions. The range is between 1:1 and 10:1. For example, if the air-to-fuel ratio is set to 10:1, and the sensor detects that the ratio has changed to 8:1, then it is necessary to reduce the fuel inflow and adjust the valve opening by an estimated 5% to restore the ideal combustion ratio.

[0135] In this embodiment, the ideal performance index output by the combustion model is a target set based on the optimal combustion theory and historical operating data, including ideal combustion efficiency, temperature and emission standards. For example, the combustion efficiency in the ideal case is 95%. If the actual result is 90%, the premix valve needs to be adjusted to improve the efficiency.

[0136] In this embodiment, the current performance index refers to the actual performance of the combustion process as monitored in real time, including indicators such as combustion efficiency, temperature, and emissions. For example, if the currently monitored efficiency is 90%, measures (such as adjusting the premix valve or flow rate) should be taken immediately to improve the combustion effect if it is lower than the ideal value.

[0137] The working principle and beneficial effects of the above technical solution are as follows: By setting three adjustment rules, the opening degree of the premixing valve is dynamically adjusted to cope with different environmental conditions and operational changes. Based on the identification results and causes of environmental conditions, combined with the characteristics of combustion process data, a predictive model is constructed to calculate the optimal premixing valve opening degree, thereby optimizing the combustion process, improving efficiency, and reducing emissions. The flexible rules can better respond to environmental changes and ensure that the gas turbine operates in the best condition.

[0138] Example 8:

[0139] This invention provides a combustion optimization method for combined cycle units, comprising: adjusting a premixing valve according to the optimal opening degree; collecting real-time combustion data after premixing valve adjustment and comparing it with predicted combustion data; forming a feedback mechanism based on the comparison results; and updating the combustion model, including:

[0140] The predicted combustion situation is based on the output of the prediction model after the optimal opening degree is executed;

[0141] The real-time combustion situation is compared with the predicted combustion situation in multiple dimensions to obtain the multidimensional deviation. The deviation is judged. If the absolute value of the deviation is lower than the deviation threshold, it is marked as normal, the optimization effect is obtained, and positive feedback is given to the combustion model. If the absolute value of the deviation is higher than the deviation threshold, it is marked as abnormal, the negative adjustment result is obtained, and negative feedback is given to the combustion model.

[0142] The combustion model is updated based on the feedback mechanism.

[0143] In this embodiment, the combustion performance is predicted by using a combustion model and data analysis to predict the expected performance of the combustion process at a specific premixing valve opening, including indicators such as temperature, pressure, combustion efficiency, and emissions. For example, when the premixing valve opening is set to 30%, the prediction model may output an ideal combustion efficiency of 95%, a temperature of 1200°C, and NOx emissions of 50ppm.

[0144] In this embodiment, multidimensional deviation refers to the difference between actual combustion and predicted combustion, typically a comprehensive comparison of multiple performance indicators. These indicators may include efficiency, temperature, pressure, and emissions, forming a multidimensional deviation data set. For example, if the actual combustion efficiency is 90%, the temperature is 1250℃, and the NOx emission is 70ppm, while the predicted values ​​are 95%, 1200℃, and 50ppm, then the deviations are: efficiency deviation = 90% - 95% = -5%, temperature deviation = 1250℃ - 1200℃ = +50℃, and NOx deviation = 70ppm - 50ppm = +20ppm.

[0145] In this embodiment, the deviation threshold is a preset critical value used to determine whether the actual operating conditions are within an acceptable range. If the absolute value of the multidimensional deviation exceeds this threshold, the system will mark it as abnormal and trigger negative feedback. For example, the set deviation threshold is 5% (percentage indicator), 50°C (temperature indicator), and 10ppm (emission indicator). If the deviation in efficiency is 5% or greater, or the temperature deviation exceeds 50°C, an abnormal labeling is triggered.

[0146] In this embodiment, negative feedback refers to the negative adjustment signal applied to the model when the performance index output by the combustion model deviates significantly from the actual situation, so as to prompt it to make adjustments and improvements. For example, if the actual combustion efficiency deviation is greater than the threshold and is marked as abnormal, the negative feedback to the model may be to reduce the fuel supply and increase the air inflow to lower the temperature, improve efficiency, and reduce emissions.

[0147] In this embodiment, positive feedback refers to the support signal given to the model when the performance index output by the combustion model is within an acceptable range, so as to promote its continued good operation. For example, if the actual effect matches the predicted effect and the deviation is less than the threshold, positive feedback will be given to the model, such as maintaining the current fuel flow and premix valve settings to ensure efficient system operation.

[0148] The working principle and beneficial effects of the above technical solution are as follows: By comparing the combustion status in real time with the optimal opening degree output by the prediction model, multidimensional deviation is calculated, and the operation is judged by the set deviation threshold. If the deviation is less than the threshold, it is marked as normal and positive feedback is given to the combustion model to improve the prediction accuracy; if the deviation is greater than the threshold, it is marked as abnormal and negative feedback is given to prompt the combustion model to adjust and update, thereby optimizing the combustion process, improving efficiency and reducing emissions.

[0149] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A combustion optimization method based on combined cycle units, characterized in that, include: Step 1: Monitor the operating parameters of the gas turbine through sensors, evaluate the operating parameters of the gas turbine in combination with the first combustion law, and extract important features from the combustion evaluation results based on the combustion optimization target; Step 2: Obtain real-time calorific value and density data of the combustion source, integrate them with the operating parameters of the gas turbine to obtain the first dataset, construct features from the first dataset to obtain a derived feature set, and establish a combustion model by combining the second combustion law; Step 3: Combining the combustion model with historical operating data of the combined cycle unit, construct a predictive model for the optimal opening of the premixed valve, and obtain the optimal opening of the premixed valve; Step 4: Adjust the premixing valve according to the optimal opening, collect the real-time combustion data after the premixing valve adjustment, compare it with the predicted combustion data, form a feedback mechanism based on the comparison results, and update the combustion model; By monitoring the gas turbine's operating parameters through sensors and evaluating these parameters in conjunction with the first combustion law, key features are extracted from the combustion evaluation results based on combustion optimization objectives, including: Obtain basic equipment information of the combined cycle unit, and obtain the operating parameters of the gas turbine based on the basic equipment information; Based on combustion theory, the combustion relationship between the operating parameters and combustion conditions of different gas engines is established, and the first combustion law of the combustion relationship under different operating conditions is evaluated in combination with combustion characteristics to obtain combustion evaluation results; Define combustion optimization objectives, and extract key features related to combustion optimization from the combustion evaluation results based on the combustion optimization objectives; Real-time calorific value and density data of the combustion source are acquired and integrated with the operating parameters of the gas turbine to obtain the first dataset. A derived feature set is then constructed from the first dataset. Combined with the second combustion law, a combustion model is established, including: Based on the aforementioned key features and the first dataset, corresponding domain knowledge and combustion theory are applied to construct features and derive a derived feature set. Based on the combustion optimization objective, a target variable is derived. The derived feature set and the target variable are discretized, and the discretized derived feature set and the target variable are randomly combined to obtain a category combination. A frequency table is created based on the category combination. The category combinations are tested according to the frequency table, and the corresponding category combinations with a higher correlation between the target variable and the derived features than the preset correlation are selected from the test results to generate a significant derived feature set. The combustion impact of each significant derived feature set on each stage of the combustion process is determined by combining the second combustion law, and a combustion model is constructed based on the combustion impact and combustion evaluation results.

2. The combustion optimization method based on a combined cycle unit according to claim 1, characterized in that, Based on the aforementioned key features and the first dataset, corresponding domain knowledge and combustion theory are applied to construct features, resulting in a derived feature set, including: Based on the combustion optimization objective, a domain analysis is performed on the important features and the first dataset to determine the domains involved in the important features and the first dataset. By combining the relevant domain knowledge from the combustion knowledge base, and obtaining the combustion theory consistent with the combustion optimization goal, and based on the important features and the first dataset, hierarchical feature construction of the relevant domain is carried out. By combining the feature construction results of all relevant domains, a derived feature set is obtained.

3. The combustion optimization method based on a combined cycle unit according to claim 1, characterized in that, The combustion impact of each significant derived feature set on each stage of the combustion process is determined by combining the second combustion law. A combustion model is then constructed based on these combustion impacts and the combustion assessment results, including: By establishing a connection between the set of significant derived features and each stage of the combustion process through the second combustion law, the combustion impact of each significant derived feature at each stage is quantified. At the same time, the contribution of significant derived features to combustion optimization objectives was quantified; A combustion model is constructed based on the combustion impact and contribution combined with the combustion assessment results.

4. The combustion optimization method based on a combined cycle unit according to claim 1, characterized in that, By combining combustion models with historical operating data of combined cycle units, a predictive model for the optimal opening of the premixing valve is constructed, yielding the optimal opening of the premixing valve, including: By combining combustion models with historical operating data of combined cycle units, various environmental conditions affecting the opening degree of the premixed valve are determined, baseline data of the premixed valve opening degree of the combined cycle unit under stable operating conditions are obtained, and the expected value range of the premixed valve opening degree under various environmental conditions is determined based on the baseline data. Based on the expected value range, changes in various environmental conditions are identified, and the causes of these changes are analyzed to determine the reasons for the changes. By matching the identification results of various environmental conditions and their corresponding changes with the causes of the changes, and setting the opening degree of the premixed valve according to the corresponding results, a prediction model for the optimal opening degree of the premixed valve is obtained, which predicts the optimal opening degree of the premixed valve under different environmental conditions.

5. The combustion optimization method based on a combined cycle unit according to claim 4, characterized in that, By mapping each environmental condition and its corresponding change to its cause, and then setting the opening degree of the premixing valve based on the corresponding results, a prediction model is derived to predict the optimal opening degree of the premixing valve under different environmental conditions, including: Based on the change identification results and the cause of the change, a first rule for adjusting the opening of the premixing valve is set; A second rule for adjusting the opening of the premixing valve is set based on the environmental conditions and changes identified. A third rule is set to adjust the opening of the premixing valve based on environmental conditions and the reasons for changes; Based on the first, second, and third rules, the characteristics of fuel chamber temperature, combustion source flow rate, and air-to-combustion source ratio during the combustion process are determined, and a prediction model is constructed, setting the calculation formula for the optimal opening of the premixing valve: ； in, A This represents the weighted sum of deviations for each environmental condition; m This represents the total number of environmental conditions; k Indicates the first k Environmental conditions; No. k Weighting coefficients for each environmental condition; Indicates the first k The actual variables of each environmental condition; Indicates the first k Reference variables for each environmental condition; This indicates the optimal opening degree of the premixed valve under the current conditions; Indicates the reference opening degree of the premix valve; This indicates the first in the historical operation data. i One premixing valve opening value; n This indicates the number of premixed valve opening values ​​in the historical operating data; Indicates the current performance indicators of the combustion process; This represents the ideal performance index output by the combustion model; Indicates the dynamic adjustment factor; HV Indicates the real-time calorific value of the combustion source; D Indicates the density of the combustion source; Indicates the supply of combustion source; These represent the interaction term coefficient, the nonlinear term coefficient, and the exponential correlation coefficient, respectively. X1、 X2, X3 These represent the characteristics of fuel chamber temperature, combustion source flow rate, and air-to-combustion source ratio during the combustion process, respectively.

6. The combustion optimization method based on a combined cycle unit according to claim 1, characterized in that, The premixing valve is adjusted according to the optimal opening degree. Real-time combustion data after premixing valve adjustment is collected and compared with predicted combustion data. A feedback mechanism is formed based on the comparison results to update the combustion model, including: The predicted combustion situation is based on the output of the prediction model after the optimal opening degree is executed; The real-time combustion situation is compared with the predicted combustion situation in multiple dimensions to obtain the multidimensional deviation. The deviation is judged. If the absolute value of the deviation is lower than the deviation threshold, it is marked as normal, the optimization effect is obtained, and positive feedback is given to the combustion model. If the absolute value of the deviation is higher than the deviation threshold, it is marked as abnormal, the negative adjustment result is obtained, and negative feedback is given to the combustion model. The combustion model is updated based on the feedback mechanism.

Images