Fuel characteristic and thermal efficiency regression analysis method of gas turbine combined cycle system
By combining the multivariate linear regression and support vector regression models and adaptively adjusting the weight values, the problem of thermal efficiency prediction error caused by changes in fuel composition in the gas turbine combined cycle system is solved, and a more efficient and accurate fuel property and thermal efficiency analysis is achieved.
Patent Information
- Application Number
- CN202510833461.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-10-10
AI Technical Summary
The existing thermal efficiency analysis method of gas turbine combined cycle system relies on a single linear or nonlinear model, which has limitations and cannot effectively capture the linear and nonlinear changes of fuel composition, resulting in large prediction errors. In particular, the overfitting problem is serious when the sample size is small or the fuel changes suddenly.
The multivariate linear regression model and the support vector regression model are combined, and the prediction results of the two are fused by adaptively adjusting the weight values. The fuel characteristics and system operating parameters are used for regression analysis to improve the prediction accuracy.
The efficiency and accuracy of regression analysis of gas turbine combined cycle systems have been significantly improved, and the impact of fuel composition on thermal efficiency can be quantified more accurately, reducing prediction deviations.
Smart Images
Figure CN120763447A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present invention relate to the technical field of gas turbine state monitoring, and in particular to a method for regression analysis of fuel characteristics and thermal efficiency of a gas turbine combined cycle system. Background Art
[0002] The thermal efficiency of a gas turbine combined cycle system is highly dependent on the physicochemical properties of the fuel. The significant differences in the main components of the fuel directly affect the combustion temperature, calorific value, and thermodynamic properties of the combustion products. Currently, thermal efficiency analysis and optimization methods for gas turbine combined cycle systems typically rely on single linear or nonlinear models for prediction. However, single-model predictions have many limitations. For example, linear models can only characterize the explicit linear relationship between fuel composition and thermal efficiency and cannot capture the nonlinear effects caused by dynamic changes in fuel composition. Nonlinear models require large amounts of data and have poor interpretability. They are prone to overfitting when using small samples or when the fuel undergoes sudden changes. Summary of the Invention
[0003] Based on the above situation of the prior art, the purpose of the embodiment of the present invention is to provide a regression analysis method for fuel characteristics and thermal efficiency of a gas turbine combined cycle system, which organically combines linear models and nonlinear models to improve the efficiency and accuracy of the regression analysis of the gas turbine combined cycle system.
[0004] To achieve the above object, according to one aspect of the present invention, a method for regression analysis of fuel characteristics and thermal efficiency of a gas turbine combined cycle system is provided, the method comprising the steps of:
[0005] Acquiring input characteristic data of a gas turbine combined cycle system, wherein the input characteristic data includes fuel characteristic parameters and system operating parameters;
[0006] Inputting the input feature data into a first linear regression model and a second nonlinear regression model respectively to obtain a first linear prediction result and a second nonlinear prediction result;
[0007] determining weight values of the first linear prediction result and the second nonlinear prediction result according to the fuel type;
[0008] A regression analysis result is obtained based on the first linear prediction result, the second nonlinear prediction result and the weight value.
[0009] Furthermore, the first linear regression model includes a multiple linear regression model, and the second nonlinear regression model includes a support vector regression model.
[0010] Furthermore, the first linear prediction result includes a linear prediction value of thermal efficiency, and the second nonlinear prediction result includes a nonlinear prediction value of thermal efficiency.
[0011] Furthermore, the fuel characteristic parameters include methane volume fraction, carbon dioxide volume fraction and fuel calorific value; the system operation parameters include gas turbine inlet temperature and gas turbine exhaust temperature.
[0012] Furthermore, determining the weight values of the first linear prediction result and the second nonlinear prediction result according to the fuel type includes:
[0013] Obtain the current fuel type and the preset weight value-fuel type mapping relationship table;
[0014] Based on the current fuel type, the corresponding weight value is searched in the preset weight value-fuel type mapping relationship table.
[0015] Furthermore, the preset weight value-fuel type mapping relationship table is obtained by the following steps:
[0016] The historical data is divided into several training sets and several validation sets based on the fuel type. Each set of training sets and validation sets corresponds to the historical data of different fuel types.
[0017] For each fuel type, the historical data corresponding to the first linear regression model and the second nonlinear regression model are trained using the training set to obtain the trained first linear regression model and the second nonlinear regression model;
[0018] Inputting the input feature data of the validation set corresponding to the fuel type into the trained first linear regression model and the second nonlinear regression model respectively to obtain a first linear prediction result and a second nonlinear prediction result respectively;
[0019] The first linear prediction result, the second nonlinear prediction result and the actual output value of the validation set are used to obtain a preset weight value-fuel type mapping relationship table by minimizing the mean square error of the prediction value.
[0020] Furthermore, the mean square error of the predicted value is expressed as:
[0021]
[0022] Among them, η i represents the actual output value of the validation set, ω MLR Represents the weight value of the first linear prediction result, ω SVR represents the weight value of the second linear prediction result, η MLR,i represents the first linear prediction result, η SVR,i Represents the second linear prediction result, and N represents the number of samples in the validation set.
[0023] Furthermore, by minimizing the mean square error of the predicted value, a preset weight value-fuel type mapping relationship table is obtained, including:
[0024] Based on the constraints, the mean square error of the predicted values is minimized to obtain the optimal weight value corresponding to the fuel type, wherein the optimal weight value includes the optimal weight value of the first linear prediction result and the optimal weight value of the second linear prediction result;
[0025] The constraints include: MLR +ω SVR =1, and the weight value is non-negative.
[0026] Furthermore, the method further comprises:
[0027] The weight value is dynamically adjusted based on the confidence of the first linear regression model and the second nonlinear regression model. If the confidence of the first linear regression model is high, the weight value of the first linear regression model is increased; if the confidence of the second nonlinear regression model is high, the weight value of the second nonlinear regression model is increased.
[0028] Furthermore, the confidence level is determined according to the following formula:
[0029]
[0030] Where λ represents the decay factor and MSE[tk:t] represents the mean square error of the regression model in the past k time steps.
[0031] In summary, an embodiment of the present invention provides a method for regression analysis of fuel characteristics and thermal efficiency of a gas turbine combined cycle system. The method comprises the following steps: obtaining input characteristic data of the gas turbine combined cycle system, the input characteristic data including fuel characteristic parameters and system operating parameters; inputting the input characteristic data into a first linear regression model and a second nonlinear regression model, respectively, to obtain a first linear prediction result and a second nonlinear prediction result; determining weight values for the first linear prediction result and the second nonlinear prediction result based on the fuel type; and obtaining a regression analysis result based on the first linear prediction result, the second nonlinear prediction result, and the weight values. The technical solution of the embodiment of the present invention uses a linear model and a nonlinear model to predict thermal efficiency respectively, and combines the prediction results of the two models through adaptively adjusted weight values, thereby greatly improving the efficiency and accuracy of the regression analysis of the gas turbine combined cycle system. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 The present invention provides a flow chart of a method for regression analysis of fuel characteristics and thermal efficiency of a gas turbine combined cycle system. DETAILED DESCRIPTION
[0033] To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present invention.
[0034] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprising" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.
[0035] The technical solution of the present invention is described in detail below with reference to the accompanying drawings. The embodiment of the present invention provides a method for regression analysis of fuel characteristics and thermal efficiency of a gas turbine combined cycle system. Figure 1 The flowchart of the regression analysis method according to the embodiment of the present invention is shown in FIG. Figure 1 As shown, the method includes the following steps:
[0036] S202. Obtain input feature data for the gas turbine combined cycle system. The input feature data includes fuel characteristic parameters and system operating parameters. In this embodiment of the present invention, the fuel characteristic parameters include methane volume fraction, carbon dioxide volume fraction, and fuel calorific value; the system operating parameters include gas turbine inlet temperature and gas turbine exhaust temperature. Methane, the primary combustible component of the fuel, directly affects combustion efficiency and calorific value. Carbon dioxide reflects the impurity content of the fuel, affecting the specific heat capacity of the combustion products and exhaust temperature. Fuel calorific value directly determines the energy released per unit of fuel. The gas turbine inlet temperature, i.e., the combustion chamber inlet air temperature, has a significant impact on combustion efficiency. The gas turbine exhaust temperature, i.e., the turbine outlet exhaust temperature, is strongly correlated with the efficiency of the waste heat boiler. Selecting these parameters as input feature data can more comprehensively characterize the fuel combustion characteristics and core variables of the thermodynamic cycle, enabling the regression model to accurately quantify linear and nonlinear effects, significantly improving the prediction accuracy of thermal efficiency. The selected parameters are preprocessed by normalization, dimensionality reduction (for high-dimensional parameters), and outlier removal to obtain the input feature data.
[0037] S204. Input the input feature data into the first linear regression model and the second nonlinear regression model respectively to obtain a first linear prediction result and a second nonlinear prediction result. In an embodiment of the present invention, a linear model and a nonlinear model are used to predict thermal efficiency respectively, and the prediction results are fused. The first linear regression model adopts a multiple linear regression model, and the second nonlinear regression model adopts a support vector regression model. The first linear prediction result obtained based on the first linear regression model is a linear prediction value of thermal efficiency, and the second nonlinear prediction result obtained based on the second nonlinear regression model is a nonlinear prediction value of thermal efficiency. In an embodiment of the present invention, the prediction results of the multiple linear regression model and the support vector regression model are fused. The multiple linear regression model can explicitly quantify the linear influence of fuel components, while the support vector regression model can capture the nonlinear interaction between parameters. The prediction results of the linear prediction model and the nonlinear prediction model are combined so that the model can capture both linear and nonlinear relationships at the same time, thereby avoiding the prediction bias caused by a single model.
[0038] The first linear regression model and the second nonlinear regression model are trained separately using historical data. The input feature data used for training is consistent with the input feature data mentioned in step S202 above. The output is the thermal efficiency data in the historical data. The trained first linear regression model and the second nonlinear regression model are obtained. The trained first linear regression model and the second nonlinear regression model are used for prediction in step S204 above. For the multivariate linear regression model, the relationship between the output and input feature vectors can be expressed as:
[0039] η MLR =β0+β1x1+β2x2+…+β n x n
[0040] Among them, η MLR Represents the predicted value of thermal efficiency output by the multivariate linear regression model, β0…β n Represents the regression coefficient of the multiple linear regression model, x1…x n Represents the input feature data of the sample to be predicted. The multivariate linear regression model is trained and the regression coefficients are obtained by least squares fitting.
[0041] For the support vector regression model, the relationship between the output and input feature vectors can be expressed as:
[0042]
[0043] Among them, η SVR represents the predicted value of thermal efficiency output by the support vector regression model, K represents the kernel function of the support vector regression model, N represents the number of support vectors, and α iIndicates the error tolerance of the corresponding sample on the upper bound of the regression target, α i * represents the error tolerance of the corresponding sample in the lower bound of the regression target, b represents the bias term, x i The input feature vector corresponding to the i-th support vector is the key sample selected during training, and x represents the input feature vector of the sample to be predicted. The support vector regression model is trained and the kernel function parameters are optimized using grid search.
[0044] S206: Determine weight values for the first linear prediction result and the second nonlinear prediction result based on the fuel type. Obtain the current fuel type and a preset weight value-fuel type mapping relationship table, and search the preset weight value-fuel type mapping relationship table for a corresponding weight value based on the current fuel type.
[0045] S2061. Divide the historical data into several training sets and several validation sets based on fuel type, where each training set and validation set corresponds to historical data of a different fuel type. Categorize the historical data by fuel type, which may include, for example, natural gas, synthesis gas, and hydrogen-blended fuel.
[0046] S2062. Based on the historical data corresponding to each fuel type, the first linear regression model and the second nonlinear regression model are trained using the training set to obtain trained first linear regression models and second nonlinear regression models. Different sub-models are trained using the classified historical data, each sub-model corresponding to a different fuel type. For each fuel type, a trained first linear regression model and a trained second nonlinear regression model are obtained.
[0047] S2063. Input feature data from the validation set corresponding to the fuel type into the trained first linear regression model and the trained second nonlinear regression model, respectively, to obtain first linear prediction results and second nonlinear prediction results, respectively. For each fuel type, input feature data from the validation set corresponding to the fuel type into the trained first linear regression model and the trained second nonlinear regression model, respectively, to obtain first linear prediction results and second nonlinear prediction results for the fuel type.
[0048] S2064. Utilize the first linear prediction result, the second nonlinear prediction result, and the actual output value of the validation set to obtain a preset weight value-fuel type mapping relationship table by minimizing the mean square error of the prediction value. The mean square error of the prediction value can be expressed using the following objective function:
[0049]
[0050] Among them, η i represents the actual output value of the validation set, ω MLRRepresents the weight value of the first linear prediction result, ω SVR represents the weight value of the second linear prediction result, η MLR,i represents the first linear prediction result, η SVR,i Represents the second linear prediction result, and N represents the number of samples in the validation set.
[0051] Based on the constraints, the mean square error of the predicted value, i.e., the above-mentioned objective function, is minimized to obtain the optimal weight value corresponding to the fuel type, which is the optimal weight value of the first linear prediction result and the optimal weight value of the second linear prediction result; wherein, the constraints include: ω MLR +ω SVR =1, and the weight value is non-negative. The mean square error of the predicted value can be optimized by using a grid search method or a numerical optimization method (such as a gradient descent method). For example, in the grid search method, the weight value ω of the first linear prediction result can be traversed. MLR From the candidate values 0 to 1 (step size 0.01), calculate the corresponding MSE and select the weight value corresponding to the minimum MSE. For example, in the gradient descent method, the constraint condition ω can be SVR =1-ω MLR Substitute into the above objective function, thereby transforming the above optimization problem into a single variable unconstrained optimization problem, and iteratively solve the weight value ω of the optimal first linear prediction result MLR Record the weight combination (ω) that minimizes the MSE of the validation set for each fuel type. MLR ,ω SVR ), thereby obtaining a preset weight value-fuel type mapping relationship table.
[0052] According to some optional embodiments, the method further comprises the steps of:
[0053] S207. Dynamically adjust the weight value based on the confidence of the first linear regression model and the second nonlinear regression model. For example, if the confidence of the first linear regression model is high (for example, the confidence increases relative to the previous time period) within a predetermined time period, the weight value of the first linear regression model is increased, and the weight value of the second nonlinear regression model is correspondingly reduced; if the confidence of the second nonlinear regression model is high (for example, the confidence increases relative to the previous time period) within a predetermined time period, the weight value of the second nonlinear regression model is increased, and the weight value of the first linear regression model is correspondingly reduced. The weight value determined in the above steps can be further adjusted according to the confidence calculation result. In this embodiment of the present invention, the confidence of the first linear regression model and the second nonlinear regression model are respectively evaluated by the prediction error of the sliding window (for example, the past k time steps). The confidence can be determined according to the following formula:
[0054]
[0055] Wherein, λ represents an attenuation factor (usually λ>0, and 0.1 in the embodiment of the present invention), which is used to control the sensitivity of the error to the confidence level; MSE[tk:t] represents the mean square error of the regression model in the past k time steps.
[0056] S208: Obtain a regression analysis result based on the first linear prediction result, the second nonlinear prediction result, and the weight value. According to the determined weight value, the first linear prediction result and the second nonlinear prediction result are weighted and combined to obtain a final regression analysis result. For example, the weight value ω of the first linear prediction result is determined according to the above steps. MLR , the weight value ω of the second linear prediction result SVR , then the regression analysis results can be expressed as:
[0057] η=ω MLR ·η MLR +ω SVR ·η SVR
[0058] Among them, η represents the final regression analysis result, η MLR represents the first linear prediction result, i.e. the thermal efficiency prediction value output by the multiple linear regression model, η SVR represents the second linear prediction result, i.e., the predicted thermal efficiency value output by the support vector regression model. Based on the regression analysis results obtained above, closed-loop control of the gas turbine combined cycle system can be implemented. For example, if the predicted thermal efficiency value is lower than the target thermal efficiency value, the system can be adjusted by adjusting the gas turbine air-fuel ratio or steam turbine parameters.
[0059] In summary, an embodiment of the present invention relates to a method for regression analysis of fuel characteristics and thermal efficiency of a gas turbine combined cycle system. The method comprises the following steps: obtaining input characteristic data of the gas turbine combined cycle system, the input characteristic data including fuel characteristic parameters and system operating parameters; inputting the input characteristic data into a first linear regression model and a second nonlinear regression model, respectively, to obtain a first linear prediction result and a second nonlinear prediction result; determining weight values for the first linear prediction result and the second nonlinear prediction result based on the fuel type; and obtaining a regression analysis result based on the first linear prediction result, the second nonlinear prediction result, and the weight values. The technical solution of the embodiment of the present invention uses a linear model and a nonlinear model to predict thermal efficiency respectively, and combines the prediction results of the two models through adaptively adjusted weight values, thereby greatly improving the efficiency and accuracy of the regression analysis of the gas turbine combined cycle system.
[0060] It should be understood that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the present invention (including the claims) is limited to these examples; under the thinking of the present invention, the technical features in the above embodiments or different embodiments may also be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present invention as described above, which are not provided in detail for the sake of simplicity. The above specific embodiments of the present invention are merely used to illustrate or explain the principles of the present invention and do not constitute a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present invention should be included in the scope of protection of the present invention. In addition, the claims appended to the present invention are intended to cover all changes and modifications that fall within the scope and boundaries of the appended claims, or the equivalent forms of such scope and boundaries.
Claims
1. A method for regression analysis of fuel characteristics and thermal efficiency of a gas turbine combined cycle system, characterized in that: The method comprises the steps of: Acquiring input characteristic data of a gas turbine combined cycle system, wherein the input characteristic data includes fuel characteristic parameters and system operating parameters; Inputting the input feature data into a first linear regression model and a second nonlinear regression model respectively to obtain a first linear prediction result and a second nonlinear prediction result; determining weight values of the first linear prediction result and the second nonlinear prediction result according to the fuel type; A regression analysis result is obtained based on the first linear prediction result, the second nonlinear prediction result and the weight value.
2. The method according to claim 1, characterized in that The first linear regression model includes a multiple linear regression model, and the second nonlinear regression model includes a support vector regression model.
3. The method according to claim 2, characterized in that The first linear prediction result includes a linear prediction value of thermal efficiency, and the second nonlinear prediction result includes a nonlinear prediction value of thermal efficiency.
4. The method according to any one of claims 1 to 3, characterized in that The fuel characteristic parameters include methane volume fraction, carbon dioxide volume fraction and fuel calorific value; the system operation parameters include gas turbine inlet temperature and gas turbine exhaust temperature.
5. The method according to claim 4, characterized in that Determining weight values of a first linear prediction result and a second nonlinear prediction result according to the fuel type includes: Obtain the current fuel type and the preset weight value-fuel type mapping relationship table; Based on the current fuel type, the corresponding weight value is searched in the preset weight value-fuel type mapping relationship table.
6. The method according to claim 5, characterized in that The preset weight value-fuel type mapping relationship table is obtained by the following steps: The historical data is divided into several training sets and several validation sets based on the fuel type. Each set of training sets and validation sets corresponds to the historical data of different fuel types. For each fuel type, the historical data corresponding to the first linear regression model and the second nonlinear regression model are trained using the training set to obtain the trained first linear regression model and the second nonlinear regression model; Inputting the input feature data of the validation set corresponding to the fuel type into the trained first linear regression model and the second nonlinear regression model respectively to obtain a first linear prediction result and a second nonlinear prediction result respectively; The first linear prediction result, the second nonlinear prediction result and the actual output value of the validation set are used to obtain a preset weight value-fuel type mapping relationship table by minimizing the mean square error of the prediction value.
7. The method according to claim 6, characterized in that The mean square error of the predicted value is expressed as: Among them, η i represents the actual output value of the validation set, ω MLR Represents the weight value of the first linear prediction result, ω SVR represents the weight value of the second linear prediction result, η MLR,i represents the first linear prediction result, η SVR,i Represents the second linear prediction result, and N represents the number of samples in the validation set.
8. The method according to claim 7, characterized in that By minimizing the mean square error of the predicted value, a preset weight value-fuel type mapping relationship table is obtained, including: Based on the constraints, the mean square error of the predicted values is minimized to obtain the optimal weight value corresponding to the fuel type, wherein the optimal weight value includes the optimal weight value of the first linear prediction result and the optimal weight value of the second linear prediction result; The constraints include: MLR +ω SVR =1, and the weight value is non-negative.
9. The method according to claim 8, characterized in that The method further comprises: The weight value is dynamically adjusted based on the confidence of the first linear regression model and the second nonlinear regression model. If the confidence of the first linear regression model is high, the weight value of the first linear regression model is increased; if the confidence of the second nonlinear regression model is high, the weight value of the second nonlinear regression model is increased.
10. The method according to claim 9, characterized in that The confidence level is determined according to the following formula: Where λ represents the decay factor and MSE[tk:t] represents the mean square error of the regression model in the past k time steps.