Method for establishing a steam turbine associated model

Abstract

The present application relates to a kind of steam turbine associated model establishing method, comprising: 1) construct traditional steam turbine model;2) inject the input data set A in existing steam turbine operation data, obtain output data set B;3) output data set B and existing steam turbine output data set C are compared, whether need to judge according to gap delta whether traditional steam turbine model is modified, if yes, then after modification, step 4) is carried out;If no, then directly step 4) is carried out;4) to traditional steam turbine model or the input parameter data set D in modified traditional steam turbine model, obtain output parameter data set E;5) using the input parameter data set A in existing steam turbine operation data, existing steam turbine output parameter data set C, input parameter data set D and output parameter data set E are trained in steam turbine machine learning model, finally obtain steam turbine associated model.The present application is faster in the same precision, more wide in adaptive working condition range.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of safe operation of steam turbines, and relates to a method for establishing a steam turbine companion model, in particular to a method for establishing a steam turbine companion model for widening the working condition range of a steam turbine simulation model and improving the operation speed. BACKGROUND

[0002] In the operation of steam turbines in power plants, the steam turbine needs to be detected to evaluate the performance of the steam turbine, provide performance optimization direction, improve the variable working condition performance of the steam turbine, and improve the unit efficiency; and provide data for the maintenance and repair of the steam turbine. However, in the actual operation of the steam turbine, many key parameter points cannot be arranged with sensors or the arrangement of sensors will cause power and efficiency loss, or because the region where a fault may occur is not arranged with sensors in advance, the parameters cannot be directly obtained. At this time, the steam turbine model can provide fine analysis data for the optimization, maintenance and repair of the steam turbine, and provide early warning for the failure of the steam turbine.

[0003] There are three types of existing steam turbine companion simulation technologies.

[0004] One type is a technology based on a first principle model. That is, a physical simulation model established based on energy conservation equation, mass conservation equation and momentum conservation equation. The advantages are that less data is needed, only the main parameters of the steam turbine are needed; and the working condition range is wide. The disadvantages are that the accuracy is poor; the model runs slowly, and it is difficult to run synchronously with the steam turbine when time-varying parameters are involved, and it is difficult to achieve all the main purposes of the companion model.

[0005] The second type is a technology mainly based on empirical formula, which is often used in combination with the first principle technology. Compared with the pure first principle technology, the advantage is that the model runs faster than the first principle technology, but the working condition range is narrower. The first two types are generally referred to as traditional models.

[0006] The third type is a machine learning technology based on big data training, such as neural network. That is, using existing operation data to simulate and predict the input and output by machine learning. The advantages are that the running speed is fast after training, and it can still run synchronously with the steam turbine when time-varying parameters are involved, and it can be used as a companion model within the existing data range. The disadvantages are that a large amount of existing data is needed to use; and it can only simulate within the existing working condition data range, and the working condition range is narrow. SUMMARY

[0007] In order to solve the above technical problems in the background art, the present application provides a steam turbine companion model establishing method which has faster running speed, wider working condition range and can be used under existing data working condition at the same accuracy.

[0008] In order to achieve the above object, the present application adopts the following technical solutions:

[0009] A steam turbine associated model establishing method, characterized in that the steam turbine associated model establishing method comprises the following steps:

[0010] 1) Construct a traditional steam turbine model according to the design parameters of the steam turbine;

[0011] 2) Inject the input parameter data set A in the existing steam turbine operation data into the traditional steam turbine model constructed in step 1) to obtain an output parameter data set B;

[0012] 3) Compare the output parameter data set B obtained in step 2) with the existing steam turbine output parameter data set C corresponding to the input parameter data set A in the existing steam turbine operation data, and judge whether the traditional steam turbine model needs to be modified according to the gap δ between the output parameter data set B and the existing steam turbine output parameter data set C, if the traditional steam turbine model needs to be modified, modify the traditional steam turbine model to obtain a modified traditional steam turbine model and proceed to step 4); if the traditional steam turbine model does not need to be modified, use the traditional steam turbine model and directly proceed to step 4);

[0013] 4) Inject an input parameter data set D into the traditional steam turbine model or the modified traditional steam turbine model, and obtain an output parameter data set E after the traditional steam turbine model or the modified traditional steam turbine model; the input parameter data set D has a wider working condition than the working condition corresponding to the input parameter data set A in the existing steam turbine operation data and the existing steam turbine output parameter data set C;

[0014] 5) Use the input parameter data set A in the existing steam turbine operation data, the existing steam turbine output parameter data set C, the input parameter data set D and the output parameter data set E obtained in step 4) to train the steam turbine machine learning model, and finally obtain the steam turbine associated model.

[0015] The specific implementation of step 1) is to establish a traditional steam turbine model containing time-varying parameters or not containing time-varying parameters according to the design parameters of the steam turbine and the first principle.

[0016] The input parameter data set A in the existing steam turbine operation data is a direct measurement result or a parameter calculated from a direct measurement result; the direct measurement result is the steam turbine inlet flow, the steam turbine inlet pressure and / or temperature, the steam turbine outlet flow, the steam turbine outlet pressure and / or temperature, the flow at the steam turbine monitoring point and the pressure and / or temperature at the steam turbine monitoring point; the parameter calculated from the direct measurement result is enthalpy, entropy and / or dryness.

[0017] The output parameter dataset B is a directly measured result or a parameter calculated from a directly measured result; the directly measured result is pressure, temperature and / or flow rate at a certain measuring point; the parameter calculated from a directly measured result is enthalpy, entropy and / or dryness.

[0018] The specific way of the correction in step 3) is a correction coefficient method, a linear regression method and / or a polynomial fitting method.

[0019] The input parameter dataset D includes input parameters of the steam turbine under a fault condition and a failure condition.

[0020] The training method in step 5) is a supervised learning method, an unsupervised learning method, reinforcement learning and / or a semi-supervised learning method.

[0021] The advantages of the present application are:

[0022] The application discloses a steam turbine associated model establishing method, comprising the following steps: 1) constructing a traditional steam turbine model according to steam turbine design parameters; 2) injecting an input parameter data set A in existing steam turbine operation data into the traditional steam turbine model obtained in step 1) to obtain an output parameter data set B; 3) comparing the output parameter data set B obtained in step 2) with an existing steam turbine output parameter data set C corresponding to the input parameter data set A in the existing steam turbine operation data, and judging whether the traditional steam turbine model needs to be modified according to the difference delta between the output parameter data set B and the existing steam turbine output parameter data set C; if the traditional steam turbine model needs to be modified, the traditional steam turbine model is modified to obtain a modified traditional steam turbine model, and step 4) is performed simultaneously; if the traditional steam turbine model does not need to be modified, the traditional steam turbine model is used and step 4) is directly performed; 4) injecting an input parameter data set D into the traditional steam turbine model or the modified traditional steam turbine model, and obtaining an output parameter data set E after the traditional steam turbine model or the modified traditional steam turbine model; the input parameter data set D has a wider working condition than the input parameter data set A in the existing steam turbine operation data and the working condition corresponding to the existing steam turbine output parameter data set C; 5) training in a steam turbine machine learning model by using the input parameter data set A in the existing steam turbine operation data, the existing steam turbine output parameter data set C, the input parameter data set D and the output parameter data set E obtained in step 4), and finally obtaining a steam turbine associated model. The application is based on the first principle or the empirical formula to establish the traditional steam turbine model, and then the model is modified in combination with the existing input and output parameters of the steam turbine. The wider range of parameters is input into the modified traditional steam turbine model to obtain the corresponding output data. The steam turbine model input and output data and the existing real operation data of the steam turbine are used to establish the steam turbine associated model based on the machine learning algorithm. In the same high precision, the method can reduce the dependence of the model on the quality of the data, improve the running speed of the model, and widen the working condition range that the model can adapt to. Compared with the traditional model method, the precision of the method is high after training, the running speed is fast, and the method can still run with the steam turbine in the case of time-varying parameters and can be used as an associated model within the existing data range. Compared with the general machine learning model method, the application needs less data and can adapt to the working conditions outside the existing data range. In general, in the same high precision, the method can reduce the dependence of the model on the quality of the data, needs less data, can improve the running speed of the model, helps to establish an associated model, can widen the working condition range that the model can adapt to, and can be used outside the working condition range of the existing data. The application overcomes the problems of the existing technologies, such as poor precision, slow model running speed, the need of a large amount of existing data for use, narrow working condition range, and simulation within the existing working condition data range. The application combines the advantages of the existing technologies and avoids the main shortcomings.It achieves faster operating speed while maintaining the same level of accuracy; it is adaptable to a wide range of operating conditions and can be used outside the range of existing data. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the method for establishing a steam turbine associated model provided by the present invention.

[0024] Figure 2 This is a schematic diagram of data transmission in the method for establishing a steam turbine associated model provided by the present invention. Detailed Implementation

[0025] See Figure 1 as well as Figure 2 This invention provides a method for establishing a steam turbine associated model, which can broaden the operating condition range of the steam turbine simulation model and improve the operating speed. Specifically, the method includes the following steps:

[0026] 1) Construct a traditional steam turbine model based on the steam turbine design parameters. The traditional steam turbine model is a traditional steam turbine model with or without time-varying parameters, established based on the steam turbine design parameters and first principles.

[0027] 2) Input the existing turbine operating data dataset A into the traditional turbine model constructed in step 1) to obtain the output parameter dataset B. For example, the existing turbine operating data dataset A consists of directly measured results or parameters calculated from directly measured results. The directly measured results include turbine inlet flow rate, turbine inlet pressure and / or temperature, turbine outlet flow rate, turbine outlet pressure and / or temperature, flow rate at turbine monitoring points, and pressure and / or temperature at turbine monitoring points. The parameters calculated from the directly measured results are common parameters such as enthalpy, entropy, and / or dryness fraction. For example, the output parameter dataset B can also be the results of direct measurements or parameters calculated from directly measured results. The directly measured results include pressure, temperature, and / or flow rate at a certain measuring point. The parameters calculated from the directly measured results are common parameters such as enthalpy, entropy, and / or dryness fraction.

[0028] 3) comparing the output parameter data set B obtained in step 2) with the existing steam turbine output parameter data set C corresponding to the input parameter data set A in the existing steam turbine operation data, and determining whether the traditional steam turbine model needs to be corrected according to the gap δ between the output parameter data set B and the existing steam turbine output parameter data set C; if the gap δ is greater than a standard (the standard is determined according to the error of the to-be-shaped steam turbine companion model, and is generally not greater than the error required by the to-be-shaped steam turbine companion model), the traditional steam turbine model is corrected by using a correction coefficient method, a linear regression method, a polynomial fitting method, etc., until δ meets the standard, and the corrected traditional steam turbine model is obtained, and the corrected traditional steam turbine model is used as the basis for the subsequent process. If the gap δ is less than or equal to the standard, the traditional steam turbine model is directly used as the basis for the subsequent process.

[0029] 4) inputting an input parameter data set D into the traditional steam turbine model or the corrected traditional steam turbine model, and obtaining an output parameter data set E after the traditional steam turbine model or the corrected traditional steam turbine model; the input parameter data set D has a wider working condition than the working condition corresponding to the input parameter data set A in the existing steam turbine operation data and the existing steam turbine output parameter data set C. Exemplarily, the input parameter data set D includes, but is not limited to, steam turbine input parameters in a steam turbine fault working condition and a failure working condition.

[0030] 5) using the input parameter data set A in the existing steam turbine operation data, the existing steam turbine output parameter data set C, the input parameter data set D, and the output parameter data set E obtained in step 4) to train the steam turbine machine learning model, and finally obtaining the steam turbine companion model. The training method is a supervised learning method, an unsupervised learning method, reinforcement learning, and / or a semi-supervised learning method.

[0031] The application can reduce the dependence of the model on data quality, improve the running speed of the model, and broaden the working condition range of the model under the same high precision. Compared with the traditional model method, the method has high precision and fast running speed after training, and can still run simultaneously with the steam turbine in the case of time-varying parameters, and can be used as a companion model within the existing data range. Compared with the general machine learning model method, the application requires less data and can adapt to working conditions outside the existing data range. In general, under the same high precision, the method can reduce the dependence of the model on data quality, requires less data, can improve the running speed of the model, help to establish a companion model, and can broaden the working condition range of the model and be used outside the working condition range of the existing data. The application overcomes the problems of the existing technology, such as poor accuracy, slow model running speed, the need for a large amount of existing data to use, narrow working condition range, and the ability to simulate only within the existing working condition data range. The application combines the advantages of existing technologies while avoiding major drawbacks. Under the same precision, the application achieves faster running speed and a wide working condition range, and can be used outside the working condition range of the existing data.

Claims

1. A method of establishing a model of a steam turbine, characterized by: The steam turbine associated model establishment method comprises the following steps: 1) constructing a traditional steam turbine model according to the design parameters of the steam turbine; specifically, constructing a traditional steam turbine model containing time-varying parameters or not containing time-varying parameters according to the design parameters of the steam turbine and first principles; 2) injecting the input parameter data set A in the existing steam turbine operation data into the traditional steam turbine model constructed in step 1) to obtain an output parameter data set B; 3) comparing the output parameter data set B obtained in step 2) with the existing steam turbine output parameter data set C corresponding to the input parameter data set A in the existing steam turbine operation data, and determining whether the traditional steam turbine model needs to be corrected according to the difference δ between the output parameter data set B and the existing steam turbine output parameter data set C: If correction is needed, the traditional steam turbine model is corrected to obtain a corrected traditional steam turbine model, and step 4) is performed at the same time; If correction is not needed, the traditional steam turbine model is used and step 4) is directly performed; 4) injecting an input parameter data set D into the traditional steam turbine model or the corrected traditional steam turbine model, and obtaining an output parameter data set E after passing through the traditional steam turbine model or the corrected traditional steam turbine model; the input parameter data set D has a wider working condition than the working condition corresponding to the input parameter data set A in the existing steam turbine operation data and the existing steam turbine output parameter data set C; 5) training in a steam turbine machine learning model using the input parameter data set A in the existing steam turbine operation data, the existing steam turbine output parameter data set C, the input parameter data set D, and the output parameter data set E obtained in step 4), and finally obtaining a steam turbine associated model.

2. The gas turbine companion model building method according to claim 1, characterized by: In step 3), whether the traditional steam turbine model needs to be corrected is determined according to the difference δ between the output parameter data set B and the existing steam turbine output parameter data set C, specifically: If the difference δ is greater than a standard, the traditional steam turbine model is corrected until δ meets the standard, a corrected traditional steam turbine model is obtained, and the corrected traditional steam turbine model is used as the basis for the subsequent process; If the difference δ is less than or equal to the standard, the traditional steam turbine model is not corrected.

3. The method of claim 1, wherein: The input parameter data set A in the existing steam turbine operation data is a direct measurement result; the direct measurement result is the steam turbine inlet flow, the steam turbine inlet pressure and / or temperature, the steam turbine outlet flow, the steam turbine outlet pressure and / or temperature, the flow at the steam turbine monitoring point, and the pressure and / or temperature at the steam turbine monitoring point.

4. The gas turbine companion model development method of claim 1, wherein: The input parameter data set A in the existing steam turbine operation data is a parameter calculated from a direct measurement result; the direct measurement result is the steam turbine inlet flow, the steam turbine inlet pressure and / or temperature, the steam turbine outlet flow, the steam turbine outlet pressure and / or temperature, the flow at the steam turbine monitoring point, and the pressure and / or temperature at the steam turbine monitoring point; the parameter calculated from the direct measurement result is enthalpy, entropy, and / or dryness.

5. A gas turbine companion model building method according to claim 1 or 2 or 3 or 4, characterised in that: The specific way of correction in step 3) is the correction coefficient method, the linear regression method, and / or the polynomial fitting method.

6. The gas turbine companion model development method according to claim 5, characterized by: The input parameter data set D comprises input parameters of the steam turbine in the event of a fault condition and in the event of a failure condition.

7. The gas turbine companion model building method according to claim 6, characterized by: The manner in which the step 5) is trained is a supervised learning method, an unsupervised learning method, reinforcement learning and / or a semi-supervised learning method.

Images