Monitoring method for wide-load high-control-valve flow of photo-thermal steam turbine

Abstract

The invention discloses a monitoring method for wide-load high-control-valve flow of a photo-thermal steam turbine. The monitoring method specifically comprises the steps that 1, the operation mode of the photo-thermal steam turbine is set; 2, flow characteristic testing is conducted on each control valve of the photo-thermal steam turbine in a single-valve mode and a downstream-valve mode; 3, collecting the flow characteristics of each control valve in a single valve mode and the flow characteristics of each control valve in a sequential valve mode by taking the time interval as n, and taking the collected data as training data; 4, establishing a valve flow characteristic curve prediction model; 5, training a valve flow characteristic curve prediction model; step 6, obtaining a predicted flow characteristic curve; and 7, calculating the deviation between the predicted flow characteristic curve and the ideal flow characteristic curve of the corresponding control valve, and optimizing the flow of the corresponding control valve according to the deviation. By means of real-time evaluation and circulation optimization, adjustment is carried out when flow characteristics deviate, energy distribution is regulated and controlled in time, and steam heat energy loss is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of steam turbine flow management methods, and relates to a method for monitoring the wide-load high-valve flow of a solar thermal steam turbine. Background Art

[0002] The operation of a CSP steam turbine relies on a concentrator to collect solar energy into a heat collector, which then heats the heat-collecting medium to complete the conversion. Ultimately, the heat exchanger provides steam, which in turn drives the steam turbine to generate electricity. However, the operation of CSP steam turbines is susceptible to various factors, such as diurnal variations, climate change, and rain and snow. Therefore, to improve turbine efficiency, CSP steam turbines must be able to quickly start and stop, rapidly change loads, operate continuously at low loads, and operate efficiently.

[0003] Conventional steam turbines typically use single-valve control when the load is below 50%, and sequential valve control when the load exceeds 50%. In single-valve mode, all regulating valves open or close simultaneously, allowing steam to flow evenly into the turbine through multiple valves, with each valve maintaining a consistent opening. In sequential valve mode, the valves open or close one by one in sequence, with each valve individually adjusted within a specific load range, while the other valves remain fully open or fully closed. Both control modes have their own advantages and disadvantages. However, simply switching between single and sequential valves is no longer sufficient to meet the unique operating requirements of solar thermal steam turbines.

[0004] The flow characteristic of a steam turbine valve, which refers to the relationship between the valve opening and the steam flow through it, exhibits typical nonlinear characteristics. In particular, the flow characteristics of a high-pressure regulating valve vary significantly under varying loads, directly impacting the unit's thermal economy and operating efficiency. Due to the dynamic nature of steam turbine valve flow characteristics, a single optimization test cannot consistently maintain ideal valve flow characteristics. Therefore, to accommodate the wide load range of CSP steam turbines, a high-pressure regulating valve flow management method that can adapt to these special requirements is needed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for monitoring the wide-load high-throttle flow of a solar thermal steam turbine. By using real-time evaluation and cycle optimization means, adjustments are made when the flow characteristics deviate, thereby achieving timely regulation of energy distribution and reducing steam heat energy loss.

[0006] The technical solution adopted by the present invention is a method for monitoring the wide-load high-speed valve flow of a solar thermal steam turbine, which is specifically implemented according to the following steps:

[0007] Step 1: Set the operation mode of the solar thermal turbine;

[0008] Step 2: Test the flow characteristics of each regulating valve of the CSP steam turbine in single valve mode and sequential valve mode respectively;

[0009] Step 3: At a time interval of n, the flow characteristics of each throttle valve in the single valve mode and the flow characteristics of each throttle valve in the sequential valve mode are collected, and the collected data are used as training data;

[0010] Step 4: Establish a valve flow characteristic curve prediction model;

[0011] Step 5: Using the training data to train the valve flow characteristic curve prediction model established in step 4;

[0012] Step 6: Input the operating data of the corresponding valve during the operation of the CSP steam turbine unit into the trained valve flow characteristic curve prediction model to obtain a predicted flow characteristic curve;

[0013] Step 7: Calculate the deviation between the predicted flow characteristic curve and the ideal flow characteristic curve of the corresponding throttle, and optimize the flow of the corresponding throttle according to the deviation.

[0014] Preferably, in step 1, the operation mode of the solar thermal steam turbine is divided into a start-stop mode, a low-load long-term operation mode, a rapid adjustment mode, and a high-load mode;

[0015] In the start-stop mode, the CSP steam turbine is in the start-stop state. In this mode, the valve is in single-valve mode, and the four regulating valves GV1, GV2, GV3, and GV4 maintain the same opening;

[0016] In the low-load long-term operation mode, the load of the CSP steam turbine is 15%-25%Pe, where Pe is the rated load of the CSP steam turbine. In this mode, the valve is in the sequential valve mode, that is, two regulating valves are opened and two regulating valves are closed;

[0017] In the fast regulation mode, when the load of the CSP steam turbine is below 50%Pe, the single valve mode is adopted; when the load of the CSP steam turbine is above 50%Pe, including 50%Pe, the sequential valve mode is adopted, and the regulating valves are opened or closed one by one during the mode conversion;

[0018] In high load mode, that is, when the load of the solar thermal steam turbine is greater than 50% Pe, the valve sequence mode is adopted.

[0019] Preferably, in step 2, the flow characteristic test of each regulating valve of the CST turbine in the single valve mode is performed as follows:

[0020] Step 2.1: When the load of the CSP steam turbine is stable, start the single valve operation and maintain the main steam pressure and temperature stable;

[0021] Step 2.2: Gradually close GV1 by 3% opening through forced output. Every time GV1 is closed by 3%, the corresponding data is recorded, including valve command, flow command, valve position opening, regulating stage pressure, main steam pressure, actual load, and high exhaust pressure, until GV1 is completely closed.

[0022] Step 2.3, then gradually open GV1 by 3% opening, and record the data every 3% opening until GV1 is fully open. After GV1 is fully open and stable, gradually close it to the position before the test and release the forced output. At this point, the GV1 test is completed;

[0023] Step 2.4: After completing the flow characteristic test of GV1, the operator adjusts the unit to the pre-test parameters, that is, the stable operation state of step 2.1, and then continues the flow characteristic test of GV2, GV3, and GV4 according to steps 2.2-2.2;

[0024] In step 2.5, based on the data collected by GV1, GV2, GV3, and GV4 during the experiment, the corresponding equivalent actual flow rate is calculated according to the corresponding data each time, and then the multiple equivalent actual flow rates of each throttle valve form the ideal characteristic curve of the throttle valve in single valve mode.

[0025] Preferably, in step 2, the flow characteristic test of each regulating valve of the CST turbine in the sequential valve mode is specifically performed as follows:

[0026] Start the unit sequence valve operation mode, put into power control, keep the main steam pressure setting value unchanged, reduce the unit load from 90% Pe at a rate of 3MW / min, and close the four regulating valves by setting the target load until the unit load drops to 40% Pe. End the test, collect the valve position opening, main steam pressure, and regulating stage pressure of GV1, GV2, GV3, and GV4 during the experiment, calculate the corresponding equivalent actual flow rate based on the corresponding data each time, and then form the ideal characteristic curve of each regulating valve in the sequence valve mode based on multiple equivalent actual flow rates.

[0027] Preferably, step 3 is specifically:

[0028] When testing the flow characteristics of each throttle valve of a CSP steam turbine in the sequential valve mode, the main steam flow, main steam pressure, regulating stage pressure, throttle valve opening, main steam temperature, and actual unit load are collected at 5-second intervals, and the corresponding equivalent actual flow is calculated to form a training sample data set.

[0029] When testing the flow characteristics of each throttle valve of the solar thermal steam turbine in the sequential valve mode, the main steam flow, main steam pressure, regulating stage pressure, throttle valve opening, main steam temperature, and actual load of the unit are collected at a time interval of 5 seconds, and the corresponding equivalent actual flow is calculated to form a training sample data.

[0030] Preferably, in step 3, all the collected training sample data are annotated, and then the annotated training sample data are annotated. Each training sample data x t = {main steam flow, main steam pressure, regulating stage pressure, regulating valve opening, main steam temperature, actual load of the unit, equivalent actual flow}, t = 1-N, N is the total training sample data, N is a multiple of 30;

[0031] The training sample data is used to generate sequence data x=[x t ,x t+1 ,…,X t+29 ], that is, each sequence includes 30 sets of training sample data.

[0032] Preferably, step 4 of establishing a valve flow characteristic curve prediction model includes sequentially connecting a KAN network model and an LSTM model;

[0033] The KAN network model consists of three KAN layers connected in sequence, with 4, 6, and 8 nodes in each layer respectively;

[0034] The input of the LSTM model is the output of the last KAN layer. The LSTM model is set to three LSTM layers and one linear layer connected in sequence. The number of nodes in each of the three LSTM layers is 32, 64, and 128, respectively. The input of the first LSTM layer is the output of the last KAN layer.

[0035] Preferably, step 5 is specifically:

[0036] Set the loss function to Huber Loss, optimizer to Adam, learning rate to 0.001, batch size to 32, epochs to 100, and then start training. Specifically:

[0037] Input the sample data X into the KAN network model, and the output of the KAN network model is KAN out ;

[0038] Will KAN out In the input LSTM model, the final linear layer converts the output of the last LSTM layer into equivalent actual traffic.

[0039] Preferably, step 6 is specifically:

[0040] The CSP steam turbine unit operates according to the operating mode divided in step 1, and monitors the control and operating parameters of each throttle in real time. The main steam flow, main steam pressure, regulating stage pressure, throttle opening, and main steam temperature data of the corresponding throttle in single valve or sequential valve mode are collected at certain time intervals and input into the valve flow characteristic curve prediction model trained in step 5 to obtain the corresponding predicted flow characteristic curve.

[0041] Preferably, step 7 specifically uses the determination coefficient to calculate the deviation R between the corresponding predicted flow characteristic curve and the ideal characteristic curve of the corresponding throttle valve in single valve or sequential valve mode generated in step 2. 2 , if the coefficient of determination R 2 If the deviation is greater than or equal to 0.98, it is considered that there is no deviation between the predicted flow characteristic curve and the ideal characteristic curve, that is, in the corresponding mode, the corresponding valve flow does not need to be optimized;

[0042] If the coefficient of determination R 2 If the deviation is less than 0.98, it is considered that the corresponding valve flow characteristic curve of the CST turbine deviates from the ideal flow characteristic curve in the corresponding mode. Then the valve flow in the corresponding mode is optimized, and the corresponding predicted flow characteristic curve after optimization is obtained according to the method in step 6. Then, the deviation R between the optimized and ideal characteristic curves is calculated. 2 , until R 2 Satisfies the condition of being greater than or equal to 0.98.

[0043] Coefficient of determination R 2 The specific calculation is based on the following formula:

[0044]

[0045] Among them, y A,i is the value of the ordinate in the ideal flow characteristic curve when the abscissa is i, y B,i is the value of the ordinate in the predicted flow characteristic curve when the abscissa is i, For all y A,i The average value of , n is the number of values.

[0046] The beneficial effects of the present invention are:

[0047] The present invention switches the single valve or sequential valve mode in real time under different load conditions, thereby taking into account the safety, response speed and efficiency of the steam turbine.

[0048] The present invention utilizes real-time evaluation and cycle optimization to monitor flow rates. When flow characteristics deviate, valve control logic can be quickly adjusted to achieve refined regulation of energy distribution, reduce steam heat energy loss, and improve overall thermal economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a comparison diagram of the flow characteristic curves in the single valve mode in the method for monitoring the wide-load high-valve flow of a solar thermal steam turbine according to the present invention;

[0050] Figure 2 This is a comparison chart of flow characteristic curves in the forward valve mode in the method for monitoring the wide-load high-valve flow of a solar thermal steam turbine. DETAILED DESCRIPTION

[0051] The following describes it in detail with reference to specific implementation methods.

[0052] Example 1

[0053] The present invention is used for monitoring the wide load high-speed valve flow of a solar thermal steam turbine, and is specifically implemented according to the following steps:

[0054] Step 1: Set the operation mode of the solar thermal turbine;

[0055] Step 2: Test the flow characteristics of each regulating valve of the CSP steam turbine in single valve mode and sequential valve mode respectively;

[0056] Step 3: At a time interval of n, the flow characteristics of each throttle valve in the single valve mode and the flow characteristics of each throttle valve in the sequential valve mode are collected, and the collected data are used as training data;

[0057] Step 4: Establish a valve flow characteristic curve prediction model;

[0058] Step 5: Using the training data to train the valve flow characteristic curve prediction model established in step 4;

[0059] Step 6: Input the operating data of the corresponding valve during the operation of the CSP steam turbine unit into the trained valve flow characteristic curve prediction model to obtain a predicted flow characteristic curve;

[0060] Step 7: Calculate the deviation between the predicted flow characteristic curve and the ideal flow characteristic curve of the corresponding throttle, and optimize the flow of the corresponding throttle according to the deviation.

[0061] Example 2

[0062] Based on Example 1, in step 1, the operation mode of the solar thermal steam turbine is divided into a start-stop mode, a low-load long-term operation mode, a fast adjustment mode, and a high-load mode;

[0063] In the start-stop mode, the CSP steam turbine is in the start-stop state. In this mode, the valve is in single-valve mode, and the four regulating valves GV1, GV2, GV3, and GV4 maintain the same opening;

[0064] In the low-load long-term operation mode, the load of the CSP steam turbine is 15%-25%Pe, where Pe is the rated load of the CSP steam turbine. In this mode, the valve is in the sequential valve mode, that is, two regulating valves are opened and two regulating valves are closed;

[0065] In the fast regulation mode, when the load of the CSP steam turbine is below 50%Pe, the single valve mode is adopted; when the load of the CSP steam turbine is above 50%Pe, including 50%Pe, the sequential valve mode is adopted, and the regulating valves are opened or closed one by one during the mode conversion;

[0066] In high load mode, that is, when the load of the solar thermal steam turbine is greater than 50% Pe, the valve sequence mode is adopted.

[0067] Example 3

[0068] Based on Example 2, the flow characteristic test of each throttle valve of the CST in single valve mode in step 2 is specifically as follows:

[0069] Step 2.1: When the load of the CSP steam turbine is stable, start the single valve operation and maintain the main steam pressure and temperature stable;

[0070] Step 2.2: Gradually close GV1 by 3% opening through forced output. Every time GV1 is closed by 3%, the corresponding data is recorded, including valve command, flow command, valve position opening, regulating stage pressure, main steam pressure, actual load, and high exhaust pressure, until GV1 is completely closed.

[0071] Step 2.3, then gradually open GV1 by 3% opening, and record the data every 3% opening until GV1 is fully open. After GV1 is fully open and stable, gradually close it to the position before the test and release the forced output. At this point, the GV1 test is completed;

[0072] Step 2.4: After completing the flow characteristic test of GV1, the operator adjusts the unit to the pre-test parameters, that is, the stable operation state of step 2.1, and then continues the flow characteristic test of GV2, GV3, and GV4 according to steps 2.2-2.2;

[0073] Step 2.5, according to the data collected by GV1, GV2, GV3, and GV4 during the experiment, calculate the corresponding equivalent actual flow rate according to the corresponding data each time. The equivalent flow rate calculated by the single valve test is Figure 1 The test flow in, and then multiple equivalent actual flow rates of each throttle valve form the ideal characteristic curve of the throttle valve in single valve mode.

[0074] In step 2, the flow characteristic test of each regulating valve of the solar thermal turbine in the sequential valve mode is as follows: open the sequential valve operation mode of the unit, put into power control, keep the main steam pressure setting value unchanged, reduce the unit load from 90% Pe at a rate of 3MW / min, and close the four regulating valves by setting the target load until the unit load drops to 40% Pe. End the test, collect the valve position opening, main steam pressure, and regulating stage pressure of GV1, GV2, GV3, and GV4 during the experiment, and calculate the corresponding equivalent actual flow based on the corresponding data each time. The calculation result is Figure 2 The test flow rate and then multiple equivalent actual flow rates of each throttle valve form the ideal characteristic curve of the throttle valve in the sequential valve mode.

[0075] Example 4

[0076] Based on Example 3, step 3 is specifically as follows:

[0077] When testing the flow characteristics of each throttle valve of a CSP steam turbine in the sequential valve mode, the main steam flow, main steam pressure, regulating stage pressure, throttle valve opening, main steam temperature, and actual unit load are collected at 5-second intervals, and the corresponding equivalent actual flow is calculated to form a training sample data set.

[0078] When testing the flow characteristics of each throttle valve of the solar thermal steam turbine in the sequential valve mode, the main steam flow, main steam pressure, regulating stage pressure, throttle valve opening, main steam temperature, and actual load of the unit are collected at a time interval of 5 seconds, and the corresponding equivalent actual flow is calculated to form a training sample data.

[0079] In step 3, all the collected training sample data are annotated, and then the annotated training sample data is annotated. Each training sample data x t = {main steam flow, main steam pressure, regulating stage pressure, regulating valve opening, main steam temperature, actual load of the unit, equivalent actual flow}, t = 1-N, N is the total training sample data, N is a multiple of 30;

[0080] The training sample data is used to generate sequence data x=[x t ,x t+1 ,…,X t+29 ], that is, each sequence includes 30 sets of training sample data.

[0081] Example 5

[0082] On the basis of Example 4, step 4 establishes a valve flow characteristic curve prediction model including a KAN network model and an LSTM model connected in sequence;

[0083] The KAN network model consists of three KAN layers connected in sequence, with 4, 6, and 8 nodes in each layer respectively;

[0084] The input of the LSTM model is the output of the last KAN layer. The LSTM model is set to three LSTM layers and one linear layer connected in sequence. The number of nodes in each of the three LSTM layers is 32, 64, and 128, respectively. The input of the first LSTM layer is the output of the last KAN layer.

[0085] Set the loss function to Huber Loss, optimizer to Adam, learning rate to 0.001, batch size to 32, epochs to 100, and then start training. Specifically:

[0086] Input the sample data X into the KAN network model, and the output of the KAN network model is KAN out ;

[0087] Will KAN out In the input LSTM model, the final linear layer converts the output of the last LSTM layer into equivalent actual traffic.

[0088] The mathematical basis of the KAN (Kolmogorov-Arnold Network) model comes from the Kolmogorov-Arnold Representation Theorem (KART), which states that any multivariate continuous function f(x1,...,x) can be represented as a combination of a finite number of single-variable continuous functions and addition:

[0089] f(x1,...,x)=∑ i Φ i (∑g i (x))

[0090] g i j(x j )=∑ k c ijk B k (x j )

[0091] Φ i (x)=∑ l d il B l (x)

[0092] Input sample X into the KAN network model, that is, input X t ∈R (30×10) , output KAN out ∈R (30×H) .

[0093] The KAN network model is input into the LSTM model. The LSTM model is a recurrent neural network that uses a gating mechanism and memory units to solve the problem of long-term information preservation and short-term input loss. The input of the LSTM model is consistent with the output of the KAN model. Assume that the LSTM model has (h) hidden units, a batch size of (n), and the number of inputs is (d). The input is (Xt∈R n×d ), the hidden state of the previous time step is (Ht-1∈R n×h ). Accordingly, the gate at time step (t) is defined as: Input gate (I t ∈R n×h ), forget gate (F t ∈R n×h ), the output gate is (O t ∈R n×h ).

[0094] Input gate: I t =σ(X t W xi +H t-1 W hi +b i )

[0095] Forget Gate:F t =σ(X t W xf +H t-1 W hf +b f )

[0096] Output gate: O t =σ(X t W xo +H t-1 W ho +b o )

[0097] Where: W xi ,W xf ,W xo ∈R d×h and W hi ,W hf ,W ho ∈R h×h is the weight parameter, b i ,b f ,b o ∈R 1×h is the paranoid parameter.

[0098] Example 6

[0099] Based on Example 5, step 6 is specifically as follows:

[0100] The CSP steam turbine unit operates according to the operating mode divided in step 1, and monitors the control and operating parameters of each throttle in real time. The main steam flow, main steam pressure, regulating stage pressure, throttle opening, and main steam temperature data of the corresponding throttle in single valve or sequential valve mode are collected at certain time intervals and input into the valve flow characteristic curve prediction model trained in step 5 to obtain the corresponding predicted flow characteristic curve.

[0101] Step 7 specifically uses the determination coefficient to calculate the deviation R between the corresponding predicted flow characteristic curve and the ideal characteristic curve of the corresponding throttle valve in single valve or sequential valve mode generated in step 2. 2 , if the coefficient of determination R 2 If the deviation is greater than or equal to 0.98, it is considered that there is no deviation between the predicted flow characteristic curve and the ideal characteristic curve, that is, in the corresponding mode, the corresponding valve flow does not need to be optimized;

[0102] If the coefficient of determination R 2 If the deviation is less than 0.98, it is considered that the corresponding valve flow characteristic curve of the CST turbine deviates from the ideal flow characteristic curve in the corresponding mode. Then the valve flow in the corresponding mode is optimized, and the corresponding predicted flow characteristic curve after optimization is obtained according to the method in step 6. Then, the deviation R between the optimized and ideal characteristic curves is calculated. 2 , until R 2 Satisfies the condition of being greater than or equal to 0.98.

[0103] Coefficient of determination R 2 The specific calculation is based on the following formula:

[0104]

[0105] Among them, y A,i is the value of the ordinate in the ideal flow characteristic curve when the abscissa is i, y B,i is the value of the ordinate in the predicted flow characteristic curve when the abscissa is i, For all y A,i The average value of , n is the number of values.

[0106] Example 7

[0107] On the basis of Example 6, the method of the present invention is adopted with a time interval n of 5s to obtain the predicted flow characteristic curve, the ideal characteristic curve of the single valve and the ideal characteristic curve of the sequential valve, as shown in FIG. Figure 1 and Figure 2 As shown in the figure, the valve flow characteristic curves predicted under the single valve and sequential valve modes of the test unit deviate seriously from the ideal valve flow characteristic. When the deviation between the predicted flow characteristic curve and the single valve ideal characteristic curve or the sequential valve ideal characteristic curve is greater than the evaluation function, it is considered that the current valve flow characteristic curve is not applicable and needs to be optimized.

Claims

1. A method for monitoring the flow rate of a wide-load high-throttle valve of a solar thermal steam turbine, characterized in that: The specific implementation steps are as follows: Step 1: Set the operation mode of the solar thermal turbine; Step 2: Test the flow characteristics of each regulating valve of the CSP steam turbine in single valve mode and sequential valve mode respectively; Step 3: At a time interval of n, the flow characteristics of each throttle valve in the single valve mode and the flow characteristics of each throttle valve in the sequential valve mode are collected, and the collected data are used as training data; Step 4: Establish a valve flow characteristic curve prediction model; Step 5: Using the training data to train the valve flow characteristic curve prediction model established in step 4; Step 6: Input the operating data of the corresponding valve during the operation of the CSP steam turbine unit into the trained valve flow characteristic curve prediction model to obtain a predicted flow characteristic curve; Step 7: Calculate the deviation between the predicted flow characteristic curve and the ideal flow characteristic curve of the corresponding throttle, and optimize the flow of the corresponding throttle according to the deviation.

2. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 1, characterized in that: In step 1, the operation mode of the solar thermal steam turbine is divided into a start-stop mode, a low-load long-term operation mode, a fast adjustment mode, and a high-load mode; In the start-stop mode, the solar thermal turbine is in the start-stop state. In this mode, the valve is in the single valve mode, and the four regulating valves GV1, GV2, GV3, and GV4 maintain the same opening; In the low-load long-term operation mode, the load of the CSP steam turbine is 15%-25%Pe, where Pe is the rated load of the CSP steam turbine. In this mode, the valve is in the forward valve mode, that is, two regulating valves are opened and two regulating valves are closed; In the fast adjustment mode, when the load of the CSP steam turbine is below 50%Pe, the single valve mode is adopted; when the load of the CSP steam turbine is above 50%Pe, including 50%Pe, the sequential valve mode is adopted, and the regulating valves are opened or closed one by one during the mode conversion; In the high load mode, that is, when the load of the solar thermal steam turbine is greater than 50% Pe, the sequential valve mode is adopted.

3. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 2, characterized in that: In step 2, the flow characteristic test of each regulating valve of the CST turbine in the single valve mode is specifically performed as follows: Step 2.1: When the load of the CSP steam turbine is stable, start the single valve operation and maintain the main steam pressure and temperature stable; Step 2.2: Gradually close GV1 by 3% opening through forced output. Every time GV1 is closed by 3%, the corresponding data is recorded, including valve command, flow command, valve position opening, regulating stage pressure, main steam pressure, actual load, and high exhaust pressure, until GV1 is completely closed. Step 2.3, then gradually open GV1 by 3% opening, and record the data every 3% opening until GV1 is fully open. After GV1 is fully open and stable, gradually close it to the position before the test and release the forced output. At this point, the GV1 test is completed; Step 2.4: After completing the flow characteristic test of GV1, the operator adjusts the unit to the pre-test parameters, that is, the stable operation state of step 2.1, and then continues the flow characteristic test of GV2, GV3, and GV4 according to steps 2.2-2.2; In step 2.5, based on the data collected by GV1, GV2, GV3, and GV4 during the experiment, the corresponding equivalent actual flow rate is calculated according to the corresponding data each time, and then the multiple equivalent actual flow rates of each throttle valve form the ideal characteristic curve of the throttle valve in single valve mode.

4. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 3 is characterized in that: In step 2, the flow characteristic test of each regulating valve of the CST turbine in the sequential valve mode is specifically performed as follows: Start the unit sequence valve operation mode, put into power control, keep the main steam pressure setting value unchanged, reduce the unit load from 90% Pe at a rate of 3MW / min, and close the four regulating valves by setting the target load until the unit load drops to 40% Pe. End the test, collect the valve position opening, main steam pressure, and regulating stage pressure of GV1, GV2, GV3, and GV4 during the experiment, calculate the corresponding equivalent actual flow rate based on the corresponding data each time, and then form the ideal characteristic curve of each regulating valve in the sequence valve mode based on multiple equivalent actual flow rates.

5. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 4, characterized in that: The step 3 is specifically as follows: When testing the flow characteristics of each throttle valve of a CSP steam turbine in the sequential valve mode, the main steam flow, main steam pressure, regulating stage pressure, throttle valve opening, main steam temperature, and actual unit load are collected at 5-second intervals, and the corresponding equivalent actual flow is calculated to form a training sample data set. When testing the flow characteristics of each throttle valve of a CSP steam turbine in the sequential valve mode, the main steam flow, main steam pressure, regulating stage pressure, throttle valve opening, main steam temperature, and actual unit load are collected at 5-second intervals, and the corresponding equivalent actual flow is calculated to form a training sample data set. In step 3, all the collected training sample data are annotated, and then the annotated training sample data are annotated. Each training sample data X t = {main steam flow, main steam pressure, regulating stage pressure, regulating valve opening, main steam temperature, actual load of the unit, equivalent actual flow}, t = 1-N, N is the total training sample data, N is a multiple of 30; The training sample data is used to generate sequence data x=[x t ,x t+1 ,…,X t+29 ], that is, each sequence includes 30 sets of training sample data.

6. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 5, characterized in that: The step 4 establishes a valve flow characteristic curve prediction model including a KAN network model and an LSTM model connected in sequence; The KAN network model includes three KAN layers connected in sequence, with 4, 6, and 8 nodes in each layer respectively; The input of the LSTM model is the output of the last KAN layer. The LSTM model is configured to include three LSTM layers and one linear layer connected in sequence. The number of nodes in each of the three LSTM layers is 32, 64, and 128, respectively. The input of the first LSTM layer is the output of the last KAN layer.

7. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 6, characterized in that: The step 5 is specifically as follows: Set the loss function to Huber Loss, optimizer to Adam, learning rate to 0.001, batch size to 32, epochs to 100, and then start training. Specifically: Input the sample data X into the KAN network model, and the output of the KAN network model is KAN out ; Will KAN out In the input LSTM model, the final linear layer converts the output of the last LSTM layer into equivalent actual traffic.

8. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 7, characterized in that: The step 6 is specifically as follows: The CSP steam turbine unit operates according to the operating mode divided in step 1, and monitors the control and operating parameters of each throttle in real time. The main steam flow, main steam pressure, regulating stage pressure, throttle opening, and main steam temperature data of the corresponding throttle in single valve or sequential valve mode are collected at certain time intervals and input into the valve flow characteristic curve prediction model trained in step 5 to obtain the corresponding predicted flow characteristic curve.

9. The method for monitoring the wide load high-speed valve flow of a solar thermal steam turbine according to claim 8, characterized in that: The step 7 specifically uses the determination coefficient to calculate the deviation r between the corresponding predicted flow characteristic curve and the ideal characteristic curve of the corresponding throttle valve generated in the single valve or sequential valve mode generated in step 2. 2 , if the coefficient of determination r 2 If the deviation is greater than or equal to 0.98, it is considered that there is no deviation between the predicted flow characteristic curve and the ideal characteristic curve, that is, in the corresponding mode, the corresponding valve flow does not need to be optimized; If the coefficient of determination R 2 If the deviation is less than 0.98, it is considered that the corresponding valve flow characteristic curve of the CST in the corresponding mode deviates from the ideal flow characteristic curve. Then the valve flow in the corresponding mode is optimized, and the corresponding predicted flow characteristic curve after optimization is obtained according to the method in step 6. Then, the deviation r from the ideal characteristic curve is calculated. 2 , until r 2 Satisfies the condition of being greater than or equal to 0.

98.

10. The method for monitoring the wide load high-speed regulating valve flow of a solar thermal steam turbine according to claim 9, characterized in that: The coefficient of determination r 2 The specific calculation is based on the following formula: Among them, y A,i is the value of the ordinate in the ideal flow characteristic curve when the abscissa is i, y B,i is the value of the ordinate in the predicted flow characteristic curve when the abscissa is i, For all y A,i The average value of , n is the number of values.

Images