Method for monitoring and diagnosing internal leakage of main steam valve of steam turbine

Abstract

The embodiment of the invention provides a method for monitoring and diagnosing inner leakage of a main steam valve of a steam turbine. According to the method, the XGBoost gradient elevator algorithm is adopted to establish the inner leakage diagnosis model, and the XGBoost algorithm is adopted to establish the inner leakage diagnosis model, so that the method has remarkable beneficial effects of improving the accuracy and efficiency of diagnosis. The XGBoost is a powerful machine learning algorithm, and can effectively process a complex nonlinear relationship and improve the prediction accuracy by constructing a plurality of decision trees and carrying out integrated learning. Meanwhile, factors which have the most influence on internal leakage diagnosis can be screened out from a large number of features, so that the interpretability and predictive ability of the model are improved; secondly, the high efficiency of the XGBoost algorithm means that the model completes training in a short time, and the XGBoost algorithm is suitable for a real-time monitoring system; through the iterative optimization model, the prediction error is minimized, misdiagnosis and missed diagnosis are reduced, it is ensured that the monitoring system can accurately and reliably recognize the inner leakage fault, and powerful guarantee is provided for safe production of power generation enterprises.

Description

Technical Field

[0001] The embodiment of the present invention relates to the technical field of valve monitoring in power generation enterprises, and specifically to a method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine. Background Art

[0002] With the continuous growth of energy demand, power generation companies play a vital role in power generation. As a key equipment in power generation companies, the safe and stable operation of steam turbines is of great significance to ensuring power supply. In the steam turbine system, the main steam valve is an important component for controlling the flow direction and pressure of steam. Its normal operation is crucial to the efficiency and safety of the steam turbine. During the long-term operation of the main steam valve, internal leakage may occur due to various reasons, which will not only affect the efficiency of the steam turbine, but also cause equipment damage and even cause safety accidents.

[0003] However, the monitoring methods commonly used by power generation companies are mainly manual inspections and simple temperature monitoring. These methods have the following shortcomings: manual inspections are inefficient, difficult to achieve real-time monitoring, and easily affected by subjective factors, and the accuracy is not high. Simple temperature monitoring cannot accurately determine the degree of valve internal leakage and it is difficult to achieve accurate diagnosis of the internal leakage status. Existing monitoring methods lack in-depth analysis of historical data and cannot predict potential internal leakage risks. Summary of the invention

[0004] The embodiment of the present invention aims to solve at least one of the technical problems existing in the prior art, and provides a method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine.

[0005] An embodiment of the present invention provides a method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine, the method comprising:

[0006] Step S1: Determine the monitoring point: select the inlet and outlet pipes of the main steam valve of the steam turbine as the monitoring point to ensure that the monitoring point can accurately reflect the internal leakage of the valve;

[0007] Step S2: Installing temperature sensors: Install high-precision temperature sensors at monitoring points to collect temperature data of valve inlet and outlet pipes in real time;

[0008] Step S3: data acquisition, setting a data acquisition system to collect temperature sensor data at a fixed frequency of once per second and store it in a database;

[0009] Step S4: perform data preprocessing, preprocessing the collected temperature data, including removing outliers, filtering, and normalizing to ensure data quality;

[0010] Step S5: Establishing a temperature reference model. According to the temperature data during normal operation, an autoregressive moving average model algorithm is used to establish a temperature reference model during normal operation of the valve.

[0011] Step S6: temperature change analysis, real-time analysis of the temperature changes of the valve inlet and outlet pipes, using the sliding window method to compare with the temperature benchmark model to determine whether there is an abnormal temperature change;

[0012] Step S7: Perform preliminary diagnosis of valve internal leakage: When abnormal temperature changes are detected, perform preliminary diagnosis to determine whether it is valve internal leakage; if it is initially determined to be internal leakage, proceed to the next step;

[0013] Step S8: extracting features from abnormal temperature data, including statistical features of temperature change rate, temperature difference, kurtosis, and skewness, and extracting frequency domain features using short-time Fourier transform;

[0014] Step S9: Using the XGBoost gradient boosting algorithm, an endoleak diagnosis model is established, and the loss function is minimized by iteratively building the model:

[0015] Step S10: Input the extracted fault features into the internal leakage diagnosis model for diagnosis; if it is confirmed to be an internal leakage, issue an early warning immediately and notify the operation and maintenance personnel to take measures to deal with it; at the same time, regularly use the cross-validation method to evaluate the model performance and adjust the model parameters according to the actual situation.

[0016] In some possible embodiments, in step S1, a monitoring point is selected on each of the straight pipe sections before and after the valve to preliminarily determine whether the valve has internal leakage by comparing the temperature difference between the inlet and outlet; if the valve is designed with a bypass pipe, a monitoring point is also installed here to more comprehensively grasp the operating status of the valve;

[0017] In step S2, a platinum resistance temperature sensor is selected. During installation, ensure that the temperature sensor is in close contact with the pipeline surface and is fixed with high-temperature resistant insulating material to avoid measurement errors caused by temperature changes. At the same time, appropriate protective measures should be taken for the wiring of the temperature sensor to prevent electromagnetic interference.

[0018] In some possible embodiments, in step S3, a data acquisition card is connected to the temperature sensor and is programmed to collect data once a second; during the data collection process, the system automatically records and stores the readings of the temperature sensor in a database, including an SQL or NoSQL database, to ensure data integrity and traceability.

[0019] In some possible embodiments, in step S4, in the collected temperature data, abnormal values ​​are automatically identified and removed by setting a reasonable threshold; and a digital filter is applied to remove high-frequency noise to ensure the smoothness of the data.

[0020] In some possible embodiments, in step S5, the specific algorithm steps of using the ARMA model to establish a temperature reference model for normal operation of the valve include:

[0021] Step 1: Data stationarity test: Collect the temperature data sequence when the valve is operating normally; use the ADF test to test the stationarity of the data. The calculation formula of the ADF test is:

[0022]

[0023] Where Δ represents the first-order difference; is the estimated value of the time series; α is the constant term; β is the coefficient of the time trend term; γ is the coefficient of the lag term; δ is the coefficient of the difference lag term; ∈t is the white noise error term;

[0024] Step 2: Determine the order of the ARMA model, and use the autocorrelation function ACF and partial autocorrelation function PACF to identify the order of the AR and MA models; observe the ACF graph to determine the order q of the MA, and the PACF graph to determine the order p of the AR;

[0025] Step 3: Model estimation: According to the determined orders p and q, an ARMA (p, q) model is established; the maximum likelihood estimation (MLE) optimization method is used to estimate the model parameters;

[0026] y t =c+φ1y t-1 +...+φ p y t-p +θ1∈ t-1 +...+θ q ∈ t-q +∈ t

[0027] where y t is the value of the time series at time t; c is the constant term; φ is the coefficient of the autoregressive term; θ is the coefficient of the moving average term; ∈t is the white noise error term;

[0028] Step 4: Model test, use the Ljung-Box Q test to test whether the residual of the model is white noise; if the residual is not white noise, you may need to reselect the order of the model or consider other models:

[0029] The formula for the Ljung-Box Q test is:

[0030] Where n is the length of the time series; m is the number of lag periods; is the sample autocorrelation coefficient delayed by k periods;

[0031] Step 5: Model prediction, use the estimated ARMA model to predict future temperature values; these predicted values ​​will serve as the temperature benchmark for normal valve operation, and the prediction formula is:

[0032] in is the predicted value for the next h steps based on current and past information.

[0033] In some possible embodiments, in step S6, the specific algorithm steps of using the sliding window method to determine whether there is an abnormal temperature change are as follows:

[0034] Step 1: Determine the window size and select an appropriate time window size N, which should be large enough to contain enough temperature data points;

[0035] Step 2: Initialize the window and load the temperature data in the initial window into the window to prepare for calculation;

[0036] Step 3: Calculate the statistics within the window, perform statistical analysis on the temperature data within the window, and calculate the mean and standard deviation statistics;

[0037] The window mean calculation formula is: [mu frac1N sumi=1N xi];

[0038] The formula for calculating the window standard deviation is: [sigma sqrtfrac1N sumi=1N(xi-mu)^2;

[0039] N is the window size;

[0040] xi is the i-th data point in the window;

[0041] mu is the mean of the data in the window;

[0042] sigma is the standard deviation of the data within the window;

[0043] Step 4: Set a threshold value. Set a threshold value for anomaly detection based on historical data or expert experience.

[0044] Step 5: Sliding window. Each time new temperature data is collected, the window is slid forward by one data point, the oldest data point in the window is removed, and the latest data point is added.

[0045] Step 6: Update the statistics, using the new data in the sliding window to update the statistics, that is, update the mean and standard deviation;

[0046] Step 7: Anomaly detection, for each new data point in the window, calculate its deviation from the current window statistic, the calculation formula is:

[0047] The calculation formula of the deviation is: [delta xi=xi-mu]

[0048] Where: xi is the new data point;

[0049] mu is the mean of the current window;

[0050] The calculation formula for judging whether a new data point is abnormal is: [textIf|delta xi|>k*sigma,textthen textit{abnormal}];

[0051] Where: k is the set threshold multiple; sigma is the standard deviation of the current window;

[0052] Step 8: Record and report anomalies. If an anomaly is detected, record the anomaly data and timestamp and report it to the operation and maintenance personnel.

[0053] In some possible embodiments, in step S7, if the temperature of the monitoring point suddenly rises while other relevant parameters remain unchanged, this indicates that there is internal leakage in the valve; the system will perform preliminary internal leakage diagnosis based on preset rules or simple logical judgments, such as when the temperature difference exceeds a set threshold.

[0054] In some possible embodiments, in step S8, the temperature change rate is extracted from the abnormal temperature data by calculating the ratio of the temperature difference between consecutive time points to time.

[0055] In some possible embodiments, in step S9, the specific algorithm steps of using XGBoost to establish the endoleak diagnosis model include:

[0056] Step 1: Data preparation: collect and organize the temperature data of the main steam valve of the steam turbine under normal operation and internal leakage conditions; preprocess the data, including missing value processing, outlier processing, and feature engineering; divide the data set into a training set and a test set;

[0057] Step 2: Define the loss function and use the logarithmic loss function to evaluate the performance of the model. The calculation formula is:

[0058] [L(y,y')=-[y log(y')+(1-y)log(1-y')]]; where y is the true label; y' is the probability predicted by the model;

[0059] Step 3: Initialize the model and model parameters, including learning rate, number of trees, and tree depth;

[0060] Step 4: Build a decision tree. For each tree, select the optimal features and split points to minimize the loss function. The objective function calculation formula is:

[0061] [Obj(t)=sum(L(yi,y'i)+Omega(f_t))];

[0062] Where Obj(t) is the objective function of the tth tree; L is the loss function; Omega is the regularization term used to control the complexity of the model;

[0063] Step 5: Calculate the gradient of the loss function for each sample to guide the growth of the decision tree. The gradient calculation formula is:

[0064] [gi=partialfracpartialy'L(yi,y'i)|y'_i=current prediction]; where gi is the gradient of the ith sample;

[0065] Step 6: Construct leaf nodes, find the optimal segmentation point according to the gradient, and assign data to leaf nodes;

[0066] Step 7: Calculate the value of the leaf node. For each leaf node, calculate a value to minimize the objective function. The leaf node value calculation formula is:

[0067] [wj=-frac1sum_{iin Ij}fracpartialpartialy'L(yi,y'i)|y'i=currentprediction];

[0068] Where wj is the value of the jth leaf node; Ij is the sample set assigned to the jth leaf node;

[0069] Step 8: Update the model, add the newly built tree to the model, update the predicted value,

[0070] The updated forecast calculation formula is:

[0071] [y'i=y'i+eta*wj]; where eta is the learning rate; wj is the leaf node value of the new tree;

[0072] Step 9: Iterate optimization and repeat steps 4 to 8 until the predetermined number of trees is reached or the loss function no longer decreases significantly. Through the above steps, an XGBoost model for internal leakage diagnosis of main steam valves of steam turbines is established.

[0073] In some possible embodiments, in step S10, the model will output an internal leakage probability or classification result based on the characteristic data; if the model output result shows that the internal leakage probability exceeds the set threshold, the system will immediately issue an early warning and notify the operation and maintenance personnel via SMS, email or control system alarm.

[0074] In summary, due to the adoption of the above technical solution, the beneficial effects of the embodiments of the present invention are:

[0075] 1. The method for monitoring and diagnosing internal leakage of the main steam valve of a steam turbine according to an embodiment of the present invention adopts an autoregressive moving average model algorithm to construct a benchmark model based on the temperature data of the valve during normal operation. The model can capture the temperature change pattern of the valve under normal working conditions. Such a benchmark model provides a reference standard for subsequent temperature change analysis, allowing the system to identify temperature anomalies that are inconsistent with normal conditions. In addition, the establishment of a benchmark model also helps to reduce false alarms because it can distinguish between temperature fluctuations caused by normal operation or environmental factors and true signs of internal leakage, thereby improving the accuracy and reliability of the monitoring system.

[0076] 2. The method for monitoring and diagnosing internal leakage of the main steam valve of the steam turbine of the embodiment of the present invention analyzes the temperature changes of the inlet and outlet pipelines of the valve in real time, and uses the sliding window method to compare the temperature reference model. This step provides the monitoring system with the ability of real-time monitoring and early warning. By continuously monitoring and analyzing the temperature data, the system can promptly detect any abnormal situation that deviates from the reference model, which may be an early sign of internal leakage. The beneficial effect of this real-time analysis is that it can quickly respond to temperature changes and provide timely decision support for operation and maintenance personnel, thereby avoiding or reducing equipment damage and downtime caused by failure to timely detect internal leakage. In addition, through comparative analysis, it also helps operation and maintenance personnel to better understand the operating characteristics of the valve and provide data support for equipment maintenance and optimization. Real-time analysis of the temperature changes of the inlet and outlet pipelines of the valve, using the sliding window method to compare the temperature reference model, this step provides the monitoring system with the ability of real-time monitoring and early warning. By continuously monitoring and analyzing the temperature data, the system can promptly detect any abnormal situation that deviates from the reference model, which may be an early sign of internal leakage. The beneficial effect of this real-time analysis is that it can quickly respond to temperature changes and provide timely decision support for operation and maintenance personnel, thereby avoiding or reducing equipment damage and downtime caused by failure to timely detect internal leakage. In addition, comparative analysis can help operation and maintenance personnel better understand the operating characteristics of valves and provide data support for equipment maintenance and optimization.

[0077] 3. The internal leakage monitoring and diagnosis method of the main steam valve of the steam turbine according to the embodiment of the present invention adopts the XGBoost gradient boosting algorithm to establish the internal leakage diagnosis model. The internal leakage diagnosis model is established by using the XGBoost algorithm, which has a significant beneficial effect on improving the accuracy and efficiency of diagnosis. XGBoost is a powerful machine learning algorithm. By constructing multiple decision trees and performing ensemble learning, it can effectively handle complex nonlinear relationships and improve the accuracy of prediction. The beneficial effects of this step are reflected in several aspects: first, it can filter out the most influential factors for internal leakage diagnosis from a large number of features, thereby improving the interpretability and predictive ability of the model; secondly, the high efficiency of the XGBoost algorithm means that the model completes training in a shorter time, which is suitable for real-time monitoring systems; finally, by iteratively optimizing the model, minimizing the prediction error, reducing misdiagnosis and missed diagnosis, ensuring that the monitoring system can accurately and reliably identify internal leakage faults, and providing strong protection for the safe production of power generation enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0079] Figure 1 The present invention is a flowchart of a method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to an embodiment of the present invention. DETAILED DESCRIPTION

[0080] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0081] Unless otherwise specified, the technical terms or scientific terms used in the embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. "Including" or "comprising" used in the embodiments of the present invention neither limit the shapes, numbers, steps, actions, operations, components, originals and / or their groups mentioned, nor exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, originals and / or their groups, or the addition of these. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number and order of the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0082] Unless otherwise specifically stated, the relative arrangement of the components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention. At the same time, it should be understood that for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship, and the techniques, methods and devices known to ordinary technicians in the relevant fields may not be discussed in detail, but where appropriate, the techniques, methods and devices shown should be considered as part of the authorized specification. In all examples shown and discussed here, any specific other examples may have different values. It should be noted that similar symbols and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0083] In the description of the embodiments of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in the embodiments of the present invention and the features of the different embodiments or examples, without contradiction.

[0084] Below, the exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0085] A method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine, the method comprising:

[0086] Step S1: Determine the monitoring point: Select the inlet and outlet pipelines of the main steam valve of the steam turbine as the monitoring point to ensure that the monitoring point can accurately reflect the internal leakage of the valve.

[0087] Step S2: Installing temperature sensors: Install high-precision temperature sensors at monitoring points to collect temperature data of valve inlet and outlet pipes in real time.

[0088] Step S3: Data acquisition: setting up a data acquisition system to collect temperature sensor data at a fixed frequency of once per second and store it in a database.

[0089] Step S4: perform data preprocessing to preprocess the collected temperature data, including removing outliers, filtering, and normalizing to ensure data quality.

[0090] Step S5: Establishing a temperature reference model. According to the temperature data during normal operation, an autoregressive moving average model algorithm is used to establish a temperature reference model during normal operation of the valve.

[0091] Step S6: Temperature change analysis: Real-time analysis of the temperature changes of the valve inlet and outlet pipes, using the sliding window method to compare with the temperature benchmark model to determine whether there is an abnormal temperature change.

[0092] Step S7: Perform preliminary diagnosis of valve internal leakage: When abnormal temperature changes are monitored, perform preliminary diagnosis to determine whether there is internal leakage of the valve; if it is preliminarily determined to be internal leakage, proceed to the next step.

[0093] Step S8: extracting features from abnormal temperature data, including statistical features of temperature change rate, temperature difference, kurtosis, and skewness, and extracting frequency domain features using short-time Fourier transform.

[0094] Step S9: Using the XGBoost gradient boosting algorithm, an endoleak diagnosis model is established, and the loss function is minimized by iteratively building the model.

[0095] Step S10: Input the extracted fault features into the internal leakage diagnosis model for diagnosis; if it is confirmed to be an internal leakage, issue an early warning immediately and notify the operation and maintenance personnel to take measures to deal with it; at the same time, regularly use the cross-validation method to evaluate the model performance and adjust the model parameters according to the actual situation.

[0096] In some embodiments, in step S1, a monitoring point is selected on each of the straight pipe sections before and after the valve to preliminarily determine whether the valve has internal leakage by comparing the temperature difference between the inlet and outlet; if the valve is designed with a bypass pipe, a monitoring point is also installed here to more comprehensively grasp the operating status of the valve;

[0097] In step S2, platinum resistance temperature sensors are selected, which have high precision and good linearity. During installation, it is necessary to ensure that the sensor is in close contact with the pipeline surface and is fixed with high temperature resistant insulating materials to avoid measurement errors caused by temperature changes. At the same time, appropriate protective measures should be taken for the wiring of the sensor to prevent electromagnetic interference.

[0098] In some embodiments, in step S3, a data acquisition card (DAQ) is connected to the temperature sensor, and data is collected once per second through programming. During the data collection process, the system automatically records and stores the readings of the temperature sensor in a database, such as an SQL or NoSQL database, to ensure the integrity and traceability of the data. In addition, the data acquisition system should also have a power-off protection function to prevent data loss.

[0099] In some embodiments, in step S4, abnormal values ​​may appear in the collected temperature data due to sensor failure or environmental interference. These abnormal values ​​are automatically identified and removed by setting a reasonable threshold. In addition, a digital filter, such as a low-pass filter, is also applied to remove high-frequency noise to ensure the smoothness of the data.

[0100] In some embodiments, in step S5, the specific algorithm steps of using the ARMA model to establish a temperature reference model for normal valve operation include:

[0101] Step 1: Data stationarity test: Collect the temperature data sequence when the valve is operating normally; use the ADF test to test the stationarity of the data. The calculation formula of the ADF test is:

[0102]

[0103] Where Δ represents the first-order difference; is the estimated value of the time series; α is the constant term; β is the coefficient of the time trend term; γ is the coefficient of the lag term; δ is the coefficient of the difference lag term; ∈t is the white noise error term.

[0104] Step 2: Determine the order of the ARMA model, and use the autocorrelation function ACF and partial autocorrelation function PACF to identify the order of the AR and MA models; observe the ACF graph to determine the order q of the MA, and the PACF graph to determine the order p of the AR.

[0105] Step 3: Model estimation: According to the determined orders p and q, an ARMA (p, q) model is established; the maximum likelihood estimation (MLE) optimization method is used to estimate the model parameters;

[0106] y t =c+φ1y t-1 +...+φ p yt-p +θ1∈ t-1 +...+θ q ∈ t-q +∈ t

[0107] where y t is the value of the time series at time t; c is the constant term; φ is the coefficient of the autoregressive term; θ is the coefficient of the moving average term; ∈t is the white noise error term.

[0108] Step 4: Model test, use the Ljung-Box Q test to test whether the residual of the model is white noise; if the residual is not white noise, you may need to reselect the order of the model or consider other models:

[0109] The formula for the Ljung-Box Q test is:

[0110] Where n is the length of the time series; m is the number of lag periods; is the sample autocorrelation coefficient with a lag of k periods.

[0111] Step 5: Model prediction, use the estimated ARMA model to predict future temperature values; these predicted values ​​will serve as the temperature benchmark for normal valve operation, and the prediction formula is:

[0112] in is the predicted value for the next h steps based on current and past information.

[0113] In some embodiments, in step S6, the specific algorithm steps of using the sliding window method to determine whether there is an abnormal temperature change are as follows:

[0114] Step 1: Determine the window size. Select an appropriate time window size N, which should be large enough to contain enough temperature data points.

[0115] Step 2: Initialize the window and load the temperature data in the initial window into the window in preparation for calculation.

[0116] Step 3: Calculate the statistics within the window, perform statistical analysis on the temperature data within the window, and calculate the mean and standard deviation statistics;

[0117] The window mean calculation formula is: [mu frac1N sumi=1N xi];

[0118] The formula for calculating the window standard deviation is: [sigma sqrtfrac1N sumi=1N(xi-mu)^2;

[0119] N is the window size;

[0120] xi is the i-th data point in the window;

[0121] mu is the mean of the data in the window;

[0122] sigma is the standard deviation of the data within the window.

[0123] Step 4: Set a threshold. Set a threshold for anomaly detection based on historical data or expert experience. This threshold is a multiple of the statistic, for example, the mean plus or minus several times the standard deviation.

[0124] Step 5: Sliding window. Each time new temperature data is collected, slide the window forward by one data point, remove the oldest data point in the window, and add the latest data point.

[0125] Step 6: Update the statistics using the new data in the sliding window, that is, update the mean and standard deviation.

[0126] Step 7: Anomaly detection, for each new data point in the window, calculate its deviation from the current window statistic, the calculation formula is:

[0127] The calculation formula of the deviation is: [delta xi=xi-mu]

[0128] Where: xi is the new data point;

[0129] mu is the mean of the current window;

[0130] The calculation formula for judging whether a new data point is abnormal is: [textIf|delta xi|>k*sigma,textthen textit{abnormal}];

[0131] Where: k is the set threshold multiple; sigma is the standard deviation of the current window.

[0132] Step 8: Record and report anomalies. If an anomaly is detected, record the anomaly data and timestamp and report it to the operation and maintenance personnel.

[0133] In some embodiments, in step S7, if the temperature of the monitoring point suddenly rises while other relevant parameters remain unchanged, this indicates that the valve has internal leakage; the system will perform preliminary internal leakage diagnosis based on preset rules or simple logical judgments, such as the temperature difference exceeding the set threshold. This step helps operation and maintenance personnel quickly identify potential problems and provides a basis for further feature extraction and model diagnosis.

[0134] In some embodiments, in step S8, the temperature change rate is extracted from the abnormal temperature data by calculating the ratio of the temperature difference between consecutive time points to the time. The temperature difference feature is obtained by comparing the temperature difference values ​​of different monitoring points. The kurtosis and skewness statistical features describe the shape of the temperature distribution, and the short-time Fourier transform (STFT) is used to analyze the frequency domain characteristics of the temperature signal to reveal potential periodic changes.

[0135] In some embodiments, in step S9, the specific algorithm steps of using XGBoost to establish the endoleak diagnosis model include:

[0136] Step 1: Data preparation: collect and organize the temperature data of the main steam valve of the turbine under normal operation and internal leakage conditions; preprocess the data, including missing value processing, outlier processing, and feature engineering; divide the data set into a training set and a test set.

[0137] Step 2: Define the loss function and use the logarithmic loss function to evaluate the performance of the model. The calculation formula is:

[0138] [L(y,y')=-[y log(y')+(1-y)log(1-y')]]; where y is the true label; y' is the probability predicted by the model.

[0139] Step 3: Initialize the model and model parameters, including learning rate, number of trees, and tree depth.

[0140] Step 4: Build a decision tree. For each tree, select the optimal features and split points to minimize the loss function. The objective function calculation formula is:

[0141] [Obj(t)=sum(L(yi,y'i)+Omega(f_t))];

[0142] Where Obj(t) is the objective function of the tth tree; L is the loss function; Omega is the regularization term used to control the complexity of the model.

[0143] Step 5: Calculate the gradient of the loss function for each sample to guide the growth of the decision tree. The gradient calculation formula is:

[0144] [gi=partialfracpartialy'L(yi,y'i)|y'_i=current prediction]; where gi is the gradient of the i-th sample.

[0145] Step 6: Construct leaf nodes, find the optimal segmentation point based on the gradient, and assign data to leaf nodes.

[0146] Step 7: Calculate the value of the leaf node. For each leaf node, calculate a value to minimize the objective function. The leaf node value calculation formula is:

[0147] [wj=-frac1sum_{iin Ij}fracpartialpartialy'L(yi,y'i)|y'i=currentprediction];

[0148] Where wj is the value of the j-th leaf node; Ij is the set of samples assigned to the j-th leaf node.

[0149] Step 8: Update the model, add the newly built tree to the model, update the predicted value,

[0150] The updated forecast calculation formula is:

[0151] [y'i=y'i+eta*wj]; where eta is the learning rate; wj is the leaf node value of the new tree;

[0152] Step 9: Iterative optimization, repeating steps 4 to 8 until the predetermined number of trees is reached or the loss function no longer decreases significantly. Through the above steps, an XGBoost model for internal leakage diagnosis of the main steam valve of a steam turbine is established. This model minimizes the loss function by iteratively constructing a decision tree, thereby achieving accurate prediction of the internal leakage state.

[0153] In some embodiments, in step S10, the model will output an internal leakage probability or classification result based on the feature data; if the model output result shows that the internal leakage probability exceeds the set threshold, the system will immediately issue an early warning and notify the operation and maintenance personnel through SMS, email or control system alarm. At the same time, in order to maintain the accuracy of the model, it is necessary to regularly use the cross-validation method to evaluate the performance of the model and adjust the model parameters, such as the learning rate and the number of trees, according to the actual operation situation.

[0154] The method for monitoring and diagnosing internal leakage of the main steam valve of a steam turbine according to an embodiment of the present invention adopts an autoregressive moving average model (ARMA) algorithm to construct a benchmark model based on the temperature data of the valve during normal operation. The model can capture the temperature variation law of the valve under normal working conditions. Such a benchmark model provides a reference standard for subsequent temperature variation analysis, allowing the system to identify temperature anomalies that are inconsistent with normal conditions. In addition, the establishment of a benchmark model also helps to reduce false alarms because it can distinguish between temperature fluctuations caused by normal operation or environmental factors and real signs of internal leakage, thereby improving the accuracy and reliability of the monitoring system.

[0155] The method for monitoring and diagnosing internal leakage of the main steam valve of the steam turbine of the embodiment of the present invention analyzes the temperature changes of the inlet and outlet pipelines of the valve in real time, and uses the sliding window method to compare the temperature reference model. This step provides the monitoring system with the ability of real-time monitoring and early warning. By continuously monitoring and analyzing the temperature data, the system can promptly detect any abnormal situation that deviates from the reference model, which may be an early sign of internal leakage. The beneficial effect of this real-time analysis is that it can quickly respond to temperature changes and provide timely decision support for operation and maintenance personnel, thereby avoiding or reducing equipment damage and downtime caused by failure to timely detect internal leakage. In addition, through comparative analysis, it also helps operation and maintenance personnel to better understand the operating characteristics of the valve and provide data support for equipment maintenance and optimization. Real-time analysis of the temperature changes of the inlet and outlet pipelines of the valve, using the sliding window method to compare the temperature reference model, this step provides the monitoring system with the ability of real-time monitoring and early warning. By continuously monitoring and analyzing the temperature data, the system can promptly detect any abnormal situation that deviates from the reference model, which may be an early sign of internal leakage. The beneficial effect of this real-time analysis is that it can quickly respond to temperature changes and provide timely decision support for operation and maintenance personnel, thereby avoiding or reducing equipment damage and downtime caused by failure to timely detect internal leakage. In addition, comparative analysis can help operation and maintenance personnel better understand the operating characteristics of valves and provide data support for equipment maintenance and optimization.

[0156] The internal leakage monitoring and diagnosis method of the main steam valve of the steam turbine of the embodiment of the present invention adopts the XGBoost gradient boosting algorithm to establish the internal leakage diagnosis model. The internal leakage diagnosis model is established by using the XGBoost algorithm, which has a significant beneficial effect on improving the accuracy and efficiency of diagnosis. XGBoost is a powerful machine learning algorithm. By constructing multiple decision trees and performing ensemble learning, it can effectively handle complex nonlinear relationships and improve the accuracy of prediction. The beneficial effects of this step are reflected in several aspects: first, it can filter out the most influential factors for internal leakage diagnosis from a large number of features, thereby improving the interpretability and predictive ability of the model; secondly, the high efficiency of the XGBoost algorithm means that the model completes training in a shorter time, which is suitable for real-time monitoring systems; finally, by iteratively optimizing the model, minimizing the prediction error, reducing the situation of misdiagnosis and missed diagnosis, ensuring that the monitoring system can accurately and reliably identify internal leakage faults, and providing strong protection for the safe production of power generation enterprises.

[0157] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine, characterized in that: The method comprises: Step S1: Determine the monitoring point: select the inlet and outlet pipes of the main steam valve of the steam turbine as the monitoring point to ensure that the monitoring point can accurately reflect the internal leakage of the valve; Step S2: Installing temperature sensors: Install high-precision temperature sensors at monitoring points to collect temperature data of valve inlet and outlet pipes in real time; Step S3: data acquisition, setting a data acquisition system to collect temperature sensor data at a fixed frequency of once per second and store it in a database; Step S4: perform data preprocessing, preprocessing the collected temperature data, including removing outliers, filtering, and normalizing to ensure data quality; Step S5: Establishing a temperature reference model. According to the temperature data during normal operation, an autoregressive moving average model algorithm is used to establish a temperature reference model during normal operation of the valve. Step S6: temperature change analysis, real-time analysis of the temperature changes of the valve inlet and outlet pipes, using the sliding window method to compare with the temperature benchmark model to determine whether there is an abnormal temperature change; Step S7: Perform preliminary diagnosis of valve internal leakage: When abnormal temperature changes are detected, perform preliminary diagnosis to determine whether it is valve internal leakage; if it is initially determined to be internal leakage, proceed to the next step; Step S8: extracting features from abnormal temperature data, including statistical features of temperature change rate, temperature difference, kurtosis, and skewness, and extracting frequency domain features using short-time Fourier transform; Step S9: Using the XGBoost gradient boosting algorithm, an endoleak diagnosis model is established, and the loss function is minimized by iteratively building the model: Step S10: Input the extracted fault features into the internal leakage diagnosis model for diagnosis; if it is confirmed to be an internal leakage, issue an early warning immediately and notify the operation and maintenance personnel to take measures to deal with it; at the same time, regularly use the cross-validation method to evaluate the model performance and adjust the model parameters according to the actual situation.

2. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S1, a monitoring point is selected on each of the straight pipe sections before and after the valve to preliminarily determine whether the valve has internal leakage by comparing the temperature difference between the inlet and outlet; if the valve is designed with a bypass pipe, a monitoring point is also installed here to more comprehensively grasp the operating status of the valve; In step S2, a platinum resistance temperature sensor is selected. During installation, ensure that the temperature sensor is in close contact with the pipeline surface and is fixed with high-temperature resistant insulating material to avoid measurement errors caused by temperature changes. At the same time, appropriate protective measures should be taken for the wiring of the temperature sensor to prevent electromagnetic interference.

3. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S3, a data acquisition card is connected to the temperature sensor and programmed to collect data once a second; during the data collection process, the system automatically records and stores the readings of the temperature sensor in a database, including an SQL or NoSQL database, to ensure data integrity and traceability.

4. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S4, in the collected temperature data, abnormal values ​​are automatically identified and removed by setting a reasonable threshold; and a digital filter is applied to remove high-frequency noise to ensure the smoothness of the data.

5. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S5, the specific algorithm steps of using the ARMA model to establish a temperature reference model for normal valve operation include: Step 1: Data stationarity test: Collect the temperature data sequence when the valve is operating normally; use the ADF test to test the stationarity of the data. The calculation formula of the ADF test is: Where Δ represents the first-order difference; is the estimated value of the time series; α is a constant term; β is the coefficient of the time trend term; γ is the coefficient of the lag term; δ is the coefficient of the difference lag term; ∈t is the white noise error term; Step 2: Determine the order of the ARMA model, and use the autocorrelation function ACF and partial autocorrelation function PACF to identify the order of the AR and MA models; observe the ACF graph to determine the order q of the MA, and the PACF graph to determine the order p of the AR; Step 3: Model estimation: According to the determined orders p and q, an ARMA (p, q) model is established; the maximum likelihood estimation (MLE) optimization method is used to estimate the model parameters; y t =c+φ1y t-1 +...+φ p y t-p +θ1∈ t-1 +...+θ q-q +∈ t where y t is the value of the time series at time t; c is the constant term; φ is the coefficient of the autoregressive term; θ is the coefficient of the moving average term; ∈t is the white noise error term; Step 4: Model test, use the Ljung-Box Q test to test whether the residual of the model is white noise; if the residual is not white noise, you may need to reselect the order of the model or consider other models: The formula for the Ljung-Box Q test is: Where n is the length of the time series; m is the number of lag periods; is the sample autocorrelation coefficient delayed by k periods; Step 5: Model prediction, use the estimated ARMA model to predict future temperature values; these predicted values ​​will serve as the temperature benchmark for normal valve operation, and the prediction formula is: in is the predicted value for the next h steps based on current and past information.

6. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S6, the specific algorithm steps of using the sliding window method to determine whether there is an abnormal temperature change are as follows: Step 1: Determine the window size and select an appropriate time window size N, which should be large enough to contain enough temperature data points; Step 2: Initialize the window and load the temperature data in the initial window into the window to prepare for calculation; Step 3: Calculate the statistics within the window, perform statistical analysis on the temperature data within the window, and calculate the mean and standard deviation statistics; The window mean calculation formula is: [mu frac1N sumi=1N xi]; The formula for calculating the window standard deviation is: [sigma sqrtfrac1N sumi=1N(xi-mu)^2; N is the window size; xi is the i-th data point in the window; mu is the mean of the data in the window; sigma is the standard deviation of the data within the window; Step 4: Set a threshold value. Set a threshold value for anomaly detection based on historical data or expert experience. Step 5: Sliding window. Each time new temperature data is collected, the window is slid forward by one data point, the oldest data point in the window is removed, and the latest data point is added. Step 6: Update the statistics, using the new data in the sliding window to update the statistics, that is, update the mean and standard deviation; Step 7: Anomaly detection, for each new data point in the window, calculate its deviation from the current window statistic, the calculation formula is: The calculation formula of the deviation is: [delta xi=xi-mu] Where: xi is the new data point; mu is the mean of the current window; The calculation formula for judging whether a new data point is abnormal is: [textIf|delta xi|>k*sigma,text thentextit{abnormal}]; Where: k is the set threshold multiple; sigma is the standard deviation of the current window; Step 8: Record and report anomalies. If an anomaly is detected, record the anomaly data and timestamp and report it to the operation and maintenance personnel.

7. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S7, if the temperature of the monitoring point suddenly rises while other relevant parameters remain unchanged, this indicates that there is internal leakage in the valve; the system will perform preliminary internal leakage diagnosis based on preset rules or simple logical judgments, such as when the temperature difference exceeds a set threshold.

8. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S8, the temperature change rate is extracted from the abnormal temperature data by calculating the ratio of the temperature difference between consecutive time points to time.

9. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S9, the specific algorithm steps of using XGBoost to establish the endoleak diagnosis model include: Step 1: Data preparation: collect and organize the temperature data of the main steam valve of the steam turbine under normal operation and internal leakage conditions; preprocess the data, including missing value processing, outlier processing, and feature engineering; divide the data set into a training set and a test set; Step 2: Define the loss function and use the logarithmic loss function to evaluate the performance of the model. The calculation formula is: [L(y,y')=-[ylog(y')+(1-y)log(1-y')]]; where y is the true label; y' is the probability predicted by the model; Step 3: Initialize the model and model parameters, including learning rate, number of trees, and tree depth; Step 4: Build a decision tree. For each tree, select the optimal features and split points to minimize the loss function. The objective function calculation formula is: [Obj(t)=sum(L(yi,y'i)+Omega(f_t))]; Where Obj(t) is the objective function of the tth tree; L is the loss function; Omega is the regularization term used to control the complexity of the model; Step 5: Calculate the gradient of the loss function for each sample to guide the growth of the decision tree. The gradient calculation formula is: [gi=partialfracpartialy'L(yi,y'i)|y'_i=current prediction]; where gi is the gradient of the ith sample; Step 6: Construct leaf nodes, find the optimal segmentation point according to the gradient, and assign data to leaf nodes; Step 7: Calculate the value of the leaf node. For each leaf node, calculate a value to minimize the objective function. The leaf node value calculation formula is: [wj=-frac1sum_{iin Ij}fracpartialpartialy'L(yi,y'i)|y'i=currentprediction]; Where wj is the value of the jth leaf node; Ij is the sample set assigned to the jth leaf node; Step 8: Update the model, add the newly built tree to the model, update the predicted value, The updated forecast calculation formula is: [y'i=y'i+eta*wj]; where eta is the learning rate; wj is the leaf node value of the new tree; Step 9: Iterate optimization and repeat steps 4 to 8 until the predetermined number of trees is reached or the loss function no longer decreases significantly. Through the above steps, an XGBoost model for internal leakage diagnosis of main steam valves of steam turbines is established.

10. The method for monitoring and diagnosing internal leakage of a main steam valve of a steam turbine according to claim 1, characterized in that: In step S10, the model will output an internal leakage probability or classification result based on the characteristic data; if the model output result shows that the internal leakage probability exceeds the set threshold, the system will immediately issue an early warning and notify the operation and maintenance personnel via SMS, email or control system alarm.

Images