A generator set turbine start-up and shutdown risk prediction system and method
By using multi-dimensional data monitoring and risk level assessment models, the problem of diagnosing equipment failures under frequent start-stop of thermal power units has been solved, enabling early warning and fault tracing, reducing the risk of unplanned shutdowns, and improving equipment safety and operational reliability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUADIAN LAIZHOU POWER GENERATION
- Filing Date
- 2026-03-10
- Publication Date
- 2026-05-26
AI Technical Summary
Existing single vibration monitoring systems cannot integrate thermal parameters for multi-dimensional diagnosis, making it difficult to meet the safety requirements of complex operating conditions under frequent start-stop of thermal power units, resulting in a high risk of equipment failure and unplanned shutdown.
By monitoring multidimensional data of generator turbines in real time, a risk level assessment model is constructed. Anomaly data analysis and risk prediction are performed by combining the start-up and shutdown process feature database, and control suggestions are provided. The risk level is determined by using the multi-source state estimation method, and a dynamic assessment model is constructed through machine learning for real-time display and optimization.
It enables multi-dimensional risk monitoring of the turbine start-up and shutdown process, accurately quantifies risk factors, provides early warning and fault tracing, reduces the probability of unplanned shutdowns, and improves equipment safety and operational reliability.
Smart Images
Figure CN122089081A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of thermal power unit monitoring technology, specifically to a generator unit turbine start-up and shutdown risk prediction system and method. Background Technology
[0002] In the past two years, the number of starts-up operations of coal-fired power units has increased significantly, with the annual average rising from about 5 times in 2022 to nearly 14 times in 2024. With the development of new energy sources, this high frequency of start-ups and shutdowns is expected to become the norm. This seriously threatens the reliability and lifespan of the unit equipment, especially older units, which are prone to a series of problems such as jamming of the steam inlet mechanism during start-ups and shutdowns, increasing maintenance costs and downtime. With the construction of new power systems and the advancement of "dual carbon" goals, frequent start-ups and shutdowns for peak shaving of thermal power units have become the norm, highlighting the risks of equipment failures and unplanned shutdowns caused by abnormal shaft vibration. Existing single vibration monitoring systems cannot integrate thermodynamic parameters for multi-dimensional diagnosis, making it difficult to meet the safety requirements under complex operating conditions. Summary of the Invention
[0003] To address the above problems, this invention provides a method for predicting the start-up and shutdown risks of a generator turbine, comprising the following steps: S1. Real-time monitoring of the operating status data of the generator set turbine, and collection of historical start-up and shutdown data to generate a start-up and shutdown process feature database; S2. Analyze and process the monitored and collected data, and transform it into structured data; S3. Generate a constraint system for monitoring the start-up and shutdown status of the unit based on structured data, and analyze the correlation between various data in the constraint system for monitoring the start-up and shutdown status of the unit; S4. Construct a risk level assessment model, combine the start-up and shutdown process characteristic database to predict the risks of abnormal monitoring data, determine the risk level of the predicted potential risks, and provide relevant control recommendations; The specific operations for analyzing and processing the monitored data in step S2 include: handling missing data values, data denoising, numerical filtering, and rationality judgment; wherein, the rationality judgment specifically involves: performing logical pattern judgment on the monitored data based on the principle of thermal balance, and identifying and alarming data that violates the principle of thermal balance.
[0004] Furthermore, the operating status data of the generator turbine in step S1 specifically includes: status signals; start-stop sequence control signals; start-stop process parameters; key bearing parameters: bearing metal temperature, vibration, frequency doubling, expansion difference, displacement, and eccentricity; core indicators of the oil system: lubricating oil pressure, temperature, and control oil quality; and operating data of the thermal system: main steam parameters and cylinder wall temperature.
[0005] Furthermore, the collection of historical start-up and shutdown data in step S1 specifically includes: historical operating parameter data of generator set start-up and shutdown, handling methods when a fault occurs, and operating parameter data of the generator set after handling.
[0006] Furthermore, the unit start-up and shutdown status monitoring constraint system in step S3 specifically includes: Generator set status monitoring: Real-time recording of start-up and shutdown alarm information and trend analysis based on historical start-up and shutdown data to identify safety risk factors; Start-stop process curve record comparison: Generate start-stop curves based on start-stop process parameters and compare and analyze them with historical curves; Shaft system, oil system and main unit system monitoring: Real-time monitoring based on key bearing parameters, core indicators of the oil system and operating data of the thermal system, recording trends and issuing early warnings for exceeding limits.
[0007] Furthermore, the analysis of the correlation between various data in the unit start-up and shutdown status monitoring constraint system in step S3 specifically involves: analyzing the correlation between each pair of monitored data or between one and many data, and calculating the correlation coefficient between indicators.
[0008] Furthermore, it also includes a feedback optimization step: S5. The user selects and executes control suggestions, and optimizes based on user feedback and real-time monitoring of changes in generator start-up and shutdown parameter data.
[0009] Furthermore, the risk prediction of the monitored abnormal data and the determination of the risk level of the predicted potential risks in step S4 are specifically determined by using a multi-source state estimation method, which includes the following steps: S41. Select m observation vectors from historical data during normal operation of the generator turbine and construct a memory matrix D; S42. Obtain the observation vector of the generator turbine operation at a certain moment; S43. Using Euclidean distance as a nonlinear operator, calculate the m-dimensional weight vector W of the observation vector obtained in step S42 using the least squares method; S44. Estimate the residuals of the observation point at this moment in step S42, and classify the risk level according to the size of the residuals: the larger the residuals, the higher the risk level of the point.
[0010] A generator set turbine start-up and shutdown risk prediction system includes a multi-dimensional data monitoring module, a data processing and analysis module, a fault signal tracking and location module, a decision-making and feedback optimization module, and an interaction and visualization module; The multidimensional data monitoring module is used to monitor and collect the start-up and shutdown parameter data of the generator set. This module is configured to perform the steps of real-time monitoring of the operating status data of the generator set turbine and collecting historical start-up and shutdown data. The data processing and analysis module is used to preprocess the collected data and perform correlation analysis. This module is configured to perform the steps of analyzing and processing the monitored and collected data and converting it into structured data, as well as analyzing the correlation between various data in the unit start-up and shutdown status monitoring constraint system. The fault signal tracking and location module is used to locate and track abnormal signals monitored in the data processing and analysis module. The decision and feedback optimization module is used to generate control suggestions for the monitored abnormal signals and to optimize decisions based on user feedback. This module is configured to execute the steps of building a risk level assessment model and making risk predictions, as well as the optimization based on user feedback. The interaction and visualization module is used to display the monitored data and its correlation, locate abnormal signals, and generate control suggestions, and to interact with the user.
[0011] This invention provides a generator set turbine start-up and shutdown risk prediction system and method, which has the following beneficial effects: This invention integrates multi-dimensional risk monitoring data to construct a machine learning-based dynamic assessment model. This model displays the safety risk levels of turbine generator sets during start-up, shutdown, and operation in real time, accurately quantifies the weights of risk factors, and enables early warning and tracing of potential hazards in key components. The system effectively identifies hidden fault chains under complex operating conditions, providing operators with closed-loop decision support based on "risk threshold - adjustment direction - priority ranking," significantly improving equipment safety and operational reliability, and reducing the probability of unplanned outages. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.
[0013] Figure 1 The method flowchart provided by the present invention. Detailed Implementation
[0014] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.
[0015] The following detailed description of the implementation method of the present invention is in conjunction with the accompanying drawings. The description is only a partial embodiment and not all embodiments. For clarity, representations and descriptions unrelated to the present invention are omitted in the drawings and description.
[0016] To provide a clearer understanding of the technical features, objectives, and beneficial effects of this invention, the following detailed description of the technical solution is provided. Obviously, the described embodiments are only a portion of the embodiments of this invention, not all of them, and should not be construed as limiting the scope of implementation of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the protection scope of this invention.
[0017] During unit start-up and shutdown phases and peak-shaving operations, artificial intelligence (AI) technology is used to analyze historical data and identify key safety risk factors, such as equipment fatigue, abnormal parameter fluctuations, and improper operation. Deep learning models predict potential faults, issue early warnings, and prevent safety accidents. Simultaneously, a risk level assessment model is established to identify operational deviations in real time. Combined with AI's decision support capabilities, the model automatically determines the risk level based on the severity and probability of occurrence of risk factors and provides corresponding control suggestions or automatically executes control measures.
[0018] like Figure 1 As shown, the present invention provides a method for predicting the start-up and shutdown risks of a generator turbine, comprising the following steps: S1. Real-time monitoring of the turbine's operating status data (status signals; start-stop sequence control signals; start-stop process parameters; key bearing parameters: bearing metal temperature, vibration, frequency doubling, expansion difference, displacement, eccentricity; core indicators of the oil system: lubricating oil pressure, temperature, control oil quality; operating data of the thermal system: main steam parameters, cylinder wall temperature), and collection of historical start-stop data (historical operating parameter data of generator set start-stop, handling methods when faults occur, and operating parameter data of the generator set after handling) to generate a start-stop process feature database.
[0019] S2. Analyze and process the monitored and collected data, and transform it into structured data using knowledge extraction technology. The specific operations for analyzing and processing the monitored data include: handling missing data values (noisy data replacement; discrete point removal; data cleaning through binning, etc., to remove data unsuitable for data mining and analysis), data denoising, numerical filtering, and rationality judgment; specifically, the rationality judgment involves: based on the principle of thermal balance, performing logical pattern judgment on the monitored data, identifying and issuing alarms for data that violates the principle of thermal balance.
[0020] S3. Based on structured data, generate a constraint system for monitoring the start-up and shutdown status of the unit, and analyze the correlation between various data in the constraint system: analyze the correlation between pairs or one-to-many of the monitored data, and calculate the correlation coefficient between indicators. By integrating Pearson correlation analysis, cosine similarity, grey relational analysis and mutual information, perform data correlation analysis on the collected data. The calculation formula for Pearson correlation analysis is as follows: In the formula, R is the Pearson correlation coefficient, and Xi is the i-th sample value of parameter X. Let X be the mean of parameter X, and Yi be the i-th sample value of parameter Y. Let Y be the mean value of parameter Y.
[0021] The formula for calculating cosine similarity is: , where cos(θ) is the cosine similarity between parameters A and B, Ai is the i-th sample value of parameter A, Bi is the i-th sample value of parameter B, n is the total number of samples, i=1,2,...,n.
[0022] The formula for calculating grey relational analysis is: , Where x0(k) is the k-th sample value of the reference sequence x0, x i (k) is the comparison sequence x i The kth sample value; ρ is the correlation coefficient for comparing sequence i; ρ is the resolution coefficient, 0 < ρ < 1; r oi To compare the correlation of sequence i, m is the sample size.
[0023] The formula for calculating mutual information is: , where I(X;Y) is the mutual information of parameters X and Y, P(x,y) is the joint probability distribution of parameters X and Y, and P(x) and P(y) are the marginal probability distributions of parameters X and Y, respectively.
[0024] The specific constraints of the unit start-up and shutdown status monitoring system include: Generator set status monitoring: Real-time recording of start-up and shutdown alarm information and trend analysis based on historical start-up and shutdown data to identify safety risk factors; Start-stop process curve record comparison: Generate start-stop curves based on start-stop process parameters and compare and analyze them with historical curves; Shaft system, oil system and main unit system monitoring: Real-time monitoring based on key bearing parameters, core indicators of the oil system and operating data of the thermal system, recording trends and issuing early warnings for exceeding limits.
[0025] S4. Construct a risk level assessment model, combine the start-up and shutdown process characteristic database to predict the risks of monitored abnormal data, determine the risk level of the predicted potential risks, and provide relevant control recommendations.
[0026] Among them, the detection of abnormal data adopts the multi-source state estimation method, and the MSET modeling principle is as follows: 1. Observation Vector X(i): First, a set of time-series data of the power plant unit operating normally is given. At time i, data from n measuring points of the system can be observed, that is: .
[0027] 2. Historical Observation Vector Set: During this time period, all observation vectors in normal operating state constitute the normal historical working space, which contains o observation vectors, i.e.: .
[0028] 3. Memory Matrix: From the historical observation vector set of normal operation, use re-clustering to find m observation vectors that can cover the characteristics of all operating conditions. Record the m operating condition data into the memory matrix, i.e.: ; The memory matrix is constructed by selecting m representative data points from historical data under normal conditions, which are the data features extracted when the equipment is running normally.
[0029] 4. State estimation: The input to the model is the observation vector X at a certain moment. obs The output is the prediction result X for this observation vector. est For any observation vector, the model generates an m-dimensional weight vector W, i.e.: ; so: ; Therefore, the prediction vector output by the multivariate state estimation is a linear combination of the m observation vectors in the memory matrix constructed during normal operation.
[0030] The residuals between the observation vectors and the prediction results in the model are: ; By minimizing the residuals using the least squares method, the weight vector is obtained as follows: ; in, To represent non-linear operators, there are many methods; here, we mainly use the Euclidean distance, i.e.: ; The predicted values for each measuring point of the equipment are obtained using the multivariate state estimation model MSET: ; Calculate the estimated residuals: ; where X obs For the observed value, X est Its estimated value; X obs,i Let X be the value of the i-th measurement point of the observed values. est,i This is the estimated value for the i-th measurement point.
[0031] Risk level classification: Risk levels are classified based on the magnitude of residual values. When the equipment is operating normally, the new input observation vector of MSET will be close to the historical vector in the memory matrix, resulting in a small residual between the model prediction value and the input observation vector. If the equipment malfunctions, its operating pattern changes, and the observation vector of the equipment will deviate from the normal sample distribution space during equipment operation, resulting in a large residual between the model prediction value and the input observation vector.
[0032] It also includes a feedback optimization step: S5. The user selects and executes control suggestions, and optimizes based on user feedback and real-time monitoring of changes in generator start-up and shutdown parameter data.
[0033] A generator set turbine start-up and shutdown risk prediction system includes a multi-dimensional data monitoring module, a data processing and analysis module, a fault signal tracking and location module, a decision-making and feedback optimization module, and an interaction and visualization module.
[0034] The multi-dimensional data monitoring module is used to monitor and collect start-up and shutdown parameter data of the generator set.
[0035] The data processing and analysis module is used to preprocess the collected data and perform correlation analysis.
[0036] The fault signal tracking and location module is used to locate and track abnormal signals detected by the data processing and analysis module.
[0037] The decision and feedback optimization module is used to generate control suggestions for monitored abnormal signals and to optimize decisions based on user feedback.
[0038] The interaction and visualization module is used to display the monitored data and its correlation, abnormal signal location, and generated control suggestions and interact with the user: real-time monitoring of operating status signals and start-stop sequence control signals, dynamically presenting key parameters through a visual interface; tracking the source of fault signal conditions, accurately locating and reporting abnormal state trigger points; integrating a fault / abnormal information collection module, and combining it with a knowledge base to generate standardized operation suggestions.
[0039] This invention integrates multi-dimensional risk monitoring data to construct a machine learning-based dynamic assessment model. This model displays the safety risk levels of turbine generator sets during start-up, shutdown, and operation in real time, accurately quantifies the weights of risk factors, and enables early warning and tracing of potential hazards in key components. The system effectively identifies hidden fault chains under complex operating conditions, providing operators with closed-loop decision support based on "risk threshold - adjustment direction - priority ranking," significantly improving equipment safety and operational reliability, and reducing the probability of unplanned outages.
[0040] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.
Claims
1. A method for predicting the start-up and shutdown risks of a generator set turbine, characterized in that, Includes the following steps: S1. Real-time monitoring of the operating status data of the generator set turbine, and collection of historical start-up and shutdown data to generate a start-up and shutdown process feature database; S2. Analyze and process the monitored and collected data, and transform it into structured data; S3. Generate a constraint system for monitoring the start-up and shutdown status of the unit based on structured data, and analyze the correlation between various data in the constraint system for monitoring the start-up and shutdown status of the unit; S4. Construct a risk level assessment model, combine the start-up and shutdown process characteristic database to predict the risks of abnormal monitoring data, determine the risk level of the predicted potential risks, and provide relevant control recommendations; The specific operations for analyzing and processing the monitored data in step S2 include: handling missing data values, data denoising, numerical filtering, and rationality judgment; wherein, the rationality judgment specifically involves: performing logical pattern judgment on the monitored data based on the principle of thermal balance, and identifying and alarming data that violates the principle of thermal balance.
2. The generator set turbine start-up and shutdown risk prediction method according to claim 1, characterized in that, The specific operating status data of the generator set turbine in step S1 includes: status signals; start-stop sequential control signals; start-stop process parameters; key bearing parameters: bearing metal temperature, vibration, frequency doubling, expansion difference, displacement, and eccentricity; core indicators of the oil system: lubricating oil pressure, temperature, and control oil quality; and operating data of the thermal system: main steam parameters and cylinder wall temperature.
3. The generator set turbine start-up and shutdown risk prediction method according to claim 2, characterized in that, The collection of historical start-up and shutdown data in step S1 specifically includes: historical operating parameter data of generator set start-up and shutdown, handling methods when a fault occurs, and operating parameter data of the generator set after handling.
4. The generator set turbine start-up and shutdown risk prediction method according to claim 2, characterized in that, The unit start-up and shutdown status monitoring constraint system in step S3 specifically includes: Generator set status monitoring: Real-time recording of start-up and shutdown alarm information and trend analysis based on historical start-up and shutdown data to identify safety risk factors; Start-stop process curve record comparison: Generate start-stop curves based on start-stop process parameters and compare and analyze them with historical curves; Shaft system, oil system and main unit system monitoring: Real-time monitoring based on key bearing parameters, core indicators of the oil system and operating data of the thermal system, recording trends and issuing early warnings for exceeding limits.
5. The generator set turbine start-up and shutdown risk prediction method according to claim 4, characterized in that, The specific steps in step S3, which involve analyzing the correlation between various data in the unit start-up and shutdown status monitoring constraint system, are as follows: analyzing the correlation between each pair of monitored data or between one and many data, and calculating the correlation coefficient between indicators.
6. The generator set turbine start-up and shutdown risk prediction method according to claim 1, characterized in that, It also includes a feedback optimization step: S5. The user selects and executes control suggestions, and optimizes based on user feedback and real-time monitoring of changes in generator start-up and shutdown parameter data.
7. The generator set turbine start-up and shutdown risk prediction method according to claim 1, characterized in that, The S4 step, which involves risk prediction of the monitored abnormal data and risk level determination of the predicted potential risks, specifically utilizes a multi-source state estimation method for detection and determination, and includes the following steps: S41. Select m observation vectors from historical data during normal operation of the generator turbine and construct a memory matrix D; S42. Obtain the observation vector of the generator turbine operation at a certain moment; S43. Using Euclidean distance as a nonlinear operator, calculate the m-dimensional weight vector W of the observation vector obtained in step S42 using the least squares method; S44. Estimate the residuals of the observation point at this moment in step S42, and classify the risk level according to the size of the residuals: the larger the residuals, the higher the risk level of the point.
8. A generator set turbine start-up and shutdown risk prediction system, used to implement the generator set turbine start-up and shutdown risk prediction method according to any one of claims 1 to 7, characterized in that, It includes a multi-dimensional data monitoring module, a data processing and analysis module, a fault signal tracking and location module, a decision-making and feedback optimization module, and an interaction and visualization module; The multidimensional data monitoring module is used to monitor and collect the start-up and shutdown parameter data of the generator set. The module is configured to perform the steps described in claim 1: real-time monitoring of the operating status data of the generator set turbine and collection of historical start-up and shutdown data. The data processing and analysis module is used to preprocess the collected data and perform correlation analysis. The module is configured to perform the steps described in claim 1: analyzing and processing the monitored and collected data and converting it into structured data, and analyzing the correlation between various data in the unit start-up and shutdown status monitoring constraint system. The fault signal tracking and location module is used to locate and track abnormal signals monitored in the data processing and analysis module. The decision and feedback optimization module is used to generate control suggestions for the monitored abnormal signals and to optimize decisions based on user feedback. This module is configured to perform the steps of constructing a risk level assessment model and making risk prediction as described in claim 1, and the steps of optimizing based on user feedback as described in claim 6. The interaction and visualization module is used to display the monitored data and its correlation, locate abnormal signals, and generate control suggestions, and to interact with the user.