Method and system for monitoring and analyzing steam leakage amount of steam turbine shaft seal

Abstract

The invention relates to the field of steam turbine shaft seal steam leakage monitoring, and provides a steam turbine shaft seal steam leakage monitoring analysis method and system.In the system, distributed monitoring sensors are used for collecting multi-source heterogeneous shaft seal working parameters of a steam turbine in real time and conducting fusion processing on the shaft seal working parameters; the monitoring cloud platform is used for generating a steam leakage trend curve, correcting the steam leakage trend curve, generating a fitting prediction curve in combination with a rotor abnormal amplitude, performing prediction analysis on shaft seal working parameters in combination with a pre-trained shaft seal prediction model, and outputting a shaft seal state evaluation result; the early warning feedback module is used for triggering the classification early warning instruction, encrypting the classification early warning instruction and pushing the classification early warning instruction to the visual user side; and the visual user side is used for analyzing the classification early warning instruction into classification early warning information and visually presenting the classification early warning information. The influence of interference factors such as noise is reduced, the data quality is improved, and the shaft seal steam leakage amount can be monitored more accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of steam turbine shaft seal steam leakage monitoring, and in particular to a steam turbine shaft seal steam leakage monitoring and analysis method and system. Background Art

[0002] Steam turbines are core equipment in thermal power generation, chemical engineering, and industrial drive applications. The sealing performance of their shaft seal systems directly impacts unit efficiency, safety, and operational economy. As a critical component for ensuring proper operation, the steam turbine shaft seal system plays a vital role in preventing high-pressure steam leakage and air intrusion.

[0003] Shaft seal steam leakage is a key indicator of a steam turbine's shaft seal system performance. Appropriate shaft seal steam leakage ensures the turbine's internal sealing effectiveness, maintains normal steam pressure and temperature distribution, and thus ensures efficient operation of the turbine. Excessive shaft seal steam leakage results in significant steam waste. Excessive steam leakage can also generate localized high temperatures and high pressures in the shaft seal area, causing erosion and wear to seal components (such as seal teeth and sleeves), shortening the seal system's service life and increasing equipment maintenance costs.

[0004] Chinese patent application CN102798502A discloses a method for obtaining the leakage amount of a turbine shaft seal system, comprising: obtaining the thermal design parameters of the sealing system of each leakage end of a pressure cylinder in the turbine shaft seal system; wherein the thermal design parameters of the sealing system include: the theoretical leakage amount of each leakage end, the theoretical exhaust driving back pressure of each leakage end, and the theoretical exhaust driving enthalpy of each leakage end; obtaining the total flow coefficient of each leakage end of the pressure cylinder based on the thermal design parameters of the sealing system of each leakage end of the pressure cylinder; obtaining the real-time exhaust driving back pressure and real-time exhaust driving enthalpy of each leakage end of the pressure cylinder; and obtaining the real-time leakage amount of each leakage end of the pressure cylinder using the total flow coefficient of each leakage end of the pressure cylinder, the real-time exhaust driving back pressure, and the real-time exhaust driving enthalpy of the leakage end of the pressure cylinder. However, the above method mainly focuses on a few thermal design parameters at the leakage end of the pressure cylinder when obtaining real-time leakage. It lacks effective fusion processing methods for the multi-source heterogeneous parameters involved in the turbine shaft seal system (such as temperature, pressure, and vibration at different locations), and is unable to filter and perform anti-interference processing on the collected data. As a result, interference factors such as noise may affect the accurate measurement of steam leakage, resulting in limited accuracy of the measurement results. Summary of the Invention

[0005] The present invention aims to solve at least one of the problems existing in the prior art and provides a method and system for monitoring and analyzing steam leakage of a steam turbine shaft seal.

[0006] In one aspect of the present invention, a steam turbine shaft seal steam leakage monitoring and analysis system is provided, the system comprising a distributed monitoring sensor, a monitoring cloud platform, an early warning feedback module, and a visual user terminal;

[0007] The distributed monitoring sensor is deployed on the shaft seal side of the steam turbine and is used to collect the multi-source heterogeneous shaft seal operating parameters of the steam turbine in real time, and fuse the shaft seal operating parameters to extract the rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal operating parameters;

[0008] The monitoring cloud platform is used to obtain the fused shaft seal operating parameters, generate a steam leakage trend curve based on the steam leakage monitoring value, correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, generate a fitting prediction curve based on the correction curve combined with the rotor abnormal amplitude, perform prediction analysis on the shaft seal operating parameters in combination with a pre-trained shaft seal prediction model, and output a shaft seal status assessment result;

[0009] The early warning feedback module is used to obtain the fitting prediction curve and the shaft sealing status assessment result, trigger a classification early warning instruction based on the fitting prediction curve and the shaft sealing status assessment result, and encrypt and push the classification early warning instruction to the visualization user terminal;

[0010] The visualization user terminal is used to parse the classification warning instruction into classification warning information and visually present the classification warning information.

[0011] Optionally, the distributed monitoring sensor includes:

[0012] At least one set of data collectors for real-time collection of the shaft seal operating parameters; wherein the shaft seal operating parameters include steam parameters, pressure parameters, flow parameters, temperature parameters, shaft seal vibration displacement parameters, shaft seal mechanical wear parameters, and turbine power output parameters;

[0013] a signal processing unit, configured to load the shaft seal operating parameters in real time, perform multiple rounds of iterative fusion on the shaft seal operating parameters based on an iterative consensus algorithm, and output the fused shaft seal operating parameters;

[0014] A pre-processing unit is used to obtain the shaft seal working parameters after fusion processing, and extract the rotor abnormal amplitude, the operating condition value, the steam leakage monitoring value, and the steam parameter fluctuation value from the shaft seal working parameters after fusion processing.

[0015] Optionally, the signal processing unit is configured to perform multiple rounds of iterative fusion on the shaft seal operating parameters based on an iterative consensus algorithm, including:

[0016] The signal processing unit is used for:

[0017] Obtaining the shaft seal working parameters, processing missing values ​​in the shaft seal working parameters using a filling method, and deleting duplicate values ​​to obtain a data cleaning set;

[0018] Performing filtering processing on the data cleaning set based on an unscented Kalman filter, and outputting the filtered data cleaning set;

[0019] The data cleaning set after filtering is expressed as:

[0020]

[0021] Among them, s L (n) represents the data cleaning set after filtering, are the maximum and minimum amplitudes of the hth harmonic when processed by the unscented Kalman filter, L and D are the filtering times and filter orders of the unscented Kalman filter, respectively. s ,θ h Represent the parameter sampling frequency, the phase of the hth harmonic, Δω h is the frequency sampling angle difference of the hth harmonic and G(ω) represents the magnitude response of the unscented Kalman filter, They are respectively the sampling angle difference of the positive frequency component and the negative frequency component of the hth harmonic, f r 、f h Respectively represent the rated frequency and fundamental frequency;

[0022] Capture the historical parameters of the shaft seal, and based on the historical parameters, mechanical wear parameters, and pressure parameters of the shaft seal, initialize the dynamic weight of the parameter type in the consensus mechanism, reset the dynamic weight of the parameter type to fusion priority data, and retain the parameter type with high fusion priority data;

[0023] The dynamic weight of parameter types, the number of types, and parameter signal attributes are used as fusion constraints to determine the iteration rounds and verification parameters in the consensus mechanism;

[0024] After each iteration, the shaft seal operating parameters are checked to see if they meet the fusion constraints and verification parameters. If so, the iterative fusion is stopped and the fusion consensus value is output.

[0025] Optionally, the monitoring cloud platform includes:

[0026] a curve generating unit, configured to obtain the shaft seal operating parameters after fusion processing, and generate the steam leakage trend curve based on the steam leakage monitoring value;

[0027] a steam leakage prediction unit, configured to correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, and generate the fitting prediction curve based on the correction curve in combination with the rotor abnormal amplitude;

[0028] A shaft sealing prediction unit is used to predict and analyze the shaft sealing working parameters based on the fitting prediction curve and the pre-trained shaft sealing prediction model, and output the shaft sealing status evaluation result.

[0029] Optionally, the steam leakage amount prediction unit is configured to correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, including:

[0030] The steam leakage prediction unit is used to:

[0031] Loading the steam leakage trend curve, extracting an interpolation vector composed of interpolation points in the steam leakage trend curve, and constructing a continuous distribution model based on the interpolation vector, wherein the continuous distribution model includes the interpolation vector, an operating condition coefficient matrix, and a fluctuation coefficient matrix;

[0032] Obtaining the continuous distribution model, obtaining an operating condition influence constraint and a fluctuation influence constraint based on the operating condition coefficient matrix and the fluctuation coefficient matrix of the continuous distribution model, respectively, aggregating the operating condition influence constraint and the fluctuation influence constraint, solving a min-max problem when the continuous distribution model is 0, and obtaining an optimal correction coefficient for the interpolation point constraint on the steam leakage trend curve;

[0033] Obtain the optimal correction coefficient corresponding to at least one set of interpolation points, iterate the interpolation vector of the interpolation point based on the optimal correction coefficient, the operating condition influence constraint, and the fluctuation influence constraint, update the interpolation vector after iteration, update the interpolation vector of the interpolation point, and output the correction curve.

[0034] Optionally, the steam leakage prediction unit is configured to generate the fitting prediction curve based on the correction curve in combination with the abnormal amplitude of the rotor, including:

[0035] The steam leakage prediction unit is used to:

[0036] Loading the correction curve, identifying the interpolation vector after iterative update in the correction curve, constructing a static adjacency matrix based on a fuzzy clustering algorithm combined with a graph neural network, using the rotor abnormal amplitude as a constraint, and the operating condition impact constraint and the fluctuation impact constraint as prior probabilities, transforming the static adjacency matrix to obtain a dynamic adjacency matrix;

[0037] Extracting the temporal and spatial features of the dynamic adjacency matrix based on a time series convolutional neural network to obtain corresponding feature extraction results, fusing the feature extraction results using a BP neural network combined with an ELM algorithm to output a temporal interpolation prediction value;

[0038] At least one set of interpolation prediction quantities and their corresponding prediction time points are obtained, and the fitting prediction curve is generated based on the interpolation prediction quantities and their corresponding prediction time points.

[0039] Optionally, the interpolation vector after updating iteration is expressed as:

[0040]

[0041] Among them, T(x) t represents the interpolation vector at time t before the update iteration, T(x) t+1 Represents T(x) t The interpolation vector at time t+1 after the update iteration, G(x), U (l) are the working condition impact constraint value and the fluctuation impact constraint value respectively, σ(·) represents the activation function of the continuous distribution model, w x is the weight matrix of the interpolation vector before updating iteration, b x represents the bias term of the activation function, q1 and q2 are the weight values ​​of the working condition influence constraint and the fluctuation influence constraint respectively, and λ best is the optimal correction coefficient constrained by the interpolation point and j is the interpolation point number, J is the number of interpolation points, Δt represents the sampling time difference, is the steam leakage at the current time t, The steam leakage in the previous sampling period is also the steam leakage at time t-1, max(T(x) t ,0) means taking T(x) t and the maximum value of 0, Indicates taking and 0.

[0042] Optionally, the shaft sealing prediction unit is configured to perform prediction analysis on the shaft sealing operating parameters based on the fitted prediction curve and a pre-trained shaft sealing prediction model, and output the shaft sealing status assessment result, including:

[0043] The shaft seal prediction unit is used for:

[0044] Acquiring the shaft seal operating parameters, and calculating the shaft seal brittleness index and the shaft seal imbalance within the adoption cycle based on the shaft seal operating parameters;

[0045] Establishing a characteristic vector for a periodic time period based on the shaft seal brittleness index, the shaft seal imbalance, and the interpolation vector after updating and iteration;

[0046] Using the shaft seal prediction model to perform band periodicity detection and time segment interception on the feature vector, obtaining at least one group of intercepted time segments, performing convolution fusion on the feature vectors within the time segments based on the CFCs method, and performing global pooling and average pooling processing on the convolution fused feature vectors through the global pooling layer and average pooling layer in the shaft seal prediction model;

[0047] Obtaining global pooling and average pooling results, forming a residual connection between the global pooling and average pooling results and the feature vector to obtain a feature evaluation matrix;

[0048] Taking the feature evaluation matrix as input, the feature evaluation matrix is ​​linearly transformed by a weight matrix, and then nonlinearly transformed by a nonlinear activation function to obtain a shaft seal evaluation value.

[0049] Another aspect of the present invention provides a method for monitoring and analyzing steam leakage of a steam turbine shaft seal, the method comprising:

[0050] Real-time collection of multi-source heterogeneous shaft seal operating parameters of the steam turbine, fusion processing of the shaft seal operating parameters, and extraction of rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal operating parameters;

[0051] Acquire the fused shaft seal operating parameters, generate a steam leakage trend curve based on the steam leakage monitoring value, correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, generate a fitting prediction curve based on the correction curve and the rotor abnormal amplitude, perform prediction analysis on the shaft seal operating parameters in combination with a pre-trained shaft seal prediction model, and output a shaft seal status assessment result;

[0052] acquiring the fitting prediction curve and the shaft seal status assessment result, and triggering a classification warning instruction based on the fitting prediction curve and the shaft seal status assessment result;

[0053] The classified warning instructions are parsed into classified warning information, and the classified warning information is visually presented.

[0054] Optionally, the real-time collection of multi-source heterogeneous shaft seal operating parameters of the steam turbine and the fusion processing of the shaft seal operating parameters to extract rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal operating parameters include:

[0055] Real-time collection of the shaft seal operating parameters; wherein the shaft seal operating parameters include steam parameters, pressure parameters, flow parameters, temperature parameters, shaft seal vibration displacement parameters, shaft seal mechanical wear parameters, and turbine power output parameters;

[0056] Loading the shaft seal working parameters in real time, performing multiple rounds of iterative fusion on the shaft seal working parameters based on an iterative consensus algorithm, and outputting the fused shaft seal working parameters;

[0057] The shaft seal working parameters after fusion processing are acquired, and the rotor abnormal amplitude, the operating condition value, the steam leakage monitoring value, and the steam parameter fluctuation value are extracted from the shaft seal working parameters after fusion processing.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. By setting up distributed monitoring sensors, on the one hand, distributed fusion processing of multi-source heterogeneous shaft seal operating parameters is possible, significantly reducing system load. On the other hand, filtering and anti-interference processing are performed on the shaft seal operating parameters generated on the shaft seal side of the turbine, thereby reducing the impact of noise on shaft seal steam leakage monitoring and analysis. Distributed monitoring sensors can simultaneously perform filtering and anti-interference processing when fusing the collected multi-source heterogeneous parameters. By filtering the data, the impact of interference factors such as noise is reduced, data quality is improved, and shaft seal steam leakage is more accurately monitored, resolving the insufficient anti-interference capability of existing technologies.

[0060] 2. By setting up a monitoring cloud platform, in-depth analysis of real-time data can be performed. Combining historical data with pre-set analysis models, this allows for a comprehensive assessment of the shaft seal system's performance, providing operators with accurate monitoring information. By setting up a curve generation unit, a steam leakage prediction unit, and a shaft seal prediction unit within the monitoring cloud platform, the curve generation unit is used to generate leakage trend curves to provide real-time data feedback. The steam leakage prediction unit is used to generate fitted prediction curves for dynamic prediction of leakage trends. The shaft seal prediction unit is used to output shaft seal status assessment results. This enables the early warning feedback module to trigger classified early warning instructions, thus forming a closed loop of "monitoring-prediction-decision-execution."

[0061] 3. By performing multiple rounds of iterative fusion of the shaft seal operating parameters based on an iterative consensus algorithm, and verifying and adjusting the fusion results with each iteration, the fusion results gradually approach the true values, significantly improving the accuracy of the fusion results. Furthermore, by combining filtering techniques such as the unscented Kalman filter, noise and outliers in the data can be effectively removed, reducing the impact of noise on the fusion results and improving data reliability. By performing data preprocessing and fusion at the sensor end, the computational burden of the monitoring cloud platform is reduced, improving the processing efficiency of the entire system.

[0062] 4. During the correction of the steam leakage trend curve, the optimal correction coefficient for the interpolation point constraints is obtained by solving the min-max problem. This allows for precise correction of the steam leakage curve, improving its accuracy and reliability. By updating the interpolation vector based on the optimal correction coefficient, the shape of the steam leakage trend curve can be dynamically adjusted to better align with actual trends. In the process of generating a fitted prediction curve based on the correction curve combined with the rotor abnormal amplitude, the static adjacency matrix is ​​converted into a dynamic adjacency matrix by introducing the rotor abnormal amplitude as a constraint. This allows for real-time reflection of changes in system status and enhances the model's dynamic adaptability. Furthermore, by utilizing a temporal convolutional neural network to extract the temporal and spatial features of the dynamic adjacency matrix, it is possible to capture temporal changes and spatial correlation information in the data, providing rich feature support for prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] One or more embodiments are exemplarily illustrated by pictures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0064] Figure 1 A schematic structural diagram of a steam turbine shaft seal steam leakage monitoring and analysis system provided in one embodiment of the present invention;

[0065] Figure 2 A flowchart of steps performed by the signal processing unit 120 according to another embodiment of the present invention;

[0066] Figure 3 A flowchart of the steps performed by the steam leakage prediction unit 220 according to another embodiment of the present invention;

[0067] Figure 4 A flowchart of the steps performed by the shaft seal prediction unit 230 according to another embodiment of the present invention;

[0068] Figure 5 A flow chart of a method for monitoring and analyzing steam leakage of a steam turbine shaft seal provided in another embodiment of the present invention;

[0069] Figure 6 Another embodiment of the present invention provides Figure 5 Flowchart of step S10 in FIG. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, each embodiment of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be understood by those skilled in the art that in each embodiment of the present invention, many technical details are provided to enable the reader to better understand the present invention. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present invention can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined and referenced with each other under the premise that there is no contradiction.

[0071] When obtaining the real-time leakage rate of a steam turbine, existing methods mainly focus on a few thermal design parameters at the leakage end of the pressure cylinder. They lack effective fusion processing methods for the multi-source heterogeneous parameters involved in the steam turbine shaft seal system (such as temperature, pressure, and vibration at different locations), and are unable to filter and perform anti-interference processing on the collected data. This makes it possible for interference factors such as noise to affect the accurate measurement of the leakage rate, resulting in limited accuracy of the measurement results. To this end, one embodiment of the present invention provides a steam turbine shaft seal leakage monitoring and analysis system, whose structure is as follows: Figure 1 As shown, it includes distributed monitoring sensors 100, a monitoring cloud platform 200, an early warning feedback module 300, and a visualization user terminal 400.

[0072] The distributed monitoring sensor 100 is deployed on the shaft seal side of the steam turbine to collect the multi-source heterogeneous shaft seal working parameters of the steam turbine in real time, and to fuse the shaft seal working parameters to extract the rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal working parameters.

[0073] It should be noted that the distributed monitoring sensor 100 can be deployed at different locations of the turbine shaft seal system to achieve comprehensive monitoring of the shaft seal working parameters, ensure the integrity and accuracy of data acquisition, and timely reflect the operating status of the turbine shaft seal system by real-time collection of shaft seal working parameters, providing timely data support for subsequent analysis and prediction.

[0074] By implementing distributed monitoring sensors, the system can not only integrate and process multiple, heterogeneous shaft seal operating parameters, significantly reducing system load, but also filter and perform anti-interference processing on the shaft seal operating parameters generated on the turbine's shaft seal side, thereby reducing the impact of noise on shaft seal steam leakage monitoring and analysis. Distributed monitoring sensors can simultaneously perform filtering and anti-interference processing when integrating and processing the collected, heterogeneous, multi-source parameters. By filtering the data, the impact of interference factors such as noise is reduced, improving data quality and enabling more accurate monitoring of shaft seal steam leakage, addressing the insufficient anti-interference capabilities of existing technologies.

[0075] For example, Figure 1 As shown, the distributed monitoring sensor 100 includes at least one set of data collector 110 , a signal processing unit 120 , and a pre-processing unit 130 .

[0076] At least one data collector 110 is used to collect shaft seal operating parameters in real time. These include steam parameters, pressure parameters, flow parameters, temperature parameters (including but not limited to cylinder temperature, rotor temperature, and shaft seal steam temperature), shaft seal vibration displacement parameters, shaft seal mechanical wear parameters (such as the wear and deformation of components such as the shaft seal teeth and shaft seal sleeve), and turbine power output parameters. Steam parameters include steam pressure parameters and steam temperature parameters. Steam pressure parameters include the shaft seal steam supply pressure, turbine extraction pressure at each stage, and exhaust pressure. The shaft seal steam supply pressure must be maintained within an appropriate range to ensure sufficient steam enters the shaft seal system and form an effective seal. If the shaft seal steam supply pressure is too low, the shaft seal will experience insufficient steam leakage, failing to effectively prevent air ingress. Conversely, if the shaft seal steam supply pressure is too high, the shaft seal will experience excessive steam leakage, resulting in steam waste and erosion of related components. Steam temperature parameters include the shaft seal steam supply temperature and the metal temperature of the rotor and cylinder. The shaft seal steam supply temperature affects the steam state at the seal gap and the thermal expansion of the seal components. The distribution of metal temperature has an important influence on the thermal stress and deformation of the turbine, and also indirectly affects the change of the shaft seal clearance.

[0077] The signal processing unit 120 is used to load the shaft seal working parameters in real time, perform multiple rounds of iterative fusion on the shaft seal working parameters based on an iterative consensus algorithm, and output the fused shaft seal working parameters.

[0078] Preprocessing unit 130 is used to obtain the fused shaft seal operating parameters and extract the rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the fused shaft seal operating parameters. Preprocessing unit 130 can extract key features such as rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from a large number of shaft seal operating parameters, thereby providing key information for subsequent analysis and diagnosis.

[0079] It should be noted that the data collector 110, the signal processing unit 120, and the pre-processing unit 130 can be connected by Bluetooth, 5G or local area network to realize the interactive transmission of data. The data collector 110 can be integrated with multiple types of sensors such as micro-pressure, temperature and humidity, vibration displacement, and mechanical wear, and synchronously obtain heterogeneous data such as steam parameters (pressure, temperature, flow), shaft seal vibration displacement (accuracy ±0.01mm), and mechanical wear, covering the full state characteristics of the shaft seal system. In addition, the data collector 110 can also use industrial-grade sensors (such as QBM2030-1U micro-pressure sensor) with an error of less than or equal to ±0.1kPa, combined with the windproof cover and metal reinforcement mesh design to isolate high-temperature steam corrosion and mechanical vibration interference, ensuring data reliability under complex working conditions.

[0080] The distributed monitoring sensor 100 accurately extracts key features such as abnormal rotor amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from a large number of shaft seal working parameters. It can effectively focus on the core factors that affect the performance and safety of the shaft seal system, and provide high-value data support for subsequent analysis and diagnosis. Among them, the flow sensor or differential pressure flowmeter and other equipment in the distributed monitoring sensor 100 can be used to directly measure the steam leakage of the turbine shaft seal system. It can also indirectly calculate the steam leakage value, i.e., the steam leakage monitoring value, by monitoring parameters such as pressure and temperature at the shaft seal and combining them with the thermal characteristics of steam. By collecting key operating parameters during the operation of the turbine, such as speed, power output, steam flow, pressure, and temperature, and through data fusion and processing, a comprehensive indicator reflecting the current operating conditions, i.e., the operating condition value, can be calculated.

[0081] The monitoring cloud platform 200 is used to obtain the fusion-processed shaft seal working parameters, generate a steam leakage trend curve based on the steam leakage monitoring value, correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, generate a fitting prediction curve based on the correction curve combined with the abnormal amplitude of the rotor, and predict and analyze the shaft seal working parameters in combination with the pre-trained shaft seal prediction model to output the shaft seal status evaluation results.

[0082] For example, Figure 1 As shown, the monitoring cloud platform 200 includes a curve generating unit 210 , a steam leakage prediction unit 220 , and a shaft seal prediction unit 230 .

[0083] The curve generating unit 210 is used to obtain the fused shaft seal operating parameters and generate a steam leakage trend curve based on the steam leakage monitoring value.

[0084] The steam leakage prediction unit 220 is used to correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, and generate a fitting prediction curve based on the correction curve and the rotor abnormal amplitude.

[0085] The shaft seal prediction unit 230 is used to perform prediction analysis on the shaft seal operating parameters based on the fitted prediction curve and the pre-trained shaft seal prediction model, and output a shaft seal status assessment result.

[0086] It should be noted that the curve generation unit 210, steam leakage prediction unit 220, and shaft seal prediction unit 230 can be connected via a Data Transfer Unit (DTU). By integrating the steam leakage monitoring data from the processed shaft seal operating parameters into a steam leakage trend curve, the curve generation unit 210 can intuitively display the temporal trend of the steam turbine shaft seal steam leakage, allowing operators to promptly understand the operating status of the shaft seal system.

[0087] The monitoring cloud platform 200 is the core processing unit of the entire turbine shaft seal steam leakage monitoring and analysis system. It deeply analyzes and processes the data collected by the distributed monitoring sensor 100, provides accurate warning information to the warning feedback module 300, and provides detailed evaluation results to the visualization user terminal 400.

[0088] By setting up a monitoring cloud platform, it is possible to conduct in-depth analysis of real-time data. Combining historical data with pre-set analysis models, it comprehensively evaluates the performance of the shaft seal system and provides operators with accurate monitoring information. By setting up a curve generation unit, a steam leakage prediction unit, and a shaft seal prediction unit within the monitoring cloud platform, the curve generation unit is used to generate leakage trend curves to provide real-time data feedback, the steam leakage prediction unit is used to generate fitted prediction curves for dynamic prediction of leakage trends, and the shaft seal prediction unit is used to output shaft seal status assessment results. This enables the early warning feedback module to trigger classified early warning instructions, thus forming a closed loop of "monitoring-prediction-decision-execution."

[0089] The early warning feedback module 300 is used to obtain the fitted prediction curve and the shaft seal status assessment results, trigger a classified early warning instruction based on the fitted prediction curve and the shaft seal status assessment results, and encrypt and push the classified early warning instruction to the visualization client. When encrypting and pushing the classified early warning instruction to the visualization client 400, the early warning feedback module 300 uses an encrypted transmission protocol (such as MQTT / HTTPS) to ensure the security of the instruction and prevent data tampering and leakage.

[0090] The visualization client 400 is used to parse the classification warning instruction into classification warning information and visually present the classification warning information.

[0091] Specifically, the visualization user terminal 400 can support multi-platform adaptation such as personal computers (PCs), mobile terminals and large screens, and present classified warning information (such as leakage location heat maps and trend curves) in real time.

[0092] Compared with the prior art, the turbine shaft seal steam leakage monitoring and analysis system provided in the embodiment of the present invention fuses the multi-source heterogeneous shaft seal operating parameters of the turbine, and performs filtering and anti-interference processing during the data fusion process, thereby reducing the influence of interference factors such as noise, improving data quality, and being able to more accurately monitor the shaft seal steam leakage, thereby solving the defect of insufficient anti-interference ability of the prior art.

[0093] Exemplarily, the signal processing unit 120 is used to perform multiple rounds of iterative fusion of the shaft seal working parameters based on the iterative consensus algorithm, including: the signal processing unit 120 is used to perform the following steps S101 to S105. Figure 2 , steps S101 to S105 are explained in detail.

[0094] Step S101: Acquire shaft seal working parameters, use a filling method to process missing values ​​in the shaft seal working parameters, and delete duplicate values ​​to obtain a data cleaning set.

[0095] It should be noted that step S101 can use the filling method to fill in the missing values. Depending on the specific situation, methods such as mean, median or interpolation can be selected, and duplicate values ​​can be deleted through the data deduplication algorithm to obtain a clean data set, namely a data cleaning set.

[0096] Step S102: Filter the cleaned data set using an unscented Kalman filter, and output the filtered cleaned data set. Using an unscented Kalman filter to filter the cleaned data, i.e., the cleaned data set, removes noise interference from the data, smoothes the data curve, and makes the data more stable and reliable, further improving data quality.

[0097] The data cleaning set after filtering is expressed as:

[0098]

[0099] Among them, s L (n) represents the data cleaning set after filtering, are the maximum and minimum amplitudes of the hth harmonic when processed by the unscented Kalman filter, L and D are the filtering times and filter orders of the unscented Kalman filter, respectively. s ,θ h Represent the parameter sampling frequency, the phase of the hth harmonic, Δω h is the frequency sampling angle difference of the hth harmonic and G(ω) represents the magnitude response of the unscented Kalman filter, They are respectively the sampling angle difference of the positive frequency component and the negative frequency component of the hth harmonic, f r 、f hRepresent the rated frequency and fundamental frequency respectively.

[0100] Step S103 captures historical shaft seal parameters. Based on these historical shaft seal parameters, mechanical seal wear parameters, and pressure parameters, the dynamic weights of the parameter types in the consensus mechanism are initialized. The dynamic weights of the parameter types are reset to fusion priority data, retaining parameter types with high fusion priority data. By introducing historical shaft seal parameters, mechanical seal wear parameters, and pressure parameters and properly initializing the dynamic weights of parameter types, the role of important parameters in the fusion process can be highlighted, improving the reliability and accuracy of the fusion results.

[0101] Step S104: Determine the iteration rounds and verification parameters in the consensus mechanism using the dynamic weight of the parameter type, the number of types, and the parameter signal attributes as fusion constraints.

[0102] Step S105: After each iteration, the shaft seal operating parameters are verified to see if they meet the fusion constraints and verification parameters. If so, the iterative fusion process stops and the fusion consensus value is output. Through multiple rounds of iterative fusion, the parameter weights and fusion results are continuously adjusted until the fusion constraints and verification parameters are met. This fully exploits the information in the data and improves the accuracy and stability of the fusion results.

[0103] By performing multiple rounds of iterative fusion of shaft seal operating parameters based on an iterative consensus algorithm, with the fusion results verified and adjusted at each iteration, the fusion results gradually approach the true values, significantly improving the accuracy of the fusion results. Furthermore, by combining filtering techniques such as the unscented Kalman filter, noise and outliers in the data can be effectively removed, reducing the impact of noise on the fusion results and improving data reliability. Data preprocessing and fusion performed at the sensor end reduces the computational burden on the monitoring cloud platform and improves the processing efficiency of the entire system.

[0104] Exemplarily, the steam leakage prediction unit 220 is used to correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain the correction curve, including: the steam leakage prediction unit 220 is used to perform the following steps S201 to S203. Figure 3 Steps S201 to S203 are described in detail.

[0105] Step S201 loads the steam leakage trend curve, extracts an interpolation vector consisting of interpolation points from the steam leakage trend curve, and constructs a continuous distribution model based on the interpolation vector. The continuous distribution model includes the interpolation vector, an operating coefficient matrix, and a fluctuation coefficient matrix. By constructing a multi-dimensional parameter-coupled continuous distribution model based on the interpolation vector, the operating coefficient matrix, and the fluctuation coefficient matrix, a dynamic correlation between steam leakage and multiple physical field parameters can be established, overcoming the limitations of traditional single-parameter models. Compared with traditional linear fitting methods, the continuous distribution model can more accurately describe the complex nonlinear changes in steam leakage.

[0106] Step S202, obtain the continuous distribution model, obtain the operating condition influence constraint and the fluctuation influence constraint based on the operating condition coefficient matrix and the fluctuation coefficient matrix of the continuous distribution model respectively, after aggregating the operating condition influence constraint and the fluctuation influence constraint, solve the min-max problem with the continuous distribution model being 0, and obtain the optimal correction coefficient of the interpolation point constraint on the steam leakage trend curve. The min-max problem is the minimum-maximum problem. By using the min-max optimization problem to solve the optimal correction coefficient, the influence of sudden operating condition changes (such as load drops) and fluctuation interference (such as valve jitter) on the steam leakage can be effectively eliminated. For example, sequential quadratic programming (SQP) can be used to solve the constrained min-max problem.

[0107] Step S203, obtaining the optimal correction coefficient corresponding to at least one group of interpolation points, iterating the interpolation vector of the interpolation point based on the optimal correction coefficient, the working condition influence constraint, and the fluctuation influence constraint, updating the interpolation vector after iteration, updating the interpolation vector of the interpolation point, and outputting the correction curve.

[0108] For example, the interpolation vector after the update iteration is expressed as:

[0109]

[0110] Where T(x) t represents the interpolation vector at time t before the update iteration, T(x) t+1 Represents T(x) t The interpolation vector at time t+1 after the update iteration, G(x), U (l) are the working condition impact constraint value and the fluctuation impact constraint value respectively, σ(·) represents the activation function of the continuous distribution model, w x is the weight matrix of the interpolation vector before updating iteration, b x represents the bias term of the activation function, q1 and q2 are the weight values ​​of the working condition influence constraint and the fluctuation influence constraint respectively, and λ best is the optimal correction coefficient constrained by the interpolation point and j is the interpolation point number, J is the number of interpolation points, Δt represents the sampling time difference, is the steam leakage at the current time t, The steam leakage in the previous sampling period is also the steam leakage at time t-1, max(T(x) t ,0) means taking T(x) t and the maximum value of 0, Indicates taking and 0.

[0111] By iteratively optimizing the interpolation vector and updating the interpolation vector based on the optimal correction coefficient, the shape of the steam leakage curve can be dynamically adjusted to make it more consistent with the actual change trend and gradually approach the real steam leakage distribution.

[0112] Exemplarily, the steam leakage prediction unit 220 is used to generate the fitting prediction curve based on the correction curve combined with the rotor abnormal amplitude, including: the steam leakage prediction unit 220 is used to perform the following steps S204 to S206. Figure 3 Steps S204 to S206 are described in detail.

[0113] Step S204: load the correction curve, identify the interpolation vector after the update iteration in the correction curve, construct a static adjacency matrix based on the fuzzy clustering algorithm combined with the graph neural network, use the rotor abnormal amplitude as a constraint, and the operating condition impact constraint and the fluctuation impact constraint as prior probabilities, transform the static adjacency matrix to obtain a dynamic adjacency matrix.

[0114] Specifically, in step S204, when constructing a static adjacency matrix based on a fuzzy clustering algorithm combined with a graph neural network, an improved Fuzzy C-Means (FCM) algorithm (introducing a Gaussian kernel into the membership function) can be used to divide vibration mode clusters (e.g., normal / abnormal vibration zones). Combining fuzzy clustering (to handle the uncertainty of rotor amplitude) with a graph neural network (to capture the spatiotemporal correlation between vibration and steam leakage) allows for rapid identification of the impact of abnormal vibration modes (e.g., high-frequency resonance) on steam leakage.

[0115] In step S205, the temporal and spatial features of the dynamic adjacency matrix are extracted using a time-series convolutional neural network to obtain corresponding feature extraction results. These feature extraction results are then fused using a BP neural network combined with an ELM algorithm to output a temporal interpolation prediction. The BP neural network refers to a back propagation network, and the ELM algorithm refers to an extreme learning machine (ELM).

[0116] Step S206: Obtain at least one set of interpolated prediction quantities and their corresponding prediction time points, and generate a fitted prediction curve based on the interpolated prediction quantities and their corresponding prediction time points. This smooth fitted prediction curve can visually display the steam leakage trend, helping maintenance personnel quickly locate abnormal periods (such as a surge in leakage during low-load periods at night).

[0117] During the correction of the steam leakage trend curve, the optimal correction coefficients for the interpolation point constraints are obtained by solving the min-max problem. This allows for precise correction of the steam leakage curve, improving its accuracy and reliability. By updating the interpolation vector based on the optimal correction coefficients, the shape of the steam leakage trend curve can be dynamically adjusted to better align with actual trends. When generating a fitted prediction curve based on the correction curve combined with the rotor abnormal amplitude, the static adjacency matrix is ​​converted into a dynamic adjacency matrix by introducing the rotor abnormal amplitude as a constraint. This allows for real-time reflection of system state changes and enhances the model's dynamic adaptability. Furthermore, by utilizing a time-series convolutional neural network to extract the temporal and spatial features of the dynamic adjacency matrix, the model captures temporal variations and spatial correlations in the data, providing rich feature support for prediction.

[0118] Exemplarily, the shaft seal prediction unit 230 is used to predict and analyze the shaft seal working parameters based on the fitting prediction curve and the pre-trained shaft seal prediction model, and output the shaft seal status evaluation result, including: the shaft seal prediction unit 230 is used to perform the following steps S301 to S305. Figure 4 Steps S301 to S305 are described in detail.

[0119] Step S301: Obtain shaft seal operating parameters and calculate the shaft seal brittleness index and shaft seal imbalance within the adoption cycle based on the shaft seal operating parameters. By calculating the shaft seal brittleness index and shaft seal imbalance, the current state of the shaft seal system can be comprehensively assessed, providing key status characteristics for subsequent predictive analysis.

[0120] Step S302: Establish a periodic feature vector based on the shaft seal brittleness index, shaft seal imbalance, and the updated interpolated vector. By integrating the shaft seal brittleness index, shaft seal imbalance, and the updated interpolated vector into a periodic feature vector, the changes in shaft seal operating parameters at different time scales can be comprehensively considered.

[0121] In step S303, the shaft seal prediction model is used to perform band periodicity detection and time segment interception on the feature vector, obtaining at least one set of intercepted time segments. The feature vectors within the time segment are convolutionally fused based on the CFCs method, and the convolutionally fused feature vectors are globally pooled and averaged using the global pooling layer and average pooling layer in the shaft seal prediction model. Band periodicity detection can identify the periodic variation patterns in the shaft seal operating parameters, providing important periodic features for subsequent feature fusion. Time segment interception of the feature vector can focus on key time segments, reduce data redundancy, and improve analysis efficiency.

[0122] In step S304, global pooling and average pooling results are obtained and residual connections are formed between the global pooling and average pooling results and the feature vector to obtain a feature evaluation matrix. By residually connecting the global pooling and average pooling results with the original feature vector to form a feature evaluation matrix, feature information at different levels can be integrated, enhancing the model's feature expression capabilities. Residual connections help preserve the details of the original features, avoid information loss, and improve the model's prediction accuracy.

[0123] Step S305 , taking the feature evaluation matrix as input, performing a linear transformation on the feature evaluation matrix through a weight matrix, and then performing a nonlinear transformation using a nonlinear activation function to obtain a shaft seal evaluation value.

[0124] By transforming the feature evaluation matrix using a weight matrix and a nonlinear activation function, we ultimately obtain the shaft seal evaluation value, enabling a quantitative assessment of the shaft seal system's status. The nonlinear activation function captures the nonlinear relationships between features, improving the model's ability to model complex operating conditions and resulting in more accurate evaluation results.

[0125] Another embodiment of the present invention provides a method for monitoring and analyzing steam leakage of a steam turbine shaft seal, the process of which is as follows: Figure 5 As shown, it includes steps S10 to S40. Figure 5 Steps S10 to S40 will be described in detail.

[0126] Step S10: Real-time collection of multi-source heterogeneous shaft seal working parameters of the steam turbine, fusion processing of the shaft seal working parameters, and extraction of rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal working parameters.

[0127] Step S20, obtain the fused shaft seal working parameters, generate a steam leakage trend curve based on the steam leakage monitoring value, correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, generate a fitting prediction curve based on the correction curve combined with the rotor abnormal amplitude, predict and analyze the shaft seal working parameters in combination with the pre-trained shaft seal prediction model, and output the shaft seal status assessment result.

[0128] Step S30: obtaining a fitting prediction curve and a shaft seal status evaluation result, and triggering a classification warning instruction based on the fitting prediction curve and the shaft seal status evaluation result.

[0129] Step S40: parsing the classified warning instruction into classified warning information, and visually presenting the classified warning information.

[0130] Exemplarily, step S10 includes steps S401 to S403. Figure 6 Steps S401 to S403 are described in detail.

[0131] Step S401 , collecting shaft seal operating parameters in real time; wherein the shaft seal operating parameters include steam parameters, pressure parameters, flow parameters, temperature parameters, shaft seal vibration displacement parameters, shaft seal mechanical wear parameters, and turbine power output parameters.

[0132] Step S402 : Load the shaft seal working parameters in real time, perform multiple rounds of iterative fusion on the shaft seal working parameters based on an iterative consensus algorithm, and output the fused shaft seal working parameters.

[0133] Step S403: Acquire the fused shaft seal working parameters, and extract the rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the fused shaft seal working parameters.

[0134] The method for monitoring and analyzing the steam leakage of a steam turbine shaft seal provided in the embodiment of the present invention can be implemented by the system for monitoring and analyzing the steam leakage of a steam turbine shaft seal provided in the embodiment of the present invention.

[0135] Compared with the prior art, the method for monitoring and analyzing the steam leakage of a turbine shaft seal provided in the embodiment of the present invention fuses the multi-source heterogeneous shaft seal operating parameters of the turbine, and performs filtering and anti-interference processing during the data fusion process, thereby reducing the influence of interference factors such as noise, improving data quality, and being able to more accurately monitor the steam leakage of the shaft seal, thereby solving the defect of insufficient anti-interference ability of the prior art.

[0136] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present invention, and that in actual applications, various changes may be made in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A steam turbine shaft seal steam leakage monitoring and analysis system, characterized in that: The system includes distributed monitoring sensors, a monitoring cloud platform, an early warning feedback module, and a visual user terminal; The distributed monitoring sensor is deployed on the shaft seal side of the steam turbine and is used to collect the multi-source heterogeneous shaft seal operating parameters of the steam turbine in real time, and fuse the shaft seal operating parameters to extract the rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal operating parameters; The monitoring cloud platform is used to obtain the fused shaft seal operating parameters, generate a steam leakage trend curve based on the steam leakage monitoring value, correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, generate a fitting prediction curve based on the correction curve combined with the rotor abnormal amplitude, perform prediction analysis on the shaft seal operating parameters in combination with a pre-trained shaft seal prediction model, and output a shaft seal status assessment result; The early warning feedback module is used to obtain the fitting prediction curve and the shaft sealing status assessment result, trigger a classification early warning instruction based on the fitting prediction curve and the shaft sealing status assessment result, and encrypt and push the classification early warning instruction to the visualization user terminal; The visualization user terminal is used to parse the classification warning instruction into classification warning information and visually present the classification warning information.

2. The system according to claim 1, wherein: The distributed monitoring sensor includes: At least one set of data collectors for real-time collection of the shaft seal operating parameters; wherein the shaft seal operating parameters include steam parameters, pressure parameters, flow parameters, temperature parameters, shaft seal vibration displacement parameters, shaft seal mechanical wear parameters, and turbine power output parameters; a signal processing unit, configured to load the shaft seal operating parameters in real time, perform multiple rounds of iterative fusion on the shaft seal operating parameters based on an iterative consensus algorithm, and output the fused shaft seal operating parameters; A pre-processing unit is used to obtain the shaft seal working parameters after fusion processing, and extract the rotor abnormal amplitude, the operating condition value, the steam leakage monitoring value, and the steam parameter fluctuation value from the shaft seal working parameters after fusion processing.

3. The system according to claim 2, characterized in that The signal processing unit is used to perform multiple rounds of iterative fusion of the shaft seal working parameters based on an iterative consensus algorithm, including: The signal processing unit is used for: Obtaining the shaft seal working parameters, processing missing values ​​in the shaft seal working parameters using a filling method, and deleting duplicate values ​​to obtain a data cleaning set; Performing filtering processing on the data cleaning set based on an unscented Kalman filter, and outputting the filtered data cleaning set; The data cleaning set after filtering is expressed as: Among them, s L (n) represents the data cleaning set after filtering, are the maximum and minimum amplitudes of the hth harmonic when processed by the unscented Kalman filter, L and D are the filtering times and filter orders of the unscented Kalman filter, respectively. s ,θ h Represent the parameter sampling frequency, the phase of the hth harmonic, Δω h is the frequency sampling angle difference of the hth harmonic and G(ω) represents the magnitude response of the unscented Kalman filter, They are respectively the sampling angle difference of the positive frequency component and the negative frequency component of the hth harmonic, f r 、f h Respectively represent the rated frequency and fundamental frequency; Capture the historical parameters of the shaft seal, and based on the historical parameters, mechanical wear parameters, and pressure parameters of the shaft seal, initialize the dynamic weight of the parameter type in the consensus mechanism, reset the dynamic weight of the parameter type to fusion priority data, and retain the parameter type with high fusion priority data; The dynamic weight of parameter types, the number of types, and parameter signal attributes are used as fusion constraints to determine the iteration rounds and verification parameters in the consensus mechanism; After each iteration, the shaft seal operating parameters are checked to see if they meet the fusion constraints and verification parameters. If so, the iterative fusion is stopped and the fusion consensus value is output.

4. The system according to claim 1, wherein: The monitoring cloud platform includes: a curve generating unit, configured to obtain the shaft seal operating parameters after fusion processing, and generate the steam leakage trend curve based on the steam leakage monitoring value; a steam leakage prediction unit, configured to correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, and generate the fitting prediction curve based on the correction curve in combination with the rotor abnormal amplitude; A shaft sealing prediction unit is used to predict and analyze the shaft sealing working parameters based on the fitting prediction curve and the pre-trained shaft sealing prediction model, and output the shaft sealing status evaluation result.

5. The system according to claim 4, characterized in that The steam leakage prediction unit is configured to correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, including: The steam leakage prediction unit is used to: Loading the steam leakage trend curve, extracting an interpolation vector composed of interpolation points in the steam leakage trend curve, and constructing a continuous distribution model based on the interpolation vector, wherein the continuous distribution model includes the interpolation vector, an operating condition coefficient matrix, and a fluctuation coefficient matrix; Obtaining the continuous distribution model, obtaining an operating condition influence constraint and a fluctuation influence constraint based on the operating condition coefficient matrix and the fluctuation coefficient matrix of the continuous distribution model, respectively, aggregating the operating condition influence constraint and the fluctuation influence constraint, solving a min-max problem when the continuous distribution model is 0, and obtaining an optimal correction coefficient for the interpolation point constraint on the steam leakage trend curve; Obtain the optimal correction coefficient corresponding to at least one set of interpolation points, iterate the interpolation vector of the interpolation point based on the optimal correction coefficient, the operating condition influence constraint, and the fluctuation influence constraint, update the interpolation vector after iteration, update the interpolation vector of the interpolation point, and output the correction curve.

6. The system according to claim 5, characterized in that The steam leakage prediction unit is configured to generate the fitting prediction curve based on the correction curve and the abnormal rotor amplitude, and includes: The steam leakage prediction unit is used to: Loading the correction curve, identifying the interpolation vector after iterative update in the correction curve, constructing a static adjacency matrix based on a fuzzy clustering algorithm combined with a graph neural network, using the rotor abnormal amplitude as a constraint, and the operating condition impact constraint and the fluctuation impact constraint as prior probabilities, transforming the static adjacency matrix to obtain a dynamic adjacency matrix; Extracting the temporal and spatial features of the dynamic adjacency matrix based on a time series convolutional neural network to obtain corresponding feature extraction results, fusing the feature extraction results using a BP neural network combined with an ELM algorithm to output a temporal interpolation prediction value; At least one set of interpolation prediction quantities and their corresponding prediction time points are obtained, and the fitting prediction curve is generated based on the interpolation prediction quantities and their corresponding prediction time points.

7. The system according to claim 5, characterized in that The interpolation vector after updating iteration is expressed as: Among them, T(x) t represents the interpolation vector at time t before the update iteration, T(x) t+1 Represents T(x) t The interpolation vector at time t+1 after the update iteration, G(x), U (l) are the working condition impact constraint value and the fluctuation impact constraint value respectively, σ(·) represents the activation function of the continuous distribution model, w x is the weight matrix of the interpolation vector before updating iteration, b x represents the bias term of the activation function, q1 and q2 are the weight values ​​of the working condition influence constraint and the fluctuation influence constraint respectively, and λ best is the optimal correction coefficient constrained by the interpolation point and j is the interpolation point number, J is the number of interpolation points, Δt represents the sampling time difference, is the steam leakage at the current time t, The steam leakage in the previous sampling period is also the steam leakage at time t-1, max(T(x) t ,0) means taking T(x) t and the maximum value of 0, Indicates taking and 0.

8. The system according to claim 7, characterized in that The shaft seal prediction unit is configured to perform prediction analysis on the shaft seal operating parameters based on the fitted prediction curve and the pre-trained shaft seal prediction model, and output the shaft seal status assessment result, including: The shaft seal prediction unit is used for: Acquiring the shaft seal operating parameters, and calculating the shaft seal brittleness index and the shaft seal imbalance within the adoption cycle based on the shaft seal operating parameters; Establishing a characteristic vector for a periodic time period based on the shaft seal brittleness index, the shaft seal imbalance, and the interpolation vector after updating and iteration; Using the shaft seal prediction model to perform band periodicity detection and time segment interception on the feature vector, obtaining at least one group of intercepted time segments, performing convolution fusion on the feature vectors within the time segments based on the CFCs method, and performing global pooling and average pooling processing on the convolution fused feature vectors through the global pooling layer and average pooling layer in the shaft seal prediction model; Obtaining global pooling and average pooling results, forming a residual connection between the global pooling and average pooling results and the feature vector to obtain a feature evaluation matrix; Taking the feature evaluation matrix as input, the feature evaluation matrix is ​​linearly transformed by a weight matrix, and then nonlinearly transformed by a nonlinear activation function to obtain a shaft seal evaluation value.

9. A method for monitoring and analyzing steam leakage of a steam turbine shaft seal, characterized in that: The method comprises: Real-time collection of multi-source heterogeneous shaft seal operating parameters of the steam turbine, fusion processing of the shaft seal operating parameters, and extraction of rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal operating parameters; Acquire the fused shaft seal operating parameters, generate a steam leakage trend curve based on the steam leakage monitoring value, correct the steam leakage trend curve based on the steam parameter fluctuation value and the operating condition value to obtain a correction curve, generate a fitting prediction curve based on the correction curve and the rotor abnormal amplitude, perform prediction analysis on the shaft seal operating parameters in combination with a pre-trained shaft seal prediction model, and output a shaft seal status assessment result; acquiring the fitting prediction curve and the shaft seal status assessment result, and triggering a classification warning instruction based on the fitting prediction curve and the shaft seal status assessment result; The classified warning instructions are parsed into classified warning information, and the classified warning information is visually presented.

10. The method according to claim 9, characterized in that The real-time collection of multi-source heterogeneous shaft seal working parameters of the steam turbine and the fusion processing of the shaft seal working parameters are performed to extract the rotor abnormal amplitude, operating condition value, steam leakage monitoring value, and steam parameter fluctuation value from the shaft seal working parameters, including: Real-time collection of the shaft seal operating parameters; wherein the shaft seal operating parameters include steam parameters, pressure parameters, flow parameters, temperature parameters, shaft seal vibration displacement parameters, shaft seal mechanical wear parameters, and turbine power output parameters; Loading the shaft seal working parameters in real time, performing multiple rounds of iterative fusion on the shaft seal working parameters based on an iterative consensus algorithm, and outputting the fused shaft seal working parameters; The shaft seal working parameters after fusion processing are acquired, and the rotor abnormal amplitude, the operating condition value, the steam leakage monitoring value, and the steam parameter fluctuation value are extracted from the shaft seal working parameters after fusion processing.

Images