Integrated clutch state combined cycle unit start-stop vibration control method and system

Abstract

The present application belongs to the technical field of combined cycle unit start-stop control, and provides a combined cycle unit start-stop vibration control method and system integrated with clutch state, which solves the problem of limited control precision caused by passive control strategy based on operation parameter optimization of the existing method, and the method comprises the following steps: preprocessing of multi-source correlation parameters, analysis of SSS clutch health state based on clutch state parameters, determination of start-stop control strategy combination based on the unit start-stop control model combined with the SSS clutch health state value, and screening of the start-stop control strategy combination based on the particle swarm optimization algorithm; by embedding the clutch health state as a key feedforward variable into the control decision process, the present application can effectively compensate for the system characteristic drift caused by environmental factor changes and unit component aging, while ensuring vibration suppression effect, taking into account start-stop efficiency and equipment life, and significantly improving the safety, stability and control precision of the combined cycle unit start-stop process.

Description

Technical Field

[0001] This invention relates to the field of start-up and shutdown control technology for combined cycle power units, and in particular to a method and system for start-up and shutdown vibration control of combined cycle power units with integrated clutch status. Background Technology

[0002] Gas-steam combined cycle power generation technology has become the main power source for peak shaving and frequency regulation in modern power systems due to its high efficiency, low emissions and fast start-stop characteristics. In single-shaft combined cycle units, the gas turbine and steam turbine are usually arranged coaxially to drive the generator. The SSS (Synchro-Self-Shifting) clutch, as a key transmission component, is widely used to connect the turbine rotor and the generator rotor.

[0003] The SSS clutch is an overrunning clutch that achieves automatic engagement and disengagement entirely through its own mechanical structure. Its working principle is based on the engagement mechanism of helical teeth: when the driving and driven ends rotate synchronously, the clutch automatically completes axial displacement and engages; when the input speed is lower than the output speed, the clutch automatically disengages. The SSS clutch provides the following operational advantages to combined cycle units: it enables flexible switching between simple cycle and combined cycle operating conditions of the gas turbine; the steam turbine does not need to start synchronously with the gas turbine, reducing start-up power consumption and cooling steam requirements; and it supports online switching between various heating modes such as pure condensing, extraction condensing, and back pressure, significantly improving the operational flexibility of the cogeneration unit.

[0004] However, as the unit's service life increases, the SSS clutch inevitably experiences performance degradation. For example, wear on the SSS clutch friction plates alters the friction coefficient and torque transmission characteristics during engagement, leading to increased engagement shock. Slight leakage in the hydraulic cylinder can affect the clutch's responsiveness and positional accuracy, resulting in unexpected dynamic loads. These disturbances introduced by the deterioration of clutch health make it difficult for traditional start-stop control strategies based solely on fixed parameter models or empirical rules to consistently and effectively control vibration at optimal levels. In cases of poor clutch health, using the original control parameters or logic may exacerbate vibration and fail to achieve the desired control effect.

[0005] Chinese patent CN110284930B discloses a method for controlling the engagement of a constant exhaust cylinder temperature clutch in a Siemens 9F gas turbine combined cycle unit. The method includes the following steps: an SSS clutch is used to engage between the generator and the turbine; the unit's lubricating oil temperature is T, the absolute vacuum corresponding to T is P1, and the ambient atmospheric pressure is P0. Before the SSS clutch engages with the turbine shaft, the relative vacuum setting value of the turbine is P2; the relative vacuum setting value P2 of the turbine is calculated, and the turbine's relative vacuum is set to P2. However, existing methods are typically based on a passive control strategy that optimizes operating parameters. This makes the aforementioned control method unable to adapt to changes in ambient temperature and parameter drift caused by unit aging, and it requires manual intervention by operators, resulting in limited control accuracy. Summary of the Invention

[0006] The present invention aims to solve at least one of the problems existing in the prior art, and provides a method and system for controlling the start-up and shutdown vibration of a combined cycle unit with integrated clutch status.

[0007] One aspect of the present invention provides a method for controlling the start-up and shutdown vibration of a combined cycle power unit with an integrated clutch state, comprising: Real-time acquisition of multi-source correlation parameters during the start-up and shutdown of combined cycle units; preprocessing of multi-source correlation parameters; output of preprocessed multi-source correlation dataset. Identify clutch state parameters that represent the health status of SSS clutches in multi-source association datasets, analyze the health status of SSS clutches based on clutch state parameters, and output the health status value of SSS clutches. Vibration target parameters during the start-up and shutdown process of the combined cycle unit are determined based on the multi-source correlation dataset. Using the vibration target parameters and the multi-source correlation dataset as inputs, the start-up and shutdown control strategy combination is determined based on the unit start-up and shutdown control model and the health status value of the SSS clutch.

[0008] Optionally, the combined cycle unit start-up and shutdown vibration control method with integrated clutch further includes: A combination of start-stop control strategies is obtained. With the optimization objectives of maximum vibration suppression, shortest start-stop time, and minimum active torque compensation, a multi-objective optimization function for the strategy is constructed. The combination of start-stop control strategies is screened based on the particle swarm optimization algorithm, and the optimal vibration control strategy during the start-stop process is output. In response to the optimal vibration control strategy during start-up and shutdown, the control adjustment parameters of the unit start-up and shutdown actuators are automatically triggered, and control adjustment commands are generated based on the control adjustment parameters of the unit start-up and shutdown actuators.

[0009] Optionally, the preprocessing of the multi-source correlation parameters includes: The multi-source correlation parameters during the start-up and shutdown process of the combined cycle unit are obtained. Using speed signal and clutch status signal as labels, the multi-source correlation parameters are traversed and marked. The marked multi-source correlation parameters are then standardized to obtain the standardized multi-source correlation parameters. The standardized multi-source association parameters are loaded, and an adaptive time threshold is preset. Using the adaptive time threshold as a priori, the multi-source association parameters are adaptively filtered by the improved temporal convolutional clustering method to obtain continuous multi-source association parameters within the adaptive time threshold. Obtain continuous multi-source correlation parameters within an adaptive time threshold, and normalize the multi-source correlation parameters based on the batch normalization method to obtain normalized multi-source correlation parameters. The speed signal in the multi-source correlation parameters is defined as the observation vector, and the clutch state signal in the multi-source correlation parameters is defined as the state vector. The extended Kalman filter algorithm is used to recursively estimate the minimum variance of the state vector with the observation vector as input, and outputs a clutch health feature vector that characterizes the clutch health characteristics. Obtain normalized multi-source correlation parameters and clutch health feature vectors, and generate a unified encapsulated multi-source correlation dataset based on the multi-source correlation parameters and clutch health feature vectors.

[0010] Optionally, the health status analysis of the SSS clutch based on clutch state parameters includes: The clutch state parameters characterizing the health state of the SSS clutch are obtained, and the clutch state parameters are processed by short-time Fourier transform to extract the frequency domain features of the clutch state. The Hilbert-Huang transform method is used to extract the clutch state time-frequency domain features of the clutch state parameters, and obtain the clutch state frequency domain features, clutch state time-frequency domain features and clutch health feature vector. Correlation analysis is used to calculate the feature sensitivity of the clutch state frequency domain features, clutch state time-frequency domain features and clutch health feature vector. Determine whether the feature sensitivity exceeds a preset sensitivity threshold; If the feature sensitivity exceeds the preset sensitivity threshold, the corresponding clutch state frequency domain features, clutch state time frequency domain features, and clutch health feature vector are retained. A health state feature set is generated based on the retained clutch state frequency domain features, clutch state time frequency domain features, and clutch health feature vector. Obtain a health status feature set, and adaptively fuse the features within the health status feature set based on the heterogeneous graph attention mechanism of DS-GNN to output a comprehensive fusion vector; Load the integrated fusion vector, use the integrated fusion vector as input, calculate the SSS clutch health status value based on the fusion vector-health status value function, and output the SSS clutch health status value.

[0011] Optionally, determining the vibration target parameters during the start-up and shutdown process of the combined cycle unit based on a multi-source correlation dataset includes: Obtain a multi-source associated dataset, and retrieve the associated multi-source historical dataset based on the multi-source associated dataset. Extract time-domain, frequency-domain, and time-frequency-domain features from the multi-source associated dataset and the multi-source historical dataset respectively to obtain a multi-source feature set. Using multi-source historical datasets as prior working condition label information, dynamic time warping algorithm combined with historical working condition labeling method is used to establish the spatiotemporal correlation of multi-source features within the multi-source feature set, resulting in a multi-source feature set containing spatiotemporal relationships of multi-source features. The nonlinear dependence between multi-source features and vibration target parameters in the multi-source feature set is determined based on the vine copula model, and the conditional correlation coefficient between multi-source features and vibration target parameters is calculated. Using the conditional correlation coefficient between multi-source features and vibration target parameters as constraints, the contribution of multi-source features to vibration target parameters is calculated based on the fusion method of permutation importance and SHAP value. The initial safety limits of the vibration target parameters are obtained. The target margin of the vibration target parameters is dynamically adjusted based on the contribution of multi-source characteristics to the vibration target parameters and the health status value of the SSS clutch. The target margin of the adjusted vibration target parameters is used as the vibration target parameters during the start-up and shutdown process of the combined cycle unit.

[0012] Optionally, the determination of the start-stop control strategy combination based on the unit start-stop control model and the SSS clutch health status value includes: Obtain vibration target parameters and multi-source correlation datasets. Based on the vibration target parameters and multi-source correlation datasets, establish a multi-physics field coupled dynamic model during the start-up and shutdown process of the combined cycle unit. Establish a clutch engagement function that includes the SSS clutch health state value and embed the clutch engagement function into the multi-physics field coupled dynamic model. A multiphysics coupled dynamics model with embedded clutch engagement function is loaded, and the order of the multiphysics coupled dynamics model is reduced based on the intrinsic orthogonal decomposition method to obtain a multiphysics reduced-order model.

[0013] Optionally, the step of determining the start-stop control strategy combination based on the unit start-stop control model and the SSS clutch health status value further includes: A multi-dimensional strategy control space is constructed using speed control strategy, temperature control strategy, clutch control strategy, and torque control strategy, and a unit start-up and shutdown control model is constructed based on the multi-dimensional strategy control space; Based on the learning of the mapping relationship between the multi-dimensional strategy control space and the multi-physics field reduced-order model of the unit start-stop control model, the multi-physics field reduced-order model is solved by the sequential quadratic programming (SQP) method to obtain the start-stop control strategy combination.

[0014] Optionally, the clutch engagement function is expressed as: ; ; ; in, This represents the clutch engagement function. These represent the clutch friction constraint term, the clutch equivalent stiffness constraint term, and the slip friction stage function, respectively, where C represents the clutch equivalent stiffness and t represents time. These are the basic equivalent stiffness and the stiffness influence coefficient, respectively. These are the friction coefficient in the brand-new state and the wear influence coefficient, respectively. Indicates the health status value of the SSS clutch and , This represents the activation function of the fusion vector-health state value function. Represents the i-th feature component The physical degradation function, where n represents the total number of characteristic components. Let represent the adaptive weight, weight coefficient, and bias term of the i-th element in the fusion vector, respectively. This is a deep discriminative network composed of LSTM networks, where the superscript T indicates transpose.

[0015] In another aspect, the present invention provides a combined cycle unit start-up and shutdown vibration control system with integrated clutch state, for implementing the combined cycle unit start-up and shutdown vibration control method with integrated clutch state described above, the combined cycle unit start-up and shutdown vibration control system with integrated clutch state comprising: The data preprocessing module is used to collect multi-source correlation parameters in real time during the start-up and shutdown process of the combined cycle unit, preprocess the multi-source correlation parameters, and output the preprocessed multi-source correlation dataset. The clutch condition assessment module is used to identify clutch condition parameters that represent the health status of the SSS clutch in a multi-source associated dataset, analyze the health status of the SSS clutch based on the clutch condition parameters, and output the SSS clutch health status value. The control strategy generation module is used to determine the vibration target parameters during the start-up and shutdown process of the combined cycle unit based on the multi-source correlation dataset. Taking the vibration target parameters and the multi-source correlation dataset as input, the module determines the start-up and shutdown control strategy combination based on the unit start-up and shutdown control model and the SSS clutch health status value. The optimal strategy screening module is used to obtain the start-stop control strategy combination. With the optimization objectives of maximum vibration suppression, shortest start-stop time, and minimum active torque compensation, a multi-objective optimization function of strategy is constructed. The start-stop control strategy combination is screened based on the particle swarm optimization algorithm, and the optimal vibration control strategy during the start-stop process is output. The start-stop vibration control module is used to respond to the optimal vibration control strategy during the start-stop process, automatically trigger the control adjustment parameters of the unit's start-stop actuator, and generate control adjustment commands based on the control adjustment parameters of the unit's start-stop actuator.

[0016] Optionally, the data preprocessing module includes: The parameter labeling unit is used to acquire multi-source associated parameters during the start-up and shutdown process of the combined cycle unit. Using speed signal and clutch status signal as labels, it traverses the multi-source associated parameters and labels them. The labeled multi-source associated parameters are then processed for format standardization to obtain standardized multi-source associated parameters. An adaptive filtering unit is used to load standardized multi-source association parameters, preset an adaptive time threshold, use the adaptive time threshold as a priori, and combine an improved temporal convolutional clustering method to adaptively filter the multi-source association parameters to obtain continuous multi-source association parameters within the adaptive time threshold. The feature vector derivation unit is used to obtain continuous multi-source correlation parameters within the adaptive time threshold. The multi-source correlation parameters are normalized based on the batch normalization method to obtain normalized multi-source correlation parameters. The speed signal in the multi-source correlation parameters is defined as the observation vector, and the clutch state signal in the multi-source correlation parameters is defined as the state vector. Through the extended Kalman filter algorithm, the state vector is recursively estimated with the observation vector as input, and the clutch health feature vector representing the clutch health characteristics is output. The dataset encapsulation unit is used to obtain normalized multi-source correlation parameters and clutch health feature vectors, and to generate a unified encapsulated multi-source correlation dataset based on the multi-source correlation parameters and clutch health feature vectors.

[0017] Compared with the prior art, the present invention has the following advantages: 1. By real-time acquisition and preprocessing of multi-source correlation parameters, the health status of the SSS clutch is accurately identified and analyzed. Then, combined with multi-source data, vibration target parameters are dynamically determined. Finally, a combination of start-stop control strategies is generated based on the unit start-stop control model and health status values, realizing a shift from passive response to active prediction and adaptive optimization. This method embeds the clutch health status as a key feedforward variable into the control decision process, effectively compensating for system characteristic drift caused by changes in environmental factors and aging of unit components. It ensures vibration suppression while balancing start-stop efficiency and equipment lifespan. Furthermore, the entire process is automated without manual intervention, significantly improving the safety, stability, and control accuracy of the combined cycle unit's start-stop process. 2. By combining short-time Fourier transform and Hilbert-Huang transform, multi-dimensional feature extraction of clutch state parameters from the frequency domain to the time-frequency domain was achieved, comprehensively capturing the steady-state characteristics and transient dynamic characteristics during clutch engagement. Correlation analysis was then used for feature sensitivity screening, retaining highly sensitive features based on preset sensitivity thresholds, effectively eliminating redundant information and reducing the complexity of subsequent fusion calculations. Furthermore, the DS-GNN heterogeneous graph attention mechanism was introduced, modeling the complex relationships between different features through a heterogeneous graph structure, and using a graph attention network to achieve adaptive weighted fusion of features, significantly improving the ability of the comprehensive fusion vector to represent the clutch health state. Finally, the high-dimensional fusion features were mapped to quantified health state values ​​through a fusion vector-health state value function, realizing the intelligent conversion from multi-source heterogeneous data to interpretable and comparable health indicators. 3. When determining the vibration target parameters during the start-up and shutdown of combined cycle units based on multi-source associated datasets, multi-dimensional feature extraction in the time domain, frequency domain, and time-frequency domain is performed on the multi-source associated dataset and multi-source historical datasets. The spatiotemporal correlation of multi-source features is established by combining dynamic time warping algorithm and historical operating condition labeling method. This effectively solves the problem of inconsistent time scales and blurred operating condition boundaries of multi-source heterogeneous data. The introduction of the Copula model to characterize the nonlinear dependence between multi-source features and vibration target parameters, and the quantification of tail correlation under extreme operating conditions by conditional correlation coefficient significantly improve the ability to capture complex correlations. Finally, the target margin of vibration target parameters is dynamically adjusted based on the contribution of multi-source features and the SSS clutch health status value. This allows the vibration target parameters to adaptively tighten or loosen according to the real-time health status of the clutch, overcoming the limitation of traditional fixed thresholds that cannot adapt to the aging and degradation of the unit. This provides a safe, controllable, and economically optimized dynamic vibration boundary for the start-up and shutdown process of combined cycle units. Attached Figure Description

[0018] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0019] Figure 1 A schematic flowchart of a combined cycle unit start-up and shutdown vibration control method with integrated clutch state provided as an embodiment of the present invention; Figure 2 A schematic diagram illustrating the implementation process of preprocessing multi-source correlation parameters is provided for another embodiment of the present invention; Figure 3 A schematic diagram illustrating the implementation process of SSS clutch health status analysis based on clutch state parameters, provided for another embodiment of the present invention; Figure 4A schematic diagram illustrating the implementation process of determining vibration target parameters during the start-up and shutdown of a combined cycle unit based on a multi-source correlation dataset, as provided in another embodiment of the present invention; Figure 5 A schematic diagram illustrating the implementation process of determining the start-stop control strategy combination based on the unit start-stop control model and the SSS clutch health status value, as provided in another embodiment of the present invention; Figure 6 A schematic diagram of the structure of a combined cycle unit start-stop vibration control system with integrated clutch state provided for another embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the embodiments of the present invention to facilitate a better understanding of the invention. However, the technical solutions claimed in the present invention can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the following embodiments is for ease of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined with and referenced by each other without contradiction.

[0021] Existing methods rely on passive control strategies based on optimized operating parameters. These methods cannot adapt to parameter drift caused by changes in ambient temperature and unit aging, and require manual intervention from operators, resulting in limited control accuracy. To address these issues, this invention proposes a combined cycle unit start-up and shutdown vibration control method and system integrating clutch status. In short, the proposed method first preprocesses multi-source correlated parameters, analyzes the health status of the SSS clutch based on clutch status parameters, determines the start-up and shutdown control strategy combination based on the unit start-up and shutdown control model and the SSS clutch health status value, filters the start-up and shutdown control strategy combination using a particle swarm optimization algorithm, outputs the optimal vibration control strategy for the start-up and shutdown process, and finally, responds to the optimal vibration control strategy by automatically triggering the control adjustment parameters of the unit start-up and shutdown actuators, and generates control adjustment commands based on these parameters. This invention accurately identifies and analyzes the health status of the SSS clutch by real-time acquisition and preprocessing of multi-source correlation parameters. It then dynamically determines vibration target parameters based on multi-source data, and finally generates a combination of start-stop control strategies based on the unit start-stop control model and health status values, achieving a shift from passive response to active prediction and adaptive optimization. By embedding the clutch health status as a key feedforward variable into the control decision-making process, it effectively compensates for system characteristic drift caused by changes in environmental factors and aging of unit components. While ensuring vibration suppression, it also considers start-stop efficiency and equipment lifespan. Furthermore, the entire process is automated, requiring no manual intervention, significantly improving the safety, stability, and control accuracy of the combined cycle unit's start-stop process.

[0022] One embodiment of the present invention provides a method for vibration control during start-up and shutdown of a combined cycle power unit with integrated clutch status, the process of which is as follows: Figure 1 As shown, it includes steps S10 to S50.

[0023] S10 collects multi-source correlation parameters during the start-up and shutdown of the combined cycle unit in real time, preprocesses the multi-source correlation parameters, and outputs the preprocessed multi-source correlation dataset. These multi-source correlation parameters include, but are not limited to, gas turbine speed, steam turbine speed, generator speed, SSS clutch engagement / disengagement status signal, clutch engagement time, torque impact amplitude during engagement, helical tooth surface wear, synchronization mechanism spring fatigue coefficient, and historical engagement count.

[0024] S20 identifies clutch state parameters that represent the health status of the SSS clutch in a multi-source associated dataset, analyzes the health status of the SSS clutch based on the clutch state parameters, and outputs the health status value of the SSS clutch.

[0025] S30: Based on a multi-source correlation dataset, the vibration target parameters during the start-up and shutdown process of the combined cycle unit are determined. Using the vibration target parameters and the multi-source correlation dataset as inputs, the start-up and shutdown control strategy combination is determined based on the unit start-up and shutdown control model and the SSS clutch health status value. Among them, the vibration target parameters include the target amplitude of shaft torsional vibration and the peak value of bearing vibration response.

[0026] S40: Obtain the start-stop control strategy combination. With the optimization objectives of maximum vibration suppression, shortest start-stop time, and minimum active torque compensation, construct a multi-objective optimization function for the strategy. Based on the particle swarm optimization algorithm, screen the start-stop control strategy combination and output the optimal vibration control strategy for the start-stop process.

[0027] S50, in response to the optimal vibration control strategy during start-up and shutdown, automatically triggers the control adjustment parameters of the unit start-up and shutdown actuators, and generates control adjustment commands based on these parameters. These control adjustment parameters include, but are not limited to, the gas turbine fuel supply rate, the steam turbine bypass valve opening, the SSS clutch engagement / disengagement control pressure or speed, and the rate of change of rotational speed.

[0028] This embodiment accurately identifies and analyzes the health status of the SSS clutch by real-time acquisition and preprocessing of multi-source correlation parameters. It then dynamically determines vibration target parameters based on multi-source data, and finally generates a combination of start-stop control strategies based on the unit start-stop control model and health status values, achieving a shift from passive response to active prediction and adaptive optimization. By embedding the clutch health status as a key feedforward variable into the control decision-making process, it effectively compensates for system characteristic drift caused by changes in environmental factors and aging of unit components. While ensuring vibration suppression, it also considers start-stop efficiency and equipment lifespan. Furthermore, the entire process is automated without manual intervention, significantly improving the safety, stability, and control accuracy of the combined cycle unit's start-stop process.

[0029] For example, combined Figure 2 The preprocessing of multi-source correlation parameters includes steps S101 to S105.

[0030] S101: Acquire multi-source correlation parameters during the start-up and shutdown process of the combined cycle unit. Using speed signals and clutch status signals as labels, traverse and mark the multi-source correlation parameters. Perform format standardization processing on the marked multi-source correlation parameters to obtain standardized multi-source correlation parameters. The traversal and marking of multi-source correlation parameters, and the format standardization processing of the marked data, can eliminate the heterogeneity between data from different sources, with different sampling rates, and different dimensions, ensuring that the subsequent analysis module receives a data stream with a unified structure and clear semantics.

[0031] S102: Load the standardized multi-source correlation parameters, preset an adaptive time threshold, and use the adaptive time threshold as a prior. Combine this with an improved temporal convolutional clustering method to adaptively filter the multi-source correlation parameters, obtaining continuous multi-source correlation parameters within the adaptive time threshold. Introducing an adaptive time threshold as a prior and combining it with the improved temporal convolutional clustering method to adaptively filter multi-source correlation parameters can effectively extract parameter segments that are continuous in time and logically correlated, filtering out isolated noise points, communication packet loss, or abnormal jump data.

[0032] For example, the formula for adaptive filtering of multi-source correlation parameters is as follows: ; in, Multi-source correlation parameters representing continuity For the standardized multi-source correlation parameters, For the i-th type of original association parameters, These represent the system standard time and the acquisition time of the i-th type of original correlation parameter, respectively. This is an adaptive time threshold.

[0033] S103, obtain the continuous multi-source correlation parameters within the adaptive time threshold, and normalize the multi-source correlation parameters based on the batch normalization method to obtain the normalized multi-source correlation parameters.

[0034] S104: Using the speed signal from the multi-source correlation parameters as the observation vector and the clutch state signal from the multi-source correlation parameters as the state vector, the extended Kalman filter algorithm, with the observation vector as input, performs recursive minimum variance estimation on the state vector, outputting a clutch health feature vector characterizing the clutch's health features. By using the extended Kalman filter algorithm for recursive minimum variance estimation, the clutch health feature vector reflecting the clutch's health evolution can be extracted with high accuracy even in the presence of measurement noise and process uncertainties.

[0035] S105: Obtain normalized multi-source correlation parameters and clutch health feature vectors, and generate a unified encapsulated multi-source correlation dataset based on the multi-source correlation parameters and clutch health feature vectors.

[0036] To address the problems of inconsistent multi-source data quality, weak spatiotemporal correlation, and insufficient accuracy in clutch health feature extraction in existing technologies, which lead to poor vibration control reliability, this embodiment proposes a method for preprocessing multi-source correlation parameters. This method achieves unified representation and structured organization of heterogeneous multi-source data by labeling and standardizing data formats using speed signals and clutch state signals. Furthermore, it introduces an adaptive time threshold combined with an improved temporal convolutional clustering method for adaptive data filtering, effectively ensuring the temporal continuity and operational condition correlation of the data, and avoiding information loss or redundancy caused by traditional fixed window truncation. Based on this, batch normalization is used to eliminate dimensional differences, and an extended Kalman filter algorithm is used to recursively estimate the clutch state vector using the speed observation vector, significantly improving the accuracy and anti-interference capability of clutch health feature extraction. Finally, a unified and encapsulated multi-source correlation dataset is generated, providing a high-quality and highly reliable data foundation for subsequent clutch health state analysis, vibration target parameter determination, and start-stop control strategy optimization.

[0037] For example, combined Figure 3 The health status analysis of the SSS clutch is based on the clutch status parameters, including steps S201 to S206.

[0038] S201, obtain the clutch state parameters characterizing the health state of the SSS clutch, perform short-time Fourier transform on the clutch state parameters, and extract the frequency domain features of the clutch state.

[0039] S202 employs the Hilbert-Huang Transform (HHT) method to extract the time-frequency domain features of clutch state parameters, obtaining the clutch state frequency domain features, clutch state time-frequency domain features, and clutch health feature vector. Correlation analysis is then used to calculate the feature sensitivity of these features. The HHT extraction of the clutch state parameters' time-frequency domain features accurately characterizes the instantaneous frequency and energy variations of non-stationary signals, making it suitable for the impact and slippage processes during clutch engagement / disengagement.

[0040] S203, determine whether the feature sensitivity exceeds a preset sensitivity threshold. By determining whether the feature sensitivity exceeds the preset sensitivity threshold, automated decision-making for feature selection is achieved.

[0041] S204. If the feature sensitivity exceeds the preset sensitivity threshold, retain the corresponding clutch state frequency domain features, clutch state time-frequency domain features, and clutch health feature vector, and generate a health state feature set based on the retained clutch state frequency domain features, clutch state time-frequency domain features, and clutch health feature vector.

[0042] S205: Obtain the health status feature set, and adaptively fuse the features within the health status feature set based on the heterogeneous graph attention mechanism of DS-GNN to output a comprehensive fusion vector.

[0043] S206, Load the integrated fusion vector, take the integrated fusion vector as input, calculate the SSS clutch health status value based on the fusion vector-health status value function, and output the SSS clutch health status value.

[0044] The fusion vector-health status value function is expressed as: ; in, This indicates the health status value of the SSS clutch. This represents the activation function of the fusion vector-health state value function. Represents the i-th feature component The physical degradation function, where n represents the total number of characteristic components. Let represent the adaptive weight, weight coefficient, and bias term of the i-th element in the fusion vector, respectively. This is a deep discriminative network composed of LSTM networks, where the superscript T indicates transpose.

[0045] In this embodiment, by combining short-time Fourier transform and Hilbert-Huang transform, multi-dimensional feature extraction of clutch state parameters from the frequency domain to the time-frequency domain is achieved, comprehensively capturing the steady-state characteristics and transient dynamic characteristics during clutch engagement. Correlation analysis is then used for feature sensitivity screening, retaining highly sensitive features based on preset sensitivity thresholds, effectively eliminating redundant information and reducing the complexity of subsequent fusion calculations. Furthermore, the DS-GNN heterogeneous graph attention mechanism is introduced, modeling the complex relationships between different features through a heterogeneous graph structure, and using a graph attention network to achieve adaptive weighted fusion of features, significantly improving the ability of the comprehensive fusion vector to represent the clutch health state. Finally, the high-dimensional fusion features are mapped to quantified health state values ​​through a fusion vector-health state value function, realizing the intelligent conversion from multi-source heterogeneous data to interpretable and comparable health indicators.

[0046] For example, combined Figure 4 The vibration target parameters during the start-up and shutdown process of the combined cycle unit are determined based on the multi-source correlation dataset, including steps S301 to S305.

[0047] S301, obtain the multi-source associated dataset, and retrieve the associated multi-source historical dataset based on the multi-source associated dataset. Perform time domain, frequency domain, and time-frequency domain feature extraction on the multi-source associated dataset and the multi-source historical dataset respectively to obtain the multi-source feature set.

[0048] S302 uses multi-source historical datasets as prior operating condition label information and employs a dynamic time warping algorithm combined with historical operating condition labeling to establish spatiotemporal relationships among multi-source features within a multi-source feature set, resulting in a multi-source feature set containing spatiotemporal relationships among the features. By using operating condition label information from multi-source historical datasets as prior information and combining the dynamic time warping (DTW) algorithm with historical operating condition labeling to establish spatiotemporal relationships between multi-source features, the problem of inconsistent feature sequence lengths and time misalignments under different operating conditions can be solved.

[0049] S303, based on the Teng copula model, determines the nonlinear dependence between multi-source features and vibration target parameters in the multi-source feature set, and calculates the conditional correlation coefficient between multi-source features and vibration target parameters.

[0050] The formula for calculating the conditional correlation coefficient is as follows: ; in, Let i and j represent the i-th feature vector and the vibration target parameter, respectively, in the multi-source feature set. The weights represent the nonlinear correlation between multi-source features and vibration target parameters. This represents the partial correlation coefficient based on the Fujicopula model. These are the parameters for the copula model.

[0051] S304 uses the conditional correlation coefficient between multi-source features and vibration target parameters as a constraint, and calculates the contribution of multi-source features to vibration target parameters based on the fusion method of permutation importance and SHAP value.

[0052] The formula for calculating the contribution of multi-source features to the vibration target parameters is as follows: ; in, This indicates the contribution of multi-source features to the vibration target parameters. Let represent the normalized permutation importance and the normalized SHAP importance of the i-th feature vector in the multi-source feature set, respectively. This represents the number of multi-source features, which is also the total number of feature vectors. The weight coefficient of multi-source feature fusion is the weight coefficient corresponding to the i-th feature vector in the multi-source feature set.

[0053] S305: Obtain the initial safety limit of the vibration target parameter, dynamically adjust the target margin of the vibration target parameter based on the contribution of multi-source characteristics to the vibration target parameter and the health status value of the SSS clutch, and use the adjusted target margin of the vibration target parameter as the vibration target parameter during the start-up and shutdown process of the combined cycle unit.

[0054] The formula for calculating the target margin of the adjusted vibration target parameters is as follows: ; in, This indicates the target margin of the adjusted vibration target parameters. As the initial safety limit, Based on the basic margin factor, These are the health degradation coefficient, thermal state correction coefficient, and grid demand correction coefficient, respectively, where e represents the base of the natural logarithm.

[0055] In this embodiment, when determining the vibration target parameters during the start-up and shutdown of a combined cycle unit based on a multi-source associated dataset, multi-dimensional feature extraction in the time domain, frequency domain, and time-frequency domain is performed on the multi-source associated dataset and the multi-source historical dataset. Combined with a dynamic time warping algorithm and a historical operating condition labeling method, a spatiotemporal correlation of multi-source features is established, effectively solving the problems of inconsistent time scales and ambiguous operating condition boundaries in heterogeneous multi-source data. Furthermore, the introduction of the Copula model characterizes the nonlinear dependency between multi-source features and vibration target parameters, and the quantification of tail correlation under extreme operating conditions through conditional correlation coefficients significantly improves the ability to capture complex correlations. Finally, the target margin of the vibration target parameters is dynamically adjusted based on the contribution of multi-source features combined with the SSS clutch health status value, enabling the vibration target parameters to adaptively tighten or loosen according to the real-time health status of the clutch. This overcomes the limitation of traditional fixed thresholds being unable to adapt to unit aging and degradation, providing a safe, controllable, and economically optimized dynamic vibration boundary for the start-up and shutdown process of the combined cycle unit.

[0056] For example, combined Figure 5 The start-stop control strategy combination is determined based on the unit start-stop control model and the health status value of the SSS clutch, including steps S401 to S404.

[0057] S401: Obtain vibration target parameters and multi-source correlation datasets. Based on these parameters, establish a multi-physics coupled dynamic model for the start-up and shutdown process of the combined cycle unit. Also, establish a clutch engagement function that includes the health status value of the SSS clutch, and embed this function into the multi-physics coupled dynamic model. This multi-physics coupled dynamic model can simultaneously describe the dynamic mathematical model of multiple physical fields (such as mechanical motion fields, thermodynamic fields, hydraulic / pneumatic transmission fields, electromagnetic fields, etc.) and their mutual coupling during the start-up and shutdown of the combined cycle unit. It typically includes physical effects such as mass, stiffness, damping, friction, thermal expansion, and hydraulic pressure, and expresses the dynamic behavior of subsystems such as the gas turbine, steam turbine, generator, shaft system, and SSS clutch in the form of differential equations or algebraic-differential hybrid equations.

[0058] The clutch engagement function is expressed as: ; ; ; in, This represents the clutch engagement function. These represent the clutch friction constraint term, the clutch equivalent stiffness constraint term, and the slip friction stage function, respectively, where C represents the clutch equivalent stiffness and t represents time. These are the basic equivalent stiffness and the stiffness influence coefficient, respectively. These are the friction coefficient in the brand-new state and the wear influence coefficient, respectively.

[0059] It should be noted that the clutch engagement function takes the SSS clutch health state value as the core input and is dynamically adjusted according to changes in operating conditions such as time, speed, and load. It describes the key dynamic parameters (friction coefficient, equivalent stiffness, maximum permissible transmitted torque, and response time) of the clutch during engagement / disengagement. Its mathematical form can be embedded into a multiphysics coupled dynamics model, enabling the model to automatically adjust its dynamic characteristics based on the actual health level of the clutch, achieving health state feedforward control.

[0060] S402, a multi-physics coupled dynamics model with embedded clutch engagement function is loaded. The multi-physics coupled dynamics model is reduced in order based on the intrinsic orthogonal decomposition method to obtain a multi-physics reduced-order model.

[0061] S403 constructs a multi-dimensional strategy control space using speed control strategy, temperature control strategy, clutch control strategy, and torque control strategy, and builds a unit start-stop control model based on the multi-dimensional strategy control space.

[0062] S404, based on the learning of the mapping relationship between the multi-dimensional strategy control space and the multi-physics reduced-order model of the unit start-up and shutdown control model, uses the Sequential Quadratic Programming (SQP) method to solve the multi-physics reduced-order model, obtaining a combination of start-up and shutdown control strategies. SQP is characterized by fast convergence and high accuracy in handling constrained nonlinear optimization problems. Combined with the computational efficiency of the reduced-order model, the entire method can generate or update control strategies in real time at different stages of the start-up and shutdown process, achieving online rolling optimization and adaptive control, thereby significantly improving the safety, economy, and vibration control level of the combined cycle unit start-up and shutdown process.

[0063] In this embodiment, a multiphysics coupled dynamic model is constructed by embedding a clutch engagement function containing SSS clutch health state values. This achieves a unified representation of the rotor-heat-clutch multiphysics coupling mechanism and health state degradation characteristics, enabling the model to accurately reflect the nonlinear influence of clutch engagement impact on shaft vibration under different health states. Furthermore, the intrinsic orthogonal decomposition method is used to reduce the order of the high-dimensional complex model, improving computational efficiency by more than two orders of magnitude while preserving key dynamic characteristics, thus meeting the timeliness requirements of real-time optimization decision-making. Based on this, a multidimensional strategy control space covering four dimensions—speed, temperature, clutch, and torque—is constructed, breaking through the limitations of traditional single-parameter adjustment and realizing the collaborative optimization of multiple control variables during start-up and shutdown. Finally, the mapping relationship between the multidimensional strategy control space and the multiphysics reduced-order model is learned through the unit start-up and shutdown control model, and the sequential quadratic programming method is used for efficient solution, yielding the optimal control strategy combination that balances dynamic suppression, start-up and shutdown speed, and energy consumption economy.

[0064] Another embodiment of the present invention provides a combined cycle unit start-up and shutdown vibration control system with integrated clutch state, used to implement the combined cycle unit start-up and shutdown vibration control method with integrated clutch state provided in the above embodiments. Figure 6 As shown, the integrated clutch status combined cycle unit start-stop vibration control system includes a data preprocessing module 100, a clutch status evaluation module 200, a control strategy generation module 300, an optimal strategy screening module 400, and a start-stop vibration control module 500.

[0065] The data preprocessing module 100 is used to collect multi-source correlation parameters in real time during the start-up and shutdown process of the combined cycle unit, preprocess the multi-source correlation parameters, and output the preprocessed multi-source correlation dataset.

[0066] The data preprocessing module 100 includes a parameter labeling unit 110, an adaptive filtering unit 120, a feature vector derivation unit 130, and a dataset encapsulation unit 140.

[0067] The parameter marking unit 110 is used to acquire multi-source associated parameters during the start-up and shutdown process of the combined cycle unit. It uses speed signal and clutch status signal as labels to traverse the multi-source associated parameters and mark them. It then performs format standardization processing on the marked multi-source associated parameters to obtain standardized multi-source associated parameters. The adaptive filtering unit 120 is used to load the standardized multi-source association parameters, preset the adaptive time threshold, use the adaptive time threshold as a prior, and combine the improved temporal convolution clustering method to adaptively filter the multi-source association parameters to obtain continuous multi-source association parameters within the adaptive time threshold. The feature vector derivation unit 130 is used to obtain continuous multi-source correlation parameters within the adaptive time threshold. The multi-source correlation parameters are normalized based on the batch normalization method to obtain normalized multi-source correlation parameters. The speed signal in the multi-source correlation parameters is defined as the observation vector, and the clutch state signal in the multi-source correlation parameters is defined as the state vector. Through the extended Kalman filter algorithm, the observation vector is used as the input to perform recursive minimum variance estimation on the state vector, and the clutch health feature vector representing the clutch health characteristics is output. The dataset encapsulation unit 140 is used to obtain normalized multi-source correlation parameters and clutch health feature vectors, and to generate a unified encapsulated multi-source correlation dataset based on the multi-source correlation parameters and clutch health feature vectors.

[0068] The clutch condition assessment module 200 is used to identify clutch condition parameters that represent the health status of the SSS clutch in a multi-source associated dataset, analyze the health status of the SSS clutch based on the clutch condition parameters, and output the health status value of the SSS clutch.

[0069] The control strategy generation module 300 is used to determine the vibration target parameters during the start-up and shutdown process of the combined cycle unit based on the multi-source correlation dataset. Taking the vibration target parameters and the multi-source correlation dataset as input, the start-up and shutdown control strategy combination is determined based on the unit start-up and shutdown control model and the SSS clutch health status value.

[0070] The optimal strategy screening module 400 is used to obtain the start-stop control strategy combination. With the optimization objectives of maximum vibration suppression, shortest start-stop time, and minimum active torque compensation, a multi-objective optimization function of strategy is constructed. The start-stop control strategy combination is screened based on the particle swarm optimization algorithm, and the optimal vibration control strategy during the start-stop process is output. The start-stop vibration control module 500 is used to respond to the optimal vibration control strategy during the start-stop process, automatically trigger the control adjustment parameters of the unit start-stop actuator, and generate control adjustment commands based on the control adjustment parameters of the unit start-stop actuator.

[0071] The specific implementation method of the combined cycle unit start-up and shutdown vibration control system with integrated clutch state provided in the embodiments of the present invention can be found in the description of the combined cycle unit start-up and shutdown vibration control method with integrated clutch state provided in the embodiments of the present invention, and will not be repeated here.

[0072] In summary, this invention provides a method and system for vibration control during the start-up and shutdown of a combined cycle power unit, integrating clutch status. By real-time acquisition and preprocessing of multi-source correlated parameters, it accurately identifies and analyzes the health status of the SSS clutch, then dynamically determines vibration target parameters based on multi-source data. Finally, it generates a combination of start-up and shutdown control strategies based on the unit's start-up and shutdown control model and health status values, achieving a shift from passive response to active prediction and adaptive optimization. By embedding the clutch health status as a key feedforward variable into the control decision-making process, it effectively compensates for system characteristic drift caused by changes in environmental factors and aging of unit components. It ensures vibration suppression while balancing start-up and shutdown efficiency and equipment lifespan. Furthermore, the entire process is automated, requiring no manual intervention, significantly improving the safety, stability, and control accuracy of the combined cycle power unit's start-up and shutdown process.

[0073] It should be noted that, for the sake of simplicity, the foregoing embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to the present invention. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0074] In the embodiments provided by this invention, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or communication connections shown or discussed may be through some interfaces; the indirect coupling or communication connections between devices or units may be telecommunications or other forms.

[0075] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0076] Those skilled in the art will understand that the above embodiments are specific implementations of the present invention, and in practical applications, various changes can be made in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A method for integrated clutch status vibration control of start-stop of a combined cycle unit, characterized in that, include: Real-time acquisition of multi-source correlation parameters during the start-up and shutdown of combined cycle units; preprocessing of multi-source correlation parameters; output of preprocessed multi-source correlation dataset. Identify clutch state parameters that represent the health status of SSS clutches in multi-source association datasets, analyze the health status of SSS clutches based on clutch state parameters, and output the health status value of SSS clutches. Vibration target parameters during the start-up and shutdown process of the combined cycle unit are determined based on the multi-source correlation dataset. Using the vibration target parameters and the multi-source correlation dataset as inputs, the start-up and shutdown control strategy combination is determined based on the unit start-up and shutdown control model and the health status value of the SSS clutch.

2. The integrated clutch status combined cycle unit start-stop vibration control method of claim 1, wherein, The vibration control method for start-up and shutdown of combined cycle units in the integrated clutch state further includes: A combination of start-stop control strategies is obtained. With the optimization objectives of maximum vibration suppression, shortest start-stop time, and minimum active torque compensation, a multi-objective optimization function for the strategy is constructed. The combination of start-stop control strategies is screened based on the particle swarm optimization algorithm, and the optimal vibration control strategy during the start-stop process is output. In response to the optimal vibration control strategy during start-up and shutdown, the control adjustment parameters of the unit start-up and shutdown actuators are automatically triggered, and control adjustment commands are generated based on the control adjustment parameters of the unit start-up and shutdown actuators.

3. The integrated clutch status combined cycle unit start-stop vibration control method of claim 2, wherein, The preprocessing of multi-source correlation parameters includes: The multi-source correlation parameters during the start-up and shutdown process of the combined cycle unit are obtained. Using speed signal and clutch status signal as labels, the multi-source correlation parameters are traversed and marked. The marked multi-source correlation parameters are then standardized to obtain the standardized multi-source correlation parameters. The standardized multi-source association parameters are loaded, and an adaptive time threshold is preset. Using the adaptive time threshold as a priori, the multi-source association parameters are adaptively filtered by the improved temporal convolutional clustering method to obtain continuous multi-source association parameters within the adaptive time threshold. Obtain continuous multi-source correlation parameters within an adaptive time threshold, and normalize the multi-source correlation parameters based on the batch normalization method to obtain normalized multi-source correlation parameters. The speed signal in the multi-source correlation parameters is defined as the observation vector, and the clutch state signal in the multi-source correlation parameters is defined as the state vector. The extended Kalman filter algorithm is used to recursively estimate the minimum variance of the state vector with the observation vector as input, and outputs a clutch health feature vector that characterizes the clutch health characteristics. Obtain normalized multi-source correlation parameters and clutch health feature vectors, and generate a unified encapsulated multi-source correlation dataset based on the multi-source correlation parameters and clutch health feature vectors.

4. The integrated clutch status combined cycle unit start-stop vibration control method of claim 3, wherein, The SSS clutch health status analysis based on clutch state parameters includes: The clutch state parameters characterizing the health state of the SSS clutch are obtained, and the clutch state parameters are processed by short-time Fourier transform to extract the frequency domain features of the clutch state. The Hilbert-Huang transform method is used to extract the clutch state time-frequency domain features of the clutch state parameters, and obtain the clutch state frequency domain features, clutch state time-frequency domain features and clutch health feature vector. Correlation analysis is used to calculate the feature sensitivity of the clutch state frequency domain features, clutch state time-frequency domain features and clutch health feature vector. Determine whether the feature sensitivity exceeds a preset sensitivity threshold; If the feature sensitivity exceeds the preset sensitivity threshold, the corresponding clutch state frequency domain features, clutch state time frequency domain features, and clutch health feature vector are retained. A health state feature set is generated based on the retained clutch state frequency domain features, clutch state time frequency domain features, and clutch health feature vector. Obtain a health status feature set, and adaptively fuse the features within the health status feature set based on the heterogeneous graph attention mechanism of DS-GNN to output a comprehensive fusion vector; Load the integrated fusion vector, use the integrated fusion vector as input, calculate the SSS clutch health status value based on the fusion vector-health status value function, and output the SSS clutch health status value.

5. The integrated clutch status combined cycle unit start-stop vibration control method of claim 4, wherein, The determination of vibration target parameters during the start-up and shutdown process of a combined cycle unit based on a multi-source correlation dataset includes: Obtain a multi-source associated dataset, and retrieve the associated multi-source historical dataset based on the multi-source associated dataset. Extract time-domain, frequency-domain, and time-frequency-domain features from the multi-source associated dataset and the multi-source historical dataset respectively to obtain a multi-source feature set. Using multi-source historical datasets as prior working condition label information, dynamic time warping algorithm combined with historical working condition labeling method is used to establish the spatiotemporal correlation of multi-source features within the multi-source feature set, resulting in a multi-source feature set containing spatiotemporal relationships of multi-source features. The nonlinear dependence between multi-source features and vibration target parameters in the multi-source feature set is determined based on the vine copula model, and the conditional correlation coefficient between multi-source features and vibration target parameters is calculated. Using the conditional correlation coefficient between multi-source features and vibration target parameters as constraints, the contribution of multi-source features to vibration target parameters is calculated based on the fusion method of permutation importance and SHAP value. The initial safety limits of the vibration target parameters are obtained. The target margin of the vibration target parameters is dynamically adjusted based on the contribution of multi-source characteristics to the vibration target parameters and the health status value of the SSS clutch. The target margin of the adjusted vibration target parameters is used as the vibration target parameters during the start-up and shutdown process of the combined cycle unit.

6. The integrated clutch status combined cycle unit start-stop vibration control method of claim 5, wherein, The method for determining the start-stop control strategy combination based on the unit start-stop control model and the SSS clutch health status value includes: Obtain vibration target parameters and multi-source correlation datasets. Based on the vibration target parameters and multi-source correlation datasets, establish a multi-physics field coupled dynamic model during the start-up and shutdown process of the combined cycle unit. Establish a clutch engagement function that includes the SSS clutch health state value and embed the clutch engagement function into the multi-physics field coupled dynamic model. A multiphysics coupled dynamics model with embedded clutch engagement function is loaded, and the order of the multiphysics coupled dynamics model is reduced based on the intrinsic orthogonal decomposition method to obtain a multiphysics reduced-order model.

7. The integrated clutch status combined cycle unit start-stop vibration control method of claim 6, wherein, The method for determining the start-stop control strategy combination based on the unit start-stop control model and the SSS clutch health status value also includes: A multi-dimensional strategy control space is constructed using speed control strategy, temperature control strategy, clutch control strategy, and torque control strategy, and a unit start-up and shutdown control model is constructed based on the multi-dimensional strategy control space; Based on the learning of the mapping relationship between the multi-dimensional strategy control space and the multi-physics field reduced-order model of the unit start-stop control model, the multi-physics field reduced-order model is solved by the sequential quadratic programming (SQP) method to obtain the start-stop control strategy combination.

8. The integrated clutch status combined cycle unit start-stop vibration control method of claim 6, wherein, The clutch engagement function is expressed as follows: ； ； ； in, This represents the clutch engagement function. These represent the clutch friction constraint term, the clutch equivalent stiffness constraint term, and the slip friction stage function, respectively, where C represents the clutch equivalent stiffness and t represents time. These are the basic equivalent stiffness and the stiffness influence coefficient, respectively. These are the friction coefficient in the brand-new state and the wear influence coefficient, respectively. Indicates the health status value of the SSS clutch and , This represents the activation function of the fusion vector-health state value function. Represents the i-th feature component The physical degradation function, where n represents the total number of eigencomponents. Let represent the adaptive weight, weight coefficient, and bias term of the i-th element in the fusion vector, respectively. This is a deep discriminative network composed of LSTM networks, where the superscript T indicates transpose.

9. A combined cycle power unit start-up and shutdown vibration control system with integrated clutch status, characterized in that, A method for controlling the start-up and shutdown vibration of a combined cycle unit in an integrated clutch state according to any one of claims 1 to 8, wherein the integrated clutch state combined cycle unit start-up and shutdown vibration control system comprises: The data preprocessing module is used to collect multi-source correlation parameters in real time during the start-up and shutdown process of the combined cycle unit, preprocess the multi-source correlation parameters, and output the preprocessed multi-source correlation dataset. The clutch condition assessment module is used to identify clutch condition parameters that represent the health status of the SSS clutch in a multi-source associated dataset, analyze the health status of the SSS clutch based on the clutch condition parameters, and output the SSS clutch health status value. The control strategy generation module is used to determine the vibration target parameters during the start-up and shutdown process of the combined cycle unit based on the multi-source correlation dataset. Taking the vibration target parameters and the multi-source correlation dataset as input, the module determines the start-up and shutdown control strategy combination based on the unit start-up and shutdown control model and the SSS clutch health status value. The optimal strategy screening module is used to obtain the start-stop control strategy combination. With the optimization objectives of maximum vibration suppression, shortest start-stop time, and minimum active torque compensation, a multi-objective optimization function of strategy is constructed. The start-stop control strategy combination is screened based on the particle swarm optimization algorithm, and the optimal vibration control strategy during the start-stop process is output. The start-stop vibration control module is used to respond to the optimal vibration control strategy during the start-stop process, automatically trigger the control adjustment parameters of the unit start-stop actuator, and generate control adjustment commands based on the control adjustment parameters of the unit start-stop actuator.

10. The combined cycle unit start-up and shutdown vibration control system with integrated clutch state according to claim 9, characterized in that, The data preprocessing module includes: The parameter labeling unit is used to acquire multi-source associated parameters during the start-up and shutdown process of the combined cycle unit. Using speed signal and clutch status signal as labels, it traverses the multi-source associated parameters and labels them. The labeled multi-source associated parameters are then processed for format standardization to obtain standardized multi-source associated parameters. An adaptive filtering unit is used to load standardized multi-source association parameters, preset an adaptive time threshold, use the adaptive time threshold as a priori, and combine an improved temporal convolutional clustering method to adaptively filter the multi-source association parameters to obtain continuous multi-source association parameters within the adaptive time threshold. The feature vector derivation unit is used to obtain continuous multi-source correlation parameters within the adaptive time threshold. The multi-source correlation parameters are normalized based on the batch normalization method to obtain normalized multi-source correlation parameters. The speed signal in the multi-source correlation parameters is defined as the observation vector, and the clutch state signal in the multi-source correlation parameters is defined as the state vector. Through the extended Kalman filter algorithm, the state vector is recursively estimated with the observation vector as input, and the clutch health feature vector representing the clutch health characteristics is output. The dataset encapsulation unit is used to obtain normalized multi-source correlation parameters and clutch health feature vectors, and to generate a unified encapsulated multi-source correlation dataset based on the multi-source correlation parameters and clutch health feature vectors.

Images