Water washing prediction method of gas turbine and related device
By evaluating the multiplicative noise level and performing precise washing operations through wavelet transformation and model prediction of gas flow parameters of gas turbines, the gas turbine scale problem is solved and the reliability and safety of the equipment is improved.
Patent Information
- Application Number
- CN202410918087.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-10
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-10
Smart Images

Figure CN118916688A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to the field of data processing. More specifically, the embodiments of the present application relate to a water washing prediction method and related devices for a gas turbine. Background Art
[0002] Gas-steam combined cycle units are widely used in aviation, energy supply, industrial manufacturing and other fields because of their low emissions, high thermal efficiency, fast start-up, flexible operation, and suitability for peak load regulation.
[0003] The performance of the gas turbine, as the core component of the gas-steam combined cycle unit, will directly affect the operation of the entire unit. The gas turbine needs to continuously inhale air during operation, and the inhaled air may contain pollutants such as dust, powder, and insects.
[0004] In the related technology, before the air enters the gas turbine through the compressor, most of the pollutants in the air have been removed by the air intake filter, but a small amount of pollutants still enter the compressor and continue to be deposited in the compressor, reducing the efficiency of the compressor and even causing serious accidents such as surge, affecting the safety and reliability of the gas turbine.
[0005] In summary, there is an urgent need to provide a prediction scheme for compressor fouling degree to solve the above-mentioned technical problems existing in related technologies. Summary of the invention
[0006] In this context, the embodiments of the present application hope to provide a water washing prediction method and related devices for a gas turbine, which are used to implement water washing operations on the gas turbine that match the multiplicative noise level in the operating parameters, which can effectively reduce the damage caused by scaling, maintain the normal operating state of the equipment, and improve the reliability and safety of the gas turbine.
[0007] In a first aspect of an embodiment of the present application, a water wash prediction method for a gas turbine is provided, comprising:
[0008] Obtaining operating parameters of the gas turbine; the operating parameters at least include: gas flow parameters;
[0009] The gas flow parameters of the gas turbine are subjected to wavelet transformation to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band;
[0010] The wavelet coefficient characteristics under each frequency band are input into the frequency band transition probability matrix in the frequency band state prediction model for state prediction, and the frequency band state sequence corresponding to the wavelet coefficient characteristics under each frequency band is obtained; the frequency band transition probability matrix is used to describe the transition probability between states of the wavelet coefficient characteristics under different frequency bands;
[0011] The wavelet coefficient features corresponding to each predicted frequency band state are input into the noise intensity extraction model based on the frequency band state sequence to obtain the multiplicative noise intensity parameter; wherein the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient features corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters;
[0012] A water washing operation matching the multiplicative noise intensity parameter is performed on the gas turbine to reduce the fouling degree of the gas turbine.
[0013] In a second aspect of the embodiments of the present application, a water washing prediction device for a gas turbine is provided, the device comprising:
[0014] An acquisition module is used to acquire operating parameters of the gas turbine; the operating parameters at least include: gas flow parameters;
[0015] A transformation module is used to perform wavelet transformation on the gas flow parameters of the gas turbine to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band;
[0016] The prediction module is used to input the wavelet coefficient characteristics under each frequency band into the frequency band transition probability matrix in the frequency band state prediction model to perform state prediction, and obtain the frequency band state sequence corresponding to the wavelet coefficient characteristics under each frequency band; the frequency band transition probability matrix is used to describe the transition probability between the states of the wavelet coefficient characteristics under different frequency bands;
[0017] A noise extraction module is used to input the wavelet coefficient features corresponding to each predicted frequency band state into the noise intensity extraction model based on the frequency band state sequence to obtain a multiplicative noise intensity parameter; wherein the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient features corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters;
[0018] The execution module is used to perform a water washing operation on the gas turbine that matches the multiplicative noise intensity parameter to reduce the fouling degree of the gas turbine.
[0019] In a third aspect of the embodiments of the present application, a computing device is provided, the computing device comprising:
[0020] at least one processor, memory, and input-output unit;
[0021] The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the water washing prediction method for a gas turbine of the first aspect.
[0022] In a fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided, which includes instructions. When the instructions are executed on a computer, the computer executes the water washing prediction method for a gas turbine of the first aspect.
[0023] In an embodiment of the present application, a water washing prediction method and a related device for a gas turbine are provided. In an embodiment of the present application, the operating parameters of the gas turbine are obtained, wherein the operating parameters at least include gas flow parameters. Then, the gas flow parameters of the gas turbine are subjected to wavelet transformation to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band. Then, the wavelet coefficient characteristics in each frequency band are input into the band transition probability matrix in the band state prediction model for state prediction to obtain the band state sequence corresponding to the wavelet coefficient characteristics in each frequency band. Then, the wavelet coefficient characteristics corresponding to each predicted band state are respectively input into the noise intensity extraction model based on the band state sequence to obtain the multiplicative noise intensity parameter, and the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine. Finally, a water washing operation matching the multiplicative noise intensity parameter is performed on the gas turbine to reduce the fouling degree of the gas turbine. In the implementation mode of the present application, the fouling degree of the gas turbine is evaluated in combination with the multiplicative noise level in the gas flow parameters, so that the corresponding water washing operation is performed on the gas turbine to reduce the fouling degree of the gas turbine, assist the gas turbine to maintain a normal operating state, and improve the performance of the gas turbine. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easily understood. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, wherein:
[0025] Figure 1 A schematic flow chart of a water wash prediction method for a gas turbine provided in one embodiment of the present application;
[0026] Figure 2 A schematic diagram of the principle of a water wash prediction method for a gas turbine provided in one embodiment of the present application;
[0027] Figure 3 A schematic diagram of the principle of a water wash prediction method for a gas turbine provided in another embodiment of the present application;
[0028] Figure 4 A schematic diagram of the structure of a water wash prediction device for a gas turbine provided in one embodiment of the present application;
[0029] Figure 5 A schematic diagram of the structure of a medium in an embodiment of the present application is schematically shown;
[0030] Figure 6 A schematic diagram of the structure of a computing device according to an embodiment of the present application is shown schematically.
[0031] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION
[0032] The principles and spirit of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present application, and are not intended to limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.
[0033] Those skilled in the art know that the embodiments of the present application can be implemented as a system, device, equipment, method or computer program product. Therefore, the present disclosure can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.
[0034] Gas-steam combined cycle units are widely used in aviation, energy supply, industrial manufacturing and other fields because of their low emissions, high thermal efficiency, fast start-up, flexible operation, and suitability for peak load regulation.
[0035] The performance of the gas turbine, as the core component of the gas-steam combined cycle unit, will directly affect the operation of the entire unit. The gas turbine needs to continuously inhale air during operation, and the inhaled air may contain pollutants such as dust, powder, and insects.
[0036] In the related technology, before the air enters the gas turbine through the compressor, most of the pollutants in the air have been removed by the air intake filter, but a small amount of pollutants still enter the compressor and continue to be deposited in the compressor, reducing the efficiency of the compressor and even causing serious accidents such as surge, affecting the safety and reliability of the gas turbine.
[0037] In the related art, in order to ensure the safe and economical operation of the gas-steam combined cycle unit, the gas turbine is usually washed offline to remove the scaling part, and the cleaning cycle is generally fixed according to the operating time of the gas turbine. Although regular water washing of the compressor can restore its performance, the scaling degree of the gas turbine is also affected by environmental factors, changes in the operating conditions of the gas-steam combined cycle unit, etc. The method of periodically cleaning the compressor in the gas turbine cannot maintain the efficient operation of the unit. In addition, frequent water washing will waste a large amount of desalted water and plant electricity, causing environmental pollution. If the water washing is not timely, the efficiency of the compressor will decrease, affecting the safety and reliability of the gas turbine.
[0038] In summary, there is an urgent need to provide a prediction scheme for compressor fouling degree to solve the above-mentioned technical problems existing in related technologies.
[0039] In order to overcome the above-mentioned technical problems existing in the existing solutions, according to the implementation mode of the present application, a water washing prediction method and related devices for a gas turbine are proposed.
[0040] In the implementation mode of the present application, first, it is necessary to obtain the operating parameters of the gas turbine, so that by obtaining the operating parameters of the gas turbine, the current working state and performance of the gas turbine can be understood, and the necessary data basis can be provided for subsequent analysis. In order to facilitate the convenience of the operating parameter acquisition process, the gas flow parameters can be used as the operating parameters for analysis. Furthermore, the gas flow parameters of the gas turbine are subjected to wavelet transformation, and the gas flow parameters can be converted into frequency domain information, thereby extracting the wavelet coefficient characteristics of the gas flow parameters in each frequency band. Next, the wavelet coefficient characteristics in each frequency band are input into the frequency band transition probability matrix in the frequency band state prediction model for state prediction, and the frequency band state sequence corresponding to the wavelet coefficient characteristics in each frequency band is obtained. The frequency band transition probability matrix used here is used to describe the transition probability between states of the wavelet coefficient characteristics in different frequency bands. Through the above steps, the gas flow characteristics inside the gas turbine can be identified and analyzed in different frequency ranges by combining the wavelet coefficient characteristics, thereby predicting the most likely gas flow state inside the gas turbine. Then, the wavelet coefficient features corresponding to each predicted frequency band state are input into the noise intensity extraction model based on the frequency band state sequence to obtain the multiplicative noise intensity parameter. The multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient features corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters. In this way, the noise intensity extraction model can be combined with the gas flow state inside the gas turbine to evaluate the multiplicative noise level in the gas flow parameters, so as to evaluate the fouling degree inside the gas turbine through the multiplicative noise level. Finally, a water washing operation matching the multiplicative noise intensity parameter is performed on the gas turbine to reduce the fouling degree of the gas turbine. In this way, by performing a water washing operation matching the multiplicative noise level in the operating parameters on the gas turbine, the damage caused by fouling can be effectively reduced, the normal operation state of the equipment can be maintained, and the reliability and safety of the gas turbine can be improved.
[0041] The technical solution provided in the embodiments of the present application can be implemented by a server and / or a terminal device. It should be noted that the server involved in the embodiments of the present application can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, big data and artificial intelligence platforms.
[0042] The terminal device involved in the embodiments of the present application may be a device that provides voice and / or data connectivity to a user, a handheld device with a wireless connection function, or other processing devices connected to a wireless modem. For example, a mobile phone (or "cellular" phone) and a computer with a mobile terminal, for example, may be a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device that exchanges voice and / or data with a wireless access network. For example, Personal Communication Service (PCS) phones, cordless phones, Session Initiation Protocol (SIP) phones, Wireless Local Loop (WLL) stations, Personal Digital Assistants (PDAs) and other devices.
[0043] It should be understood herein that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction rather than having any limiting meaning.
[0044] The principles and spirit of the present application are explained in detail below with reference to several representative implementations of the present application.
[0045] Reference below Figure 1 , Figure 1 A flow chart of a method for predicting water washing of a gas turbine provided in an embodiment of the present application. A method for predicting water washing of a gas turbine provided in an embodiment of the present application comprises:
[0046] Step S101, obtaining the operating parameters of the gas turbine.
[0047] The gas turbine involved in this application is a thermal device. The gas turbine mainly generates mechanical power by driving the rotating blades on the bearings through high-temperature gas. For example, the gas turbine is mainly composed of a compressor, a combustion chamber, a turbine, a load drive, and an exhaust system. The gas turbine mainly includes the following parts:
[0048] Compressor: Air is compressed through a compressor, increasing the pressure and temperature of the gas.
[0049] Combustion chamber: The compressed air is mixed with fuel and ignited in the combustion chamber, releasing high-temperature and high-pressure combustion gas.
[0050] Turbine: High-temperature and high-pressure combustion gas is sprayed onto the turbine through the gas turbine blades, causing the turbine to rotate at high speed.
[0051] Load drive: The rotating blades on the turbine drive shaft output high-speed rotating shaft power. The power generated by this part can be used to drive the power equipment in which it is located, such as the gas turbine in the gas-steam combined cycle unit, the turbofan in the jet engine (used as the power for aircraft propulsion), the compressor (used as the power for compressed natural gas), etc.
[0052] Exhaust system: The exhaust gas from the turbine is cooled through the exhaust system and discharged into the atmosphere.
[0053] As can be seen from the above structure, if scaling occurs in the gas turbine, it will affect the fluidity of air inside the gas turbine. Therefore, by analyzing the gas flow parameters inside the gas turbine, the degree of scaling inside the gas turbine can be obtained.
[0054] Since fouling occurs in every stage of the gas turbine operation, the gas flow noise (referred to as aerodynamic noise) caused by fouling is a multiplicative noise. Due to the different degrees of fouling and the locations of fouling (such as compressors, combustion chambers, turbines, load drives, and exhaust systems), aerodynamic noise presents different signal characteristics. For example, aerodynamic noise is distributed in different frequency bands and presents different noise intensities in different frequency bands.
[0055] In a gas turbine, taking the compressor as an example, the compressor is one of the key components of the gas turbine, and its rotating gears and blades will generate airflow noise and mechanical noise when rotating at high speed. Among them, airflow noise is one of the gas flow parameters involved in this application. Taking the turbine as an example, when the turbine blades and the turbine rotate, noise will be generated due to the gas flow and the interaction between the blades and the airflow. These noises also belong to the gas flow parameters involved in this application. Taking the exhaust system as an example, the components in the exhaust system of the gas turbine, such as the exhaust pipe, muffler, bell mouth, etc., may generate noise due to gas exhaust and flow noise. These noises also belong to the gas flow parameters involved in this application. It should be noted that the specific noise level and spectral characteristics also depend on factors such as the type, operating status, design and installation of the gas turbine, which can be excluded by pre-sampling.
[0056] It can be understood that multiplicative noise is a type of noise. Compared with additive noise (that is, noise added directly to the signal), multiplicative noise is noise that is doped into the signal through the multiplication process, so multiplicative noise can change the amplitude of the signal without changing its frequency and phase. Multiplicative noise plays an important role in many signal processing applications, such as modulation and demodulation in the field of noise processing. For these signal processing applications, understanding the characteristics and generation mechanism of multiplicative noise is crucial to effectively reduce noise and optimize signal processing methods.
[0057] Based on the above principles, in the present application, the multiplicative noise level can be evaluated from the operating parameters of the gas turbine, and the fouling degree of the gas turbine can be estimated based on the multiplicative noise level, so as to match the gas turbine with a water washing operation that matches the estimated fouling degree, making the water washing operation of the gas turbine more accurate, avoiding equipment performance loss caused by fouling, and reducing resource waste and damage to equipment caused by excessive water washing, thereby further ensuring the safety and reliability of the gas turbine.
[0058] Step S102, performing wavelet transformation on the gas flow parameters of the gas turbine to obtain wavelet coefficient characteristics of the gas flow parameters in each frequency band.
[0059] Among them, wavelet transform is a mathematical method that transforms and decomposes signals at different scales to extract the frequency, time and amplitude characteristics of signals. Wavelet transform can perform multi-scale analysis on signals through a series of basis functions (wavelets). In this application, the gas flow parameters are decomposed and reconstructed at different scales by wavelet transform, so as to extract the local characteristics of the gas flow parameters and realize the signal frequency analysis and time domain feature extraction of the gas flow parameters.
[0060] Specifically, in an optional implementation of step S102, the gas flow parameters are decomposed based on the preset number of decomposition layers by using the preset wavelet transform basis functions to obtain the wavelet coefficients of the gas flow parameters in each frequency band. Then, the wavelet coefficients are feature extracted to obtain the wavelet coefficient features of the gas flow parameters in each frequency band. The wavelet coefficient features include at least one of the following: aerodynamic noise energy, aerodynamic noise variance, aerodynamic noise mean, aerodynamic noise peak, aerodynamic noise skewness, and aerodynamic noise wavelet packet energy ratio.
[0061] For example, the energy of aerodynamic noise can be obtained by calculating the sum of the squares of the wavelet coefficients in each frequency band to represent the energy of the frequency band. The variance of aerodynamic noise can be obtained by calculating the variance of the wavelet coefficients in each frequency band to measure the degree of discreteness of the wavelet coefficients. The mean of aerodynamic noise can be calculated by calculating the mean of the wavelet coefficients in each frequency band to reflect the concentration trend of the wavelet coefficients. The peak value of aerodynamic noise and the skewness of aerodynamic noise are used to evaluate the non-Gaussian characteristics of gas flow parameters by measuring the shape characteristics of the distribution of wavelet coefficients. The energy ratio of aerodynamic noise wavelet packets is obtained by calculating the ratio of the energy of each frequency band to the energy of the entire signal in the wavelet packet decomposition to analyze the contribution of each frequency band. Of course, there are many other ways to obtain the above-mentioned wavelet coefficient characteristics, which are not limited in this application.
[0062] In the embodiment of the present application, the wavelet transform basis function is determined by the attribute type of the gas turbine. Further optionally, the matching wavelet transform basis function is determined according to the signal characteristics of the type of gas turbine. In practical applications, different types of wavelet transform basis functions, such as Daubechies wavelet, Haar wavelet, Morlet wavelet, etc., can be selected according to specific application requirements for feature extraction and signal analysis of different types of gas flow parameters.
[0063] In an optional example, specifically, the expression of continuous wavelet transform (CWT) is as follows:
[0064]
[0065] Among them, x(t) is the input signal, x(n) is the discrete signal, ψ is the wavelet function, j and k are the scale and translation parameters used to adjust the frequency and position of the wavelet.
[0066] Optionally, after step S102, the wavelet coefficient features to be analyzed may be further selected. Specifically, the most representative feature subset may be selected from the wavelet coefficient features using a feature selection method or a dimensionality reduction technique to reduce feature dimensions and redundant information.
[0067] In an optional embodiment, the wavelet coefficient features of the gas flow parameters in each frequency band are subjected to feature selection by principal component analysis (PCA) to obtain the wavelet coefficient feature subset most relevant to the degree of scaling. Specifically, during feature selection, the wavelet coefficient features of the gas flow parameters in each frequency band are constructed as a matrix to be selected. Next, the mean of each wavelet coefficient feature in the matrix to be selected is calculated, and the corresponding mean is subtracted from each wavelet coefficient feature to obtain the mean processing result of each wavelet coefficient feature. Then, the covariance matrix of the matrix to be selected after mean processing is calculated. Then, the covariance matrix is subjected to eigenvalue decomposition to obtain the eigenvalues to be selected and the corresponding eigenvectors to be selected; the eigenvectors to be selected whose eigenvalues to be selected meet the eigenvalue threshold are selected as reference eigenvector matrices, and the reference eigenvector matrix is subjected to projection changes to obtain the target wavelet coefficient feature matrix of the gas flow parameters. Therefore, the most representative feature subset is selected from the wavelet coefficient features to reduce feature dimensions and redundant information, further improve the model calculation efficiency, reduce the risk of overfitting, and improve the model generalization ability.
[0068] In addition to the feature selection methods in the above examples, other methods can also be used to select features from wavelet coefficient features. For example, filter method, wrapper method, embedded method, linear discriminant analysis (LDA) method. No matter which feature selection method is chosen, important features should be retained during feature selection, and the model performance and computational efficiency in the subsequent processing process should be taken into account.
[0069] Step S103: input the wavelet coefficient features in each frequency band into the band transition probability matrix in the band state prediction model to perform state prediction, and obtain the band state sequence corresponding to the wavelet coefficient features in each frequency band.
[0070] Among them, the band transition probability matrix in the band state prediction model is used to describe the transition probability of the wavelet coefficient feature between states in different bands. The rows of the matrix represent the current state, the columns represent the next state, and each element represents the probability of transitioning from the current band to the next state. The elements in the band transition probability matrix can be obtained by statistically analyzing the frequency of the corresponding band transitions in the training set, or by estimating them by using the algorithm of the band state prediction model. It can be understood that the band transition probability matrix is one of the key components of the band state prediction model, and the band transition probability matrix reflects the transition law of the wavelet coefficient features between different states. Through the establishment of the training stage, the band transition probability matrix will be used as part of the establishment of the multiplicative noise model for state prediction and multiplicative noise modeling of the new wavelet coefficient features in the next prediction stage.
[0071] Specifically, an optional embodiment of step S103 is shown in Figure 2 As shown, it can be implemented as follows:
[0072] Step S201 : normalizing the wavelet coefficient features in each frequency band so that the wavelet coefficient features in each frequency band are within a feature value range consistent with the training sample.
[0073] In this way, the ranges of different eigenvalues can be mapped to the same scale range through normalization processing, avoiding the deviation of model training due to the difference in scales of different eigenvalues.
[0074] Step S202, input the normalized wavelet coefficient features in each frequency band into the band state prediction model, and perform state prediction on the wavelet coefficient features in each frequency band through the band transition probability matrix pre-constructed in the band state prediction model to obtain the next frequency band state with the highest probability as the predicted frequency band state corresponding to the wavelet coefficient features.
[0075] Step S203: compose a frequency band state sequence with the predicted frequency band states corresponding to the wavelet coefficient features in each frequency band.
[0076] Thus, through the above steps S201 to S203, the frequency band state prediction model is used to identify and analyze the gas flow characteristics inside the gas turbine in different frequency ranges based on the wavelet coefficient characteristics, so as to predict the most likely gas flow state inside the gas turbine.
[0077] In practical applications, in some optional embodiments, the frequency band state prediction model can be implemented as a Markov Switching Model (MSM). The Markov Switching Model is a special Markov model used to model the process of a system switching between different states. In the analysis of aerodynamic noise, the Markov Switching Model can be used to describe the mode switching of aerodynamic noise under different working conditions or operating states, thereby providing more comprehensive modeling and analysis capabilities.
[0078] Alternatively, in some other embodiments, the band state prediction model can be implemented as a hidden Markov model (HMM). The hidden Markov model is a statistical model used to model sequence data with potential hidden states. In the analysis of aerodynamic noise, the state sequence of aerodynamic noise can be regarded as a hidden state sequence, and the observed noise feature sequence can be regarded as a visible state sequence. The state of multiplicative noise in the gas flow parameters can be modeled and predicted by the hidden Markov model.
[0079] In the embodiment of the present application, combined with the above-mentioned model principle, the band transition probability matrix in the band state prediction model can also be constructed in the following manner, namely: clustering the wavelet coefficient feature samples under each frequency band to divide each wavelet coefficient feature sample into different frequency band states; counting the transition probability of each wavelet coefficient feature sample between different frequency band states, and calculating the band transition probability of each wavelet coefficient feature sample based on the transition probability; constructing the band transition probability of each wavelet coefficient feature sample into a band transition probability matrix.
[0080] Based on the model construction principle, in step S202 of the embodiment of the present application, the normalized wavelet coefficient features under each frequency band are input into the frequency band state prediction model, and the wavelet coefficient features under each frequency band are predicted by the frequency band transition probability matrix pre-constructed in the frequency band state prediction model to obtain the next frequency band state with the highest probability as the predicted frequency band state corresponding to the wavelet coefficient feature, see Figure 3 , which can be implemented as:
[0081] Step S301, matching the normalized wavelet coefficient features in each frequency band to the corresponding initial state;
[0082] Step S302, calculating the probability of the next frequency band state of the normalized wavelet coefficient feature in each frequency band through the frequency band transition probability matrix.
[0083] Among them, the next band state S j Probability It is expressed as the following formula:
[0084]
[0085] Where P(S) is the probability of occurrence of band state S in the band transition probability matrix, S j The frequency band transition probability matrix corresponds to the frequency band state S at time t+1. t+1 , S i The frequency band transition probability matrix corresponds to the frequency band state S at time t. t , Indicates that the initial state is S i The probability of Indicates that from the initial state S i Transition to next band state S j The probability of Indicates the state of the frequency band S i Transition to next band state S i2 The probability of Indicates that the state of the frequency band is s i2 Transfer to band state s i3 The probability of Indicates that the state of the frequency band is st-1 Transition to band state S t The probability of , t is a positive integer.
[0086] Step S104: inputting the wavelet coefficient features corresponding to each predicted frequency band state into the noise intensity extraction model based on the frequency band state sequence to obtain a multiplicative noise intensity parameter.
[0087] In step S104, the correlation between aerodynamic noise (such as multiplicative noise in gas flow parameters) and the fouling degree of the gas turbine is analyzed in combination with the previously obtained frequency band state sequence.
[0088] In an optional embodiment of step S104, based on the frequency band state sequence, the wavelet coefficient features under the same prediction frequency band state are divided into the same group to obtain multiple groups of wavelet coefficient features. Then, each group of wavelet coefficient features under each prediction frequency band state is input into the noise intensity extraction model, and the probability that each group of wavelet coefficient features belongs to multiplicative noise is calculated as the multiplicative noise intensity probability value of each group of wavelet coefficient features.
[0089] In the embodiment of the present application, the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine. The multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient feature corresponding to each predicted frequency band state.
[0090] It is worth noting that the sources of the correlation between aerodynamic noise and the degree of fouling of gas turbines include the following aspects: On the one hand, the fouling of gas turbines will cause changes in the geometry of the internal flow channel and increase the disturbance of gas flow. These changes may cause turbulence and non-uniform flow of the fluid, leading to the generation of aerodynamic noise and the increase of aerodynamic noise intensity. On the other hand, the fouling inside the gas turbine may change the vibration characteristics and sound transmission path of the gas turbine components, thereby affecting the conduction and radiation of sound waves inside the gas turbine, resulting in the propagation and amplification of aerodynamic noise. On the other hand, the fouling of the gas turbine forms roughness on the surface of the gas turbine blades, rotors and stator rings, increasing the number and intensity of noise sources, which essentially increases the generation of aerodynamic noise. In addition, from the perspective of changes in operating conditions, the degree of fouling of the gas turbine will affect its operating conditions and efficiency during operation, which may cause changes in the aerodynamic noise characteristics of the gas turbine at different loads and speeds.
[0091] In the embodiment of the present application, multiplicative noise modeling, i.e., the modeling process of the noise intensity extraction model based on multiplicative noise, is to associate each state with the corresponding multiplicative noise model through the state sequence obtained by prediction. A specific multiplicative noise modeling method can be to use GMM to perform density modeling on the wavelet coefficient characteristics under each state. Through GMM, the probability density function of the wavelet coefficients under each state can be estimated to obtain a multiplicative noise model. In multiplicative noise modeling, the partial data corresponding to each state obtained in the training phase can be used to train the Gaussian mixture model to obtain model parameters. Then, by using the trained Gaussian mixture model, the wavelet coefficients under each state are probability calculated in the prediction phase to determine whether it belongs to multiplicative noise.
[0092] Based on the above principle, the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters. Further optionally, the noise intensity extraction model can be constructed based on a Gaussian mixture model (GMM). GMM is a commonly used probabilistic model that can be used to analyze the correlation between aerodynamic noise and the fouling degree of a gas turbine.
[0093] In practical applications, for example, a Gaussian mixture model can be used to model the nonlinear relationship between the fouling degree and aerodynamic noise of a gas turbine. The model is based on a combination of multiple Gaussian distributions, and the parameters of each Gaussian distribution are estimated by fitting data, and these distributions are used to predict the degree of correlation between aerodynamic noise and the degree of fouling. Alternatively, in another example, a Gaussian mixture clustering model can be used to divide the wavelet coefficient feature set into multiple clusters and assign a Gaussian distribution to each cluster. By using the wavelet coefficient feature data as input, a Gaussian mixture clustering model can be used for cluster analysis to explore the distribution of aerodynamic noise under different fouling degrees and the correlation between the two.
[0094] Based on the above model building principle, further optionally, the noise intensity extraction model is expressed as the following formula:
[0095] P(S|j)=Σπ j *N(S|μ j ,Σ j )
[0096] Wherein, P(S|j) represents the probability of observing the frequency band state S corresponding to the wavelet coefficient feature from the gas flow parameter in the case of multiplicative noise, ∑ represents the weighted sum of the Gaussian distributions of the frequency band states corresponding to all wavelet coefficient features, each Gaussian distribution includes a mean vector μ and a covariance matrix Σ, and the weight of each Gaussian distribution is π, π j represents the weight of the j-th Gaussian distribution, μ j represents the mean vector of the jth Gaussian distribution, Σj represents the covariance matrix of the jth Gaussian distribution, N(S|μ j ,Σ j ) represents the probability density function of the j-th Gaussian distribution.
[0097] Step S105, performing a water washing operation matching the multiplicative noise intensity parameter on the gas turbine to reduce the fouling degree of the gas turbine.
[0098] Therefore, based on the correlation between aerodynamic noise (such as multiplicative noise in gas flow parameters) and the fouling degree of the gas turbine, a water washing operation matching the multiplicative noise intensity parameters is performed for the gas turbine, thereby alleviating (or removing) the corresponding fouling phenomenon.
[0099] Continuing with the above example, in step S105, if the multiplicative noise intensity probability value of the current wavelet coefficient feature is greater than the multiplicative noise probability threshold, it is determined that the gas flow parameter corresponding to the current wavelet coefficient feature contains multiplicative noise, and a water washing operation matching the current multiplicative noise intensity probability value is started. The multiplicative noise intensity probability value is positively correlated with the water pressure in the water washing operation.
[0100] In the implementation mode of the present application, through the above steps S101 to S105, the fouling degree of the gas turbine is evaluated in combination with the multiplicative noise level in the gas flow parameters, so that the corresponding water washing operation is performed on the gas turbine to reduce the fouling degree of the gas turbine, assist the gas turbine to maintain a normal operating state, and improve the performance of the gas turbine.
[0101] After introducing the method of the exemplary embodiment of the present application, next, refer to Figure 4 A water washing prediction device for a gas turbine according to an exemplary embodiment of the present application is described. The water washing prediction device for a gas turbine can be arranged in a gas turbine or a corresponding cleaning device. The device includes:
[0102] The acquisition module 401 is used to acquire the operating parameters of the gas turbine; the operating parameters at least include: gas flow parameters;
[0103] A transformation module 402 is used to perform wavelet transformation on the gas flow parameters of the gas turbine to obtain wavelet coefficient characteristics of the gas flow parameters in each frequency band;
[0104] Prediction module 403, used for inputting the wavelet coefficient characteristics under each frequency band into the frequency band transition probability matrix in the frequency band state prediction model for state prediction, and obtaining the frequency band state sequence corresponding to the wavelet coefficient characteristics under each frequency band; the frequency band transition probability matrix is used for describing the transition probability between states of the wavelet coefficient characteristics under different frequency bands;
[0105] The noise extraction module 404 is used to input the wavelet coefficient features corresponding to each predicted frequency band state into the noise intensity extraction model based on the frequency band state sequence to obtain a multiplicative noise intensity parameter; wherein the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient features corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters;
[0106] The execution module 405 is used to perform a water washing operation on the gas turbine that matches the multiplicative noise intensity parameter to reduce the fouling degree of the gas turbine.
[0107] As an optional implementation manner, the transformation module 402 performs wavelet transformation on the gas flow parameters of the gas turbine to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band, which is specifically used for:
[0108] By means of a preset wavelet transform basis function, the gas flow parameter is subjected to signal decomposition based on a preset number of decomposition layers to obtain the wavelet coefficients of the gas flow parameter in each frequency band; the wavelet transform basis function is determined by the attribute type of the gas turbine; feature extraction is performed on the wavelet coefficient to obtain the wavelet coefficient characteristics of the gas flow parameter in each frequency band; wherein the wavelet coefficient characteristics include at least one of the following: aerodynamic noise energy, aerodynamic noise variance, aerodynamic noise mean, aerodynamic noise peak, aerodynamic noise skewness, and aerodynamic noise wavelet packet energy ratio.
[0109] As an optional implementation, it also includes a feature selection module, which is specifically used to:
[0110] After the transformation module 402 performs feature extraction on the wavelet coefficients to obtain the wavelet coefficient features of the gas flow parameters in each frequency band, the wavelet coefficient features of the gas flow parameters in each frequency band are constructed as a matrix to be selected; the mean of each wavelet coefficient feature in the matrix to be selected is calculated, and the corresponding mean is subtracted from each wavelet coefficient feature to obtain the mean processing result of each wavelet coefficient feature; the covariance matrix of the matrix to be selected after the mean processing is calculated; the covariance matrix is decomposed by eigenvalues to obtain candidate eigenvalues and corresponding candidate eigenvectors; the candidate eigenvectors whose candidate eigenvalues meet the eigenvalue threshold are selected as reference eigenvector matrices, and the reference eigenvector matrix is projected and transformed to obtain the target wavelet coefficient feature matrix of the gas flow parameters.
[0111] As an optional implementation, the prediction module 403 inputs the wavelet coefficient characteristics under each frequency band into the frequency band transition probability matrix in the frequency band state prediction model to perform state prediction, and obtains the frequency band state sequence corresponding to the wavelet coefficient characteristics under each frequency band, which is specifically used for:
[0112] Normalize the wavelet coefficient features in each frequency band so that the wavelet coefficient features in each frequency band are within the characteristic value range consistent with the training samples;
[0113] The normalized wavelet coefficient features under each frequency band are input into the frequency band state prediction model, and the state of the wavelet coefficient features under each frequency band is predicted by the frequency band transition probability matrix pre-constructed in the frequency band state prediction model, so as to obtain the next frequency band state with the highest probability as the predicted frequency band state corresponding to the wavelet coefficient features;
[0114] The frequency band state sequence is composed of the predicted frequency band states corresponding to the wavelet coefficient features in each frequency band.
[0115] As an optional implementation, it also includes a building module, specifically used for:
[0116] The wavelet coefficient characteristic samples under each frequency band are clustered to divide each wavelet coefficient characteristic sample into different frequency band states; the transition probability of each wavelet coefficient characteristic sample between different frequency band states is counted, and the frequency band transition probability of each wavelet coefficient characteristic sample is calculated according to the transition probability; the frequency band transition probability of each wavelet coefficient characteristic sample is constructed as the frequency band transition probability matrix.
[0117] As an optional implementation, the prediction module 403 inputs the normalized wavelet coefficient features under each frequency band into the frequency band state prediction model, and performs state prediction on the wavelet coefficient features under each frequency band through the frequency band transition probability matrix pre-constructed in the frequency band state prediction model, so as to obtain the next frequency band state with the highest probability as the predicted frequency band state corresponding to the wavelet coefficient feature, which is specifically used for:
[0118] The initial state corresponding to the normalized wavelet coefficient feature matching in each frequency band;
[0119] The probability of the next band state of the normalized wavelet coefficient feature in each band is calculated through the band transition probability matrix; wherein the probability of the next band state Sj is It is expressed as the following formula:
[0120]
[0121]
[0122] Where P(S) is the probability of occurrence of band state S in the band transition probability matrix, and Sj represents the band state S corresponding to the band transition probability matrix at time t+1. t+1 , S i The frequency band transition probability matrix corresponds to the frequency band state S at time t. t , Indicates that the initial state is S i The probability of Indicates that from the initial state S i Transition to next band state S j The probability of Indicates the state of the frequency band S i Transition to next band state S i2 The probability of Indicates that the state of the frequency band is s i2 Transfer to band state s i3 The probability of Indicates that the state of the frequency band is s t-1 Transition to band state S t probability.
[0123] As an optional implementation, the noise extraction module 404 inputs the wavelet coefficient features corresponding to each predicted frequency band state into the noise intensity extraction model based on the frequency band state sequence, and obtains the multiplicative noise intensity parameter, which is specifically used for:
[0124] Based on the frequency band state sequence, the wavelet coefficient features under the same predicted frequency band state are divided into the same group to obtain multiple groups of wavelet coefficient features;
[0125] Each group of wavelet coefficient features under each prediction frequency band state is input into the noise intensity extraction model, and the probability that each group of wavelet coefficient features belongs to multiplicative noise is calculated as the multiplicative noise intensity probability value of each group of wavelet coefficient features;
[0126] When the water washing operation matching the multiplicative noise intensity parameter is performed on the gas turbine to reduce the fouling degree of the gas turbine, it is specifically used for:
[0127] If the multiplicative noise intensity probability value of the current wavelet coefficient feature is greater than the multiplicative noise probability threshold, it is determined that the gas flow parameter corresponding to the current wavelet coefficient feature contains multiplicative noise; and
[0128] A water washing operation matching the current multiplicative noise intensity probability value is started; wherein the multiplicative noise intensity probability value is positively correlated with the water pressure in the water washing operation.
[0129] As an optional implementation, the noise intensity extraction model is expressed as the following formula:
[0130] P(S|j)=Σπ j *N(S|μ j ,Σ j )
[0131] Wherein, P(S|j) represents the probability of observing the frequency band state S corresponding to the wavelet coefficient feature from the gas flow parameter in the case of multiplicative noise, ∑ represents the weighted sum of the Gaussian distributions of the frequency band states corresponding to all wavelet coefficient features, each Gaussian distribution includes a mean vector μ and a covariance matrix Σ, and the weight of each Gaussian distribution is π, π j represents the weight of the j-th Gaussian distribution, μ j represents the mean vector of the jth Gaussian distribution, Σ j represents the covariance matrix of the jth Gaussian distribution, N(S|μ j ,Σ j ) represents the probability density function of the j-th Gaussian distribution.
[0132] In the embodiment of the present application, the fouling degree of the gas turbine is evaluated by a water washing prediction device of the gas turbine in combination with the multiplicative noise level in the gas flow parameters, so that the corresponding water washing operation is performed on the gas turbine to reduce the fouling degree of the gas turbine, assist the gas turbine to maintain a normal operating state, and improve the performance of the gas turbine.
[0133] After introducing the devices, methods and apparatus of the exemplary embodiments of the present application, next, reference is made to Figure 5 The computer-readable storage medium of the exemplary embodiment of the present application is described. The computer-readable storage medium can be set in the water washing prediction device of the gas turbine. Please refer to Figure 5, the computer-readable storage medium shown is a CD 110, on which a computer program (i.e., a program product) is stored. When the computer program is executed by the processor, each step recorded in the above method implementation is implemented, for example, obtaining the operating parameters of the gas turbine; the operating parameters at least include: gas flow parameters; performing wavelet transform on the gas flow parameters of the gas turbine to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band; inputting the wavelet coefficient characteristics in each frequency band into the frequency band transition probability matrix in the frequency band state prediction model to perform state prediction, and obtaining the frequency band state sequence corresponding to the wavelet coefficient characteristics in each frequency band; the frequency band transition probability matrix is used to describe The transition probability of the wavelet coefficient feature between states in different frequency bands; the wavelet coefficient features corresponding to each predicted frequency band state are respectively input into the noise intensity extraction model based on the frequency band state sequence to obtain the multiplicative noise intensity parameter; wherein the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient feature corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters; a water washing operation matching the multiplicative noise intensity parameter is performed on the gas turbine to reduce the fouling degree of the gas turbine; the specific implementation method of each step will not be repeated here.
[0134] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.
[0135] After introducing the devices, methods, media and apparatus of the exemplary embodiments of the present application, reference is now made to Figure 6 The computing device for water wash prediction of a gas turbine according to the exemplary embodiment of the present application may be arranged in a water wash prediction device of the gas turbine.
[0136] Figure 6 A block diagram of an exemplary computing device 120 suitable for implementing embodiments of the present application is shown. The computing device 120 may be a computer system or a server. Figure 6 The computing device 120 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0137] like Figure 6As shown, the components of computing device 120 may include, but are not limited to: one or more processors or processing units 1201, a system memory 1202, and a bus 1203 connecting different system components (including system memory 1202 and processing unit 1201).
[0138] The computing device 120 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computing device 120, including volatile and nonvolatile media, removable and non-removable media.
[0139] System memory 1202 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 12021 and / or cache memory 12022. Computing device 120 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, ROM 12023 may be used to read and write non-removable, non-volatile magnetic media ( Figure 6 is not shown in the Figure 6 As shown in FIG. 1 , a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) can be provided. In these cases, each drive can be connected to bus 1203 via one or more data medium interfaces. System memory 1202 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of each embodiment of the present application.
[0140] A program / utility 12025 having a set (at least one) of program modules 12024 may be stored, for example, in system memory 1202, and such program modules 12024 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. Program modules 12024 generally perform the functions and / or methods of the embodiments described herein.
[0141] The computing device 120 may also communicate with one or more external devices 1204 (e.g., a keyboard, a pointing device, a display, etc.). Such communication may be performed via an input / output (I / O) interface 605. In addition, the computing device 120 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 1206. Figure 6As shown, the network adapter 1206 communicates with other modules (such as the processing unit 1201, etc.) of the computing device 120 via the bus 1203. It should be understood that although Figure 6 Not shown, other hardware and / or software modules may be used in conjunction with computing device 120 .
[0142] The processing unit 1201 executes various functional applications and data processing by running the program stored in the system memory 1202, for example, obtaining the operating parameters of the gas turbine; the operating parameters at least include: gas flow parameters; performing wavelet transform on the gas flow parameters of the gas turbine to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band; inputting the wavelet coefficient characteristics in each frequency band into the frequency band transition probability matrix in the frequency band state prediction model to perform state prediction, and obtaining the frequency band state sequence corresponding to the wavelet coefficient characteristics in each frequency band; the frequency band transition probability matrix is used to describe the wavelet coefficient characteristics. The transition probability between states under different frequency bands is calculated; the wavelet coefficient features corresponding to each predicted frequency band state are input into the noise intensity extraction model based on the frequency band state sequence to obtain the multiplicative noise intensity parameter; wherein the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient features corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters; a water washing operation matching the multiplicative noise intensity parameter is performed on the gas turbine to reduce the fouling degree of the gas turbine. The specific implementation method of each step is not repeated here. It should be noted that although several units / modules or sub-units / sub-modules of the water washing prediction device for the gas turbine are mentioned in the above detailed description, this division is only exemplary and not mandatory. In fact, according to the embodiment of the present application, the features and functions of two or more units / modules described above can be concretized in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be concretized.
[0143] In the description of the present application, it should be noted that the terms "first", "second" and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0144] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0145] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.
[0146] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0147] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0148] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0149] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
[0150] In addition, although the operations of the method of the present application are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.
Claims
1. A method for predicting water washing of a gas turbine, characterized in that: The method comprises: Acquiring operating parameters of the gas turbine; the operating parameters at least include: gas flow parameters; Performing wavelet transformation on the gas flow parameters of the gas turbine to obtain wavelet coefficient characteristics of the gas flow parameters in each frequency band; Inputting the wavelet coefficient characteristics under each frequency band into the frequency band transition probability matrix in the frequency band state prediction model for state prediction, and obtaining the frequency band state sequence corresponding to the wavelet coefficient characteristics under each frequency band; the frequency band transition probability matrix is used to describe the transition probability between states of the wavelet coefficient characteristics under different frequency bands; The wavelet coefficient features corresponding to each predicted frequency band state are input into the noise intensity extraction model based on the frequency band state sequence to obtain a multiplicative noise intensity parameter; wherein the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes the multiplicative noise intensity probability value of the wavelet coefficient features corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters; A water washing operation matching the multiplicative noise intensity parameter is performed on the gas turbine to reduce the fouling degree of the gas turbine.
2. The method according to claim 1, characterized in that: The wavelet transform is performed on the gas flow parameters of the gas turbine to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band, including: By using a preset wavelet transform basis function, the gas flow parameter is subjected to signal decomposition based on a preset number of decomposition layers to obtain wavelet coefficients of the gas flow parameter in each frequency band; the wavelet transform basis function is determined by the attribute type of the gas turbine; Extracting features of the wavelet coefficients to obtain wavelet coefficient features of the gas flow parameters in each frequency band; The wavelet coefficient feature includes at least one of the following: aerodynamic noise energy, aerodynamic noise variance, aerodynamic noise mean, aerodynamic noise peak, aerodynamic noise skewness, and aerodynamic noise wavelet packet energy ratio.
3. The method according to claim 2, characterized in that After extracting the characteristics of the wavelet coefficients to obtain the wavelet coefficient characteristics of the gas flow parameters in each frequency band, the method further includes: constructing the wavelet coefficient characteristics of the gas flow parameters in each frequency band as a matrix to be selected; Calculating the mean of each wavelet coefficient feature in the matrix to be selected, and subtracting the corresponding mean from each wavelet coefficient feature to obtain a mean processing result of each wavelet coefficient feature; Calculating the covariance matrix of the matrix to be selected after mean processing; Performing eigenvalue decomposition on the covariance matrix to obtain candidate eigenvalues and corresponding candidate eigenvectors; The candidate eigenvectors whose candidate eigenvalues meet the eigenvalue threshold are selected as reference eigenvector matrices, and the reference eigenvector matrices are projected and transformed to obtain the target wavelet coefficient characteristic matrix of the gas flow parameters.
4. The method according to claim 1, characterized in that: The step of inputting the wavelet coefficient features under each frequency band into the frequency band transition probability matrix in the frequency band state prediction model to perform state prediction, and obtaining the frequency band state sequence corresponding to the wavelet coefficient features under each frequency band, comprises: Normalize the wavelet coefficient features in each frequency band so that the wavelet coefficient features in each frequency band are within the characteristic value range consistent with the training samples; The normalized wavelet coefficient features under each frequency band are input into the frequency band state prediction model, and the state of the wavelet coefficient features under each frequency band is predicted by the frequency band transition probability matrix pre-constructed in the frequency band state prediction model, so as to obtain the next frequency band state with the highest probability as the predicted frequency band state corresponding to the wavelet coefficient features; The frequency band state sequence is composed of the predicted frequency band states corresponding to the wavelet coefficient features in each frequency band.
5. The method according to claim 4, characterized in that The step of constructing a frequency band transition probability matrix in the frequency band state prediction model also includes: Clustering is performed on the wavelet coefficient feature samples in each frequency band to divide each wavelet coefficient feature sample into different frequency band states; Counting the transition probabilities of each wavelet coefficient characteristic sample between different frequency band states, and calculating the frequency band transition probability of each wavelet coefficient characteristic sample according to the transition probability; The frequency band transition probability of each wavelet coefficient feature sample is constructed as the frequency band transition probability matrix.
6. The method according to claim 5, characterized in that The normalized wavelet coefficient features in each frequency band are input into the frequency band state prediction model, and the wavelet coefficient features in each frequency band are predicted by using the frequency band transition probability matrix pre-constructed in the frequency band state prediction model to obtain the next frequency band state with the highest probability as the predicted frequency band state corresponding to the wavelet coefficient features, including: The initial state s_i corresponding to the normalized wavelet coefficient feature matching in each frequency band; The next band state S of the normalized wavelet coefficient characteristics in each band is calculated by using the band transition probability matrix. j The probability of the next band state S j Probability It is expressed as the following formula: Where P(S) is the probability of occurrence of band state S in the band transition probability matrix, and Sj represents the band state S corresponding to the band transition probability matrix at time t+1. t+1 , S i The frequency band transition probability matrix corresponds to the frequency band state S at time t. t , Indicates that the initial state is S i The probability of Represents that from the initial state S i Transition to next band state S j The probability of Indicates the state of the band S i Transition to next band state S i2 The probability of Indicates that the state of the frequency band is s i2 Transfer to band state s i3 The probability of Indicates that the state of the frequency band is s t-1 Transition to band state S t probability.
7. The method according to claim 1, characterized in that The step of inputting the wavelet coefficient features corresponding to each predicted frequency band state into the noise intensity extraction model based on the frequency band state sequence to obtain the multiplicative noise intensity parameter comprises: Based on the frequency band state sequence, the wavelet coefficient features under the same predicted frequency band state are divided into the same group to obtain multiple groups of wavelet coefficient features; Each group of wavelet coefficient features under each prediction frequency band state is input into the noise intensity extraction model, and the probability that each group of wavelet coefficient features belongs to multiplicative noise is calculated as the multiplicative noise intensity probability value of each group of wavelet coefficient features; The performing of a water washing operation on the gas turbine that matches the multiplicative noise intensity parameter to reduce the fouling degree of the gas turbine comprises: If the multiplicative noise intensity probability value of the current wavelet coefficient feature is greater than the multiplicative noise probability threshold, it is determined that the gas flow parameter corresponding to the current wavelet coefficient feature contains multiplicative noise; and A water washing operation matching the current multiplicative noise intensity probability value is started; wherein the multiplicative noise intensity probability value is positively correlated with the water pressure in the water washing operation.
8. The method according to claim 7, characterized in that The noise intensity extraction model is expressed as the following formula: P(S|j)=Sp j *N(S|μ j ,∑ j ) Wherein, P(S|j) represents the probability of observing the frequency band state S corresponding to the wavelet coefficient feature from the gas flow parameter in the case of multiplicative noise, ∑ represents the weighted sum of the Gaussian distributions of the frequency band states corresponding to all wavelet coefficient features, each Gaussian distribution includes a mean vector μ and a covariance matrix Σ, and the weight of each Gaussian distribution is π, π j represents the weight of the jth Gaussian distribution, μ j represents the mean vector of the jth Gaussian distribution, Σ j represents the covariance matrix of the jth Gaussian distribution, N(S|μ j ,Σ j ) represents the probability density function of the j-th Gaussian distribution.
9. A gas turbine water wash prediction device, characterized in that: The device comprises: An acquisition module, used for acquiring operating parameters of the gas turbine; the operating parameters at least include: gas flow parameters; A transformation module, used for performing wavelet transformation on the gas flow parameters of the gas turbine to obtain wavelet coefficient characteristics of the gas flow parameters in each frequency band; A prediction module is used to input the wavelet coefficient characteristics under each frequency band into the frequency band transition probability matrix in the frequency band state prediction model to perform state prediction, and obtain the frequency band state sequence corresponding to the wavelet coefficient characteristics under each frequency band; the frequency band transition probability matrix is used to describe the transition probability between the states of the wavelet coefficient characteristics under different frequency bands; A noise extraction module, for inputting the wavelet coefficient features corresponding to each predicted frequency band state into a noise intensity extraction model based on the frequency band state sequence to obtain a multiplicative noise intensity parameter; wherein the multiplicative noise intensity parameter is positively correlated with the fouling degree of the gas turbine; the multiplicative noise intensity parameter includes a multiplicative noise intensity probability value of the wavelet coefficient features corresponding to each predicted frequency band state; the noise intensity extraction model is pre-constructed based on the multiplicative noise samples in the gas flow parameters; An execution module is used to perform a water washing operation on the gas turbine that matches the multiplicative noise intensity parameter to reduce the fouling degree of the gas turbine.
10. A computing device, characterized in that: The computing device comprises: at least one processor, memory, and input-output unit; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the water washing prediction method for a gas turbine according to any one of claims 1 to 8.
Citation Information
Patent Citations
Gas turbine washing early warning method
CN115931049A
Power station gas turbine scaling quantitative characterization detection method
CN116773211A
Zero-resistance noise reduction method and system for gas inlet system of gas turbine
CN118030281A
Procedure for determining the noise level of an internal combustion engine
EP1209458A1
Method for controlling a gas turbine
US20220220904A1
