Heavy Gas Turbine Compressor Fouling Detection Method Based on Liquid Neural Network

Abstract

The present invention discloses a fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network, which relates to the technical field of anomaly detection and includes the following steps: collecting operation data for correlation analysis; preprocessing the selected feature data; extracting the time features of the compressor operation parameters; calculating the weights of the interactions between parameters and outputting the spatial correlation; constructing a spatio-temporal fusion model and inputting normal operation data to train the model; outputting a parameter prediction sequence, calculating the multi-dimensional residual vector of the predicted value and the actual value, and by constructing a dynamic threshold function, when the model converges and the residual distribution deviates from the confidence interval of the normal operating condition, outputting the fouling anomaly detection result of the compressor. Through the dynamic topological structure of the liquid neural network and the spatio-temporal feature collaborative mining mechanism, the present invention realizes the quantitative evaluation of the fouling degree of the compressor, and improves the accuracy and robustness of the fouling anomaly detection of the compressor.

Description

Technical Field

[0001] The present invention relates to the technical field of abnormal detection of heavy-duty gas turbine compressor systems, and particularly to a fouling detection method for heavy-duty gas turbine compressors based on a liquid neural network. Background Technique

[0002] As a core power equipment in the fields of aviation and power generation, the performance and operating safety of heavy-duty gas turbines directly affect the system reliability. The performance degradation of gas turbines is mainly due to the performance decline of compressors and turbines, especially the problem of compressor fouling. Fouling is caused by the accumulation of impurities in the air on the surfaces of compressor blades and flow channels, resulting in changes in the air flow geometry and surface quality, and thus affecting the performance of the compressor and the entire engine. The maintenance strategies for compressor fouling mainly include online / offline cleaning. Optimizing the cleaning cycle is crucial for the economy and safety of gas turbine operation because it requires a trade-off between performance improvement and corresponding costs. Therefore, how to accurately predict the degradation trend of the compressor and adopt an efficient and economical water washing strategy has become a technical bottleneck that urgently needs to be broken through for the safe and economic operation of current gas turbines.

[0003] Compressors often have difficult mechanism analysis and strong dynamics due to reasons such as changes in operating conditions, multi-variable coupling, and external environmental interference, resulting in difficult accurate modeling. Summary of the Invention

[0004] The purpose of the present invention is to provide a fouling detection method for heavy-duty gas turbine compressors based on a liquid neural network, to solve the problems raised in the above background technique, effectively simulate the dynamics of compressor operation, use a graph attention mechanism to extract spatial features, and adaptively track the drift of the residual distribution during the compressor fouling process through a dynamic threshold algorithm, so as to improve the prediction accuracy of compressor fouling anomalies.

[0005] To achieve the above purpose, the present invention provides a fouling detection method for heavy-duty gas turbine compressors based on a liquid neural network, including the following steps:

[0006] Step S1: Due to problems such as multi-variables and strong coupling in the compressor, the Spearman correlation coefficient method is used to perform a correlation analysis on the compressor operation data, and variables with strong correlations are selected for compressor modeling;

[0007] Step S2: Preprocess the selected feature data and divide the data into a training set and a test set;

[0008] Step S3: Due to the difficult mechanism analysis and strong dynamics of the compressor, it is difficult to accurately model. Therefore, a liquid neural network with good modeling ability for complex dynamic data is used as a time model to extract the time features of the compressor operation parameters;

[0009] Step S4: Considering the coupling phenomenon of compressor operation parameters, calculate the weights of the interaction between compressor operation parameters, and use the graph attention mechanism to output spatial correlation;

[0010] Step S5: Construct a spatio-temporal fusion model for training, calculate the multi-dimensional residual vector, and construct a dynamic threshold function;

[0011] Step S6: Observe whether the model tends to converge and whether the residual distribution deviates from the confidence interval of the normal operating condition. If the model converges and the residual distribution deviates from the confidence interval of the normal operating condition, output the abnormal detection result of compressor fouling; otherwise, repeat steps S3 - S5.

[0012] In the present invention, a liquid neural network is used for modeling. By converting the hierarchical depth features of a static neural network and the temporal dynamic characteristics of a recurrent neural network (RNN) into a continuous vector field, parameter sharing, adaptive calculation of non-uniformly sampled data, and function approximation are achieved. Therefore, it has important theoretical value and practical significance to use the liquid neural network method to conduct abnormal detection research on the heavy gas turbine compressor system.

[0013] Preferably, in step S1, the Spearman correlation coefficient method is used to analyze the correlation of compressor operation data, and variables with strong correlation are selected for compressor modeling. The specific formula is:

[0014] ;

[0015] where is the total number of observation objects, is the rank difference of the i th observation object in two variables. If represents a perfect positive correlation, represents a perfect negative correlation, represents no monotonic correlation relationship.

[0016] Preferably, the preprocessing of the operation data collected in step S2 specifically includes collecting the operation data of different measurement points in the normal and abnormal states of the heavy gas turbine compressor system, and preprocessing the missing values and abnormal values of the data.

[0017] Preferably, in step S3, a liquid neural network is used as a time model to extract time features. Its core lies in using a dynamically changing time constant to control the state change rate of each neuron. This structure enables each neuron to adaptively change at different time points according to the input features. Its dynamic equation formula is as follows:

[0018] ;

[0019] Among them, t is the time step, is the hidden state of the liquid time constant layer of the neuron; is the liquid time constant parameter vector, which controls the rate of neuron state update; is a neural network parameterized by ; is the exogenous input of the neural network; A is a bias vector.

[0020] The time constant parameter not only determines the change rate of the hidden state, but also is closely related to the input features. Specifically, the liquid time constant can be expressed by the following formula:

[0021] ;

[0022] Among them, is the time-varying factor; is a neural network parameterized by ; is the exogenous input of the neural network.

[0023] The approximate closed-form solution of the liquid neural network can be expressed by the following formula:

[0024] ;

[0025] is shown as:

[0026] ;

[0027] Among them, B is the decomposed parameter vector; is the liquid time constant parameter vector; is a neural network parameterized by ; is the exogenous input of the neural network; A is a bias vector; is the Hadamard product.

[0028] Preferably, the specific steps of step S4 are as follows:

[0029] Step S41: Calculate the attention value between the compressor operation parameters;

[0030] Step S42: Normalize the adjacent attention information of each node;

[0031] Step S43: Adopt the multi-head attention mechanism for feature aggregation;

[0032] Step S44: Output spatial correlation through the graph attention mechanism.

[0033] Preferably, the formula for calculating the attention value in step S41 is as follows:

[0034] ;

[0035] where, a represents the attention network; respectively represent nodes i and j ; W represents the weight sharing matrix.

[0036] Preferably, the formula for normalization processing in step S42 is as follows:

[0037] ;

[0038] where, represents the attention coefficient between nodes i and j ; represents the attention value between nodes i and j ; represents the attention value between nodes i and k ; represents the neighbor nodes of node i .

[0039] Preferably, in step S43, a multi-head attention mechanism is used for aggregation, and the formula is as follows:

[0040] ;

[0041] where, represents the feature representation of node i ; represents concatenating the outputs of heads; represents any activation function; represents the attention coefficient of the th head; represents the weight matrix of the th head; represents the feature representation of node j .

[0042] Preferably, the graph attention formula in step S44 is as follows:

[0043] ;

[0044] Among them, represents the attention coefficient between nodes i and j ; represents the neighbor nodes of node i ; represents the activation function; represents a single-layer feedforward neural network, T represents the transpose; W represents the shared weight matrix; represents the feature representation of node i ; represents the feature representation of node j ; represents the feature representation of node k ;

[0045] Preferably, the specific steps of step S5 are as follows:

[0046] Step S51: Input the data into the spatio-temporal fusion model for training using the sliding window mechanism;

[0047] Step S52: Calculate the multi-dimensional residual vector and construct a dynamic threshold function based on the historical normal condition residual samples;

[0048] Step S53: Set the trigger condition for outputting the detection result of compressor fouling abnormality.

[0049] Preferably, the formula of the sliding window function in step S51 is as follows:

[0050] ;

[0051] Among them, represents the window length; represents the selected monitoring parameter dimension; represents t the observation vector at time

[0052] Input the current window data into the spatio-temporal fusion model , and output the prediction sequence. The formula is as follows:

[0053] ;

[0054] Among them, represents the prediction length.

[0055] Preferably, the formula for calculating the multi-dimensional residual vector in step S52 is as follows:

[0056] ;

[0057] Among them, Represents the covariance matrix of the normal data residuals in the training phase; Represents the Mahalanobis distance metric.

[0058] The formula for constructing the probability distribution model using kernel density estimation is as follows:

[0059] ;

[0060] Among them, the kernel function Selects the Gaussian kernel, and the formula is as follows:

[0061] ;

[0062] The bandwidth matrix H Is determined by the Silverman criterion, and the formula is as follows:

[0063] ;

[0064] Among them, Represents the standard deviation estimate of the i -dimensional residuals; Represents the standard deviation estimate of the j -dimensional residuals; Q Represents the number of residual samples.

[0065] Set the confidence level , and calculate the dynamic threshold, and the formula is as follows:

[0066] ;

[0067] Among them, Represents the inverse function of the cumulative distribution function.

[0068] Preferably, the trigger condition formula for outputting the abnormal detection result of the compressor fouling in step S53 is:

[0069] ;

[0070] Output the abnormal detection result of the compressor fouling and give the abnormal confidence level:

[0071] ;

[0072] Among them, Represents the sensitivity adjustment factor (default value 0.5).

[0073] Therefore, the present invention adopts the above-mentioned compressor fouling detection method based on the liquid neural network. Through the dynamic topological structure of the liquid neural network and the combination of the spatio-temporal feature collaborative mining mechanism, the quantitative evaluation of the compressor fouling degree is realized, and the accuracy and robustness of the abnormal detection of the compressor fouling are improved.

[0074] The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0075] Figure 1 It is a flowchart of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention;

[0076] Figure 2 It is a Spearman correlation coefficient heat map of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention;

[0077] Figure 3 It is a schematic diagram of a liquid neural network of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention;

[0078] Figure 4 It is a schematic diagram of a graph attention network of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention;

[0079] Figure 5 It is a prediction curve graph of the output parameters of the compressor by the liquid neural network of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention, where (a) is the prediction curve graph of the compressor exhaust pressure; (b) is the prediction curve graph of the compressor exhaust temperature;

[0080] Figure 6 It is a histogram of residual distribution and a kernel density curve graph of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention;

[0081] Figure 7 It is a comparison graph of the residual value distributions with other detection methods of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention;

[0082] Figure 8 It is a schematic structural diagram of a heavy-duty gas turbine of an embodiment of the fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to the present invention. Detailed Embodiments

[0083] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0084] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the art to which the present invention pertains. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "comprising" or "including" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0085] Embodiment

[0086] Please refer to Figure 1 , the present invention provides a fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network, including the following steps:

[0087] Step S1: Use the Spearman correlation coefficient method to perform a correlation analysis on the compressor operation data, and select variables with strong correlations for compressor modeling. The specific variables are shown in Table 1 below;

[0088] Table 1 Operating-related variables of a heavy-duty gas turbine compressor

[0089]

[0090] Please refer to Figure 2 , in step S1, use the Spearman correlation coefficient method to perform a correlation analysis on the selected compressor operation data in Table 1, and select variables with strong correlations for compressor modeling. The specific formula is:

[0091] ;

[0092] Wherein, is the total number of observation objects, is the rank difference of the i th observation object in two variables. If represents a perfect positive correlation, represents a perfect negative correlation, represents no monotonic correlation relationship.

[0093] Figure 2 The results show that there is a strong coupling relationship among the selected operating parameters under the same operating conditions, which needs to be paid special attention to in subsequent modeling or analysis.

[0094] Step S2: Collect operation data for preprocessing, specifically including collecting operation data of different measuring points in the compressor system of a heavy-duty gas turbine under normal and abnormal states, and preprocessing the missing values and outliers in the data. After preprocessing, the vibration data is divided into a training set and a test set.

[0095] Step S3: Use a liquid neural network as a time model to extract the time features of the compressor operation parameters;

[0096] Step S3 uses Figure 3 the liquid neural network shown as a time model to extract time features. Its core lies in using a dynamically changing time constant to control the state change rate of each neuron. This structure enables each neuron to adaptively change according to the input features at different time points. Its dynamic equation formula is as follows:

[0097] ;

[0098] where t is the time step, is the hidden state of the liquid time constant layer of the neuron; is the liquid time constant parameter vector, which controls the rate of neuron state update; is a neural network with as the parameter; is the exogenous input of the neural network; A is a bias vector.

[0099] The liquid time constant parameter not only determines the change rate of the hidden state but also is closely related to the input features. Specifically, the liquid time constant can be expressed by the following formula:

[0100] ;

[0101] where is the time-varying factor; is a neural network with as the parameter; is the exogenous input of the neural network, t is the time step.

[0102] The approximate closed-form solution of the liquid neural network can be expressed by the following formula:

[0103] ;

[0104] Shown as:

[0105] ;

[0106] whereB is the decomposed parameter vector; is the liquid time constant parameter vector; is a neural network with as the parameter; is the exogenous input of the neural network; A is a bias vector; is the Hadamard product.

[0107] Step S4: Calculate the attention weights of the interactions between the compressor operating parameters, perform normalization processing, then adopt the multi-head attention mechanism for feature aggregation, and finally output the spatial correlation, significantly improving the accuracy and robustness of the compressor fouling anomaly detection. The specific steps are as follows:

[0108] Step S41: Calculate the attention values between the compressor operating parameters. The formula is as follows:

[0109] ;

[0110] where, a represents the attention network; respectively represent nodes i and j ; W represents the weight sharing matrix.

[0111] Step S42: Perform normalization processing on the adjacent attention information of each node. The formula is as follows:

[0112] ;

[0113] where, represents the node i and j the attention coefficient between; represents the node i and j the attention value between; represents the node i and k the attention value between; represents the neighbor nodes of the node i .

[0114] Step S43: Adopt the multi-head attention mechanism for feature aggregation. The formula is as follows:

[0115] ;

[0116] where, represents the feature representation of the node i ; represents Output cascade of the head; Denote any activation function; Denote the Attention coefficient of the head; Denote the Weight matrix of the head; Denote the feature representation of node j .

[0117] Step S44: Output spatial correlation through the graph attention mechanism, and the formula is as follows:

[0118] ;

[0119] Among them, Denote the attention coefficient between node i and j ; Denote the neighbor nodes of node i ; Denote the activation function; Denote the single-layer feedforward neural network, T Denote the transpose; W Denote the shared weight matrix; Denote the feature representation of node i ; Denote the feature representation of node j ; Denote the feature representation of node k .

[0120] The schematic diagram of the graph attention network is as Figure 4 shown. By calculating the attention weights between each node and its neighbor nodes and using the multi-head attention mechanism to aggregate the neighbor node features, the dynamic feature weighted aggregation based on the attention coefficient is realized.

[0121] Step S5: Construct a spatio-temporal fusion model for training, calculate the multi-dimensional residual vector, and adaptively track the drift of the residual distribution during the compressor fouling process through the dynamic threshold algorithm. The specific steps are as follows:

[0122] Step S51: Input the data into the spatio-temporal fusion model for training by using the sliding window mechanism. The sliding window function is defined as follows:

[0123] ;

[0124] Among them, Denote the window length; Denote the selected monitoring parameter dimension; Denote the t observation vector at time

[0125] Input the current window data into the spatio-temporal fusion model , and generate a prediction sequence as shown in Figure 5 :

[0126] ;

[0127] Among them, represents the prediction length.

[0128] Figure 5 The results show that the model integrating the liquid neural network and the graph attention mechanism achieves accurate prediction of the compressor exhaust pressure and exhaust temperature, laying a model foundation for the subsequent quantitative evaluation of the compressor fouling degree.

[0129] Step S52: Refer to Figure 6 , calculate the multi-dimensional residual vector, and the formula is as follows;

[0130] ;

[0131] Among them, represents the covariance matrix of the normal data residuals in the training stage; represents the Mahalanobis distance metric.

[0132] Based on the historical normal operating condition residual samples , the formula for constructing the probability distribution model using kernel density estimation is as follows:

[0133] ;

[0134] Among them, the kernel function selects the Gaussian kernel, and the formula is as follows:

[0135] ;

[0136] The bandwidth matrix H is determined by the Silverman criterion, and the formula is as follows:

[0137] ;

[0138] Among them, represents the standard deviation estimate of the i -th dimension of the residuals; represents the standard deviation estimate of the j -th dimension of the residuals; Q represents the number of residual samples.

[0139] Figure 6 The results show that most of the residuals are concentrated near zero and show relatively obvious peaks, indicating that the predicted values are relatively close to the true values. The residual probability density distribution constructed by kernel density estimation is approximately a normal distribution, indicating that the prediction model is generally accurate.

[0140] Set the confidence level , and calculate the dynamic threshold. The formula is as follows:

[0141] ;

[0142] Step S53: When consecutive m windows satisfy , output the abnormal detection result of compressor fouling and give the abnormal confidence level. The formula is as follows:

[0143] ;

[0144] Among them, represents the sensitivity adjustment factor (default value 0.5).

[0145] Step S6: Observe whether the model tends to converge and whether the residual distribution deviates from the normal operating condition confidence interval. If the model converges and the residual distribution deviates from the normal operating condition confidence interval, output the abnormal detection result of compressor fouling; otherwise, repeat steps S3 - S5;

[0146] Refer to Figure 7 , which is the comparison of the residual distribution of the model described in the present invention with two improved LSTM models. Figure 7 The results show that the BiLSTM model has larger and more dispersed errors, while the xLSTM and the model described in the present invention have smaller and more concentrated errors. Compared with the xLSTM model, the median of the model described in the present invention is closer to 0 and the error range is smaller. Therefore, compared with other detection methods, the diagnostic accuracy of the present invention is higher.

[0147] Figure 8 The figure shows the structural schematic diagram of a certain type of single - shaft heavy - duty gas turbine. Compressor fouling is the main reason for the performance degradation of heavy - duty gas turbines. Studying its abnormal detection technology is of great significance.

[0148] It can be seen from this embodiment that this method realizes the dynamic update of feature information through a liquid neural network, strengthens the robustness and generalization ability of the model, uses the graph attention mechanism to output the spatial correlation between operating parameters, and constructs a dynamic threshold based on a multi - dimensional residual vector, improving the diagnostic accuracy of the model.

[0149] Therefore, the present invention adopts the above - mentioned method for detecting compressor fouling of heavy - duty gas turbines based on a liquid neural network. Through the dynamic topological structure of the liquid neural network and combined with the spatio - temporal feature collaborative mining mechanism, it realizes the quantitative evaluation of the degree of compressor fouling and improves the accuracy and robustness of abnormal detection of compressor fouling.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions of the present invention or make equivalent replacements, and these modifications or equivalent replacements do not enable the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A fouling detection method for a compressor of a heavy-duty gas turbine based on a liquid neural network, characterized in that It includes the following steps: Step S1: Use the Spearman correlation coefficient method to perform correlation analysis on the compressor operation parameters to obtain characteristic data, and select the characteristic data with strong correlation for compressor modeling; Step S2: Preprocess the characteristic data with strong correlation and divide it into a training set and a test set; Step S3: Use a liquid neural network as a time model to extract the time characteristics of the compressor operation parameters; Step S4: Calculate the weights of the interactions between the extracted compressor operation parameters, and use the graph attention mechanism to output spatial correlation; Step S5: Construct a spatio-temporal fusion model for training, calculate the multi-dimensional residual vector, and construct a dynamic threshold function; Step S6: Observe whether the spatio-temporal fusion model tends to converge and whether the residual distribution deviates from the normal operating condition confidence interval. If the spatio-temporal fusion model converges and the residual distribution deviates from the normal operating condition confidence interval, output the abnormal detection result of compressor fouling, otherwise repeat steps S3 - S5; In step S1, the method of using the Spearman correlation coefficient method to perform correlation analysis on the compressor operation parameters is calculated according to the following formula: wherein, is the total number of observation objects, is the rank difference of the i th observation object in two variables. If represents a perfect positive correlation, represents a perfect negative correlation, represents no monotonic correlation relationship; In step S3, using a liquid neural network as a time model to extract the time characteristics of the compressor operation parameters controls the state change rate of each neuron by using a dynamically changing time constant. This structure enables each neuron to adaptively change according to the input characteristics at different time points. The specific formula is as follows: wherein, t is the time step, is the hidden state of the liquid time constant layer of the neuron; is the liquid time constant parameter vector that controls the rate of neuron state update; is a neural network parameterized by ; is the exogenous input of the neural network; A is a bias vector; Liquid time constant which can be expressed by the following formula: Among them, is a time-varying factor; is a neural network with as a parameter; is the exogenous input of the neural network; The approximate closed-form solution of the liquid neural network can be expressed by the following formula: It is shown as: Among them, B is the decomposed parameter vector; is the liquid time constant parameter vector; is a neural network with as the parameter; is the exogenous input of the neural network; A is a bias vector; is the Hadamard product.

2. The fouling detection method for the compressor of a heavy-duty gas turbine based on a liquid neural network according to claim 1, wherein Step S4 includes: Step S41: Calculate the attention values between the compressor operation parameters; Step S42: Normalize the adjacent attention information of each node; Step S43: Use the multi-head attention mechanism for feature aggregation; Step S44: Output the spatial correlation through the graph attention mechanism.

3. The method for detecting fouling of a heavy-duty gas turbine compressor based on a liquid neural network according to claim 2, wherein The formula for outputting the spatial correlation through the graph attention mechanism in step S44 is as follows: Among them, represents the attention coefficient between nodes i and j ; represents the neighbor nodes of node i ; represents the activation function; represents a single-layer feedforward neural network, T represents the transpose; W represents the shared weight matrix; represents the feature representation of node i ; represents the feature representation of node j ; represents the feature representation of node k ; 4. The fouling detection method for a heavy-duty gas turbine compressor based on a liquid neural network according to claim 3, characterized in that Step S5 includes: Step S51: Input the data into the spatio-temporal fusion model for training using the sliding window mechanism; Step S52: Calculate the multi-dimensional residual vector, and construct a dynamic threshold function based on the historical normal operating condition residual samples; Step S53: Set the trigger condition for outputting the abnormal detection result of compressor fouling.

5. The method for detecting fouling of a heavy-duty gas turbine compressor based on a liquid neural network according to claim 4, wherein The function definition of the sliding window mechanism in step S51 is as follows: Among them, represents the window length; represents the selected monitoring parameter dimension; represents t the observation vector at the moment; Input the current window data into the spatio-temporal fusion model , and generate a prediction sequence: Among them, represents the prediction length.

6. The method for detecting fouling of a heavy-duty gas turbine compressor based on a liquid neural network according to claim 5, wherein The expression formula for calculating the multi-dimensional residual vector in step S52 is: Among them, represents the covariance matrix of the normal data residuals in the training phase; represents the Mahalanobis distance metric; Based on historical normal operating condition residual samples The formula for constructing the dynamic threshold function is as follows: Among them, the kernel function selects the Gaussian kernel, and the formula is as follows: Bandwidth matrix H Determined by the Silverman criterion: Among them, represents the standard deviation estimate of the i -dimensional residual; represents the standard deviation estimate of the j -dimensional residual; Q represents the number of residual samples; Set the confidence level , calculate the dynamic threshold, and the formula is as follows: Among them, represents the inverse function of the cumulative distribution function.

7. The fouling detection method for the compressor of a heavy-duty gas turbine based on a liquid neural network according to claim 6, characterized in that The triggering condition for outputting the detection result of compressor fouling abnormality in step S53 is: when m consecutive windows satisfy the following formula: Output the abnormal detection result of compressor fouling and give the abnormal confidence level to represent the confidence interval: Among them, represents a sensitivity adjustment factor.

Images