On-line monitoring method for bearing state of induced draft fan of coal-fired unit based on Di-ICA

Abstract

The invention discloses a Di-ICA-based online monitoring method for the bearing state of an induced draft fan of a coal-fired unit. The online monitoring method comprises the following steps: acquiring working condition data of the bearing of the induced draft fan through a sensor arranged at the bearing of the induced draft fan of the coal-fired unit; extracting dynamic characteristics of the working condition data, and obtaining a residual error part of the working condition data; performing data modeling on the residual error part of the working condition data through an ICA method; on-line monitoring of working condition data is realized through data modeling and based on Di-ICA; and fault separation of dynamic condition working condition data is realized based on LARS reconstruction.

Description

Technical Field

[0001] The present invention relates to the technical field of coal-fired power generation treatment, and particularly relates to an online monitoring method for the bearing state of a forced draft fan of a coal-fired unit based on Di-ICA. Background Art

[0002] As an important auxiliary equipment of a thermal power plant, the operation of a forced draft fan has a direct impact on the normal and safe operation of a thermal power unit. Large thermal power units are equipped with DCS and SIS systems, which can measure the parameters of relevant measuring points of the forced draft fan operation in real time. The traditional DCS system generally sets multi-level alarm limits based on the unit design value or the thermal performance calculation method to realize the online monitoring and fault alarm of the operation state of the forced draft fan. Although this method can play a certain warning role, it is difficult to predict the occurrence of forced draft fan faults in advance.

[0003] Therefore, researchers have introduced data-driven modeling technologies such as PCA and deep learning, established a model between the relevant parameters of the forced draft fan operation based on historical data, and then judged whether the bearing of the forced draft fan is normal by monitoring the change of parameter correlation. This type of method can effectively identify the fault symptoms of the forced draft fan bearing and avoid the further deterioration of the fault. Due to factors such as the change of operation load, there are some problems in the operation process of the forced draft fan. At this time, it is difficult to effectively diagnose the faults of the forced draft fan bearing under dynamic characteristics by using traditional feature extraction methods. Therefore, fully considering the dynamic characteristics of the forced draft fan operation is of great significance for reducing model false alarms and missed alarms and improving monitoring effectiveness. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the purpose of the present invention is to provide an online monitoring method for the bearing state of a forced draft fan of a coal-fired unit based on Di-ICA to solve the problem of how to monitor the bearing of a forced draft fan of a coal-fired unit in the prior art.

[0005] The present invention first provides an online monitoring method for the bearing state of a forced draft fan of a coal-fired unit, and the online monitoring method includes:

[0006] Obtaining the working condition data of the forced draft fan bearing through a sensor arranged at the forced draft fan bearing of the coal-fired unit;

[0007] Extracting the dynamic characteristics of the working condition data and obtaining the residual part of the working condition data;

[0008] Performing data modeling on the residual part of the working condition data through the ICA method;

[0009] Realizing the online monitoring of the working condition data through data modeling and based on Di-ICA;

[0010] Realizing the fault separation of the working condition data under dynamic conditions based on LARS reconstruction.

[0011] Preferably, for extracting the dynamic characteristics of the working condition data, obtaining the residual part of the working condition data includes the following steps:

[0012] Ⅰ. Define the training data as X train , the lag order as s, and the number of dynamic latent variables as l;

[0013] Ⅱ. Standardize X train and randomly initialize the weight coefficient w;

[0014] Ⅲ. Iteratively extract the dynamic latent variables:

[0015] ⅰ. Calculate the latent variable t through the calculation formula i , where i = 1, 2, …, s + 1;

[0016] ⅱ. Calculate β = [t 1 t 2 … t s T t s+1 ;

[0017] ⅲ. Update and calculate

[0018] ⅳ. Normalize w = w / ‖w‖ and β = β / ‖β‖;

[0019] ⅴ. When Δ‖w‖ ≤ 10 -6 , convergence is achieved and the iteration stops; otherwise, return to step ⅰ to continue the iteration;

[0020] Ⅳ. Decompose the data X T through the calculation formula X := X - tp train ;

[0021] Ⅴ. Repeat steps Ⅲ and Ⅳ until l dynamic latent variables are completely extracted to obtain the residual part of the working condition data.

[0022] Preferably, for data modeling of the residual part of the working condition data by the ICA method, it includes the following steps:

[0023] Ⅰ. Predict and for T s+1 to obtain

[0024] Ⅱ. Calculate E through the calculation formula S+1 ;

[0025] Ⅲ. Standardize and whiten E s+1 to obtain the whitened data Z;​

[0026] Ⅳ. Define the number of independent principal components as q, let i = 1, and extract independent components through iteration:

[0027] ⅰ. According to the calculation formula initialize the i-th column c n of C n,i ;

[0028] ⅱ. Use the Newton iteration method to solve where g and g' are the first and second derivatives of the G function respectively;

[0029] ⅲ. According to the calculation formula orthogonalize c n,i ;

[0030] ⅳ. According to the calculation formula c n,i ←c n,i / ‖c n,i ‖normalize c n,i ;

[0031] ⅴ. If |c n,i T c n,i |→1, then output c n,i , otherwise return to step ⅱ;

[0032] ⅵ. According to the calculation formula calculate y i ;

[0033] Ⅴ. If i < q, then let i = i + 1, and return to step Ⅳ to solve the next independent component;

[0034] Ⅵ. Calculate the demixing matrix and the mixing matrix A according to the calculation formulas and ;

[0035] Ⅶ. Calculate the reconstructed value of the residual matrix according to the calculation formula ;

[0036] Ⅷ. Calculate the monitoring indices T 2 = y T D -1 y and calculate the monitoring indices T 2 and SPE of the working condition data;

[0037] Preferably, the control limits τ 2 and δ 2 of the monitoring indices T 2 and SPE of the working condition data are calculated using the following formulas:

[0038]

[0039] In Formulas (I) and (II): m is the number of samples, σ is the standard deviation of the sampled data.

[0040] Preferably, the online monitoring of working condition data through data modeling and based on Di-ICA includes the following steps:

[0041] I. Define the test sample at the current k-th moment as x k and the lagged historical data as X test = [x k x k-1 … x k-s ;

[0042] II. Standardize X test ;

[0043] III. Calculate and extract the dynamic latent variable of x at the current moment through the calculation formula k ;

[0044] IV. Calculate the residual vector e through the calculation formula k ;

[0045] V. Standardize and whiten e k to obtain the whitened data z k ;

[0046] VI. Calculate the independent components of e through the demixing matrix k and the mixing matrix A and the data z k ;

[0047] VII. Calculate the monitoring indicators T 2 (k) = y(k) T D -1 y(k) and of the working condition data at the current k-th moment through 2 ;

[0048] VIII. Compare T 2 (k) and SPE(k) with τ 2 and δ 2 for comparison and judgment. When T 2 (k) > τ 2 or SPE(k) > δ 2 , a fault warning is issued.

[0049] Preferably, the fault separation of dynamic condition data based on LARS reconstruction includes the following steps:

[0050] Ⅰ. Let i = 1, the current data is x k-i ;

[0051] II. Definition and Through formula (III) and calculation formula Initialize β = Ξ k-i f k-i ,

[0052] Wherein, formula (III) is:

[0053]

[0054] III. Use LARS to solve the sparsity of formula (III) and obtain the fault direction Ξ k-i and fault amplitude f k-i ;

[0055] IV. By calculating the formula x k-i * =x k-i -Ξ k-i f k-i For the current data x k-i make corrections;

[0056] Ⅴ. When i <s,则令i=i+1,返回步骤Ⅱ;

[0057] VI. Calculate formula (IV) through LARS to obtain x k Fault direction k and fault amplitude f k ,

[0058] Wherein, formula (IV) is:

[0059]

[0060] Ⅶ. Determine the fault parameters and analyze the fault causes;

[0061] VIII. Fault diagnosis is completed.

[0062] The present invention also provides a Di-ICA-based online monitoring system for bearing status of an induced draft fan of a coal-fired unit, the system comprising:

[0063] A first calculation unit is used to extract the dynamic characteristics of the operating condition data and obtain the residual part of the operating condition data;

[0064] A second calculation unit for performing data modeling on the residual part of the operating condition data by means of the ICA method;

[0065] A third calculation unit for performing online monitoring of the operating condition data through data modeling and based on Di-ICA;

[0066] A fourth calculation unit for performing fault separation of the dynamic condition operating condition data based on LARS reconstruction.

[0067] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the online monitoring method for the bearing state of the induced draft fan of the coal-fired unit based on Di-ICA described above is implemented.

[0068] The present invention also provides a computer-readable storage medium storing a computer program for executing the online monitoring method for the bearing state of the induced draft fan of the coal-fired unit based on Di-ICA described above.

[0069] The online monitoring method for the bearing state of the induced draft fan of the coal-fired unit based on Di-ICA provided by the present invention has the following advantages:

[0070] 1. It can reasonably handle the dynamic characteristics existing in the operating data of the induced draft fan bearing of the coal-fired unit, can more accurately reconstruct abnormal data and locate the position of abnormal variables, greatly enhancing the prevention and diagnosis capabilities of the induced draft fan bearing faults under operating conditions such as deep peak shaving of the unit, and improving the economy and safety of the power plant operation.

[0071] 2. The monitoring method provided by the present invention has a fast calculation speed, enabling the diagnostic results to be very conveniently and quickly fed back to the system, truly playing a role in real-time operation monitoring and fault diagnosis, and ensuring the effectiveness of real-time data. Description of the Drawings

[0072] Figure 1 It is a schematic flowchart of the online monitoring method for the bearing state of the induced draft fan of the coal-fired unit based on Di-ICA provided by the present invention. Detailed Embodiments

[0073] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0074] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "top part", "bottom part", "top surface", "bottom surface", "inner", "outer", "inner side", "outer side", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention.

[0075] In the description of the present invention, the meaning of "several" is one or more, the meaning of "multiple" is two or more, and understandings such as "greater than", "less than", "exceeding", etc. do not include the present number, and understandings such as "above", "below", "within", etc. include the present number. If the terms "first", "second", "third" are described, they are only for the purpose of description and distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence relationship of the indicated technical features.

[0076] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", "connected to", "set" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations. The embodiments of the present invention will be described below according to its overall structure.

[0077] Figure 1 The flow chart of the online monitoring method for the bearing state of the induced draft fan of a coal-fired unit based on Di-ICA provided by the present invention is shown in Figure 1 An online monitoring method for the bearing state of the induced draft fan of a coal-fired unit based on Di-ICA, the online monitoring method includes: obtaining the working condition data of the induced draft fan bearing through a sensor arranged at the induced draft fan bearing of the coal-fired unit; extracting the dynamic characteristics of the working condition data to obtain the residual part of the working condition data; performing data modeling on the residual part of the working condition data through the ICA method; realizing the online monitoring of the working condition data through data modeling and based on Di-ICA; and realizing the fault separation of the dynamic condition working condition data based on LARS reconstruction.

[0078] Among them, Di-ICA refers to constrained independent component analysis, which is an extension and improvement method of independent component analysis (ICA). When Di-ICA processes data, it can improve the accuracy of data separation and extract specific components.

[0079] LARS is an algorithm for sparse regression of high-dimensional data. In the present invention, when processing working condition data, LARS is used to reconstruct the sparse representation of the data, and the data is sparsely encoded by the LARS algorithm, so as to extract key features and reduce the data dimension.

[0080] In the present invention, a large amount of operation history data is collected through parameter measuring points installed at the induced draft fan bearings, and dynamic latent variables in the process are extracted by using a dynamic feature extraction method. Independent and non-Gaussian latent variables (referred to as independent components, ICs) are extracted during the extraction process to solve the non-Gaussian characteristics in the process. ICA is used to model the fault monitoring of the non-Gaussian residual part, and a fault separation method based on the reconstruction of LARS is used to achieve fault separation under dynamic conditions. The system and method provided by the present invention can perform online monitoring and fault diagnosis on the state of the induced draft fan bearings in a thermal power plant, effectively process non-Gaussian distributed data under the dynamic characteristics of the induced draft fan operation, quickly confirm the fault type and provide an alarm, and provide operation and maintenance opinions for operators and maintenance personnel.

[0081] Wherein, the residual part of the working condition data is the working condition data after extracting the dynamic characteristics.

[0082] Refer to Figure 1 , the online monitoring method for the state of the induced draft fan bearings of a coal-fired unit based on Di-ICA provided by the present invention includes:

[0083] 1. Offline training

[0084] 1.1 Extract the dynamic characteristics of the working condition data and obtain the residual part of the working condition data

[0085] Wherein, extracting the dynamic characteristics of the working condition data and obtaining the residual part of the working condition data specifically includes the following steps:

[0086] Ⅰ. Define the training data as X train , the lag order is s and the number of dynamic latent variables is l.

[0087] Ⅱ. Standardize X train and randomly initialize the weight coefficient w.

[0088] Ⅲ. Extract dynamic latent variables through iteration:

[0089] ⅰ. Calculate the latent variable t through the calculation formula i , where i = 1, 2, …, s + 1. Wherein, k is the time point, representing the sample point collected at a certain moment.

[0090] ⅱ. Calculate β = [t 1 t 2 … t s Tt s+1 。

[0091] iii. Update the calculation

[0092] iv. Normalize w = w / ‖w‖ and β = β / ‖β‖.

[0093] v. When Δ‖w‖ ≤ 10 -6 , convergence is achieved and the iteration stops; otherwise, return to step i and continue the iteration.

[0094] IV. Decompose the data X through the calculation formula X := X - tp T for the data X train . Here, := represents iterative solution, indicating that the new X is obtained from the previous X - tp T . P represents the load vector. Among them, P = X T t / t T t.

[0095] V. Repeat steps III and IV until l dynamic latent variables are extracted, obtaining the residual part of the operating condition data. l represents the number of retained dynamic latent variables, extracted from the 1st to the lth.

[0096] 1.2 Perform data modeling on the residual part of the operating condition data through the ICA method

[0097] Among them, performing data modeling on the residual part of the operating condition data through the ICA method includes the following steps:

[0098] I. Through the calculation formulas and predict T s+1 to obtain

[0099] II. Calculate E through the calculation formula . s+1

[0100] III. Standardize and whiten E s+1 to obtain the whitened data Z. Among them, the data Z is a matrix.

[0101] IV. Define the number of independent principal components as q, let i = 1, and extract independent components through iteration.

[0102] i. Initialize the i-th column c of C n according to the calculation formula n,i .

[0103] ii. Solve using the Newton iteration method, where g and g' are the first and second derivatives of the G function respectively. Among them, E is the disability matrix.​

[0104] iii. Orthogonalize c according to the calculation formula for c n,i

[0105] iv. Standardize c according to the calculation formula c n,i ← c n,i / ‖c n,i ‖ for c n,i

[0106] v. If |c n,i T c n,i | → 1, then output c n,i , otherwise return to step ii.

[0107] vi. Calculate y according to the calculation formula i .

[0108] V. If i < q, then set i = i + 1 and return to step IV to solve for the next independent component.

[0109] vii. Calculate the demixing matrix and the mixing matrix A through the calculation formulas and . D is a diagonal matrix composed of eigenvalues, U is the eigenvector, and Λ is the eigenvalue matrix.

[0110] vii. Calculate the reconstructed value of the residual matrix through the calculation formula .

[0111] viii. Calculate the monitoring indices T 2 = y T D -1 y and calculate the monitoring indices T 2 and SPE of the operating condition data.

[0112] 1.3 Calculation of the control limits τ 2 and δ 2 of the monitoring indices T 2 and SPE of the operating condition data

[0113] Among them, the calculation of the control limits τ 2 and δ 2 of the monitoring indices T 2 and SPE of the operating condition data are all calculated using the following formulas:

[0114]

[0115] In formulas (I) and (II): m is the number of samples, ​​​σ is the standard deviation of the sampled data, where n represents the number of columns of the data.

[0116] Among them, formula (Ⅰ) is the probability density function, and formula (Ⅱ) is the cumulative distribution function.

[0117] K(u) is the kernel function. Here, the standard Gaussian kernel function, the standard Gaussian (normal) distribution function, is used to calculate the kernel density estimate. h is the smoothing coefficient, which is used to control the width of the kernel function and affects the smoothness of the estimate. H(u) is the cumulative kernel function, which is the integral of the kernel function K(u). σ is the standard deviation of the sampled data, which is used to measure the degree of dispersion of the data.

[0118] Among them, x and x in formulas (Ⅰ) and (Ⅱ) i are variables in the general expression of the probability density function.

[0119] 2. Online monitoring

[0120] Online monitoring of the working condition data through data modeling and based on Di-ICA includes the following steps:

[0121] Ⅰ. Define the test sample at the current k-th moment as x k and the lagged historical data as X test = [x k x k-1 … x k-s .

[0122] Ⅱ. Standardize X test for processing.

[0123] Ⅲ. Calculate and extract the dynamic latent variable of x at the current moment through the calculation formula k

[0124] Among them, G is the dynamic model transfer function, and q -1 is the backward shift operator.

[0125] Ⅳ. Calculate the residual vector e through the calculation formula k .

[0126] Ⅴ. Standardize and whiten e k to obtain the whitened data z k .

[0127] Ⅵ. Calculate the independent components of e through the demixing matrix k and the mixing matrix A and the data z k .

[0128] Ⅶ. Through the calculation formula T 2(k) = y(k) TD-1 y(k) and Calculate the monitoring index T of the working condition data at the current k-th moment 2 (k) and SPE(k).

[0129] Ⅷ. Compare T 2 (k) and SPE(k) with τ 2 and v 2 for comparison and judgment. When T 2 (k) > τ 2 or SPE(k) > δ 2 occurs, a fault warning is issued.

[0130] 3. Fault isolation

[0131] The fault isolation of the dynamic condition working condition data based on LARS reconstruction includes the following steps:

[0132] Ⅰ. Let i = 1, and the current data is x k-i ;

[0133] Ⅱ. Define and Initialize β = Ξ through formula (Ⅲ) and the calculation formula f k-i is an empty matrix, and P is the loading matrix. C k-i C = D, where D is a diagonal matrix composed of eigenvalues. Through C T C = D, C can be calculated. T C = D can calculate C.

[0134] Among them, formula (Ⅲ) is:

[0135]

[0136]

[0137] Among them, M = VV T , where M is a symmetric positive definite matrix, and the V in formula (Ⅲ) can be obtained from M = VV T calculated.

[0138] Ⅲ. Perform sparse solution on formula (Ⅲ) through LARS to obtain the fault direction Ξ k-i and the fault amplitude f k-i . What is obtained is the fault direction Ξ k-i of the current data x k-i and the fault amplitude f k-i .

[0139] Ⅳ. Through the calculation formula x k-i* = x k-i - Ξ k-i f k-i For the current data x k-i perform correction.

[0140] Ⅴ. When i < s, let i = i + 1 and return to Step Ⅱ.

[0141] Ⅵ. Calculate formula (Ⅳ) through LARS to obtain the fault direction Ξ k of x k and the fault amplitude f k .

[0142] Among them, formula (Ⅳ) is:

[0143]

[0144] Among them, in formula (Ⅳ), R = W(P T W) -1 .

[0145] Ⅶ. Determine the fault parameters and analyze the cause of the fault.

[0146] Ⅷ. The fault diagnosis ends.

[0147] Refer to Figure 1 , when performing fault isolation, use the LARS algorithm to select the fault variables, then determine the fault direction and fault amplitude of the fault variables, judge whether the reconstruction index is less than the limit value. When the reconstruction index is less than the limit value, determine the number of fault parameters, the fault direction and the fault amplitude, and then correct the sample data. If the reconstruction index is greater than the limit value, re-select the fault variables, update the fault direction and the fault amplitude, and then judge the reconstruction index.

[0148] An embodiment of the present invention also provides an on-line monitoring system for the bearing state of a forced draft fan of a coal-fired unit based on Di-ICA. The system includes: a first calculation unit for extracting the dynamic characteristics of the working condition data and obtaining the residual part of the working condition data; a second calculation unit for performing data modeling on the residual part of the working condition data through the ICA method; a third calculation unit for realizing on-line monitoring of the working condition data through data modeling and based on Di-ICA; a fourth calculation unit for realizing fault isolation of the dynamic condition working condition data based on LARS reconstruction.

[0149] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned on-line monitoring method for the bearing state of a forced draft fan of a coal-fired unit based on Di-ICA.

[0150] An embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program for executing the online monitoring method for the bearing state of the induced draft fan of the coal-fired unit based on Di-ICA as described above.

[0151] In summary, in order to meet the requirements of real-time operation monitoring and fault diagnosis of the bearing faults of the induced draft fan of the coal-fired unit, the present invention proposes a new online monitoring and fault diagnosis method for the bearing of the induced draft fan based on data classification and data-driven modeling, which can perform online monitoring and fault diagnosis on the bearing state of the induced draft fan of the thermal power plant, effectively process the non-Gaussian distributed data under the dynamic characteristics of the operation of the induced draft fan, quickly confirm the fault type and provide an alarm, and provide opinions on operation and maintenance for the operators and maintenance personnel.

[0152] The above are only specific embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for online monitoring of the bearing status of an induced draft fan of a coal-fired unit based on Di-ICA, characterized in that: The online monitoring method includes: Obtaining the operating condition data of the induced draft fan bearing through a sensor disposed at the induced draft fan bearing of the coal-fired unit; Extracting the dynamic characteristics of the operating condition data to obtain the residual part of the operating condition data; Performing data modeling on the residual part of the operating condition data by the ICA method; Realizing online monitoring of the operating condition data through data modeling and based on Di-ICA; Realizing fault separation of the operating condition data under dynamic conditions based on LARS reconstruction.

2. The method for online monitoring of bearing status of induced draft fan of coal-fired unit based on Di-ICA ​​according to claim 1, characterized in that: The extracting the dynamic characteristics of the operating condition data to obtain the residual part of the operating condition data includes the following steps: I. Define the training data as X train , the lag order is s and the number of dynamic latent variables is l; II. X train Standardize and randomly initialize the weight coefficient w; Ⅲ. Extracting dynamic latent variables through iteration: ⅰ. By calculation Calculate the latent variable t i , where i = 1, 2, ..., s + 1; ii. Calculate β = [t1 t2 … t s ] T t s+1 ; ⅲ. Update calculation ⅳ. Normalizing w = w / ‖w‖ and β = β / ‖β‖; ⅴ, when Δ‖w‖≤10 -6 , reach convergence and stop iteration, otherwise return to step i and continue iteration; IV. By calculating the formula x: = x-tp T For data X train To decompose; Ⅴ. Repeating steps Ⅲ and Ⅳ until l dynamic latent variables are extracted, obtaining the residual part of the operating condition data.

3. The method for online monitoring of bearing status of induced draft fan of coal-fired unit based on Di-ICA ​​according to claim 2, characterized in that: The performing data modeling on the residual part of the operating condition data by the ICA method includes the following steps: Ⅰ. Through calculation formula and To T s+1 Make a prediction II. Through calculation formula Calculate E s+1 ; III. E s+1 Perform standardization and whitening processing to obtain the whitened data Z; Ⅳ. Defining the number of independent principal components as q, setting i = 1, and extracting independent components through iteration: ⅰ. According to the calculation formula C n Column i c n,i Perform initialization processing; ⅱ. Solve using Newton iteration method Where g and g' are the first and second order derivatives of the G function respectively; ⅲ. According to the calculation formula C n,i Perform orthogonalization processing; ⅳ. According to the calculation formula c n,i ←cn ,i / ‖c n,i ‖ for c n,i Standardize the process; ⅴ, if |c n,i T c n,i |→1, then output c n,i , otherwise return to step ii; ⅵ. By calculation formula According to the calculation of y i ; Ⅴ. If i < q, then setting i = i + 1 and returning to step Ⅳ to solve the next independent component; VI. By calculation formula and Calculate the unmixing matrix and the mixing matrix A; VII. By calculation formula Calculate the residual matrix reconstruction value; VIII. By calculating the formula T 2 =y T D -1 y and Calculate the monitoring index T of the working condition data 2 and SPE.

4. The method for online monitoring of bearing status of induced draft fan of coal-fired unit based on Di-ICA ​​according to claim 3 is characterized in that: The monitoring index T of the working condition data 2 and the control limit τ of SPE 2 and δ 2 The calculations are performed using the following formula: In formulas (I) and (II), m is the number of samples, σ is the standard deviation of the sampled data.

5. The method for online monitoring of bearing status of induced draft fan of coal-fired unit based on Di-ICA ​​according to claim 4, characterized in that: The realizing online monitoring of the operating condition data through data modeling and based on Di-ICA includes the following steps: I. Define the test sample at the current k moment as x k and lagged historical data as X test =[x k x k-1 … x k-s ]; II. X test Standardize the process; III. By calculation Calculate and extract the current time x k Dynamic latent variables IV. Through calculation formula Calculate the residual vector e k ; Ⅴ. For e k Perform standardization and whitening to obtain the whitened data z k ; VI. Through the unmixing matrix and the mixing matrix A and the data z k Calculate e k independent components; VII. By calculating the formula T 2 (k) = y(k) T D -1 y(k) and Calculate the monitoring index T of the current working condition data at time k 2 (k) and SPE(k); VIII. T 2 (k) and SPE(k) with τ 2 and δ 2 Compare and judge. When T appears 2 (k)>τ 2 Or SPE(k)>δ 2 When the fault occurs, an early warning is issued.

6. The method for online monitoring of bearing status of induced draft fan of coal-fired unit based on Di-ICA ​​according to claim 5, characterized in that: The realizing fault separation of the operating condition data under dynamic conditions based on LARS reconstruction includes the following steps: Ⅰ. Let i=1, the current data is x k-i ; II. Definition and Through formula (III) and calculation formula Initialize β = Ξ k-i f k-i , Wherein, formula (Ⅲ) is: Ⅲ. Use LARS to solve the sparsity of formula (Ⅲ) and obtain the fault direction Ξ k-i and fault amplitude f k-i ; IV. By calculating the formula x k-i * =x k-i -Ξ k-i f k-i For the current data x k-i make corrections; Ⅴ. When i < s, then setting i = i + 1 and returning to step Ⅱ; VI. Calculate formula (IV) through LARS to obtain x k Fault direction k and fault amplitude f k , Wherein, formula (Ⅳ) is: Ⅶ. Determining the fault parameters and analyzing the cause of the fault; Ⅷ. Ending the fault diagnosis.

7. An online monitoring system for the bearing status of the induced draft fan of a coal-fired unit based on Di-ICA, characterized in that: The system includes: A first calculation unit for extracting the dynamic characteristics of the operating condition data to obtain the residual part of the operating condition data; A second calculation unit for performing data modeling on the residual part of the operating condition data by the ICA method; A third calculation unit for realizing online monitoring of the operating condition data through data modeling and based on Di-ICA; A fourth calculation unit for realizing fault separation of the operating condition data under dynamic conditions based on LARS reconstruction.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it realizes the online monitoring method for the state of the induced draft fan bearing of the coal-fired unit based on Di-ICA according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the online monitoring method for the state of the induced draft fan bearing of the coal-fired unit based on Di-ICA according to any one of claims 1 to 6.

Images